mice experiment in water

Drowning Rats Psychology Experiment: Resilience and the Power of Hope

In the 1950s, Curt Richter, a professor at Johns Hopkins, did a famous drowning rats psychology experiment. This experiment, though cruel, demonstrated the power of hope and resilience in overcoming difficult situations. Summary by The World of Work Project

A Psychology Experiment: Drowning Rats

In a series of experiments that are fairly cruel and unpalatable, yet interesting in their findings, Curt Richter demonstrated that hope is a powerful factor in perseverance. In our view, this is also closely linked with resilience .

The Drowning Rats Psychology Experiments

Curt’s experiments focused on how long it takes rats to die from drowning. He conducted his experiments by placing rats into buckets filled with water and seeing how long they survived. He introduced a range of variables into the experiment, that yielded some interested results.

Domesticated Rats

12 domesticated rats were used in Curt’s first set of experiments. The first of these rats initially swam around the surface, then dove to the bottom of the bucket and explored what was there for a while. It lasted a total of two minutes before it drowned.

Two of the other domesticated rats did roughly the same thing, and survived for roughly the same period of time.

The other nine domesticated rats though did something completely different. After an initial exploration, the predominantly spent their time and the surface. And the just kept swimming. They survived for literally days before eventually succumbing to exhaustion and drowning.

The second set of experiments Curt undertook involved 34 wild rats. Wild rats are excellent swimmers, and these savage and aggressive ones had only recently been caught. Obviously, Curt expected them to fight hard for their survival.

Surprisingly though, this wasn’t the case at all. Despite their ferocity, fitness and swimming ability, not one of the 34 wild rats survived more than a few minutes.

The Role of Hope

Curt reflected on what caused some of the rats to give up and decided that hope a key factor in the willingness to struggle on. Where rats have perhaps been helped in the past and have hope of being saved, they will keep fighting in the believe that all is not lost. However, when they don’t have this prior experience, they will give up quickly.

In his own words he said: “ The situation of these rats scarcely seems one demanding fight or flight — it is rather one of hopelessness… the rats are in a situation against which they have no defense… they seem literally to ‘give up.’ ”

With this in mind, Curt decided to experiment further.

Introducing Hope and Support

The last set of experiments that we’ll focus on were concerned with the impact that introducing hope would have on the perseverance of the rats in buckets. In these experiments Curt’s hypothesis was roughly that introducing hope to rats would increase their survival times.

To test his hypothesis Curt selected a new cohort of rats who were all similar to each other. Again, he introduced them into buckets and observed them as they progressed towards drowning. This time though, he noted the moment at which they gave up then, just before they died, he rescued them. He saved them, held them for a while and helped them recover.

He then placed them back into the buckets and started the experiments all over again. And he discovered that his hypothesis was right. When the rats were placed back into the water they swam and swam, for much longer than they had the first time they were placed in the buckets. The only thing that had changed was that they had been saved before, so had hope this time.

Curt wrote that “ the rats quickly learn that the situation is not actually hopeless ” and that “ after elimination of hopelessness the rats do not die .”

What This May Mean For People

Humans and rats are very different beings, but there is still a belief that we can learn a lot from these experiments. Where individuals have hope, they have higher levels of perseverance. They will keep fighting when they feel these is a chance of success or rescue. When they don’t have hope, they won’t.

A range of other experiments have also supported this.

What This Means in the World of Work

From a work perspective, these findings can be taken to mean that people will remain resilient and will continue to persevere in the face of difficult situations, provided they have hope.

So, if they are rescued from time to time. If they are supported. If they believe the future will be a better place and if they feel others are there to help them, they may be able to drive themselves through difficult situations. The importance of belief here is similar to the importance of belief in the expectancy theory of motivation .

What this means for leaders is that people in your team will be strong and resilient, provided that you give them hope of a better future. If that hope is extinguished, your people will stop fighting for you.

Some Specific Quotes from Richter

“The situation of these rats scarcely seems one demanding fight or flight—it is rather one of hopelessness; whether they are restrained in the hand or confined in the swimming jar, the rats are in a situation against which they have no defense. This reaction of hopelessness is shown by some wild rats very soon after being grasped in the hand and prevented from moving; they seem literally to ‘give up’. Support for the assumption that the sudden death phenomenon depends largely on emotional reactions to restraint or immersion comes from the observation that after elimination of the hopelessness the rats do not die. This is achieved by repeatedly holding the rats briefly and then freeing them, and by immersing them in water for a few minutes on several occasions. In this way the rats quickly learn that the situation is not actually hopeless; thereafter they again become aggressive, try to escape, and show no signs of giving up. Wild rats so conditioned swim just as long as domestic rats or longer.” You can find these comments on p196 of this pdf .

Learning More

Our resilience can be an important factor in our Wellbeing in the workplace. It’s a bit of a difficult concept to pin down, but we can get a sense of how resilient we are with the Brief Resilience Scale .

There are steps we can take to improve our own wellbeing . Improving our self-awareness might also help us improve our wellbeing. Similarly, learning about different types of stress and how to manage stress can be helpful. The below podcast covers the concept of stress-buckets, which might of interest.

The World of Work Project View

These drowning rats psychology experiments are clearly abhorrent, as is most animal testing. We know that the findings of many experiments do not translate to humans. In fact, experiments of this nature are still being used by several organizations. This should stop. A good starting point for finding out which organizations still use this form of testing so that you can avoid their products is this article by PETA .

Though these experiments should no longer take place, we shouldn’t ignore what people have already discovered from them in the past. The findings from these experiments are interesting. The fact that hope leads to greater resilience comes as little surprise to us, though of course findings in rats may not translate to other species. That said, we think this is probably the case in humans as well as rats.

In fact, we believe that a large part of the role of leadership is to help individuals feel valued, respected, supported and hopeful about their futures. In doing this, individuals can have better qualities of working life, and organizations can have higher levels of productivity.

That said, we think these experiments and the lessons that can be learned from them are also very sinister. It’s clearly the case that providing people with hope, real or false, inspires them to greater effort. We are certain that many organizations and HR functions know this, and look to build this into their management approaches.

Where hope is real, it’s good. Where it’s falsely introduced to drive individuals to higher levels of perseverance in poor working situations, then it’s quite reprehensible. Which doesn’t mean it’s not profitable or that it doesn’t happen. It just means that people should not work for these organizations where they have any choice.

Interestingly, the relationship between hope and faith has been discussed many times throughout history. A good place to listen to some reflections on this is in this episode of the BBC podcast “In Our Time”.

How We Help Organizations

We provide leadership development programmes and consulting services to clients around the world to help them become high performing organizations that are great places to work. We receive great feedback, build meaningful and lasting relationships and provide reduced cost services where price is a barrier.

Learning more about who we are and what we do it easy: To hear from us, please join our mailing list . To ask about how we can help you or your organization, please contact us . To explore topics we care about, listen to our podcast . To attend a free seminar, please check out our eventbrite page .

We’re also considering creating a community for people interested in improving the world of work. If you’d like to be part of it, please contact us .

Sources and Feedback

Schulkin, Jay, and Paul Rozin. Curt Richter: A Life in the Laboratory. Baltimore: Johns Hopkins University Press, 2005., doi:10.1353/book.60340. https://www.aipro.info/wp/wp-content/uploads/2017/08/phenomena_sudden_death.pdf

We’re a small organization who know we make mistakes and want to improve them. Please contact us with any feedback you have on this post. We’ll usually reply within 72 hours. 

Related Posts

Play Personalities – What is your style of play at work?

Subjective Wellbeing – What it is and how to boost it

Flow – Moments of Optimal Experience

Maximizing Vs. Satisficing – What is good enough?

Celebrating Success – A Key Activity for Engement and Wellbeing

Musterbation – The Self-defeating habit of making everything a “Must”

Training & development.

Leadership Development

Management Development

Seminars & Workshops

Consulting & Change

Free resources.

Free Seminars

Meet the Team

What Clients Say

The WoW Mail

Privacy Policy

Copyright 2018 – 2024 | World of Work Project CIC

Privacy Overview

Learn Something Interesting

Education for the pure joy of learning, the harvard university hope experiment.

During the 1950’s, Dr. Curt Richter from Harvard University performed a series of experiments using water, buckets, and both domesticated and wild rats which resulted in a surprising discovery within the field of psychology. In the first experiment, Richter placed his test subjects into large buckets half filled with water with even those rats which were considered above average swimmers, giving up and dying within a few short minutes. In the second experiment, Richter pulled each rat out just as it was about to give up due to exhaustion and let them rest for a few moments. Upon inserting the rats back into the bucket of water, Richter found that the rats continued to struggle to survive for up to 60 hours as the rats now believed that if they continued to push forward with enough effort put forth, eventually they would be rescued once again. Richter recorded in his notes, “after elimination of hopelessness, the rats do not die”

Share This Interesting Fact:

11 thoughts on “ the harvard university hope experiment ”.

• Pingback: Pesca d'alto mare - The Unconditional Blog
• Pingback: Pesca d'alto mare - Come Don Chisciotte
• Pingback: Deep Sea Fishing – The Burning Platform
• Pingback: Deep Sea Fishing – Thailand Daily
• Pingback: The Harvard University Hope Experiment – Learn Something Interesting – Kon/Spira[l]
• Pingback: The Scotfree | Deep Sea Fishing
• Pingback: Deep Sea Fishing - United Push Back
• Pingback: Deep Sea Fishing – OffGuardian

I’m a great swimmer and I hate rats…what does that make me?

• Pingback: HOPE - The Miracle Drug » 90rollsroyces

Thanks! I will now visit this blog every day!

Leave a Reply Cancel reply

Discover more from learn something interesting.

Subscribe now to keep reading and get access to the full archive.

Type your email…

Continue reading

We have emailed you a PDF version of the article you requested.

Can't find the email?

Please check your spam or junk folder

You can also add [email protected] to your safe senders list to ensure you never miss a message from us.

Universe 25: The Mouse "Utopia" Experiment That Turned Into An Apocalypse

Complete the form below and we will email you a PDF version

Cancel and go back

IFLScience needs the contact information you provide to us to contact you about our products and services. You may unsubscribe from these communications at any time.

For information on how to unsubscribe, as well as our privacy practices and commitment to protecting your privacy, check out our Privacy Policy

Complete the form below to listen to the audio version of this article

Advertisement

Subscribe today for our  Weekly Newsletter  in your inbox!

James Felton

Senior Staff Writer

James is a published author with four pop-history and science books to his name. He specializes in history, strange science, and anything out of the ordinary.

Book View full profile

Book Read IFLScience Editorial Policy

DOWNLOAD PDF VERSION

The utopia in all its glory. Image credit: Yoichi R Okamoto, White House photographer (public domain, via Wikimedia Commons ).

Over the last few hundred years, the human population of Earth has seen an increase, taking us from an estimated one billion in 1804 to seven billion in 2017. Throughout this time, concerns have been raised that our numbers may outgrow our ability to produce food, leading to widespread famine. 

Some –  the Malthusians  – even took the view that as resources ran out, the population would "control" itself through mass deaths until a sustainable population was reached. As it happens, advances in farming, changes in farming practices, and new farming technology have given us enough food to feed  10 billion people , and it's how the food is distributed which has caused mass famines and starvation. As we use our resources and the climate crisis worsens, this could all change  – but for now, we have always been able to produce more food than we need, even if we have lacked the will or ability to distribute it to those that need it.

But while everyone was worried about a lack of resources, one behavioral researcher in the 1970s sought to answer a different question: what happens to society if all our appetites are catered for, and all our needs are met? The answer – according to his study – was an awful lot of cannibalism shortly followed by an apocalypse.

John B Calhoun set about creating a series of experiments that would essentially cater to every need of rodents, and then track the effect on the population over time. The most infamous of the experiments was named, quite dramatically, Universe 25 .

In this study, he took four breeding pairs of mice and placed them inside a "utopia". The environment was designed to eliminate problems that would lead to mortality in the wild. They could access limitless food via 16 food hoppers, accessed via tunnels, which would feed up to 25 mice at a time, as well as water bottles just above. Nesting material was provided. The weather was kept at 68°F (20°C), which for those of you who aren't mice is the perfect mouse temperature. The mice were chosen for their health, obtained from the National Institutes of Health breeding colony. Extreme precautions were taken to stop any disease from entering the universe.

As well as this, no predators were present in the utopia, which sort of stands to reason. It's not often something is described as a "utopia, but also there were lions there picking us all off one by one". 

The experiment began, and as you'd expect, the mice used the time that would usually be wasted in foraging for food and shelter for having excessive amounts of sexual intercourse. About every 55 days, the population doubled as the mice filled the most desirable space within the pen, where access to the food tunnels was of ease.

When the population hit 620, that slowed to doubling around every 145 days, as the mouse society began to hit problems. The mice split off into groups, and those that could not find a role in these groups found themselves with nowhere to go.

"In the normal course of events in a natural ecological setting somewhat more young survive to maturity than are necessary to replace their dying or senescent established associates," Calhoun wrote in 1972 . "The excess that find no social niches emigrate."

Here, the "excess" could not emigrate, for there was nowhere else to go. The mice that found themself with no social role to fill – there are only so many head mouse roles, and the utopia was in no need of a Ratatouille -esque chef – became isolated.

"Males who failed withdrew physically and psychologically; they became very inactive and aggregated in large pools near the center of the floor of the universe. From this point on they no longer initiated interaction with their established associates, nor did their behavior elicit attack by territorial males," read the paper. "Even so, they became characterized by many wounds and much scar tissue as a result of attacks by other withdrawn males."

The withdrawn males would not respond during attacks, lying there immobile. Later on, they would attack others in the same pattern. The female counterparts of these isolated males withdrew as well. Some mice spent their days preening themselves, shunning mating, and never engaging in fighting. Due to this they had excellent fur coats, and were dubbed, somewhat disconcertingly, the "beautiful ones".

The breakdown of usual mouse behavior wasn't just limited to the outsiders. The "alpha male" mice became extremely aggressive, attacking others with no motivation or gain for themselves, and regularly raped both males and females . Violent encounters sometimes ended in mouse-on-mouse cannibalism.

Despite – or perhaps because – their every need was being catered for, mothers would abandon their young or merely just forget about them entirely, leaving them to fend for themselves. The mother mice also became aggressive towards trespassers to their nests, with males that would normally fill this role banished to other parts of the utopia. This aggression spilled over, and the mothers would regularly kill their young. Infant mortality in some territories of the utopia reached 90 percent.

This was all during the first phase of the downfall of the "utopia". In the phase Calhoun termed the "second death", whatever young mice survived the attacks from their mothers and others would grow up around these unusual mouse behaviors. As a result, they never learned usual mice behaviors and many showed little or no interest in mating, preferring to eat and preen themselves, alone.

The population peaked at 2,200 – short of the actual 3,000-mouse capacity of the "universe" – and from there came the decline. Many of the mice weren't interested in breeding and retired to the upper decks of the enclosure, while the others formed into violent gangs below, which would regularly attack and cannibalize other groups as well as their own. The low birth rate and high infant mortality combined with the violence, and soon the entire colony was extinct . During the mousepocalypse, food remained ample, and their every need completely met.

Calhoun termed what he saw as the cause of the collapse "behavioral sink".

"For an animal so simple as a mouse, the most complex behaviors involve the interrelated set of courtship, maternal care, territorial defence and hierarchical intragroup and intergroup social organization," he concluded in his study.

"When behaviors related to these functions fail to mature, there is no development of social organization and no reproduction. As in the case of my study reported above, all members of the population will age and eventually die. The species will die out."

He believed that the mouse experiment may also apply to humans, and warned of a day where – god forbid – all our needs are met.

"For an animal so complex as man, there is no logical reason why a comparable sequence of events should not also lead to species extinction. If opportunities for role fulfilment fall far short of the demand by those capable of filling roles, and having expectancies to do so, only violence and disruption of social organization can follow."

At the time, the experiment and conclusion became quite popular, resonating with people's feelings about overcrowding in urban areas leading to "moral decay"  (though of course, this ignores so many factors such as poverty and prejudice).

However, in recent times, people have questioned whether the experiment could really be applied so simply to humans – and whether it really showed what we believed it did in the first place.

The end of the mouse utopia could have arisen "not from density, but from excessive social interaction," medical historian Edmund Ramsden said in 2008 . “Not all of Calhoun’s rats had gone berserk. Those who managed to control space led relatively normal lives.”

As well as this, the experiment design has been criticized for creating not an overpopulation problem, but rather a scenario where the more aggressive mice were able to control the territory and isolate everyone else. Much like with food production in the real world, it's possible that the problem wasn't of adequate resources, but how those resources are controlled.

  THIS WEEK IN IFLSCIENCE

Article posted in.

overpopulation,

weird and wonderful

More Nature Stories

link to article

What Are The Most Dangerous Seas In The World?

A Rare Hybrid Heron Is Spotted By Birdwatchers In France

Meet Hector’s Dolphins, One Of The Smallest Dolphin Species In The World

Tool-Wielding Whales, Earth’s New Ring, And Wow! Signal Solved?

IFLScience The Big Questions: How Will The Solar System End?

How Large Are Mine Spiders?

Drowning Rats – The Hope Experiment: Dr. Curt Richter’s Harvard Rat Study

No tags found.

Have you ever been in a situation where you felt dejected and hopeless?

When you approached someone to share your feelings, did you receive the impractical solution of “Don’t give up, have hope”? Well, Dr. Curt Richter’s Harvard Rat Study may just prove the power of hope to you.

Table of Contents

The Drowning Rats Experiment

Richter, an American psycho-biologist, conducted an experiment to assess the behavior of ‘drowning rats’ and how long it took them to die. While his procedure may sound cruel by today’s ethical experimentation standards, the findings that he derived from it are incredibly interesting.

The experiment involved observing the rats’ behavior when they were immersed in buckets filled with water. Richter varied different factors to come to his conclusions about the role of ‘hope’ in perseverance.

Experiment I: Domestic Rats

The first step of the experiment involved observing 12 domesticated drowning rats. A quarter of these rats began by floating around the surface of the water for some time and then plunging inside the bucket to understand the interior of the bucket. This entire process took place for two minutes following which they drowned.

However, the other nine rats displayed dissimilar behavior. They explored the bucket in its entirety and then kept swimming to stay afloat in the bucket. After days of survival, they eventually succumbed, probably due to fatigue and drowning.

Experiment II: Wild Rats

The next phase of this Harvard rat swimming experiment took place with freshly caught wild and The next phase of experiment took place with freshly caught wild and aggressive rats. Trained by the forces of nature, these 34 rats could swim very well, thus forming the hypothesis that these wild drowning rats would strive for their life. To Richter’s surprise, this was not the case. In fact, all of these untamed drowning rats died within a few minutes. Skills that they had derived from their worldly savvy were all in vain.

Hope: The Key to Perseverance

After assessing the huge difference between the reaction of the domesticated and the wild rats, Dr. Richter felt that since the domesticated drowning rats have experienced the presence of a support system (in contrast with the wild ones), they are hopeful and thus can put in the best of their efforts to save their lives. 

He expressed:

“ The situation of these rats scarcely seems one demanding fight or flight — it is rather one of hopelessness… the rats are in a situation against which they have no defense… they seem literally to ‘give up. ”

To elaborate on his findings, he further changed some settings in the experiment.

The Hope Experiment

He wanted to find out the relationship between hope and perseverance in the drowning rats. As per his earlier statement, he hypothesized that hopefulness would make the rats fight for their survival more actively. So, he began this phase of his experiment by leaving homogeneous rats in buckets filled with water. However, when the rats drowned and were on the verge of dying, they were saved by the experimenter. They were laid down on towels, dried off and made steady.

Once the rats had recovered, they were put under the previous circumstances again. This time, it was noticed that the drowning rats would swim on and on. The duration for which they could survive surpassed the earlier time lengths.

Conclusion of the Experiment

In the last condition, the only variable that had changed was that the drowning rats had been saved. Thus, they were made aware of the feeling of hope. Since they swam for a longer time, therefore, Richter’s hypothesis stood true, he thus established that “ after elimination of hopelessness the rats do not die ”.

Even though rats and humans are very different animals, these tiny creatures give us an important lesson. They teach us that when we are hopeful about the outcomes of a situation, our perseverance and willingness to put in effort are also more. So, if we don’t have hope, we can reach a position where we would not attempt to save our lives. You should always try to find inspiration to improve your perseverance .

It was rightly said by Samuel Johnson that “The natural flights of the human mind are not from pleasure to pleasure but from hope to hope”.

To understand more about yourself and your mind, begin your journey of hope with the Evolve App now. Download the app and start your free trial.

Karishma Golchha is pursuing Bachelor’s in Psychology. She is very keen about the human mind and looks forward to connect with you at [email protected] and evolve together!

Blogs - Shop Evolve

This will close in 65 seconds

This Old Experiment With Mice Led to Bleak Predictions for Humanity’s Future

From the 1950s to the 1970s, researcher John Calhoun gave rodents unlimited food and studied their behavior in overcrowded conditions

Maris Fessenden ; Updated by Rudy Molinek

What does utopia look like for mice and rats? According to a researcher who did most of his work in the 1950s through 1970s, it might include limitless food, multiple levels and secluded little condos. These were all part of John Calhoun’s experiments to study the effects of population density on behavior. But what looked like rodent paradises at first quickly spiraled into out-of-control overcrowding, eventual population collapse and seemingly sinister behavior patterns.

In other words, the mice were not nice.

Working with rats between 1958 and 1962, and with mice from 1968 to 1972, Calhoun set up experimental rodent enclosures at the National Institute of Mental Health’s Laboratory of Psychology. He hoped to learn more about how humans might behave in a crowded future. His first 24 attempts ended early due to constraints on laboratory space. But his 25th attempt at a utopian habitat, which began in 1968, would become a landmark psychological study. According to Gizmodo ’s Esther Inglis-Arkell, Calhoun’s “Universe 25” started when the researcher dropped four female and four male mice into the enclosure.

By the 560th day, the population peaked with over 2,200 individuals scurrying around, waiting for food and sometimes erupting into open brawls. These mice spent most of their time in the presence of hundreds of other mice. When they became adults, those mice that managed to produce offspring were so stressed out that parenting became an afterthought.

“Few females carried pregnancies to term, and the ones that did seemed to simply forget about their babies,” wrote Inglis-Arkell in 2015. “They’d move half their litter away from danger and forget the rest. Sometimes they’d drop and abandon a baby while they were carrying it.”

A select group of mice, which Calhoun called “the beautiful ones,” secluded themselves in protected places with a guard posted at the entry. They didn’t seek out mates or fight with other mice, wrote Will Wiles in Cabinet magazine in 2011, “they just ate, slept and groomed, wrapped in narcissistic introspection.”

Eventually, several factors combined to doom the experiment. The beautiful ones’ chaste behavior lowered the birth rate. Meanwhile, out in the overcrowded common areas, the few remaining parents’ neglect increased infant mortality. These factors sent the mice society over a demographic cliff. Just over a month after population peaked, around day 600, according to Distillations magazine ’s Sam Kean, no baby mice were surviving more than a few days. The society plummeted toward extinction as the remaining adult mice were just “hiding like hermits or grooming all day” before dying out, writes Kean.

Calhoun launched his experiments with the intent of translating his findings to human behavior. Ideas of a dangerously overcrowded human population were popularized by Thomas Malthus at the end of the 18th century with his book An Essay on the Principle of Population . Malthus theorized that populations would expand far faster than food production, leading to poverty and societal decline. Then, in 1968, the same year Calhoun set his ill-fated utopia in motion, Stanford University entomologist Paul Ehrlich published The Population Bomb . The book sparked widespread fears of an overcrowded and dystopic imminent future, beginning with the line, “The battle to feed all of humanity is over.”

Ehrlich suggested that the impending collapse mirrored the conditions Calhoun would find in his experiments. The cause, wrote Charles C. Mann for Smithsonian magazine in 2018, would be “too many people, packed into too-tight spaces, taking too much from the earth. Unless humanity cut down its numbers—soon—all of us would face ‘mass starvation’ on ‘a dying planet.’”

Calhoun’s experiments were interpreted at the time as evidence of what could happen in an overpopulated world. The unusual behaviors he observed—such as open violence, a lack of interest in sex and poor pup-rearing—he dubbed “behavioral sinks.”

After Calhoun wrote about his findings in a 1962 issue of Scientific American , that term caught on in popular culture, according to a paper published in the Journal of Social History . The work tapped into the era’s feeling of dread that crowded urban areas heralded the risk of moral decay.

Events like the murder of Kitty Genovese in 1964—in which false reports claimed 37 witnesses stood by and did nothing as Genovese was stabbed repeatedly—only served to intensify the worry. Despite the misinformation, media discussed the case widely as emblematic of rampant urban moral decay. A host of science fiction works—films like Soylent Green , comics like 2000 AD —played on Calhoun’s ideas and those of his contemporaries . For example, Soylent Green ’s vision of a dystopic future was set in a world maligned by pollution, poverty and overpopulation.

Now, interpretations of Calhoun’s work have changed. Inglis-Arkell explains that the main problem of the habitats he created wasn’t really a lack of space. Rather, it seems likely that Universe 25’s design enabled aggressive mice to stake out prime territory and guard the pens for a limited number of mice, leading to overcrowding in the rest of the world.

However we interpret Calhoun’s experiments, though, we can take comfort in the fact that humans are not rodents. Follow-up experiments by other researchers, which looked at human subjects, found that crowded conditions didn’t necessarily lead to negative outcomes like stress, aggression or discomfort.

“Rats may suffer from crowding,” medical historian Edmund Ramsden told the NIH Record ’s Carla Garnett in 2008, “human beings can cope.”

Get the latest stories in your inbox every weekday.

Maris Fessenden | | READ MORE

Maris Fessenden is a freelance science writer and artist who appreciates small things and wide open spaces.

Rudy Molinek | READ MORE

Rudy Molinek is  Smithsonian  magazine's 2024 AAAS Mass Media Fellow.

Stack Exchange Network

Stack Exchange network consists of 183 Q&A communities including Stack Overflow , the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.

Q&A for work

Connect and share knowledge within a single location that is structured and easy to search.

Curt Richter's rat hope experiment: Why did the first nine rats survive for days?

I can understand the part that the experimenter saved the rat just before it was about to die and then the rat lasted longer for the next drowning. But I do not understand what it means that before such intervention, when he first tried drowning, first 3 rats died in 2 minutes but the remaining 9 rats survived for days.Was it just a random chance that the first 3 rats happened to have little hope and the remaining 9 rats had naturally more hopes?

The first rat, Richter noted, swam around excitedly on the surface for a very short time, then dove to the bottom, where it began to swim around, nosing its way along the glass wall. It died two minutes later. Two more of the 12 domesticated rats died in much the same way. But, interestingly, the nine remaining rats did not succumb nearly so readily; they swam for days before they eventually gave up and died. ..... Richter then tweaked the experiment: He took other, similar rats and put them in the jar. Just before they were expected to die, however, he picked them up

https://www.psychologytoday.com/gb/blog/kidding-ourselves/201405/the-remarkable-power-hope

• experimental-psychology
• animal-cognition

• $\begingroup$ The first experiment used domesticated rats so they were used to having someone take care of them. The ones that survived so long probably had more confidence that someone would come. $\endgroup$ –  Just Weighinin Commented Mar 29 at 12:09

2 Answers 2

For more information on the experiment, there is Swamy (2020) :

The conclusion drawn was that since the rats BELIEVED that they would eventually be rescued, they could push their bodies way past what they previously thought impossible.

and the source ( Richter, 1957 ) can be downloaded in PDF

I had to re-read the Richter paper a couple of times to digest it, and from my understanding, the differences in swimming time was in the same (first) experiment. The tweak to the experiment was where, instead of using domesticated rats, they used hybrid rats ("crosses between domesticated and wild rats").

Richter, C. P. (1957). On the phenomenon of sudden death in animals and man. Psychosomatic Medicine, 19 , 191–198. https://doi.org/10.1097/00006842-195705000-00004

Swamy, S. (2020). The Power of Hope: A Rat Experiment by Dr Curt Richter. LinkedIn https://www.linkedin.com/pulse/power-hope-rat-experiment-dr-curt-richter-santosh-swamy

• 3 $\begingroup$ But the time that the experimenter introduced "hope" by rescuing rats was AFTER the first batch of 12 rats had died, wasn't it? I mean, the liked article says "Richter then tweaked the experiment". Did I misunderstand it and the experimenter introduced "hope" between the third and fourth rat? $\endgroup$ –  Damn Vegetables Commented Feb 12, 2022 at 15:29

It's not clear what caused this. This variation in swimming times is in fact what motivates Richter to continue tweaking his experiment:

The significance of this average curve was greatly reduced by the marked variations in individual swimming times. At all temperatures, a small number of rats died within 5 - 10 minutes after immersion, while in some instances others apparently no more healthy, swam as long as 81 hours. The elimination of these large variations presented a real problem, which for some time we could not solve. Then the solution came from an unexpected source - the finding of the phenomenon of sudden death, which constitutes the main topic of this communication.

The numbers you cite regarding the 3 versus 9 rats comes from his second run of the experiment where he tests whether trimming the whiskers in the rats would result in different times.

The first rat swam around excitedly on the surface for a very short time, then dove to the bottom, where it began to swim around nosing its way along the glass wall. Without coming to the surface a single time, it died 2 minutes after entering the tank. Two more of the twelve domesticated rats tested died in much the same way; however, the remaining 9 swam 40 to 60 hours.

It seems that it was just caused by random luck.

Your Answer

Sign up or log in, post as a guest.

Required, but never shown

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy .

Not the answer you're looking for? Browse other questions tagged experimental-psychology animal-cognition or ask your own question .

• Featured on Meta
• Bringing clarity to status tag usage on meta sites
• We've made changes to our Terms of Service & Privacy Policy - July 2024
• Announcing a change to the data-dump process

Hot Network Questions

• How to extract code into library allowing changes without workflow overhead
• Why do combinatorists care about Kazhdan–Lusztig polynomials?
• The McDonald's Option
• In TNG: the Pegasus, why is Geordi in the first meeting with the Admiral?
• Can I use rear (thru) axle with crack for a few rides, before getting a new one?
• Does my AC vent air from the room to outside?
• How does one find distinct signals in FFT transform log that have been produced by different transmitters?
• "Show single Polygon" Atlas Filter Expression
• \includegraphics not reading \newcommand
• How would Earth's plants change in this speculated Earth-like planet?
• Are there any virtues in virtue ethics that cannot be plausibly grounded in more fundamental utilitarian principles?
• Why don't we observe protons deflecting in J.J. Thomson's experiment?
• Is there anything that stops the majority shareholder(s) from destroying company value?
• How to remove a file which name seems to be "." on an SMB share?
• python stats.spearmanr and R cor.test(method='spearman') don't return the same p-value?
• Can the speed of light inhibit the synchronisation of a power grid?
• 3 Single Phase Full Bridge Rectifiers for 3 Phase Rectification
• I am a fifteen-year-old from India planning to fly to Germany alone (without my parents accompanying me) to see my relatives.What documents do I need?
• Prove Principle of Mathematical Induction and Principal of Well-ordering are equivalent.
• Should I be worried about this giant crack?
• Who said "If you don't do politics, politics will do you"?
• Should I Inform My PhD Program About Not Completing My Master's Degree Due to Personal Challenges?
• The hat-check problem
• Home water pressure higher than city water pressure?

What Does “Rat Park” Teach Us About Addiction?

How many of us, during clinical encounters with patients, focus on their families, their social communities, their sources of human contact and support?

Addressing Alcohol or Cannabis Use in Patients With Anxiety Disorders

How to Talk to Teenagers About Substance Use

Social Media and Substance Use: What Clinicians Need to Know

More Than Postpartum Depression: Addressing Maternal Mental Health Through the Life Cycle

Stress and Substance Use Among Drag Performers

WHO Addresses Gaps in Mental Health Care Delivery and Quality

2 Commerce Drive Cranbury, NJ 08512

609-716-7777

Distillations magazine

Mouse heaven or mouse hell.

Biologist John Calhoun’s rodent experiments gripped a society consumed by fears of overpopulation.

Officially, the colony was called the Mortality-Inhibiting Environment for Mice. Unofficially, it was called mouse heaven.

Biologist John Calhoun built the colony at the National Institute of Mental Health in Maryland in 1968. It was a large pen—a 4½-foot cube—with everything a mouse could ever desire: plenty of food and water; a perfect climate; reams of paper to make cozy nests; and 256 separate apartments, accessible via mesh tubes bolted to the walls. Calhoun also screened the mice to eliminate disease. Free from predators and other worries, a mouse could theoretically live to an extraordinarily old age there, without a single worry.

But the thing is, this wasn’t Calhoun’s first rodent utopia. This was the 25th iteration. And by this point he knew how quickly mouse heaven could deteriorate into mouse hell.

John Calhoun grew up in Tennessee, the son of a high school principal and an artist, and was an avid birder when young. After earning his PhD in zoology, he joined the Rodent Ecology Project in Baltimore in 1946, whose purpose was to eliminate rodent pests in cities. The project had limited success, partly because no one could figure out what aspects of rodent behavior, lifestyle, or biology to target. Calhoun set up his first utopia, involving Norway rats, in the woods behind his house to monitor rodents over time and figure out what factors drove their population growth.

Eventually Calhoun grew fascinated with the rodent behavior for its own sake and began crafting ever more elaborate and carefully controlled environments. It wasn’t just the behavior of rats that interested him. Architects and civil engineers at the time were having vigorous debates about how to build better cities, and Calhoun imagined urban design might be studied in rodents first and then extrapolated to human beings.

Calhoun’s most famous utopia, number 25, began in July 1968, when he introduced eight albino mice into the 4½-foot cube. Following an adjustment period, the first pups were born 3½ months later, and the population doubled every 55 days afterward. Eventually this torrid growth slowed, but the population continued to climb, peaking at 2,200 mice during the 19th month.

That robust growth masked some serious problems, however. In the wild, infant mortality among mice is high, as most juveniles get eaten by predators or perish of disease or cold. In mouse utopia, juveniles rarely died. As a result, there were far more youngsters than normal, which introduced several difficulties.

Rodents have social hierarchies, with dominant alpha males controlling harems of females. Alphas establish dominance by fighting—wrestling and biting any challengers. Normally a mouse that loses a fight will scurry off to some distant nook to start over elsewhere.

But in mouse utopia, the losing mice couldn’t escape. Calhoun called them “dropouts.” And because so few juveniles died, huge hordes of dropouts would gather in the center of the pen. They were full of cuts and ugly scars, and every so often huge brawls would break out—vicious free-for-alls of biting and clawing that served no obvious purpose. It was just senseless violence. (In earlier utopias involving rats, some dropouts turned to cannibalism.)

Alpha males struggled, too. They kept their harems in private apartments, which they had to defend from challengers. But given how many mice survived to adulthood, there were always a dozen hotshots ready to fight. The alphas soon grew exhausted, and some stopped defending their apartments altogether.

As a result, apartments with nursing females were regularly invaded by rogue males. The mothers fought back, but often to the detriment of their young. Many stressed-out mothers booted their pups from the nest early, before the pups were ready. A few even attacked their own young amid the violence or abandoned them while fleeing to different apartments, leaving the pups to die of neglect.

Eventually other deviant behavior emerged. Mice who had been raised improperly or kicked out of the nest early often failed to develop healthy social bonds, and therefore struggled in adulthood with social interactions. Maladjusted females began isolating themselves like hermits in empty apartments—unusual behavior among mice. Maladjusted males, meanwhile, took to grooming all day—preening and licking themselves hour after hour. Calhoun called them “the beautiful ones.” And yet, even while obsessing over their appearance, these males had zero interest in courting females, zero interest in sex.

Intriguingly, Calhoun had noticed in earlier utopias that such maladjusted behavior could spread like a contagion from mouse to mouse. He dubbed this phenomenon “the behavioral sink.”

Between the lack of sex, which lowered the birth rate, and inability to raise pups properly, which sharply increased infant mortality, the population of Universe 25 began to plummet. By the 21st month, newborn pups rarely survived more than a few days. Soon, new births stopped altogether. Older mice lingered for a while—hiding like hermits or grooming all day—but eventually they died out as well. By spring 1973, less than five years after the experiment started, the population had crashed from 2,200 to 0. Mouse heaven had gone extinct.

Universe 25 ended a half century ago, but it continues to fascinate people today—especially as a gloomy metaphor for human society. Calhoun actively encouraged such speculation, once writing, “I shall largely speak of mice, but my thoughts are on man.” As early as 1968, journalist Tom Wolfe titled an essay about New York “O Rotten Gotham—Sliding Down into the Behavioral Sink.” Oddly, though, none of the prognosticators could agree on the main lesson of Universe 25.

The first people to fret over Universe 25 were environmentalists. The same year the study began, biologist Paul Ehrlich published The Population Bomb , an alarmist book predicting imminent starvation and population crashes due to overpopulation on Earth. Pop culture picked up on this theme in movies, such as Soylent Green , where humans in crowded cities are culled and turned into food slurry. Overall, the idea of dangerous overcrowding was in the air, and some sociologists explicitly drew on Calhoun’s work, writing: “We . . . take the animal studies as a serious model for human populations.” The message was stark: Curb population growth—or else .

More recently scholars saw similarities to the Industrial Revolution and the rise of modern urban society. The 19th and 20th centuries saw population booms across the world, largely due to drops in infant mortality—similar to what the mice experienced. Recently, however, human birth rates have dropped sharply in many developed countries—often below replacement levels—and young people in those places have reportedly lost interest in sex. The parallels to Universe 25 seem spooky.

Behavioral biologists have echoed the eugenics movement in blaming the strange behaviors of the mice on a lack of natural selection, which in their view culls those they consider weak and unfit to breed. This lack of culling resulted in supposed “mutational meltdowns” that led to widespread mouse stupidity and aberrant behavior. (The researchers argued that the brain is especially susceptible to mutations because it’s so intricate and because so many of our genes influence brain function.)

Extrapolating from this work, some political agitators warn that humankind will face a similar decline. Women are supposedly falling into Calhoun’s behavioral sink by learning “maladaptive behaviors,” such as choosing not to have children, which “destroy[s] their own genetic interests.” Other critics agonize over the supposed loss of traditional gender roles, leaving effete males and hyperaggressive females, or they deplore the undermining of religions and their imperatives to “be fruitful and multiply.” In tandem, such changes will lead to the “decline of the West.”

Still others have cast Universe 25’s collapse as a parable illustrating the dangers of socialist welfare states, which, they argue, provide material goods but remove healthy challenges from people’s lives, challenges that build character and promote “personal growth.” Another school of thought viewed Universe 25 as a warning about “the city [as] a perversion of nature.” As sociologists Claude Fischer and Mark Baldassare put it, “A red-eyed, sharp-fanged obsession about urban life stalks contemporary thought.”

Most critics who’ve fretted over Calhoun’s work cluster on the conservative end of the political spectrum, but self-styled progressives have weighed in as well. Advocates for birth control repeatedly invoked Calhoun’s mice as a cautionary tale about how runaway population growth destroys family life. More recent interpretations see the mice collapse in terms of one-percenters and wealth inequality; they blame the social dysfunction on a few aggressive males hoarding precious resources (e.g., desirable apartments). In this view, said one critic, “Universe 25 had a fair distribution problem” above all.

Given these wildly varying (even contradictory) readings, it’s hard to escape the suspicion that personal and political views, rather than objective inquiry, are driving these critics’ outlooks. And indeed, a closer look at the interpretations severely undermines them.

When forecasting population crashes among human beings, Population Bomb –type environmentalists invariably predicted that overcrowding would lead to widespread shortages of food and other goods. That’s actually the opposite of what Universe 25 was like. The mice there had all the goods they wanted. This also undermines arguments about unfair resource distribution.

Perhaps, then, it was the lack of struggles and challenges that led to dysfunction, as welfare critics claimed. Except that the spiral of dysfunction began when hordes of “dropout” mice lost challenges to alpha males, couldn’t escape elsewhere, and began brawling in the middle of the pen. The alpha males in turn grew weary after too many challenges from youngsters. Indeed, most mice faced competition far in excess of what they would encounter in the wild.

The appearance of the sexless “beautiful ones” does seem decadent and echoes the reported loss of interest in sex among young people in developed countries. Except that a closer look at the survey data indicates that such worries might be overblown. And any comparison between human birth rates and Universe 25 birth rates is complicated by the fact that the mouse rates dropped partly due to infant neglect and spikes in infant mortality—the opposite of the situation in the developed world.

Then there are the warnings about the mutational meltdown and the decline of intelligence. Aside from echoing the darkest rhetoric of the eugenics movement, this interpretation runs aground on several points. The hermit females and preening, asexual males certainly acted oddly—but in doing so, they avoided the vicious, violent free-for-alls that beset earlier generations. This hardly seems dumb. Moreover, some of Calhoun’s research actually saw rodents getting smarter during experiments.

This evidence came from an earlier utopia involving rats. In that setup, dropout rats began digging new burrows into the dirt floor of their pen. Digging produces loose dirt to clear away, and most rats laboriously carried the loose dirt outside the tunnel bit by bit, to dump it there. It’s necessary but tedious work.

But some of the dropout rats did something different. Instead of carrying dirt out bit by bit, they packed it all into a ball and rolled it out the tunnel in one trip. An enthused Calhoun compared this innovation to humankind inventing the wheel. And it happened only because the rats were isolated from the main group and didn’t learn the dominant method of digging. By normal rat standards, this was deviant behavior. It was also a creative breakthrough. Overall, then, Calhoun argued that social strife can sometimes push creatures to become smarter, not dumber.

(Incidentally, after Universe 25’s collapse, Calhoun began building new utopias to encourage creative behavior by keeping mice physically and mentally nourished. This research, in turn, inspired a children’s book named after Calhoun’s workplace— Mrs. Frisby and the Rats of NIMH , wherein a group of rats escape from a colony designed to stimulate their intelligence.)

So if all these interpretations of Universe 25 miss the mark, what lesson can we draw from the experiment?

Calhoun’s big takeaway involved status. Again, the males who lost the fights for dominance couldn’t leave to start over elsewhere. As he saw it, they were stuck in pathetic, humiliating roles and lacked a meaningful place in society. The same went for females when they couldn’t nurse or raise pups properly. Both groups became depressed and angry, and began lashing out. In other words, because mice are social animals, they need meaningful social roles to feel fulfilled. Humans are social animals as well, and without a meaningful role, we too can become hostile and lash out.

Still, even this interpretation seems like a stretch. Humans have far more ways of finding meaning in life than pumping out children or dominating some little hierarchy. And while human beings and mice are indeed both social creatures, that common label papers over some major differences. Critics of Calhoun’s work argued that population density among humans—a statistical measure—doesn’t necessarily correlate with crowding —a feeling of psychological stress. In the words of one historian, “Through their intelligence, adaptability, and capacity to make the world around them, humans were capable of coping with crowding” in ways that mice simply are not.

Ultimately Calhoun’s work functions like a Rorschach blot—people see what they want to see. It’s worth remembering that not all lab experiments, especially contrived ones such as Universe 25, apply to the real world. In which case, perhaps the best lesson to learn here is a meta-lesson: that drawing lessons itself can be a dangerous thing.

Sam Kean is a best-selling science author. His latest book is  The Icepick Surgeon: Murder, Fraud, Sabotage, Piracy, and Other Dastardly Deeds Perpetrated in the Name of Science .

More from our magazine

That Time Demons Possessed the Telegraph

Solar storms from long ago have become the delight of some scientists—and the dread of others.

21 Years, 7,600 Tests

Mary Papanicolaou, the woman behind the man behind the Pap smear.

Controversy, Control, and Cosmetics in Early Modern Italy

In a society that damned women for both plainness and adornment, wearing makeup became a defiant act of survival.

Copy the above HTML to republish this content. We have formatted the material to follow our guidelines, which include our credit requirements. Please review our full list of guidelines for more information. By republishing this content, you agree to our republication requirements.

The Doomed Mouse Utopia That Inspired the ‘Rats of NIMH’

Dr. john bumpass calhoun spent the ’60s and ’70s playing god to thousands of rodents..

On July 9th, 1968, eight white mice were placed into a strange box at the National Institute of Health in Bethesda, Maryland . Maybe “box” isn’t the right word for it; the space was more like a room, known as Universe 25, about the size of a small storage unit. The mice themselves were bright and healthy, hand-picked from the institute’s breeding stock. They were given the run of the place, which had everything they might need: food, water, climate control, hundreds of nesting boxes to choose from, and a lush floor of shredded paper and ground corn cob.

This is a far cry from a wild mouse’s life—no cats, no traps, no long winters. It’s even better than your average lab mouse’s, which is constantly interrupted by white-coated humans with scalpels or syringes. The residents of Universe 25 were mostly left alone, save for one man who would peer at them from above, and his team of similarly interested assistants. They must have thought they were the luckiest mice in the world. They couldn’t have known the truth: that within a few years, they and their descendants would all be dead.

The man who played mouse-God and came up with this doomed universe was named John Bumpass Calhoun. As Edmund Ramsden and Jon Adams detail in a paper, “ Escaping the Laboratory: The Rodent Experiments of John B. Calhoun & Their Cultural Influence ,” Calhoun spent his childhood traipsing around Tennessee , chasing toads, collecting turtles, and banding birds. These adventures eventually led him to a doctorate in biology, and then a job in Baltimore , where he was tasked with studying the habits of Norway rats, one of the city’s chief pests.

In 1947, to keep a close eye on his charges, Calhoun constructed a quarter-acre “rat city” behind his house, and filled it with breeding pairs. He expected to be able to house 5,000 rats there, but over the two years he observed the city, the population never exceeded 150. At that point, the rats became too stressed to reproduce. They started acting weirdly, rolling dirt into balls rather than digging normal tunnels. They hissed and fought.

This fascinated Calhoun—if the rats had everything they needed, what was keeping them from overrunning his little city, just as they had all of Baltimore?

Intrigued, Calhoun built another, slightly bigger rat metropolis—this time in a barn, with ramps connecting several different rooms. Then he built another and another, hopping between patrons that supported his research, and framing his work in terms of population: How many individuals could a rodent city hold without losing its collective mind? By 1954, he was working under the auspices of the National Institute of Mental Health, which gave him whole rooms to build his rodentopias. Some of these featured rats, while others focused on mice instead. Like a rodent real estate developer, he incorporated ever-better amenities: climbable walls, food hoppers that could serve two dozen customers at once, lodging he described as “walk-up one-room apartments.” Video records of his experiments show Calhoun with a pleased smile and a pipe in his mouth, color-coded mice scurrying over his boots.

Still, at a certain point, each of these paradises collapsed. “There could be no escape from the behavioral consequences of rising population density,” Calhoun wrote in an early paper . Even Universe 25—the biggest, best mousetopia of all, built after a quarter century of research—failed to break this pattern. In late October, the first litter of mouse pups was born. After that, the population doubled every two months—20 mice, then 40, then 80. The babies grew up and had babies of their own. Families became dynasties, carving out and holding down the best in-cage real estate. By August of 1969, the population numbered 620.

Then, as always, things took a turn. Such rapid growth put too much pressure on the mouse way of life. As new generations reached adulthood, many couldn’t find mates, or places in the social order—the mouse equivalent of a spouse and a job. Spinster females retreated to high-up nesting boxes, where they lived alone, far from the family neighborhoods. Washed-up males gathered in the center of the Universe, near the food, where they fretted, languished, and attacked each other. Meanwhile, overextended mouse moms and dads began moving nests constantly to avoid their unsavory neighbors. They also took their stress out on their babies, kicking them out of the nest too early, or even losing them during moves.

Population growth slowed way down again. Most of the adolescent mice retreated even further from societal expectations, spending all their time eating, drinking, sleeping and grooming, and refusing to fight or to even attempt to mate. (These individuals were forever changed—when Calhoun’s colleague attempted to transplant some of them to more normal situations, they didn’t remember how to do anything.) In May of 1970, just under 2 years into the study, the last baby was born, and the population entered a swan dive of perpetual senescence. It’s unclear exactly when the last resident of Universe 25 perished, but it was probably sometime in 1973.

Paradise couldn’t even last half a decade.

In 1973, Calhoun published his Universe 25 research as “Death Squared: The Explosive Growth and Demise of a Mouse Population.” It is, to put it lightly, an intense academic reading experience. He quotes liberally from the Book of Revelation, italicizing certain words for emphasis (e.g. “to kill with the sword and with famine and with pestilence and by wild beasts ”). He gave his claimed discoveries catchy names—the mice who forgot how to mate were “the beautiful ones”’ rats who crowded around water bottles were “social drinkers”; the overall societal breakdown was the “behavioral sink.” In other words, it was exactly the kind of diction you’d expect from someone who spent his entire life perfecting the art of the mouse dystopia.

Most frightening are the parallels he draws between rodent and human society. “I shall largely speak of mice,” he begins, “but my thoughts are on man.” Both species, he explains, are vulnerable to two types of death—that of the spirit and that of the body. Even though he had removed physical threats, doing so had forced the residents of Universe 25 into a spiritually unhealthy situation, full of crowding, overstimulation, and contact with various mouse strangers. To a society experiencing the rapid growth of cities—and reacting, in various ways, quite poorly — this story seemed familiar. Senators brought it up in meetings. It showed up in science fiction and comic books. Even Tom Wolfe, never lost for description, used Calhounian terms to describe New York City, calling all of Gotham a “behavioral sink.”

Convinced that he had found a real problem, Calhoun quickly began using his mouse models to try and fix it. If mice and humans weren’t afforded enough physical space, he thought, perhaps they could make up for it with conceptual space—creativity, artistry, and the type of community not built around social hierarchies. His later Universes were designed to be spiritually as well as physically utopic, with rodent interactions carefully controlled to maximize happiness (he was particularly fascinated by some early rats who had created an innovative form of tunneling, where they rolled dirt into balls). He extrapolated this, too, to human concerns, becoming an early supporter of environmental design and H.G. Wells’s hypothetical “World Brain,” an international information network that was a clear precursor to the internet.

But the public held on hard to his earlier work—as Ramsden and Adams put it, “everyone want[ed] to hear the diagnosis, no one want[ed] to hear the cure.” Gradually, Calhoun lost attention, standing, and funding. In 1986, he was forced to retire from the National Institute of Mental Health. Nine years later, he died.

But there was one person who paid attention to his more optimistic experiments, a writer named Robert C. O’Brien. In the late ’60s, O’Brien allegedly visited Calhoun’s lab , met the man trying to build a true and creative rodent paradise, and took note of the Frisbee on the door, the scientists’ own attempt “to help when things got too stressful,” as Calhoun put it. Soon after, O’Brien wrote Ms. Frisby and the Rats of NIMH —a story about rats who, having escaped from a lab full of blundering humans, attempt to build their own utopia. Next time, maybe we should put the rats in charge.

Naturecultures is a weekly column that explores the changing relationships between humanity and wilder things. Have something you want covered (or uncovered)? Send tips to [email protected] .

This Woman's Ongoing Fight for the Right to Live in a Treehouse

Using an ad blocker?

We depend on ad revenue to craft and curate stories about the world’s hidden wonders. Consider supporting our work by becoming a member for as little as $5 a month.

Podcast: APOPO Rats

In Colonial Williamsburg, Thieving Rats Save History

Meet the Rats That Wear Protective Poison Armor

How South Jersey Keeps Muskrat on the Menu

Follow us on Twitter to get the latest on the world's hidden wonders.

Like us on Facebook to get the latest on the world's hidden wonders.

Pre-Order Atlas Obscura: Wild Life Today!

Add some wonder to your inbox, we'd like you to like us.

A History of Rodent Mazes

The end of the maze, how the rodent labyrinth fell out of favor..

Drop a mouse in some water and white paint, and it will know just what to do. Mice can swim, by whipping their tails like a flagellum, but they don't like doing it; a mouse in a tub tries to find a way out. There's no need for training, or food pellets, or annoying electric shocks: To put a mouse through a water maze, you need only to build a little platform for it, hidden somewhere just beneath the surface. The mouse will try to find that platform without any encouragement.

It's a setup that's so simple—and so useful in measuring an animal's capacity for learning and memory—it hardly seems like it would need inventing. But it took a cognitive neuroscientist at the University of St. Andrews in Scotland to come up with the tub-and-platform method. In 1979, Richard Morris built a heated pool about 4 feet and 3 inches in diameter, filled it with water and fresh milk, and then added a platform made of stones and drain piping. Within a few years, his method (designed for rats) had been adapted for smaller lab mice, and had made its way into rodent labs around the world. Now it's among the most widespread animal-testing protocols in all of biomedicine. Scientists plunge mice in murky water to test the effects of brain damage, or the functions of particular genes on learning, or the efficacy of new drugs for treating Alzheimer's. You can even buy a standard-issue " Morris Water Maze " direct from a lab-supply shop, along with specialized software for recording its results.

Daniel Engber is a columnist for Slate.

That fact that so few of us would call a tub full of milk a “maze” only goes to show that rodent mazes aren't what they used to be. Early psychologists tempted rats with tricky blind alleys and wrong turns using contraptions built by hand, of wood and wire. The modern analogs of these devices—the standard rodent mazes of today—come in a few pared-down, elementary forms, pressed from plastic and sold in bulk: A tub , a circle , a plus-sign , a T-shape . The classic implement of behavioral psychology has grown ever less convoluted since it was first invented—less intricate, less mazy . Instead of hoping to lose a rodent in a labyrinth, today's scientists try to elicit a few simple behaviors that can be measured in simple ways. How long did it take the mouse to find the platform? Did he go left or right at the fork?

Traditional mazes take a long time to learn, says Jacqueline Crawley, a mouse behavior expert at the National Institute of Mental Health. (Her friends call her " Mrs. Frisby .") If you want to increase your throughput and generate more data, you're better off with a standard rig. "There are no motivation issues in the water maze,” she explains. You simply drop an animal in the tub a few times per day, over the span of about a week, "and then you’re done.” To run a mouse through a custom-made labyrinth full of twists and turns might take many weeks of training and produce results that can't easily be compared to those from other labs. The classic, elaborated maze that ruled the lab for half the 20 th century has grown as outdated as a phonograph. What happened?

The history of the labyrinth falls in line with the history of lab rodents: Both got their start in the early 1900s and had become standard research tools by the end of the 1930s. Still, the earliest navigational puzzles were built for simpler animals. For his 1882 book on Ants, Bees and Wasps , a polymath named John Lubbock constructed simple animal mazes using objects from around the house. He tested whether Hymenopterans could negotiate obstacles like a pencil, a China cup, and a hat box. Eventually, he devised a wooden table made from rotating discs, to see how they would confuse his lines of marching ants.

From Sir John Lubbock, Ants, bees, and wasps, 1888.

Rats made their way into the research lab at the turn of the century, starting with a group of researchers in Chicago and at Clark University in Worcester, Mass. A graduate student at Clark named Willard Small was the first to use a rodent maze to study learning. In 1901, he published the details of his new contraption: A platform about 6 feet long by 8 feet wide, covered with sawdust, and divided into galleries with walls of wire netting. He took the plan from a diagram of the hedge maze at Hampton Court, with an open space at the center, and six cul-de-sacs. The layout was selected with a natural setting in mind, he wrote, so that his experiments would be "couched in a familiar language" of rodent burrows.

From Willard S. Small, "Experimental Study of the Mental Processes of the Rat. II," American Journal of Psychology, 1901.

Small released his animals into the maze every evening, two at a time, by sliding open a glass door with a pulley as if it were a palace portcullis. Then he'd observe the animals for a while, recording their every sniff and sojourn in his notebook, before leaving them to wander the maze for the rest of the night. His 1901 paper includes extravagantly detailed accounts of each run through the maze, with its turning points described in numerical or alphabetic shorthand:

B first out. Directly to 2 , pause—a very human-like indecision. After 5 or 6 abortive starts each way, finally entered 2 and proceeded slowly to end. Turned and swiftly retraced his steps. At mouth of 2 joined by A . Together they proceeded placidly to end of 3 . Turned instantly and galloped back swiftly out of 3, not slowing up until e . Here B , charmed by the odor from C , stopped to dig. A , forward soberly, hesitated at x turning now right, now left, but finally on to n  …

The Hampton Court maze, designed to resemble both a rat's wild habitat and the carefully trimmed shrubbery of a British castle, became a neat metaphor for the taming of woolly Nature in the lab. Soon other researchers adapted Small's method with varying degrees of success. In 1902, A.J. Kinnaman put a pair of rhesus monkeys in a 17-foot-wide version of the same maze. A few years later, James Porter tried running some English sparrows and a female cowbird through the setup before resorting to a dumbed-down version with fewer blind alleys.

From James P. Porter, "Further Study of the English Sparrow and Other Birds," American Journal of Psychology, 1906.

This new approach to psychology moved in strange directions. In October 1911, a graduate student at the University of Chicago named Fleming Perrin began testing his professors and colleagues in a local amusement-park attraction called the "Mouse-trap." He blindfolded each subject with a band of black silk and set them loose in a 45-foot-wide, duodecagonal maze while he watched from a catwalk. "The subjects became so engrossed in the search for openings," he wrote, "that a few unlooked for bodily contacts with the ends of blind passages resulted."

From Fleming Allen Clay Perrin, An Experimental and Introspective Study of the Human Learning Process in the Maze, 1913

Attempts to run non-rodents through mazes were short-lived. Soon the rat became the standard animal in psychology , and the maze was the standard apparatus for the rat. One crucial innovation came from a young psychologist named James B. Watson, who for his dissertation sent rodents through a Hampton Court maze while under various degrees of sensory deprivation: Some rats he blinded; others he deafened; still others had their whiskers plucked or their paws covered over. The animals could navigate by "chain reflex" alone, he found—a kinesthetic sense of how to reach the food at the center. For one famous study, known as the "kerplunk" experiment, he altered the maze slightly after the rats had learned it. The animals were so accustomed to the original layout that they ran smack into the wall, kerplunk .

From Thomas William Brockbank, Redintegration in the albino rat, 1918.

Watson moved on from the Hampton Court layout to a circular maze of his own design, with concentric, interconnected passages. Lamps hung over the setup and researchers could track and trace the paths of their rodents by means of a camera lucida. Through a series of experiments conducted on this and other mazes, he developed a theory of behavior that recognized "no dividing line between man and brute." You could learn everything you might want to know about human psychology, he suggested, from the behavior of rats in a maze.

A golden age of labyrinths soon followed. The psychologists John Connors and Richard E. Brown describe this period in an exhaustive, unpublished paper on the history of the rodent maze. A vast array of new designs made their way into the literature: Walter R. Miles, for example, perfected the elevated trestle maze—a precursor to more simple, modern variants like the Elevated Plus Maze —in his garage near Stanford University. Rats raced around an open structure that resembled a railroad bridge or roller coaster, impelled toward their goal by a fear of heights. The new design made it easier to observe the rats' movements, and the maze could be collapsed for storage.

Reproduced in Norman Munn, Handbook of Psychological Research on the Rat, 1950.

Back at Clark University, Walter S. Hunter came up with a maze that looked like an M. C. Escher drawing: the Double-Alternation Tridimensional Spatial Maze. Rats had to navigate an upward spiral that required repeating right - right and left - left turns. This setup proved extremely difficult to learn (at least for blinded animals with their whiskers plucked): Just six of the 23 rats tested learned to make it through without an error.

The golden age of maze-building would soon come to an end, however. In the 1920s, the psychologist B.F. Skinner put rats through mazes as many of his colleagues did, but by the end of the following decade his faith in the method had waned. He began testing rats and pigeons in a bare-bones, lever-pressing apparatus. As Skinner's influence grew over the next few decades, conventional maze research fell into decline. Psychologists turned their attention toward the study of reinforcement schedules and stimulus-response relationships that could be measured without having to build a double-alternating, tridimensional spiral.

Reproduced in Handbook of Psychological Research on the Rat from Skinner, 1938, p. 49.

In the 1940s, what mazes remained were often reduced to simpler forms, such as the T-maze or the Y-maze, in which an animal has only one choice to make— left or right . The pared-down, binary layout was more amenable to statistical analysis, and it may have resonated with the developing science of information theory . At a conference on cybernetics in 1951, the father of that field, Claude Shannon, presented his own, newfangled version of the rodent maze. His lab animal was an electronic rat named "Theseus," which used some rudimentary rules of machine learning to navigate a walled grid like a Roomba .

Reprinted with permission of Alcatel- Lucent USA Inc.

Some new, unfussy mazes (for live rodents) emerged in the late-1970s, including the Barnes Maze, which consisted only of a circular platform, 4 feet in diameter, with a series of holes drilled along its perimeter—like an oversized View-Master reel laid flat. The design took advantage of a rat or mouse's tendency to seek shelter from brightly lit, open spaces. An animal placed at the center of the maze checks out each hole until it finds one that leads to the safety of a dark, enclosed box. It's a task that's simple enough for mice (which tend to be a little duller than rats), and, like the Morris Water Maze, it doesn't require any electric shocks or dietary inducements.

The elaborated maze had nearly gone extinct. As mice started to replace rats for laboratory experiments in the 1980s and 1990s, rodent mazes grew even more standardized and simplified. The Barnes Maze, along with a few others—the Morris Water Maze, the Radial Maze, the Elevated Plus, the Elevated Zero, and the Y- and the T-Maze—became ubiquitous.

In 2009, a molecular biologist and physicist from Princeton named David Tank introduced the next major step in the evolution of maze research. His setup includes no walls or passageways, no platforms or tubs of water. Instead, he clamps the heads of his Black-6 mice in place and leaves their paws free to skitter across the surface of an 8-inch Styrofoam sphere suspended on a cushion of air. It works like a giant trackball, which the mouse can use to maneuver through a virtual maze (built using software from the video game Quake II ) that's projected onto a video screen.

Tank's approach allows him to image the brains of his mice through a microscope, or measure the activity of their neurons, as they move through an imaginary space. Having designed a spatial learning task that's entirely removed from real-world constraints, scientists can now create any sort of maze imaginable, no matter how convoluted or obtuse. They can even test their mice on a course that distorts or shifts in the middle of an experiment. But so far, Tank has only constructed two virtual mazes: The first is a linear track, and the second is in a basic T-shape. That's the fate of the modern maze—the research gets more advanced, but the task gets simpler. More from this series: How mouse research could be limiting our knowledge of human disease , the one mouse who rules over all other mice , and the invincible naked mole-rat .

Shawn Maust

Swimming Rats and the Power of Hope

A few weeks ago, I learned about a(n infamous) study done back in 1957 by Dr. Curt Richter. In it, he and his team did experiments on rats. They found that if the water temperature wasn’t too hot or too cold, domesticated Norwegian rats were able to swim around 40-60 hours on average. But when they put wild Norwegian rats into the exact same situation, they would would die within 15 minutes. They were the same breed of rat; the only difference was one group was domesticated and the other was not.

The disparity of results for the two groups was significant, and so they tried to track down what may have contributed to it. In further experiments, they found that if they put the wild rats into the water, and then pulled them out after a few minutes — and then repeated this a few times before the final testing —  the rats would end up lasting about the same length of time as the domesticated group when the final test was run.

Why did these few additional immersions in the water increase the endurance of these rats from 15 minutes to dozens of hours? The team postulated that the deaths were more psychological than physiological; that the real issue was one of hopelessness.

The wild rats were not used to being confined, so as soon as they were thrown into this new environment, one which seemed impossible to escape, they simply gave up. But if they had already been exposed to this same environment, and had then been removed, they knew that there was a chance the researchers could take them out of the water at any moment. It was no longer entirely hopeless, and they ended up lasting way longer than they would otherwise.

Putting the ethics of the study aside for a moment, these findings highlight the role hope can play in our lives. Being in a state of hopelessness can strip us of our energy and motivation to continue. But finding hope can provide the strength to endure far longer than we may have expected.

Difficult times are, by definition, difficult. But they’re nothing compared to the weight of hopelessness. Which means that giving the gift of hope to someone who needs it may be one of the most valuable gifts we could ever give.

Microplastics Infiltrate Every Organ, Including Brain, Study in Mice Shows

Scientists investigating the possible health effects of microplastics have uncovered some disturbing initial results in an experiment based on mice.

When old and young rodents drank microscopic fragments of plastic suspended in their water over the course of three weeks, researchers at the University of Rhode Island found traces of the pollutants had accumulated in every organ of the tiny mammal's body, including the brain.

The presence of these microplastics was also accompanied by behavioral changes akin to dementia in humans, as well as changes to immune markers in the liver and brain.

"To us, this was striking. These were not high doses of microplastics, but in only a short period of time, we saw these changes," explains neuroscientist Jaime Ross.

"Nobody really understands the life cycle of these microplastics in the body, so part of what we want to address is the question of what happens as you get older. Are you more susceptible to systemic inflammation from these microplastics as you age? Can your body get rid of them as easily? Do your cells respond differently to these toxins?"

The results may not translate directly to humans, but studies involving animal models like these are a key first step in clinical research.

Recently, scientists have found microplastics hiding in the human intestine , circulating in our bloodstream , gathering deep in the lungs , and seeping through to the placenta .

In 2021, toxicologists warned that future studies urgently need to address what these pollutants are doing to our health, especially since exposure is now all but impossible to avoid.

In the recent experiments, both old and young mice were given water that was treated with microplastics made of fluorescent polystyrene.

Some of the mice were also given normal drinking water as a control.

During the three-week trial, the mice had their behavior regularly assessed during open field tests that encourage exploratory behavior. They also undertook light-dark preference tests, which are based on a rodent's natural aversion to brightly lit areas.

Compared to the control group, mice who drank microplastic-contaminated water for three weeks showed significant behavioral changes, changes that were especially pronounced among older mice.

At the conclusion of the three weeks, red fluorescent particles of microplastics were found in every type of tissue the team examined: the brain, liver, kidney, gastrointestinal tract, heart, spleen, and lungs. The plastics were also in the poop and urine of the mice.

The fact that the pollutants were detected outside the digestive system suggests they are undergoing systemic circulation.

Their presence in the brain is especially concerning. It indicates that these potentially toxic pollutants can cross the immune barrier that separates the central nervous system from the rest of the body's bloodstream, possibly leading to neurocognitive issues.

The findings join another study from earlier this year that found microplastics in the brains of mice just two hours after eating a contaminated meal.

In 2022, a s imilar study also found that ingested polystyrene microplastics can accumulate in the brains of mice, triggering inflammation and impairing their memory. This study did not, however, identify any behavioral changes among mice during an open field test.

Despite the discrepancies between results, Ross and her colleagues argue it is now evident polystyrene microplastics can travel to the mammal brain and exert detrimental effects after absorption.

In their recent study, they found a protein called GFAP, which supports cells in the brain, had decreased in abundance following the ingestion of microplastics.

"A decrease in GFAP has been associated with early stages of some neurodegenerative diseases, including mouse models of Alzheimer's disease, as well as depression ," says Ross.

"We were very surprised to see that the microplastics could induce altered GFAP signaling."

Ross plans to investigate these concerning changes in future research.

The study was published in the International Journal of Molecular Science .

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Transl Neurosci
• v.13(1); 2022 Jan 1

Mice can recognise water depths and will avoid entering deep water

Hiroshi ueno.

Department of Medical Technology, Kawasaki University of Medical Welfare, Okayama 701-0193, Japan

Yu Takahashi

Department of Psychiatry, Kawasaki Medical School, Kurashiki 701-0192, Japan

Shunsuke Suemitsu

Shinji murakami, naoya kitamura, yosuke matsumoto.

Department of Neuropsychiatry, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University, Okayama 700-8558, Japan

Motoi Okamoto

Department of Medical Technology, Graduate School of Health Sciences, Okayama University, Okayama 700-8558, Japan

Takeshi Ishihara

Rodents are averse to bodies of water, and this aversion has been exploited in experiments designed to study stress in mice. However, a few studies have elucidated the characteristics of murine water aversion. In this study, we investigated how mice behave in and around areas filled with water. Using variants of the open field test that contained pools of water at corners or sides of the field, we recorded the movements of mice throughout the field under various conditions. When the water was 8 mm deep, the mice explored the water pool regardless of whether an object was placed within it, but when the water was 20 mm deep, the mice were less willing to enter it. When the mice were placed on a dry area surrounded by 3 mm-deep water, they explored the water, but when they were surrounded by 8 mm-deep water, they stayed within the dry area. Our results indicate that mice exhibit exploratory behaviours around water, they can recognise water depths and avoid unacceptably deep water, and their willingness to enter water may be reduced by situational anxiety. Our experimental method could be used to investigate water-related anxiety-like behaviours in mice.

1. Background

In the wild, mice exhibit a tendency to avoid water as much as possible [ 1 ], and when mice are placed in water, they fall and immediately stiffen. Although it is difficult to determine the exact feelings of mice with regard to water, these observations clearly indicate that they do not like water. The aversion of mice to water has been exploited in the design of various behavioural tests, including the forced swim test, the water T-maze test, and the Morris water maze test [ 2 ]. Furthermore, this aversion is frequently used to induce chronic stress in mice through repeated forced swimming sessions and the placement of wet rugs in breeding cages [ 3 , 4 , 5 ].

The irrational fear of water in humans is known as aquaphobia, and aquaphobia is among the common simple phobias. Phobias, which are defined as abnormal psychological and physiological fears for a specific thing [ 6 ], are classified as anxiety disorders in the tenth revision of the International Statistical Classification of Diseases and Related Health Problems and are common comorbidities in patients with other anxiety disorders [ 7 , 8 ]. A past investigation [ 9 ] reported an aquaphobia prevalence of 1.8%, or approximately 1 in 50 people, in the general Icelandic population, and the symptoms of aquaphobia, which can include headache, suffocation, panic attacks, and decreased water intake [ 10 ], can adversely affect productivity, confidence levels, and overall health. However, only 9% of patients with general phobias report having consulted a physician about their conditions [ 9 ].

The causes of phobias remain largely undetermined. Researchers have speculated that aquaphobia may arise from a combination of genetics and experiential factors (e.g. swimming ability and instances of needing to be rescued from water) [ 11 ]. It is often thought that experiential factors are the most important contributors to aquaphobia in adults, but Poulton et al. found no association between swimming experience during the first 9 years of life and aquaphobia at the age of 18 years [ 12 ]. In accordance with Darwin’s non-associative model of fear acquisition, aquaphobia may constitute a type of innate fear that can manifest without any history of distressing experiences. Such innate fears may diminish over time due to repeated safe exposures to fear-inducing stimuli [ 13 ], and people can indeed learn to overcome or manage aquaphobia.

The neural mechanisms underlying phobias and innate fears are unknown, and this lack of knowledge has prevented the development of any mechanistic therapies or diagnostic markers for aquaphobia [ 14 ]. The advancement of therapeutic strategies for aquaphobia thus depends on the acquisition of experimental evidence identifying the neurochemical and neuroanatomical pathways underlying phobias [ 15 ], and given their aversion to water, mice are a tempting model organism for investigations into aquaphobia. However, it is unclear whether the aversion of mice to water truly reflects a fear of water, and certain key parameters of murine water aversion that could clarify the matter remain unexplored. For example, researchers have not fully determined the degree to which mice will modify their behaviours to avoid water, and it remains unknown whether a mouse’s behaviour around a body of water depends on the water’s depth.

To elucidate the behavioural parameters of murine water aversion and facilitate the development of new behavioural tests that could aid the identification of relevant neural circuits, we investigated the behaviours of mice when placed in proximity to water. We examined whether and under what conditions mice placed in a box with an area of shallow water would approach and enter the water.

2.1. Tests with an object in the water

To determine whether a mouse’s interest in exploring objects could tempt it to explore a water pool containing an object, we compared the behaviours of mice in an enriched environment (i.e. one with objects present both within and outside the water pool) with their behaviours in an empty environment (i.e. one with objects only present within the water pool) ( Figures 1a and b ). We observed no significant between-condition differences in the total distance travelled ( Figures 1c and d ), the number of entries into the zone surrounding the water pool ( Figure 1e ), or the total time spent in that zone ( Figure 1f ). However, we observed that the number of entries into the water pool and the total time spent in the water were greater under the empty environment condition than under the enriched environment condition ( Figures 1g and h ; Supplementary Videos 1 and 2), which suggests that the mice were more motivated to explore the water pool when the water pool was the only area that contained an object.

Behavioural tests featuring an object in the water pool. (a) Schematics of the experimental environments. (b) Outline of the experimental protocol. (c) Sample traces of a mouse’s movements through the enriched and empty environments. (d–h) Boxplots showing the total distance travelled (d), the number of entries into the zone surrounding the water pool (e), the total time spent in that zone (f), the number of entries into the water (g), and the total time spent in the water (h) under each experimental condition. Statistical significance was defined as * p < 0.05.

2.2. Murine interest in the water pool

To determine whether mice would exhibit any interest in exploring a water pool by itself, we compared the behaviours of mice in an enriched environment (i.e. one with objects present outside the water pool) with their behaviours in an empty environment (i.e. one without any objects) ( Figure 2a ). We observed no significant between-condition difference in the total distance travelled ( Figures 2b and c ), but we observed that the number of entries into the zone surrounding the water pool and the total time spent in that zone were greater under the empty environment condition than under the enriched environment condition ( Figures 2d and e ). We also observed more entries into the water pool under the empty environment condition than under the enriched environment condition ( Figure 2f ; Supplementary Videos 3 and 4), but we observed no significant difference in the total time spent in the water ( Figure 2g ). These results indicate that the willingness of a mouse to enter the water is not dependent on the presence of objects in the water pool. However, the fact that the mice spent more time walking around the water pool than in it suggests that the mice were still hesitant to enter the water.

Behavioural tests without an object in the water pool. (a) Schematics of the experimental environments. (b) Sample traces of a mouse’s movements through the enriched and empty environments. (c–g) Boxplots showing the total distance travelled (c), the number of entries into the zone surrounding the water pool (d), the total time spent in that zone (e), the number of entries into the water (f), and the total time spent in the water (g) under each experimental condition. Statistical significance was defined as * p < 0.05.

2.3. Effects of variable water depths on mouse behaviours

To determine the effects of water depths on a mouse’s willingness to enter the water, we compared the behaviours of mice in the presence of an 8 mm-deep water pool with those in the presence of a 20 mm-deep water pool ( Figures 3a and b ). We observed no significant between-condition difference in the total distance travelled ( Figures 3c and d ), but we observed that the number of entries into the zone surrounding the water pool and the time spent within that zone were significantly greater under the 20 mm depth condition than under the 8 mm depth condition ( Figures 3e and f ). The number of entries into the water pool and the total time spent in the water were both lower under the 20 mm depth condition than under the 8 mm depth condition, with most mice not entering the water at all under the 20 mm depth condition ( Figures 3g and h ; Supplementary Video 5). These results suggest that mice can determine the depth of a pool of water.

Behavioural tests with variable water depths. (a) Schematic of the experimental environments. (b) Schematic of the difference in water depths. (c) Sample traces of a mouse’s movements under the 8 and 20 mm depth conditions. (d–h) Boxplots showing the total distance travelled (d), the number of entries into the zone surrounding the water pool (e), the total time spent in that zone (f), the number of entries into the water (g), and the total time spent in the water (h) under each experimental condition. Statistical significance was defined as * p < 0.05.

2.4. Behaviours of mice surrounded by water

To determine how a stressful situation affects a mouse’s willingness to enter the water, we compared the behaviours of mice surrounded by 3 mm-deep water with those of mice surrounded by 8 mm-deep water ( Figures 4a and b ). We observed no significant between-condition difference in the total distance travelled ( Figure 4d ), but we observed that the mice only entered the water under the 3 mm depth condition ( Figures 4c, e and f ; Supplementary Video 6). These results constitute further evidence that mice can determine the depth of a pool of water. They also indicate that being surrounded by water reduces a mouse’s willingness to enter the water.

Behavioural tests with the mice surrounded by water. (a) Schematic of the experimental environments. (b) Schematic of the difference in water depths. (c) Sample traces of a mouse’s movements under the 3 mm and 8 mm depth conditions. (d–f) Boxplots showing the total distance travelled (d), the number of entries into the water (e), and the total time spent in the water (f) under each experimental condition. Statistical significance was defined as * p < 0.05.

3. Discussion

In this study, we have shown that a strain of mice widely used in experiments can recognise water depths, they are unwilling to enter unacceptably deep bodies of water, and the distressing situation of being surrounded by water reduces their depth tolerance.

Although the mice frequently exhibited some apparent hesitancy about entering the water, with protracted periods in which they restricted themselves to exploring the zones surrounding the water, they usually exhibited an eventual willingness to enter and explore the water pools regardless of whether objects were present within the open field device. These behaviours differ markedly from those observed in situations where mice are presumably simply attempting to avoid potential predators, such as the tendency of mice in a boxed environment to focus on exploration near the wall while avoiding the less protected open areas [ 16 , 17 ]. Indeed, the water-entry behaviours that we observed are more reminiscent of the risk-accepting behaviours after initial periods of risk-avoidance observed in various behavioural tests [ 18 ]. For example, in the open field test, light/dark transition test, and elevated plus maze test, mice initially avoid open areas, highly illuminated areas, and heights, but they also exhibit an eventual willingness to explore new and potentially risky spaces [ 19 – 21 ]. The behaviours of mice when confronted with novelty are thus determined by a conflict between the willingness to explore unknown areas and objects and the motivation to avoid potential danger. The willingness of mice in our experiments to approach and enter bodies of water probably reflects an innate motivation to explore new environments and is thus analogous to the willingness of mice to enter the open arms in the elevated plus maze test, the central area in the open field test, and the illuminated area in the light/dark transition test.

Avoidance behaviours depend on an animal’s senses and are influenced by its motor activity, motivational factors, and search strategies [ 22 ]. The elevated plus maze test, light/dark transition test, and open field test are all designed to evaluate anxiety-like behaviours by taking advantage of known avoidance behaviours [ 23 ], but the results obtained from the different tests are sometimes inconsistent [ 20 , 24 – 26 ]. Discrepancies may arise from the fact that these tests are based on distinct anxiety-like behaviours [ 23 ]: avoidance of illuminated spaces in the light/dark transition test, avoidance of open spaces in the open field test, and avoidance of heights in the elevated plus maze test [ 27 ]. The observed discrepancies suggest that different forms of anxiety may involve distinct mechanisms, and this, in turn, implies that it is important for researchers to have access to diverse tests with which to assess different forms of anxiety [ 28 , 29 ]. Our findings concerning the characteristics of murine water avoidance may be used to develop a new test of anxiety-like behaviours based on water avoidance.

In this study, mice exhibited a markedly reduced willingness to enter the water when the depth was increased to 20 mm, even though they could still walk through the water without needing to swim at that depth. Mice can swim but they will normally avoid water as much as possible [ 1 ]. Our results clearly show that mice can recognise the depth of a body of water and can choose to avoid entering deep water, just as experiments with elevated plus mazes have shown that mice can recognise heights and can choose to avoid them [ 30 ]. To the best of our knowledge, our study is the first to provide empirical evidence that mice are more willing to enter shallow water than to enter deep water.

Interestingly, we found that when mice were placed in a central area surrounded by water, they exhibited an increased aversion to deep water, with no entries into the water with an 8 mm depth that had been acceptable for mice that were not surrounded by water. This suggests that being surrounded by water prompted increased feelings of anxiety in the mice and reduced their willingness to engage in exploratory behaviours, a finding that is consistent with past investigations showing that anxiety suppresses exploratory behaviours [ 31 ]. Other factors, such as situation complexity, novelty, and the animal’s baseline emotional state, can also reduce a mouse’s willingness to explore a new environment [ 32 ].

Anxiety and fear are normal emotions that are selected for in the evolutionary process because they aid an organism in avoiding dangerous situations. For example, humans can experience fear around water because of the risk of drowning. However, individuals with aquaphobia experience abnormal symptoms around water such as headaches, feelings of suffocation, panic attacks, and decreased water intake [ 10 ]. Fear occurs in response to threats, but the physiological mechanisms underlying anxious behaviours remain unclear [ 33 ]. For small rodents, entering small spaces, holes, or tunnels is an important behaviour, but mice that lack leucine-rich repeat transmembrane neuronal proteins exhibit claustrophobia-like phenotypes that involve avoidance of small enclosures [ 34 ]. Interestingly, mice with hippocampal lesions also exhibit an unwillingness to enter small holes and tunnels [ 35 ]. Investigations into social phobias have found that such phobias may be related to interactions between the noradrenalinergic and serotonergic systems and the hypothalamic–pituitary–adrenal system [ 15 , 36 ]. Collectively, these findings indicate the existence of neural mechanisms underlying innate fears and offer clues as to how therapeutic strategies for phobias and anxiety could be developed. Our results add to the existing knowledge concerning phobias and may aid efforts to elucidate the mechanisms underlying water avoidance in mice and aquaphobia in humans.

4. Conclusion

Our results clearly indicate that mice exhibit exploratory behaviours in the context of entering shallow water. Furthermore, mice can recognise water depths and can choose not to enter the water if it is too deep. We speculate that the extent of a mouse’s exploratory behaviour in the presence of bodies of water is partially determined by anxieties related to water, and we propose that the dependence of a mouse’s exploratory behaviours on water depths could be used to design new tests of anxiety-like behaviours that could aid research into aquaphobia in humans.

5. Materials and methods

5.1. animals.

All efforts were made to minimise the number of animals used and to prevent unavoidable discomfort. Male C57BL/6N mice (age: 10 weeks) were purchased from Charles River Laboratories Japan (Kanagawa, Japan) and were housed five to a cage with food and water provided ad libitum under a 12 h light/dark cycle at 23–26°C.

Ethical approval: The research related to animals’ use has complied with all the relevant national regulations and institutional policies for the care and use of animals. All animal experiments were performed in accordance with the U.S. National Institutes of Health (NIH) – Guide for the Care and Use of Laboratory Animals (NIH Publication No. 80-23, revised in 1996) and approved by the Committee for Animal Experiments at the Kawasaki Medical School Advanced Research Center.

5.2. Behavioural tests

All behavioural tests were conducted in behavioural testing rooms between 09:00 h and 16:00 h during the light phase of the light/dark cycle. After the tests, the equipment was cleaned with 70% ethanol and super hypochlorous water to eliminate olfactory cues. Hypochlorous acid is an effective odour removal agent with a weak intrinsic odour [ 30 , 37 ]. The behavioural testing rooms were illuminated at a 100 lux intensity.

For the behavioural tests, we used an open field test apparatus that consisted of a 45 cm × 45 cm square area surrounded by 40 cm-high walls. The tests involved various arrangements of water pools and objects. Prior to object placement, each mouse was placed in the box for a 10 min free exploration period to produce habituation to the environment, after which the mouse was briefly returned to its home cage while object placement occurred. Unless otherwise noted, the water pools were filled during the habituation period. During the behavioural tests, data were video-recorded.

5.3. Tests with an object in the water pool

In this experiment, the open field included a 13.0 cm × 13.0 cm pool of 3 mm-deep water that was positioned at the centre of one wall ( Figure 1a ). A tower model was placed in the centre of the water pool, and a 4.0 cm × 4.0 cm cotton square, a 4.0 cm × 4.0 cm × 1.0 cm polystyrene rectangular prism, and a 4.0 cm × 8.0 cm × 4.0 cm wire cage were placed on the side of the open field opposite the water pool.

Ten mice were used in this experiment. In test 1, all objects were placed in the box ( Figure 1a ; enriched environment), and the mouse was placed in a corner before being allowed to move freely around the box for 12 min ( Figure 1b ). The mouse was then returned to its home cage for 5 min. In test 2, all objects except the tower model in the water pool were removed ( Figure 1a ; empty environment), and the mouse was again placed in a corner before being allowed to move freely around the box for 12 min ( Figure 1b ). The same mice were used in test 1 and test 2.

5.4. Tests of murine interest in the water

In this experiment, the open field included a 13.0 cm × 13.0 cm pool of 3 mm-deep water that was positioned in a corner ( Figure 2a ). A 4.0 cm × 4.0 cm cotton square, a 4.0 cm × 4.0 cm × 1.0 cm polystyrene rectangular prism, a 4.0 cm × 7.0 cm × 4.0 cm wire cage, and a 50 mL tube without a lid were placed in the areas away from the water pool.

Ten mice were used in this experiment. In test 1, all objects were placed in the box ( Figure 2a ; enriched environment), and the mouse was placed in a corner before being allowed to move freely around the box for 12 min. The mouse was then returned to its home cage for 5 min. In test 2, all objects were removed ( Figure 2a ; empty environment), and the mouse was again placed in a corner before being allowed to move freely around the box for 12 min. The same mice were used in test 1 and test 2.

5.5. Tests with variable water depths

In this experiment, the open field included a 14.0 cm × 20.0 cm water pool with a depth of 8 mm or 20 mm positioned in a corner ( Figures 3a and b ). The same container was used for all experiments to ensure equal container heights. A 4.0 cm × 4.0 cm cotton square, a 50 mL tube without a lid, and a third object (miniature-home) were placed in areas away from the water pool. Separate sets of 10 mice were used for tests involving the 8 mm and 20 mm depths. In this test, the mouse was placed in the corner before being allowed to move freely around the box for 12 min. In this experiment, each mouse was used in only one experiment.

5.6. Tests with water surrounding a dry zone

In this experiment, the open field consisted of a 13.0 cm × 13.0 cm dry central area that was surrounded by a water pool with a depth of 3 mm or 8 mm ( Figures 4a and b ). Separate sets of 10 mice were used for tests involving the 3 mm and 8 mm depths. In contrast to the other experiments, water was not added to the box until after the 10 min habituation period. In this test, the mouse was placed on the dry central area before being allowed to move freely around the box for 12 min. In this experiment, each mouse was used in only one experiment.

5.7. Data analyses

The video-recorded data were analysed with video-tracking software (ANY-MAZE; Stoelting, Wood Dale, IL, USA). For each 12 min test period, we determined the total distance travelled, the number of entries into the zone surrounding the water pool, the amount of time spent in that zone, the number of entries into the water pool, and the amount of time spent in the water pool. For comparing two groups, Student’s t -test was used for normally distributed data, and Mann–Whitney U test was used for not normally distributed data. In addition, one-way repeated-measures analysis of variance was used for normally distributed data, and the Friedman test was used for not normally distributed data. p < 0.05 was used as the definition of statistical significance. Statistical analyses were performed with SPSS software (IBM, Armonk, NY, USA).

Acknowledgements

We thank the Kawasaki Medical School Central Research Institute for providing the instruments used to perform this study. We also thank Editage ( www.editage.jp ) for English-language editing.

Funding information: This work was supported by a Grant Aid for the Sanyo Broadcasting Foundation and the Okayama Medical Foundation. The funding source had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; and in the decision to submit the article for publication.

Author contributions: All authors had complete access to all study data and assume complete responsibility for the integrity of the data and accuracy of the data analysis. Study concept and design: H.U., Y.T., M.O., and T.I. Acquisition of data: H.U., Y.T., S.S., and Y.T. Analysis and interpretation of data: H.U., S.S., and Y.T. Drafting of the manuscript: H.U. and M.O. Critical revision of the manuscript for important intellectual content: S.M., N.K., K.W., Y.M., and T.I. Study supervision: M.O. and T.I.

Conflict of interest : The authors state no conflict of interest.

Data availability statement: The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

August 4, 2020

Altered Mice Breathe Water instead of Air

Originally published in August 1968

On supporting science journalism

If you're enjoying this article, consider supporting our award-winning journalism by subscribing . By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.

“If by some special arrangement humans could be made to breathe water instead of air, serious obstacles to attempts to penetrate deeper into the ocean and to travel in outer space might be overcome. Suppose we prepare an isotonic solution that is like blood plasma in salt composition and charge this solution with oxygen under greater than normal pressure. Can a mammal breathe such a solution? I performed the first experiments, with mice as the experimental animals, at the University of Leiden in 1961. After their initial agitation, the mice quieted down and did not seem to be in any particular distress. They made slow, rhythmic movements of respiration, apparently inhaling and exhaling the liquid. It became evident that the decisive factor limiting the mice’s survival was not lack of oxygen but the difficulty of eliminating carbon dioxide at the required rate. —Johannes A. Kylstra”

— Scientific American , August 1968

More gems from Scientific American ’s first 175 years can be found on our anniversary archive page .

Mice experiments explain how addiction changes our brains

Experiments on mice show that drug abuse leads to permanent changes in the brain. meet one of the scientists who is trying to reverse this damage and treat addictive behaviour..

Addiction to drugs such as cocaine can cause long-term changes in dopamine producing neurons in the brain, making it hard to fully recover and beat the addiction. 

Christian Lüscher, a professor in neurology at University of Geneva, Switzerland, was the first to demonstrate this. His experiments with mice (see the video above) have provided insightful knowledge on the neurobiology behind addiction.

Lüscher presented his research at Speakers Corner, hosted by ScienceNordic during the FENS Forum for Neuroscience, 2016, which took place in Copenhagen, Denmark.

You can watch his full presentation in the video below.

Mice cannot get enough dopamine

FENS Forum of Neuroscience, 2016

This year Copenhagen hosted the FENS Forum for Neuroscience, 2016.

Science Nordic was there to film and report from the conference.

Head over to  Speakers Corner  to catch up with all our activities.

Lüscher and his team have been studying addiction using mice. In one of his experiments, the mice have a lever that they can press any time they want. When they press the button, a special optical sensor, which controls specific neurons in their brain, activates their dopamine neurons.

The mouse quickly start to show the tell tale signs of addictive behaviour as they happily keep pressing the lever to receive more and more dopamine.

“If after two hours we didn’t take them out of the cage, they wouldn’t eat, they wouldn’t drink, then they’d probably die quickly, but very happily,” says Lüscher, speaking during the conference.

But what exactly happens in their brains when the mice press the lever over and over again?

Read More:  What are the hottest trends in neuroscience?

Three mechanisms lead to addiction

“It all begins when addictive drugs activate a system in the brain called the mesolimbic reward system,” says Lüscher.

The mesolimbic reward system releases dopamine in the so-called ventral tegmental area (VTA) of the brain. Addictive substances such as nicotine, cocaine, or heroin, target a different cellular process, or a different part of the VTA.

Lüscher identified three cellular mechanisms that drive the increase of dopamine and therefore addiction.

Substances such as nicotine have a direct effect on the dopamine neurons to increase dopamine. Where as substances such as cocaine, amphetamines, and ecstasy have an indirect effect on the dopamine neurons to increases dopamine.

Substances such as opioids, GHB, benzodiazapines, and cannabis also have an indirect effect, but they affect the so-called GABA neurons, which control the activity of the dopamine neurons, and in doing so also increase dopamine in the brain.

But Lüscher and his team have not only identified what leads to addiction. They have also identified the traces of so-called addiction induced plasticity--the long-term changes in the brain caused by addictive substances.

In fact the addiction-induced changes in the brain remain, even if you stop taking drugs for long enough so that the drugs have completely left their system, says Lüscher.

Read More:  Test yourself: Are you addicted to shopping?

Reversing the damage caused by drugs

Lüscher and his team have spent several years identifying the mechanisms responsible for the long-term synaptic changes in the brain caused by addiction. Now they are investigating how they can reverse these changes with a method called 'deep brain stimulation'.

“During the last three to four years we’ve started to establish a line of translational research, where we believe that deep brain stimulation could be a technique that allows us to interfere very specifically with circuits that matter in motivated behaviour and also drug addiction,” he says.

Lüscher and his colleagues demonstrated this with mice experiments in a study published in the journal Neuron.

“This is something one can take further and this may actually then lead to new forms of deep brain stimulation. A technique that we can readily use in humans,” he says.

---------------------

Read the Danish version of this article on Videnskab.dk

Scientific links

• 'Sufficiency of Mesolimbic Dopamine Neuron Stimulation for the Progression to Addiction', 2015, Neuron, DOI: http://dx.doi.org/10.1016/j.neuron.2015.10.017

External links

• Christian Luscher

Related content

A light switch in the brain illuminates neural networks

Scientists can see which cells communicate with each other in the brain by flipping a neural light switch.

A new twist in understanding the brain’s maps

The brain’s GPS would be worthless if it simply contained maps of our surroundings that were not aligned to the real world. Now we know how this is done.

A universal theory of the brain

Norwegian researchers have recently found evidence of a generalized active network for cognitive functions of the brain.

Gambling addiction can be spotted in the brain

New study reveals impaired communication across various brain regions in compulsive gamblers. This suggests that gambling addiction may be more due to a deviation in the brain than a weakness of character.

Addiction - a growing menace

Addiction is on the rise in Europe. An increasing number of young people are unable to control their use of drugs, alcohol, sex, computer games, technology, shopping, dieting or exercise.

Many adolescents get far too little sleep. Might starting the school day later help?

New report: one in four 10-12-year-olds takes painkillers weekly.

This table is the cradle of Norwegian democracy

How many adolescents are truly satisfied with their lives?

Study links pregnant women's diet to autism risk in children.

Remarkable find: Archaeologists in Norway discovered the remains of a gauntlet on the seabed

Humans have always had difficulty attending

It's not the fault of technology. The idea of 'information overload' rests on a faulty assumption about attention, researcher argues.

Cancer-seeking 'missiles' may be the future weapon against breast cancer

The pH around a tumor is often different from the rest of the body. We can use this to develop antibodies that work like cancer-seeking missiles.

• Research Article
• Computational and Systems Biology
• Neuroscience

Mice in a labyrinth show rapid learning, sudden insight, and efficient exploration

• Matthew Rosenberg

• Division of Biology and Biological Engineering, California Institute of Technology, United States ;
• Division of Engineering and Applied Science, California Institute of Technology, United States ;
• Open access
• Copyright information

Share this article

Cite this article.

• Pietro Perona
• Markus Meister
• Copy to clipboard
• Download BibTeX
• Download .RIS

Introduction

Materials and methods, data availability, article and author information.

Animals learn certain complex tasks remarkably fast, sometimes after a single experience. What behavioral algorithms support this efficiency? Many contemporary studies based on two-alternative-forced-choice (2AFC) tasks observe only slow or incomplete learning. As an alternative, we study the unconstrained behavior of mice in a complex labyrinth and measure the dynamics of learning and the behaviors that enable it. A mouse in the labyrinth makes ~2000 navigation decisions per hour. The animal explores the maze, quickly discovers the location of a reward, and executes correct 10-bit choices after only 10 reward experiences — a learning rate 1000-fold higher than in 2AFC experiments. Many mice improve discontinuously from one minute to the next, suggesting moments of sudden insight about the structure of the labyrinth. The underlying search algorithm does not require a global memory of places visited and is largely explained by purely local turning rules.

How can animals or machines acquire the ability for complex behaviors from one or a few experiences? Canonical examples include language learning in children, where new words are learned after just a few instances of their use, or learning to balance a bicycle, where humans progress from complete incompetence to near perfection after crashing once or a few times. Clearly, such rapid acquisition of new associations or of new motor skills can confer enormous survival advantages.

In laboratory studies, one prominent instance of one-shot learning is the Bruce effect ( Bruce, 1959 ). Here, the female mouse forms an olfactory memory of her mating partner that allows her to terminate the pregnancy if she encounters another male that threatens infanticide. Another form of rapid learning accessible to laboratory experiments is fear conditioning, where a formerly innocuous stimulus gets associated with a painful experience, leading to subsequent avoidance of the stimulus ( Fanselow and Bolles, 1979 ; Bourtchuladze et al., 1994 ). These learning systems appear designed for special purposes, they perform very specific associations, and govern binary behavioral decisions. They are likely implemented by specialized brain circuits, and indeed great progress has been made in localizing these operations to the accessory olfactory bulb ( Brennan and Keverne, 1997 ) and the cortical amygdala ( LeDoux, 2000 ).

In the attempt to identify more generalizable mechanisms of learning and decision making, one route has been to train laboratory animals on abstract tasks with tightly specified sensory inputs that are linked to motor outputs via arbitrary contingency rules. Canonical examples are a monkey reporting motion in a visual stimulus by saccading its eyes ( Newsome and Paré, 1988 ), and a mouse in a box classifying stimuli by moving its forelimbs or the tongue ( Burgess et al., 2017 ; Guo et al., 2014 ). The tasks are of low complexity, typically a one bit decision based on 1 or 2 bits of input. Remarkably, they are learned exceedingly slowly: a mouse typically requires many weeks of shaping and thousands of trials to reach asymptotic performance; a monkey may require many months ( Carandini and Churchland, 2013 ).

What is needed therefore is a rodent behavior that involves complex decision making, with many input variables and many possible choices. Ideally, the animals would learn to perform this task without excessive intervention by human shaping, so we may be confident that they employ innate brain mechanisms rather than circuits created by the training. Obviously, the behavior should be easy to measure in the laboratory. Finally, it would be satisfying if this behavior showed a glimpse of rapid learning.

Navigation through space is a complex behavior displayed by many animals. It typically involves integrating multiple cues to decide among many possible actions. It relies intimately on rapid learning. For example, a pigeon or desert ant leaving its shelter acquires the information needed for the homing path in a single episode. Major questions remain about how the brain stores this information and converts it to a policy for decisions during the homing path. One way to formalize the act of decision-making in the laboratory is to introduce structure in the environment in the form of a maze that defines straight paths and decision points. A maze of tunnels is in fact a natural environment for a burrowing rodent. Early studies of rodent behavior did place the animals into true labyrinths ( Small, 1901 ), but their use gradually declined in favor of linear tracks or boxes with a single choice point.

We report here on the behavior of laboratory mice in a complex labyrinth of tunnels. A single mouse is placed in a home cage from which it has free access to the maze for one night. No handling, shaping, or training by the investigators is involved. By continuous video-recording and automated tracking, we observe the animal’s entire life experience within the labyrinth. Some of the mice are water-deprived and a single location deep inside the maze offers water. We find that these animals learn to navigate to the water port after just a few reward experiences. In many cases, one can identify unique moments of 'insight' when the animal’s behavior changes discontinuously. This all happens within ~1 h. Underlying the rapid learning is an efficient mode of exploration driven by simple navigation rules. Mice that do not lack water show the same patterns of exploration. This laboratory-based navigation behavior may form a suitable substrate for studying the neural mechanisms that implement few-shot learning.

Adaptation to the maze

At the start of the experiment, a single mouse was placed in a conventional mouse cage with bedding and food. A short tunnel offered free access to a maze consisting of a warren of corridors ( Figure 1A–B ). The bottom and walls of the maze were constructed of black plastic that is transparent in the infrared. A video camera placed below the maze captured the animal’s actions continuously using infrared illumination ( Figure 1B ). The recordings were analyzed offline to track the movements of the mouse, with keypoints on the nose, mid-body, tail base, and the four feet ( Figure 1D ). All observations were made in darkness during the animal’s subjective night.

The maze environment.

Top ( A ) and side ( B ) views of a home cage, connected via an entry tunnel to an enclosed labyrinth. The animal’s actions in the maze are recorded via video from below using infrared illumination. ( C ) The maze is structured as a binary tree with 63 branch points (in levels numbered 0,…,5) and 64 end nodes. One end node has a water port that dispenses a drop when it gets poked. Blue line in A and C: path from maze entry to water port. ( D ) A mouse considering the options at the maze’s central intersection. Colored keypoints are tracked by DeepLabCut: nose, mid body, tail base, four feet.

The logical structure of the maze is a binary tree, with 6 levels of branches, leading from the single entrance to 64 endpoints ( Figure 1C ). A total of 63 T-junctions are connected by straight corridors in a design with maximal symmetry ( Figure 1A , Figure 3—figure supplement 1 ), such that all the nodes at a given level of the tree have the same local geometry. One of the 64 endpoints of the maze is outfitted with a water port. After activation by a brief nose poke, the port delivers a small drop of water, followed by a 90 s time-out period.

After an initial period of exploratory experiments, we settled on a frozen protocol that was applied to 20 animals. Ten of these mice had been mildly water-deprived for up to 24 h; they received food in the home cage and water only from the port hidden in the maze. Another ten mice had free access to food and water in the cage, and received no water from the port in the maze. Each animal’s behavior in the maze was recorded continuously for 7 h during the first night of its experience with the maze, starting the moment the connection tunnel was opened (sample videos here ). The investigator played no role during this period, and the animal was free to act as it wished including travel between the cage and the maze.

All the mice except one passed between the cage and the maze readily and frequently ( Figure 1—figure supplement 1 ). The single outlier animal barely entered the maze and never progressed past the first junction; we excluded this mouse’s data from subsequent analysis. On average over the entire period of study the animals spent 46% of the time in the maze ( Figure 1—figure supplement 2 ). This fraction was similar whether or not the animal was motivated by water rewards (47% for rewarded vs 44% for unrewarded animals). Over time the animals appeared increasingly comfortable in the maze, taking breaks for grooming and the occasional nap. When the investigator lifted the cage lid at the end of the night some animals were seen to escape into the safety of the maze.

We examined the rate of transitions from the cage to the maze and how it depends on time spent in the cage ( Figure 1—figure supplement 3A ). Surprisingly the rate of entry into the maze is highest immediately after the animal returns to the cage. Then it declines gradually by a factor of 4 over the first minute in the cage and remains steady thereafter. This is a large effect, observed for every individual animal in both the rewarded and unrewarded groups. By contrast the opposite transition, namely exit from the maze, occurs at an essentially constant rate throughout the visit ( Figure 1—figure supplement 3B ).

The nature of the animal’s forays into the maze changed over time. We call each foray from entrance to exit a ‘bout’. After a few hesitant entries into the main corridor, the mouse engaged in one or more long bouts that dove deep into the binary tree to most or all of the leaf nodes ( Figure 2A ). For a water-deprived animal, this typically led to discovery of the reward port. After ~10 bouts, the trajectories became more focused, involving travel to the reward port and some additional exploration ( Figure 2B ). At a later stage still, the animal often executed perfect exploitation bouts that led straight to the reward port and back with no wrong turns ( Figure 2C ). Even at this late stage, however, the animal continued to explore other parts of the maze ( Figure 2D ). Similarly the unrewarded animals explored the maze throughout the night ( Figure 1—figure supplement 2 ). While the length and structure of the animal’s trajectories changed over time, the speed remained remarkably constant after ~50 s of adaptation ( Figure 2—figure supplement 1 ).

Sample trajectories during adaptation to the maze.

Four sample bouts from one mouse (B3) into the maze at various times during the experiment (time markings at bottom). The trajectory of the animal’s nose is shown; time is encoded by the color of the trace. The entrance from the home cage and the water port are indicated in panel A.

While Figure 2 illustrates the trajectory of a mouse’s nose in full spatio-temporal detail, a convenient reduced representation is the ‘node sequence’. This simply marks the events when the animal enters each of the 127 nodes of the binary tree that describes the maze (see Materials and methods and Figure 3—figure supplement 1 ). Among these nodes, 63 are T-junctions where the animal has three choices for the next node, and 64 are end nodes where the animal’s only choice is to reverse course. We call the transition from one node to the next a ‘step’. The analysis in the rest of the paper was carried out on the animal’s node sequence.

Few-shot learning of a reward location

We now examine early changes in the animal’s behavior when it rapidly acquires and remembers information needed for navigation. First, we focus on navigation to the water port.

The 10 water-deprived animals had no indication that water would be found in the maze. Yet, all 10 discovered the water port in less than 2000 s and fewer than 17 bouts ( Figure 3A ). The port dispensed only a drop of water followed by a 90 s timeout before rearming. During the timeout, the animals generally left the port location to explore other parts of the maze or return home, even though they were not obliged to do so. For each of the water-deprived animals, the frequency at which it consumed rewards in the maze increased rapidly as it learned how to find the water port, then settled after a few reward experiences ( Figure 3A ).

Few-shot learning of path to water.

( A ) Time line of all water rewards collected by 10 water-deprived mice (red dots, every fifth reward has a blue tick mark). ( B ) The length of runs from the entrance to the water port, measured in steps between nodes, and plotted against the number of rewards experienced. Main panel: All individual runs (cyan dots) and median over 10 mice (blue circles). Exponential fit decays by 1 / e over 10.1 rewards. Right panel: Histogram of the run length, note log axis. Red: perfect runs with the minimum length 6; green: longer runs. Top panel: The fraction of perfect runs (length 6) plotted against the number of rewards experienced, along with the median duration of those perfect runs.

How many reward experiences are sufficient to teach the animal reliable navigation to the water port? To establish a learning curve one wants to compare performance on the identical task over successive trials. Recall that this experiment has no imposed trial structure. Yet the animals naturally segmented their behavior through discrete visits to the maze. Thus, we focused on all the instances when the animal started at the maze entrance and walked to the water port ( Figure 3B ).

On the first few occasions these paths to water can involve hundreds of steps between nodes and their length scatters over a wide range. However, after a few rewards, the animals began taking the perfect path without detours (six steps, Figure 3—figure supplement 1 ), and soon that became the norm. Note the path length plotted here is directly related to the number of ‘turning errors’: every time the mouse turns away from the shortest path to the water port that adds two steps to the path length ( Equation 7 ). The rate of these errors declined over time, by a factor of e after ~10 rewards consumed ( Figure 3B ). Late in the night ~75% of the paths to water were perfect. The animals executed them with increasing speed; eventually, these fast ‘water runs’ took as little as 2 s ( Figure 3B ). Many of these visits went unrewarded owing to the 90 s timeout period on the water port.

In summary, after ~10 reward experiences on average the mice learn to navigate efficiently to the water port, which requires making six correct decisions, each among three options. Note that even at late times, long after they have perfected the ‘water run’, the animals continue to take some extremely long paths: a subject for a later section (Figure 7).

The role of cues attached to the maze

These observations of rapid learning raise the question 'How do the animals navigate?’ In particular, does the mouse build an internal representation that guides its action at every junction? Or does it place marks in the external environment that signal the route to the water port? In an extreme version of externalized cognition, the mouse leaves behind a trail of urine marks or other secretions as it walks away from the water port, and on a subsequent bout simply sniffs its way up the odor gradient ( Figure 4A ). This would require no internal representation.

Navigation is robust to rotation of the maze.

( A ) Logic of the experiment: The animal may have deposited an odorant in the maze (shading) that is centered on the water port. After 180 degree rotation of the maze, that gradient would lead to the image of the water port (blue dot). We also measure how often the mouse goes to two control nodes (magenta dots) that are related by symmetry. ( B ) Trajectory of mouse ‘A1’ in the bouts immediately before and after maze rotation. Time coded by color from dark to light as in Figure 2 . ( C ) Left: Cumulative number of rewards as well as visits to the water port, the image of the water port, and the control nodes. All events are plotted vs time before and after the maze rotation. Average over four animals. Middle and right: Same data with the counts centered on zero and zoomed in for better resolution.

The following experiment offers some partial insights. Owing to the design of the labyrinth one can rotate the entire apparatus by 180 degrees, open one wall and close another, and obtain a maze with the same structure ( Figure 4A ). Alternatively one can also rotate only the floor. After such a modification, all the physical cues attached to the rotated parts now point in the wrong direction, namely to the end node 180 degrees opposite the water port (the 'image location’). If the animal navigated to the goal following cues previously deposited in the maze, it should end up at that image location.

We performed a maze rotation on four animals after several hours of exposure, when they had acquired the perfect route to water. Immediately after rotation, three of the four animals went to the correct water port on their first entry into the maze, and before ever visiting the image location (e.g. Figure 4B ). The fourth mouse visited the image location once and then the correct water port ( Figure 4—figure supplement 1 ). The mice continued to collect water rewards efficiently even immediately after the rotation.

Nonetheless, the maze rotation did introduce subtle changes in behavior that lasted for an hour or more ( Figure 4C ). Visits to the image location were at chance levels prior to rotation, then increased by a factor of 1.8. Visits to the water port declined in frequency, although they still exceeded visits to the image location by a factor of 5. The reward rate declined by a factor of 0.7. These effects could be verified for each animal ( Figure 4—figure supplement 1 ). The speed of the mice was not disturbed ( Figure 4—figure supplement 2 ).

In summary, for navigation to the water port the experienced animals do not strictly depend on physical cues that are attached to the maze. This includes any material they might have deposited, but also pre-existing construction details by which they may have learned to identify locations in the maze. The mice clearly notice a change in these cues, but continue to navigate effectively to the goal. This conclusion applies to the time point of the rotation, a few hours into the experiment. Conceivably, the animal’s navigation policy and its use of sensory cues changes in the course of learning. This and many other questions regarding the mechanisms of cognition will be taken up in a separate study.

Discontinuous learning

While an average across animals shows evidence of rapid learning ( Figure 3 ) one wonders whether the knowledge is acquired gradually or discontinuously, through moments of ‘sudden insight’. To explore this we scrutinized more closely the time line of individual water-deprived animals in their experience with the maze. The discovery of the water port and the subsequent collection of water drops at a regular rate is one clear change in behavior that relies on new knowledge. Indeed, the rate of water rewards can increase rather suddenly ( Figure 3A ), suggesting an instantaneous step in knowledge.

Over time, the animals learned the path to water not only from the entrance of the maze but from many locations scattered throughout the maze. The largest distance between the water port and an end node in the opposite half of the maze involves 12 steps through 11 intersections ( Figure 5A ). Thus, we included as another behavioral variable the occurrence of long direct paths to the water port which reflects how directedly the animals navigate within the maze.

Sudden changes in behavior.

( A ) An example of a long uninterrupted path through 11 junctions to the water port (drop icon). Blue circles mark control nodes related by symmetry to the water port to assess the frequency of long paths occurring by chance. ( B ) For one animal (named C1) the cumulative number of rewards (green); of long paths (>6 junctions) to the water port (red); and of similar paths to the three control nodes (blue, divided by 3). All are plotted against the time spent in the maze. Arrowheads indicate the time of sudden changes, obtained from fitting a step function to the rates. ( C ) Same as B for animal B1. ( D ) Same as B for animal C9, an example of more continuous learning.

Figure 5—source data 1

Statistics of sudden changes in behavior.

Statistics of sudden changes in behavior. Summary of the steps in the rate of long paths to water detected in 5 of the 10 rewarded animals. Mean and standard deviation of the step time are derived from maximum likelihood fits of a step model to the data.

Figure 5B shows for one animal the cumulative occurrence of water rewards and that of long direct paths to water. The animal discovers the water port early on at 75 s, but at 1380 s the rate of water rewards jumps suddenly by a factor of 5. The long paths to water follow a rather different time line. At first they occur randomly, at the same rate as the paths to the unrewarded control nodes. At 2070 s the long paths suddenly increase in frequency by a factor of 5. Given the sudden change in rates of both kinds of events there is little ambiguity about when the two steps happen and they are well separated in time ( Figure 5B ).

The animal behaves as though it gains a new insight at the time of the second step that allows it to travel to the water port directly from elsewhere in the maze. Note that the two behavioral variables are independent: The long paths don’t change when the reward rate steps up, and the reward rate doesn’t change when the rate of long paths steps up. Another animal ( Figure 5C ) similarly showed an early step in the reward rate (at 860 s) and a dramatic step in the rate of long paths (at 2580 s). In this case, the emergence of long paths coincided with a modest increase (factor of 2) in the reward rate.

Similar discontinuities in behavior were seen in at least 5 of the 10 water-deprived animals ( Figure 5B , Figure 5—figure supplement 1 , Figure 5—source data 1 ), and their timing could be identified to a precision of ~200 s. More gradual performance change was observed for the remaining animals ( Figure 5D ). We varied the criterion of performance by asking for even longer error-free paths, and the results were largely unchanged and no additional discontinuity appeared. These observations suggest that mice can acquire a complex decision-making skill rather suddenly. A mouse may have multiple moments of sudden insight that affect different aspects of its behavior. The exact time of the insight cannot be predicted but is easily identified post-hoc. Future neurophysiological studies of the phenomenon will face the interesting challenge of capturing these singular events.

One-shot learning of the home path

For an animal entering an unfamiliar environment, the most important path to keep in memory may be the escape route. In the present case that is the route to the maze entrance, from which the tunnel leads home to the cage. We expected that the mice would begin by penetrating into the maze gradually and return home repeatedly so as to confirm the escape route, a pattern previously observed for rodents in an open arena ( Tchernichovski et al., 1998 ; Fonio et al., 2009 ). This might help build a memory of the home path gradually level-by-level into the binary tree. Nothing could be further from the truth.

At the end of any given bout into the maze, there is a ‘home run’, namely the direct path without reversals that takes the animal to the exit (see Figure 3—figure supplement 1 ). Figure 6A shows the nodes where each animal started its first home run, following the first penetration into the maze. With few exceptions, that first home run began from an end node, as deep into the maze as possible. Recall that this involves making the correct choice at six successive three-way intersections, an outcome that is unlikely to happen by chance.

Homing succeeds on first attempt.

( A ) Locations in the maze where the 19 animals started their first return to the exit (home run). Some locations were used by two or three animals (darker color). ( B ) Left: The cumulative number of home runs from different levels in the maze, summed over all animals, and plotted against the bout number. Level 1 = first T-junction, level 7 = end nodes. Right: Zoom of (Left) into early bouts. ( C ) Overlap between the outbound and the home path. Histogram of the overlap for all bouts of all animals. ( D ) Same analysis for just the first bout of each animal. The length of the home run is color-coded as in panel B.

The above hypothesis regarding gradual practice of home runs would predict that short home runs should appear before long ones in the course of the experiment. The opposite is the case ( Figure 6B ). In fact, the end nodes (level 7 of the maze) are by far the favorite place from which to return to the exit, and those maximal-length home runs systematically appear before shorter ones. This conclusion was confirmed for each individual animal, whether rewarded or unrewarded.

Clearly, the animals do not practice the home path or build it up gradually. Instead they seem to possess an Ariadne’s thread ( Apollodorus, 1921 ) starting with their first excursion into the maze, long before they might have acquired any general knowledge of the maze layout. On the other hand, the mouse does not follow the strategy of Theseus, namely to precisely retrace the path that led it into the labyrinth. In that case the animal’s home path should be the reverse of the path into the maze that started the bout. Instead the entry path and the home path tend to have little overlap ( Figure 6C ). Note the minimum overlap is 1, because all paths into and out of the maze have to pass through the central junction (node 0 in Figure 3—figure supplement 1 ). This is also the most frequent overlap. The peak at overlaps 6–8 for rewarded animals results from the frequent paths to the water port and back, a sequence of at least seven nodes in each direction. The separation of outbound and return path is seen even on the very first home run ( Figure 6D ). Many home runs from the deepest level (seven nodes) have only the central junction in common with the outbound path (overlap = 1).

In summary, it appears that the animal acquires a homing strategy over the course of a single bout, and in a manner that allows a direct return home even from locations not previously encountered.

Structure of behavior in the maze

Here, we focus on rules and patterns that govern the animal’s activity in the maze on both large and small scales.

Behavioral states

Once the animal has learned to perform long uninterrupted paths to the water port, one can categorize its behavior within the maze by three states: (1) walking to the water port; (2) walking to the exit; and (3) exploring the maze. Operationally we define exploration as all periods in which the animal is in the maze but not on a direct path to water or to the exit. For the 10 sated animals this includes all times in the maze except for the walks to the exit.

Figure 7 illustrates the occupancies and transition probabilities between these states. The animals spent most of their time by far in the exploration state: 84% for rewarded and 95% for unrewarded mice. Across animals there was very little variation in the balance of the three modes ( Figure 7—source data 1 ). The rewarded mice began about half their bouts into the maze with a trip to the water port and the other half by exploring ( Figure 7A ). After a drink, the animals routinely continued exploring, about 90% of the time.

Exploration is a dominant and persistent mode of behavior.

( A ) Ethogram for rewarded animals. Area of the circle reflects the fraction of time spent in each behavioral mode averaged over animals and duration of the experiment. Width of the arrow reflects the probability of transitioning to another mode. ‘Drink’ involves travel to the water port and time spent there. Transitions from ‘Leave’ represent what the animal does at the start of the next bout into the maze. ( B ) The fraction of time spent in each mode as a function of absolute time throughout the night. Mean ± SD across the 10 rewarded animals.

Figure 7—source data 1

Three modes of behavior.

( A ) The fraction of time mice spent in each of the three modes while in the maze. Mean ± SD for 10 rewarded and nine unrewarded animals. ( B ) Probability of transitioning from the mode on the left to the mode at the top. Transitions from ‘leave’ represent what the animal does at the start of the next bout into the maze.

For water-deprived animals, the dominance of exploration persisted even at a late stage of the night when they routinely executed perfect exploitation bouts to and from the water port: Over the duration of the night the ‘explore’ fraction dropped slightly from 0.92 to 0.75, with the balance accrued to the ‘drink’ and ‘leave’ modes as the animals executed many direct runs to the water port and back. The unrewarded group of animals also explored the maze throughout the night even though it offered no overt rewards ( Figure 7—source data 1 ). One suspects that the animals derive some intrinsic reward from the act of patrolling the environment itself.

Efficiency of exploration

During the direct paths to water and to the exit the animal behaves deterministically, whereas the exploration behavior appears stochastic. Here, we delve into the rules that govern the exploration component of behavior.

One can presume that a goal of the exploratory mode is to rapidly survey all parts of the environment for the appearance of new resources or threats. We will measure the efficiency of exploration by how rapidly the animal visits all end nodes of the binary maze, starting at any time during the experiment. The optimal agent with perfect memory and complete knowledge of the maze – including the absence of any loops – could visit the end nodes systematically one after another without repeats, thus encountering all of them after just 64 visits. A less perfect agent, on the other hand, will visit the same node repeatedly before having encountered all of them. Figure 8A plots for one exploring mouse the number of distinct end nodes it encountered as a function of the number of end nodes visited. The number of new nodes rises monotonically; 32 of the end nodes have been discovered after the mouse checked 76 times; then the curve gradually asymptotes to 64. We will characterize the efficiency of the search by the number of visits N 32 required to survey half the end nodes, and define

This mouse explores with efficiency E = 32/76 = 0.42. For comparison, Figure 8A plots the performance of the optimal agent ( E = 1.0) and that of a random walker that makes random decisions at every three-way junction ( E = 0.23). Note the mouse is about half as efficient as the optimal agent, but twice as efficient as a random walker.

Exploration covers the maze efficiently.

( A ) The number of distinct end nodes encountered as a function of the number of end nodes visited for: mouse C1 (red); the optimal explorer agent (black); an unbiased random walk (blue). Arrowhead: the value N 32 = 76 by which mouse C1 discovered half of the end nodes. ( B ) An expanded section of the graph in A including curves from 10 rewarded (red) and nine unrewarded (green) animals. The efficiency of exploration, defined as E = 32 / N 32 , is 0.385 ± 0.050 (SD) for rewarded and 0.384 ± 0.039 (SD) for unrewarded mice. ( C ) The efficiency of exploration for the same animals, comparing the values in the first and second halves of the time in the maze. The decline is a factor of 0.74 ± 0.12 (SD) for rewarded and 0.81 ± 0.13 (SD) for unrewarded mice.

The different mice were remarkably alike in this component of their exploratory behavior ( Figure 8B ): across animals the efficiency varied by only 11% of the mean (0.387 ± 0.044 SD). Furthermore, there was no detectable difference in efficiency between the rewarded animals and the sated unrewarded animals. Over the course of the night, the efficiency declined significantly for almost every animal – whether rewarded or not – by an average of 23% ( Figure 8C ).

Rules of exploration

What allows the mice to search much more efficiently than a random walking agent? We inspected more closely the decisions that the animals make at each three-way junction. It emerged that these decisions are governed by strong biases ( Figure 9 ). The probability of choosing each arm of a T-junction depends crucially on how the animal entered the junction. The animal can enter a T-junction from three places and exit it in three directions ( Figure 9A ). By tallying the frequency of all these occurrences across all T-junctions in the maze one finds clear deviations from an unbiased random walk ( Figure 9B , Figure 9—source data 1 ).

Turning biases favor exploration.

( A ) Definition of four turning biases at a T-junction based on the ratios of actions taken. Top: An animal arriving from the stem of the T (shaded) may either reverse or turn left or right. P SF is the probability that it will move forward rather than reversing. Given that it moves forward, P SA is the probability that it will take an alternating turn from the preceding one (gray), that is left-right or right-left. Bottom: An animal arriving from the bar of the T may either reverse or go straight, or turn into the stem of the T. P BF is the probability that it will move forward through the junction rather than reversing. Given that it moves forward, P BS is the probability that it turns into the stem. ( B ) Scatter graph of the biases P SF and P BF (left) and P SA and P BS (right). Every dot represents a mouse. Cross: values for an unbiased random walk. ( C ) Exploration curve of new end nodes discovered vs end nodes visited, displayed as in Figure 8A , including results from a biased random walk with the four turning biases derived from the same mouse, as well as a more elaborate Markov-chain model (see Figure 11C ). ( D ) Efficiency of exploration ( Equation 1 ) in 19 mice compared to the efficiency of the corresponding biased random walk.

Figure 9—source data 1

Bias statistics.

Statistics of the four turning biases. Mean and standard deviation of the 4 biases of  Figure 9A–B  across animals in the rewarded and unrewarded groups.

First, the animals have a strong preference for proceeding through a junction rather than returning to the preceding node ( P SF and P BF in Figure 9B ). Second, there is a bias in favor of alternating turns left and right rather than repeating the same direction turn ( P SA ). Finally, the mice have a mild preference for taking a branch off the straight corridor rather than proceeding straight ( P BS ). A comparison across animals again revealed a remarkable degree of consistency even in these local rules of behavior: The turning biases varied by only 3% across the population and even between the rewarded and unrewarded groups ( Figure 9B , Figure 9—source data 1 ).

Qualitatively, one can see that these turning biases will improve the animal’s search strategy. The forward biases P SF and P BF keep the animal from re-entering territory it has covered already. The bias P BS favors taking a branch that leads out of the maze. This allows the animal to rapidly cross multiple levels during an outward path and then enter a different territory. By comparison, the unbiased random walk tends to get stuck in the tips of the tree and revisits the same end nodes many times before escaping. To test this intuition, we simulated a biased random agent whose turning probabilities at a T-junction followed the same biases as measured from the animal ( Figure 9C ). These biased agents did in fact search with much higher efficiency than the unbiased random walk. They did not fully explain the behavior of the mice ( Figure 9D ), accounting for ~87% of the animal’s efficiency (compared to 60% for the random walk). A more sophisticated model of the animal’s behavior - involving many more parameters (Figure 11C) - failed to get any closer to the observed efficiency ( Figure 9C , Figure 8—figure supplement 1C ). Clearly some components of efficient search in these mice remain to be understood.

Systematic node preferences

A surprising aspect of the animals’ explorations is that they visit certain end nodes of the binary tree much more frequently than others ( Figure 10 ). This effect is large: more than a factor of 10 difference between the occupancy of the most popular and least popular end nodes ( Figure 10A–B ). This was surprising given our efforts to design the maze symmetrically, such that in principle all end nodes should be equivalent. Furthermore, the node preferences were very consistent across animals and even across the rewarded and unrewarded groups. Note that the standard error across animals of each node’s occupancy is much smaller than the differences between the nodes ( Figure 10B ).

Preference for certain end nodes during exploration.

( A ) The number of visits to different end nodes encoded by a gray scale. Top: rewarded, bottom: unrewarded animals. Gray scale spans a factor of 12 (top) or 13 (bottom). ( B ) The fraction of visits to each end node, comparing the rewarded vs unrewarded group of animals. Each data point is for one end node, the error bar is the SEM across animals in the group. The outlier on the bottom right is the neighbor of the water port, a frequently visited end node among rewarded animals. The water port is off scale and not shown. ( C ) As in panel B but comparing the unrewarded animals to their simulated 4-bias random walks. These biases explain 51% of the variance in the observed preference for end nodes.

The nodes on the periphery of the maze are systematically preferred. Comparing the outermost ring of 26 end nodes (excluding the water port and its neighbor) to the innermost 16 end nodes, the outer ones are favored by a large factor of 2.2. This may relate to earlier reports of a ‘centrifugal tendency’ among rats patrolling a maze ( Uster et al., 1976 ).

Interestingly, the biased random walk using four bias numbers ( Figure 9 , Figure 11D ) replicates a good amount of the pattern of preferences. For unrewarded animals, where the maze symmetry is not disturbed by the water port, the biased random walk predicts 51% of the observed variance across nodes ( Figure 10C ), and an outer/inner node preference of 1.97, almost matching the observed ratio of 2.20. The more complex Markov-chain model of behavior ( Figure 11C ) performed slightly better, explaining 66% of the variance in port visits and matching the outer/inner node preference of 2.20.

Recent history constrains the mouse’s decisions.

( A ) The mouse’s trajectory through the maze produces a sequence of states s t = n o d e o c c u p i e d a f t e r s t e p t s t = n o d e o c c u p i e d a f t e r s t e p t s t = n o d e o c c u p i e d a f t e r s t e p t . From each state, up to three possible actions lead to the next state (end nodes allow only one action). We want to predict the animal’s next action, a t + 1 , based on the prior history of states or actions. ( B–D ) Three possible models to make such a prediction. ( B ) A fixed-depth Markov chain where the probability of the next action depends only on the current state s t and the preceding state s t - 1 . The branches of the tree represent all 3 × 127 possible histories ( s t - 1 , s t ) . ( C ) A variable-depth Markov chain where only certain branches of the tree of histories contribute to the action probability. Here one history contains only the current state, some others reach back three steps. ( D ) A biased random walk model, as defined in Figure 9 , in which the probability of the next action depends only on the preceding action, not on the state. ( E ) Performance of the models in ( B,C,D ) when predicting the decisions of the animal at T-junctions. In each case we show the cross-entropy between the predicted action probability and the real actions of the animal (lower values indicate better prediction, perfect prediction would produce zero). Dotted line represents an unbiased random walk with 1/3 probability of each action.

Models of maze behavior

Moving beyond the efficiency of exploration one may ask more broadly: How well do we really understand what the mouse does in the maze? Can we predict its action at the next junction? Once the predictable component is removed, how much intrinsic randomness remains in the mouse’s behavior? Here, we address these questions using more sophisticated models that predict the probability of the mouse’s future actions based on the history of its trajectory.

At a formal level, the mouse’s trajectory through the maze is a string of numbers standing for the nodes the animal visited ( Figure 11A and Figure 3—figure supplement 1 ). We want to predict the next action of the mouse, namely the step that takes it to the next node. The quality of the model will be assessed by the cross-entropy between the model’s predictions and the mouse’s observed actions, measured in bits per action. This is the uncertainty that remains about the mouse’s next action given the prediction from the model. The ultimate lower limit is the true source entropy of the mouse, namely that component of its decisions that cannot be explained by the history of its actions.

One family of models we considered are fixed-depth Markov chains ( Figure 11B ). Here, the probability of the next action a t + 1 is specified as a function of the history stretching over the k preceding nodes ( s t - k + 1 , … , s t ) . In fitting the model to the mouse’s actual node sequence one tallies how often each history leads to each action, and uses those counts to estimate the conditional probabilities p ⁢ ( a t + 1 | s t - k + 1 , … , s t ) . Given a new node sequence, the model will then use the history strings ( s t - k + 1 , … , s t ) to predict the outcome of the next action. In practice, we trained the model on 80% of the animal’s trajectory and tested it by evaluating the cross-entropy on the remaining 20%.

Ideally, the depth k of these action trees would be very large, so as to take as much of the prior history into account as possible. However, one soon runs into a problem of over-fitting: Because each T-junction in the maze has three neighboring junctions, the number of possible histories grows as 3 k . As k increases, this quickly exceeds the length of the measured node sequence, so that every history appears only zero or one times in the data. At this point, one can no longer estimate any probabilities, and cross-validation on a different segment of data fails catastrophically. In practice, we found that this limitation sets in already beyond k = 2 ( Figure 11—figure supplement 1A ). To address this issue of data-limitation, we developed a variable-depth Markov chain ( Figure 11C ). This model retains longer histories, but only if they occur frequently enough to allow a reliable probability estimate (see Materials and methods, Figure 11—figure supplement 1B–C ). In addition, we explored different schemes of pooling the counts across certain T-junctions that are related by the symmetry of the maze (see Materials and methods).

With these methods, we focused on the portions of trajectory when the mouse was in ‘explore’ mode, because the segments in ‘drink’ and ‘leave’ mode are fully predictable. Furthermore, we evaluated the models only at nodes corresponding to T-junctions, because the decision from an end node is again fully predictable. Figure 11E compares the performance of various models of mouse behavior. The variable-depth Markov chains routinely produced the best fits, although the improvement over fixed-depth models was modest. Across all 19 animals in this study the remaining uncertainty about the animal’s action at a T-junction is 1.237 ± 0.035 (SD) bits/action, compared to the prior uncertainty of log 2 3 = 1.585 bits. The rewarded animals have slightly lower entropy than the unrewarded ones (1.216 vs 1.261 bits/action). The Markov chain models that produced the best fits to the behavior used history strings with an average length of ~4.

We also evaluated the predictions obtained from the simple biased random walk model ( Figure 11D ). Recall that this attempts to capture the history-dependence with just four bias parameters ( Figure 9A ). As expected, this produced considerably higher cross-entropies than the more sophisticated Markov chains (by about 18%, Figure 11E ). Finally, we used several professional file compression routines to try and compress the mouse’s node sequence. In principle, this sets an upper bound on the true source entropy of the mouse, even if the compression algorithm has no understanding of animal behavior. The best such algorithm (bzip2 compression Seward, 2019 ) far under-performed all the other models of mouse behavior, giving 43% higher cross-entropy on average, and thus offered no additional useful bounds.

We conclude that during exploration of the maze the mouse’s choice behavior is strongly influenced by its current location and ~3 locations preceding it. There are minor contributions from states further back. By knowing the animal’s history one can narrow down its action plan at a junction from the a priori 1.59 bits (one of three possible actions) to just ~1.24 bits. This finally is a quantitative answer to the question, ‘How well can one predict the animal’s behavior?’ Whether the remainder represents an irreducible uncertainty – akin to ‘free will’ of the mouse – remains to be seen. Readers are encouraged to improve on this number by applying their own models of behavior to our published data set.

Summary of contributions

We present a new approach to the study of learning and decision-making in mice. We give the animal access to a complex labyrinth and leave it undisturbed for a night while monitoring its movements. The result is a rich data set that reveals new aspects of learning and the structure of exploratory behavior. With these methods, we find that mice learn a complex task that requires six correct three-way decisions after only ~10 experiences of success ( Figure 2 , Figure 3 ). Along the way the animal gains task knowledge in discontinuous steps that can be localized to within a few minutes of resolution ( Figure 5 ). Underlying the learning process is an exploratory behavior that occupies 90% of the animal’s time in the maze and persists long after the task has been mastered, even in complete absence of an extrinsic reward ( Figure 7 ). The decisions the animal makes at choice points in the labyrinth are constrained in part by the history of its actions ( Figure 9 , Figure 11 ), in a way that favors efficient searching of the maze ( Figure 8 ). This microstructure of behavior is surprisingly consistent across mice, with variation in parameters of only a few percent ( Figure 9 ). Our most expressive models to predict the animal’s choices still leave a remaining uncertainty of ~1.24 bits per decision ( Figure 11 ), a quantitative benchmark by which competing models can be tested. Finally, some of the observations constrain what algorithms the animals might use for learning and navigation ( Figure 4 ).

Historical context

Mazes have been a staple of animal psychology for well over 100 years. The early versions were true labyrinths. For example, Small, 1901 built a model of the maze in Hampton Court gardens scaled to rat size. Subsequent researchers felt less constrained by Victorian landscapes and began to simplify the maze concept. Most commonly the maze offered one standard path from a starting location to a food reward box. A few blind alleys would branch from the standard path, and researchers would tally how many errors the animal committed by briefly turning into a blind ( Tolman and Honzik, 1930 ). Later on, the design was further reduced to a single T-junction. After all, the elementary act of maze navigation is whether to turn left or right at a junction ( Tolman, 1938 ), so why not study that process in isolation? And reducing the concept even further, one can ask the animal to refrain from walking altogether, and instead poke its nose into a hole on the left or the right side of a box ( Uchida and Mainen, 2003 ). This led to the popular behavior boxes now found in rodent neuroscience laboratories everywhere. Each of these reductions of the ‘maze’ concept enabled a new type of experiment to study learning and decision-making, for example limiting the number of choice points allows one to better sample neural activity at each one. However, the essence of a ‘confusing network of paths’ has been lost along the way, and with it the behavioral richness of the animals navigating those decisions.

Owing in part to the dissemination of user-friendly tools for animal tracking, one sees a renaissance of experiments that embrace complex environments, including mazes with many choice points ( Alonso et al., 2020 ; Wood et al., 2018 ; Sato et al., 2018 ; Nagy et al., 2020 ; Rondi-Reig et al., 2006 ; Yoder et al., 2011 ; McNamara et al., 2014 ), 3-dimensional environments ( Grobéty and Schenk, 1992 ), and infinite mazes ( Shokaku et al., 2020 ). The labyrinth in the present study is considerably more complex than Hampton Court or most of the mazes employed by Tolman and others ( Tolman and Honzik, 1930 ; Buel, 1934 ; Munn, 1950a ). In those mazes, the blind alleys are all short and unbranched; when an animal strays from the target path it receives feedback quickly and can correct. By contrast, our binary tree maze has 64 equally deep branches, only one of which contains the reward port. If the animal makes a mistake at any level of the tree, it can find out only after traveling all the way to the last node.

Another crucial aspect of our experimental design is the absence of any human interference. Most studies of animal navigation and learning involve some kind of trial structure. For example, the experimenter puts the rat in the start box, watches it make its way through the maze, coaxes it back on the path if necessary, and picks it up once it reaches the target box. Then another trial starts. In modern experiments with two-alternative-forced-choice (2AFC) behavior boxes the animal doesn’t have to be picked up, but a trial starts with appearance of a cue, and then proceeds through some strict protocol through delivery of the reward. The argument in favor of imposing a trial structure is that it creates reproducible conditions, so that one can gather comparable data and average them suitably over many trials.

Our experiments had no imposed structure whatsoever; in fact, it may be inappropriate to call them experiments. The investigator opened the entry to the maze in the evening and did not return until the morning. A potential advantage of leaving the animals to themselves is that they are more likely to engage in mouse-like behavior, rather than constantly responding to the stress of human interference or the alienation from being a cog in a behavior machine. The result was a rich data set, with the typical animal delivering ~15,000 decisions in a single night, even if one only counts the nodes of the binary tree as decision points. Since the mice made all the choices, the scientific effort lay primarily in adapting methods of data analysis to the nature of mouse trajectories. Somewhat surprisingly, the absence of experimental structure was no obstacle to making precise and reproducible measurements of the animal’s behavior.

How fast do animals learn?

Among the wide range of phenomena of animal learning, one can distinguish easy and hard tasks by some measure of task complexity. In a simple picture of a behavioral task, the animal needs to recognize several different contexts and based on that express one of several different actions. One can draw up a contingency table between contexts and actions, and measure the complexity of the task by the mutual information in that table. This ignores any task difficulties associated with sensing the context at all or with producing the desired actions. However, in all the examples discussed here, the stimuli are discriminated easily and the actions come naturally, thus the learning difficulty lies only in forming the associations, not in sharpening the perceptual mechanisms or practicing complex motor output.

Many well-studied behaviors have a complexity of 1 bit or less, and often animals can learn these associations after a single experience. For example, in the Bruce effect ( Bruce, 1959 ), the female maps two different contexts (smell of mate vs non-mate) onto two kinds of pregnancy outcomes (carry to term vs abort). The mutual information in that contingency table is at most one bit, and may be considerably lower, for example if non-mate males are very rare or very frequent. Mice form the correct association after a single instance of mating, although proper memory formation requires several hours of exposure to the mate odor ( Rosser and Keverne, 1985 ).

Similarly fear learning under the common electroshock paradigm establishes a mapping between two contexts (paired with shock vs innocuous) and two actions (freeze vs proceed), again with an upper bound of 1 bit of complexity. Rats and mice will form the association after a single experience lasting only seconds, and alter their behavior over several hours ( Fanselow and Bolles, 1979 ; Bourtchuladze et al., 1994 ). This is an adaptive warning system to deal with life-threatening events, and rapid learning here has a clear survival value.

Animals are particularly adept at learning a new association between an odor and food. For example, bees will extend their proboscis in response to a new odor after just one pairing trial where the odor appeared together with sugar ( Bitterman et al., 1983 ). Similarly, rodents will start digging for food in a scented bowl after just a few pairings with that odor ( Cleland et al., 2009 ). Again, these are 1-bit tasks learned rapidly after one or a few experiences.

By comparison, the tasks that a mouse performs in the labyrinth are more complex. For example, the path from the maze entrance to the water port involves six junctions, each with three options. At a minimum six different contexts must be mapped correctly into one of three actions each, which involves 6·log 2 3 = 9.5 bits of complexity. The animals begin to execute perfect paths from the entrance to the water port well within the first hour ( Figure 2C , Figure 3B ). At a later stage during the night, the animal learns to walk direct paths to water from many different locations in the maze ( Figure 5 ); by this time, it has consumed 10–20 rewards. In the limit, if the animal could turn correctly towards water from each of 63 junctions in the maze, it would have learned 63·log 2 3 = 100 bits. Conservatively, we estimate that the animals have mastered 10–20 bits of complexity based on 10–20 reward experiences within an hour of time spent in the maze. Note this considers only information about the water port and ignores whatever else the animals are learning about the maze during their incessant exploratory forays. These numbers align well with classic experiments on rats in diverse mazes and problem boxes Munn, 1950a . Although those tasks come in many varieties, a common theme is that ~10 successful trials are sufficient to learn ~10 decisions ( Woodrow, 1942 ).

In a different corner of the speed-complexity space are the many 2-alternative-forced-choice (2AFC) tasks in popular use today. These tend to be 1-bit tasks, for example the monkey should flick its eyes to the left when visual motion is to the left ( Newsome and Paré, 1988 ), or the mouse should turn a steering wheel to the right when a light appears on the left ( Burgess et al., 2017 ). Yet, the animals take a long time to learn these simple tasks. For example, the mouse with the steering wheel requires about 10,000 experiences before performance saturates. It never gets particularly good, with a typical hit rate only 2/3 of the way from random to perfect. All this training takes 3–6 weeks; in the case of monkeys several months. The rate of learning, measured in task complexity per unit time, is surprisingly low: less than 1 bit/month compared to ~10 bits/h observed in the labyrinth. The difference is a factor of 6000. Similarly when measured in complexity learned per reward experience: The 2AFC mouse may need 5000 rewards to learn a contingency table with one bit complexity, whereas the mouse in the maze needs ~10 rewards to learn 10 bits. Given these enormous differences in learning rate, one wonders whether the ultra-slow mode of learning has any relevance for an animal’s natural condition. In the month that the 2AFC mouse requires to finally report the location of a light, its relative in the wild has developed from a baby to having its own babies. Along the way, that wild mouse had to make many decisions, often involving high stakes, without the benefit of 10,000 trials of practice.

Sudden insight

The dynamics of the learning process are often conceived as a continuously growing association between stimuli and actions, with each reinforcing experience making an infinitesimal contribution. The reality can be quite different. When a child first learns to balance on a bicycle, performance goes from abysmal to astounding within a few seconds. The timing of such a discontinuous step in performance seems impossible to predict but easy to recognize after the fact.

From the early days of animal learning experiments, there have been warnings against the tendency to average learning curves across subjects ( Krechevsky, 1932 ; Estes, 1956 ). The average of many discontinuous curves will certainly look continuous and incremental, but that reassuring shape may miss the essence of the learning process. A recent reanalysis of many Pavlovian conditioning experiments suggested that discontinuous steps in performance are the rule rather than the exception ( Gallistel et al., 2004 ). Here, we found that the same applies to navigation in a complex labyrinth. While the average learning curve presents like a continuous function ( Figure 3B ), the individual records of water rewards show that each animal improves rather quickly but at different times ( Figure 3A ).

Owing to the unstructured nature of the experiment, the mouse may adopt different policies for getting to the water port. In at least half the animals, we observed a discontinuous change in that policy, namely when the animal started using efficient direct paths within the maze ( Figure 5 , Figure 5—source data 1 ). This second switch happened considerably after the animal started collecting rewards, and did not greatly affect the reward rate. Furthermore, the animals never reverted to the less efficient policy, just as a child rarely unlearns to balance a bicycle.

Presumably, this switch in performance reflects some discontinuous change in the animal’s internal model of the maze, what Tolman called the ‘cognitive map’ ( Tolman, 1948 ; Behrens et al., 2018 ). In the unrewarded animals, we could not detect any discontinuous change in the use of long paths. However, as Tolman argued, those animals may well acquire a sophisticated cognitive map that reveals itself only when presented with a concrete task, like finding water. Future experiments will need to address this. The discontinuous changes in performance pose a challenge to conventional models of reinforcement learning, in which reward events are the primary driver of learning and each event contributes an infinitesimal update to the action policy. It will also be important to model the acquisition of distinct kinds of knowledge that contribute to the same behavior, like the location of the target and efficient routes to approach it.

Exploratory behavior

By all accounts, the animals spent a large fraction of the night exploring the maze ( Figure 1—figure supplement 2 ). The water-deprived animals continued their forays into the depths of the maze long after they had found the water port and learned to exploit it regularly. After consuming a water reward, they wandered off into the maze 90% of the time ( Figure 7B ) instead of lazily waiting in front of the port during the timeout period. The sated animals experienced no overt reward from the maze, yet they likewise spent nearly half their time exploring that environment. As has been noted many times, animals – like humans – derive some form of intrinsic reward from exploration ( Berlyne, 1960 ). Some have suggested that there exists a homeostatic drive akin to hunger and thirst that elicits the information-seeking activity, and that the drive is in turn sated by the act of exploration ( Hughes, 1997 ). If this were the case then the drive to explore should be weakest just after an episode of exploration, much as the drive for food-seeking is weaker after a big meal.

Our observations are in conflict with this notion. The animal is most likely to enter the maze within the first minute of its return to the cage ( Figure 1—figure supplement 3 ), a strong trend that runs opposite to the prediction from satiation of curiosity. Several possible explanations come to mind: (1) On these very brief visits to the cage the animal may just want to certify that the exit route to the safe environment still exists, before continuing with exploration of the maze. (2) The temporal contrast between the boredom of the cage and the mystery of the maze is highest right at the moment of exit from the maze, and that may exert pressure to re-enter the maze. Understanding this in more detail will require dedicated experiments. For example, one could deliberately deprive the animals of access to the maze for some hours, and test whether that results in an increased drive to explore, as observed for other homeostatic drives around eating, drinking, and sleeping.

When left to their own devices, mice choose to spend much of their time engaged in exploration. One wonders how that affects their actions when they are strapped into a rigid behavior machine, like a 2AFC choice box. Presumably the drive to explore persists, perhaps more so because the forced environment is so unpleasant. And within the confines of the two alternatives, the only act of exploration the mouse has left is to give the wrong answer. This would manifest as an unexpectedly high error rate on unambiguous stimuli, sometimes called the 'lapse rate' ( Carandini and Churchland, 2013 ; Pisupati et al., 2021 ). The fact that the lapse rate decreases only gradually over weeks to months of training ( Burgess et al., 2017 ) suggests that it is difficult to crush the animal’s drive to explore.

The animals in our experiments had never been presented with a maze environment, yet they quickly settled into a steady mode of exploration. Once a mouse progressed beyond the first intersection it typically entered deep into the maze to one or more end nodes ( Figure 6 ). Within 50 s of the first entry, the animals adopted a steady speed of locomotion that they would retain throughout the night ( Figure 2—figure supplement 1 ). Within 250 s of first contact with the maze, the average animal already spent 50% of its time there ( Figure 1—figure supplement 2 ). Contrast this with a recent study of ‘free exploration’ in an exposed arena: Those animals required several hours before they even completed one walk around the perimeter ( Fonio et al., 2009 ). Here the drive to explore is clearly pitted against fear of the open space, which may not be conducive to observing exploration per se.

The persistence of exploration throughout the entire duration of the experiment suggests that the animals are continuously surveying the environment, perhaps expecting new features to arise. These surveys are quite efficient: The animals cover all parts of the maze much faster than expected from a random walk ( Figure 8 ). Effectively they avoid re-entering territory they surveyed just recently. It is often assumed that this requires some global memory of places visited in the environment ( Nagy et al., 2020 ; Olton, 1979 ). Such memory would have to persist for a long time: Surveying half of the available end nodes typically required 450 turning decisions. However, we found that a global long-term memory is not needed to explain the efficient search. The animals seem to be governed by a set of local turning biases that require memory only of the most recent decision and no knowledge of location ( Figure 9 ). These local biases alone can explain most of the character of exploration without any global understanding or long-term memory. Incidentally, they also explain other seemingly global aspects of the behavior, for example the systematic preference that the mice have for the outer rather than the inner regions of the maze ( Figure 10 ). Of course, this argument does not exclude the presence of a long-term memory, which may reveal itself in some other feature of the behavior.

Perhaps, the most remarkable aspect of these biases is how similar they are across all 19 mice studied here, regardless of whether the animal experienced water rewards or not ( Figure 9B , Figure 9—source data 1 ), and independent of the sex of the mouse. The four decision probabilities were identical across individuals to within a standard deviation of less than 0.03. We cannot think of a trivial reason why this should be so. For example the two biases for forward motion ( Figure 9B left) are poised halfway between the value for a random walk ( p = 2 / 3 ) and certainty ( p = 1 ). At either of those extremes, simple saturation might lead to a reproducible value, but not in the middle of the range. Why do different animals follow the exact same decision rules at an intersection between tunnels? Given that tunnel systems are part of the mouse’s natural ecology, it is possible that those rules are innate and determined genetically. Indeed the rules by which mice build tunnels have a strong genetic component ( Weber et al., 2013 ), so the rules for using tunnels may be written in the genes as well. The high precision with which one can measure those behaviors even in a single night of activity opens the way to efficient comparisons across genotypes, and also across animals with different developmental experience.

Finally, after mice discover the water port and learn to access it from many different points in the maze ( Figure 5 ), they are presumably eager to discover other things. In ongoing work, we installed three water ports (visible in the videos accompanying this article) and implemented a rule that activates the three ports in a cyclic sequence. Mice discovered all three ports rapidly and learned to visit them in the correct order. Future experiments will have to raise the bar on what the mice are expected to learn in a night.

Mechanisms of navigation

How do the animals navigate when they perform direct paths to the water port or to the exit? The present study cannot resolve that, but one can gain some clues based on observations so far. Early workers already concluded that rodents in a maze will use whatever sensory cues and tricks are available to accomplish their tasks ( Munn, 1950b ). Our maze was designed to restrict those options somewhat.

To limit the opportunity for visual navigation, the floor and walls of the maze are visually opaque. The ceiling is transparent, but the room is kept dark except for infrared illuminators. Even if the animal finds enough light, the goals (water port or exit) are invisible within the maze except from the immediately adjacent corridor. There are no visible beacons that would identify the goal.

With regard to the sense of touch and kinesthetics, the maze was constructed for maximal symmetry. At each level of the binary tree, all the junctions have locally identical geometry, with intersecting corridors of the same length. In practice, the animals may well detect some inadvertent cues, like an unusual drop of glue, that could identify one node from another. The maze rotation experiment suggests that such cues are not essential for the animal’s sense of location in the maze, at least in the expert phase.

The role of odors deserves particular attention because the mouse may use them both passively and actively. Does the animal first find the water port by following the smell of water? Probably not. For one, the port only emits a single drop of water when triggered by a nose poke. Second, we observed many instances where the animal is in the final corridor adjacent to the water port yet fails to discover it. The initial discovery seems to occur via touch. The reader can verify this in the videos accompanying this article. Regarding active use of odor markings in the maze, the maze rotation experiment suggests that such cues are not required for navigation, at least once the animals have adopted the shortest path to the water port ( Figure 4 ).

Another algorithm that is often invoked for animals moving in an open arena is vector-based navigation ( Wehner et al., 1996 ). Once the animal discovers a target, it keeps track of that target’s heading and distance using a path integrator. When it needs to return to the target it follows the heading vector and updates heading and distance until it arrives. Such a strategy has limited appeal inside a labyrinth because the vectors are constantly blocked by walls. Consider, for example, the ‘home runs’ back to the exit at the end of a bout. Here the target, namely the exit, is known from the start of the bout, because the animal enters through the same hole. At the end of the bout, when the mouse decides to exit from the maze, can it follow the heading vector to the exit? Figure 6A shows the 13 locations from which mice returned in a direct path to the exit on their very first foray. None of these locations is compatible with heading-based navigation: In each case an animal following the heading to the exit would get stuck in a different end node first and would have to reverse from there, quite unlike what really happened.

Finally, a partial clue comes from errors the animals make. We found that the rotation image of the water port, an end node diametrically across the entire maze, is one of the most popular destinations for rewarded animals ( Figure 10A ). These errors would be highly unexpected if the animals navigated from the entrance to the water by odor markings, or if they used an absolute representation of heading and distance. On the other hand, if the animal navigates via a remembered sequence of turns, then it will end up at that image node if it makes a single mistake at just the first T-junction.

Future directed experiments will serve to narrow down how mice learn to navigate this environment, and how their policy might change over time. Since the animals get to perfection within an hour or so, one can test a new hypothesis quite efficiently. Understanding what mechanisms they use will then inform thinking about the algorithm for learning, and about the neuronal mechanisms that implement it.

Experimental design

The goal of the study was to observe mice as they explored a complex environment for the first time, with little or no human interference and no specific instructions. In preliminary experiments, we tested several labyrinth designs and water reward schedules. Eventually, we settled on the protocol described here, and tested 20 mice in rapid succession. Each mouse was observed only over a 7-h period during the first night it encountered the labyrinth.

Maze construction

The maze measured ~24 x 24 x 2 inches; for manufacture we used materials specified in inches, so dimensions are quoted in those non-SI units where appropriate. The ceiling was made of 0.5 inch clear acrylic. Slots of 1/8 inch width were cut into this plate on a 1.5 inch grid. Pegged walls made of 1/8 inch infrared-transmitting acrylic (opaque in the visible spectrum, ePlastics) were inserted into these slots and secured with a small amount of hot glue. The floor was a sheet of infrared-transmitting acrylic, supported by a thicker sheet of clear acrylic. The resulting corridors (1-1/8 inches wide) formed a 6-level binary tree with T-junctions and progressive shortening of each branch, ranging from ~12 inch to 1.5 inch ( Figure 1 and Figure 2 ). A single end node contained a 1.5 cm circular opening with a water delivery port (described below). The maze included provision for two additional water ports not used in the present report. Once per week the maze was submerged in cage cleaning solution. Between different animals the floor and walls were cleaned with ethanol.

Reward delivery system

The water reward port was controlled by a Matlab script on the main computer through an interface (Sanworks Bpod State Machine r1). Rewards were triggered when the animal’s nose broke the IR beam in the water port (Sanworks Port interface + valve). The interface briefly opened the water valve to deliver ~30 µL of water and flashed an infrared LED mounted outside the maze for 1 s. This served to mark reward events on the video recording. Following each reward, the system entered a time-out period for 90 s, during which the port did not provide further reward. In experiments with sated mice the water port was turned off.

Cage and connecting passage

The entrance to the maze was connected to an otherwise normal mouse cage by red plastic tubing (3 cm dia, 1 m long). The cage contained food, bedding, nesting material, and in the case of unrewarded experiments also a normal water bottle.

Animals and treatments

All mice were C57BL/6J animals (Jackson Labs) between the ages of 45 and 98 days (mean 62 days). Both sexes were used: four males and six females in the rewarded experiments, five males and four females in the unrewarded experiments. For water deprivation, the animal was transferred from its home cage (generally group-housed) to the maze cage ~22 h before the start of the experiment. Non-deprived animals were transferred minutes before the start. All procedures were performed in accordance with institutional guidelines and approved by the Caltech IACUC.

Video recording

All data reported here were collected over the course of 7 h during the dark portion of the animal’s light cycle. Video recording was initiated a few seconds prior to connecting the tunnel to the maze. Videos were recorded by an OpenCV python script controlling a single webcam (Logitech C920) located ~1 m below the floor of the maze. The maze and access tube were illuminated by multiple infrared LED arrays (center wavelength 850 nm). Three of these lights illuminated the maze from below at a 45 degree angle, producing contrast to resolve the animal’s foot pads. The remaining lights pointed at the ceiling of the room to produce backlight for a sharp outline of the animal.

Animal tracking

A version of DeepLabCut ( Nath et al., 2019 ) modified to support gray-scale processing was used to track the animal’s trajectory, using key points at the nose, feet, tail base and mid-body. All subsequent analyses were based on the trajectory of the animal’s nose, consisting of positions x ⁢ ( t ) and y ⁢ ( t ) in every video frame.

Rates of transition between cage and maze

This section relates to Figure 1—figure supplement 3 . We entertained the hypothesis that the animals become ‘thirsty for exploration’ as they spend more time in the cage. In that case one would predict that the probability of entering the maze in the next second will increase with time spent in the cage. One can compute this probability from the distribution of residency times in the cage, as follows:

Say t = 0 when the animal enters the cage. The probability density that the animal will next leave the cage at time t is

where r ⁢ ( t ) is the instantaneous rate for entering the maze. So

This relates the cumulative of the instantaneous rate function to the cumulative of the observed transition times. In this way we computed the rates 

The rate of entering the maze is highest at short times in the cage ( Figure 1—figure supplement 3A ). It peaks after ~15 s in the cage and then declines gradually by a factor of 4 over the first minute. So the mouse is most likely to enter the maze just after it returns from there. This runs opposite to the expectation from a homeostatic drive for exploration, which should be sated right after the animal returns. We found no evidence for an increase in the rate at late times. These effects were very similar in rewarded and unrewarded groups and in fact the tendency to return early was seen in every animal.

By contrast, the rate of exiting the maze is almost perfectly constant over time ( Figure 1—figure supplement 3B ). In other words, the exit from the maze appears like a constant rate Poisson process. There is a slight elevation of the rate at short times among rewarded animals ( Figure 1—figure supplement 3B top). This may come from the occasional brief water runs they perform. Another strange deviation is an unusual number of very short bouts (duration 2–12 s) among unrewarded animals ( Figure 1—figure supplement 3B bottom). These are brief excursions in which the animal runs to the central junction, turns around, and runs to the exit. Several animals exhibited these, often several bouts in a row, and at all times of the night.

Reduced trajectories

From the raw nose trajectory we computed two reduced versions. First, we divided the maze into discrete ‘cells’, namely the squares the width of a corridor that make up the grid of the maze. At any given time, the nose is in one of these cells and that time series defines the cell trajectory.

At a coarser level still one can ask when the animal passes through the nodes of the binary tree, which are the decision points in the maze. The special cells that correspond to the nodes of the tree are those at the center of a T-junction and those at the leaves of the tree. We marked all the times when the trajectory ( x ⁢ ( t ) , y ⁢ ( t ) ) entered a new node cell. If the animal leaves a node cell and returns to it before entering a different node cell, that is not considered a new node. This procedure defines a discrete node sequence s i and corresponding arrival times at those nodes t i . We call the transition between two nodes a ‘step’. Much of the analysis in this paper is derived from the animal’s node sequence. The median mouse performed 16,192 steps in the 7 h period of observation (mean = 15,257; SD = 3340).

In Figure 5 and Figure 6 , we count the occurrence of direct paths leading to the water port (a ‘water run’) or to the exit (a ‘home run’). A direct path is a node sequence without any reversals. Figure 3—figure supplement 1 illustrates some examples.

If the animal makes one wrong step from the direct path, that step needs to be backtracked, adding a total of two steps to the length of the path. If further errors occur during backtracking, they need to be corrected as well. The binary maze contains no loops, so the number of errors is directly related to the length of the path:

Maze rotation

The maze rotation experiment ( Figure 4 ) was performed on four mice, all water-deprived. Two of the animals (’D7’ and ’D9’) had experienced the maze before, and are part of the ’rewarded’ group in other sections of the report. Two additional animals (’F2’ and ’A1’) had had no prior contact with the maze.

The maze rotation occurred after at least 6 h of exposure, by which time the animals had all perfected the direct path to the water port.

For animals ’D7’ and ’D9’ we rotated only the floor of the maze, leaving the walls and ceiling in the original configuration. For ’F2’ and ’A1’ we rotated the entire maze, moving one wall segment at the central junction and the water port to attain the same shape. Navigation remained intact for all animals. Note that ’A1’ performed a perfect path to the water port and back immediately before and after a full maze rotation ( Figure 4B ).

The visits to the four locations in the maze ( Figure 4C , Figure 4—figure supplement 1 ) were limited to direct paths of length at least two steps. This avoids counting rapid flickers between two adjacent nodes. In other words, the animal has to move at least two steps away from the target node before another visit qualifies.

Statistics of sudden insight

In Figure 5 one can distinguish two events: First, the animal finds the water port and begins to collect rewards at a steady rate: this is when the green curve rises up. At a later time, the long direct paths to the water port become much more frequent than to the comparable control nodes: this is when the red and blue curves diverge. For almost all animals these two events are well separated in time ( Figure 5—figure supplement 1 ). In many cases, the rate of long paths seems to change discontinuously: a sudden change in slope of the curve.

Here, we analyze the degree of 'sudden change', namely how rapidly the rate changes in a time series of events. We modeled the rate as a sigmoid function of time during the experiment:

The rate begins at a low initial level r i , reflecting chance occurrence of the event, and saturates at a high final level r f , limited for example by the animal’s walking speed. The other two parameters are the time t s of half-maximal rate change, and the width w over which that rate change takes place. A sudden change in the event rate would correspond to w = 0 .

The data are a set of n event times t i in the observation interval [ 0 , T ] . We model the event train as an inhomogeneous Poisson point process with instantaneous rate r ⁢ ( t ) . The likelihood of the data given the rate function r ⁢ ( t ) is

and the log likelihood is

For each of the 10 rewarded mice, we maximized ln ⁡ L over the 4 parameters of the rate model, both for the reward events and the long paths to water. The resulting fits are plotted in Figure 5—figure supplement 1 .

Focusing on the learning of long paths to water, for 6 of the 10 animals the optimal width parameter w was less than 300 s: B1, B2, C1, C3, C6, C7. These are the same animals one would credit with a sudden kink in the cumulative event count based on visual inspection ( Figure 5—figure supplement 1 ).

To measure the uncertainty in the timing of this step, we refit the data for this subgroup of mice with a model involving a sudden step in the rate,

and computed the likelihood of the data as a function of the step time t s . We report the mean and standard deviation of the step time over its likelihood in Figure 5—source data 1 . Animal C6 was dropped from this 'sudden step' group, because the uncertainty in the step time was too large (∼900 s).

The goal of this analysis is to measure how effectively the animal surveys all the end nodes of the maze. The specific question is: In a string of n end nodes that the animal samples, how many of these are distinct? On average how does the number of distinct nodes d increase with n ? This was calculated as follows:

We restricted the animal’s node trajectory ( s i ) to clips of exploration mode, excluding the direct paths to the water port or the exit. All subsequent steps were applied to these clips, then averaged over clips. Within each clip we marked the sequence of end nodes ( e i ) . We slid a window of size n across this sequence and counted the number of distinct nodes d in each window. Then we averaged d over all windows in all clips. Then we repeated that for a wide range of n . The resulting d ⁢ ( n ) is plotted in the figures reporting new nodes vs nodes visited ( Figure 8A,B and Figure 9C ).

For a summary analysis, we fitted the curves of d ⁢ ( n ) with a two-parameter function:

The parameter a is the number of visits n required to survey half of the end nodes, whereas b reflects a relative acceleration in discovering the last few end nodes. This function was found by trial and error and produces absurdly good fits to the data ( Figure 8—figure supplement 1 ). The values quoted in the text for efficiency of exploration are E = 32 / a ( Equation 1 ).

The value of b was generally small (~0.1) with no difference between rewarded and unrewarded animals. It declined slightly over the night ( Figure 8—figure supplement 1B ), along with the decline in a ( Figure 8C ).

Biased random walk

For the analysis of Figure 9 , we considered only the parts of the trajectory during ‘exploration’ mode. Then we parsed every step between two nodes in terms of the type of action it represents. Note that every link between nodes in the maze is either a ‘left branch’ or a ‘right branch’, depending on its relationship to the parent T-junction. Therefore, there are four kinds of action:

a = 0 : ‘in left’, take a left branch into the maze

a = 1 : ‘in right’, take a right branch into the maze

a = 2 : ‘out left’, take a left branch out of the maze

a = 3 : ‘out right’, take a right branch out of the maze

At any given node, some actions are not available, for example from an end node one can only take one of the ‘out’ actions.

To compute the turning biases, we considered every T-junction along the trajectory and correlated the action a 0 that led into that node with the subsequent action a 1 . By tallying the action pairs ( a 0 , a 1 ) , we computed the conditional probabilities p ⁢ ( a 1 | a 0 ) . Then the four biases are defined as

For the simulations of random agents ( Figure 8 , Figure 9 ), we used trajectories long enough so the uncertainty in the resulting curves was smaller than the line width.

Models of decisions during exploration

The general approach is to develop a model that assigns probabilities to the animal’s next action, namely which node it will move to next, based on its recent history of actions. All the analyses were restricted to the animal’s ‘exploration’ mode and to the 63 nodes in the maze that are T-junctions. During the ‘drink’ and ‘leave’ modes the animal’s next action is predictable. Similarly, when it finds itself at one of the 64 end nodes it only has one action available.

For every mouse trajectory, we split the data into five segments, trained the model on 80% of the data, and tested it on 20%, averaging the resulting cross-entropy over the five possible splits. Each segment was in turn composed of parts of the trajectory sampled evenly throughout the 7 h experiment, so as to average over the small changes in the course of the night. The model was evaluated by the cross-entropy between the predictions and the animal’s true actions. If one had an optimal model of behavior, the result would reveal the animal’s true source entropy.

Fixed depth Markov chain

To fit a model with fixed history depth k to a measured node sequence ( s t ) , we evaluated all the substrings in that sequence of length ( k + 1 ) . At any given time t , the k -string 𝐡 t = ( s t - k + 1 , … , s t ) identifies the history of the animal’s k most recent locations. The current state s t is one of 63 T-junctions. Each state is preceded by one of 3 possible states. So the number of history strings is 63 ⋅ 3 k - 1 . The 2-string ( s t , s t + 1 ) identifies the next action a t + 1 , which can be ‘in left’, ‘in right’, or ‘out’, corresponding to the 3 branches of the T junction. Tallying the history strings with the resulting actions leads to a contingency table of size 63 ⋅ 3 k - 1 × 3 , containing

Based on these sample counts, we estimated the probability of each action a conditional on the history 𝐡 as

This amounts to additive smoothing with a pseudocount of 1, also known as ‘Laplace smoothing’. These conditional probabilities were then used in the testing phase to predict the action at time t based on the preceding history 𝐡 t . The match to the actually observed actions a t was measured by the cross-entropy

Variable depth Markov chain

As one pushes to longer histories, that is larger k , the analysis quickly becomes data-limited, because the number of possible histories grows exponentially with k . Soon one finds that the counts for each history-action combination drop to where one can no longer estimate probabilities correctly. In an attempt to offset this problem, we pruned the history tree such that each surviving branch had more than some minimal number of counts in the training data. As expected, this model is less prone to over-fitting and degrades more gently as one extends to longer histories ( Figure 11—figure supplement 1A ). The lowest cross-entropy was obtained with an average history length of ~4.0 but including some paths of up to length 6. Of all the algorithms we tested, this produced the lowest cross-entropies, although the gains relative to the fixed-depth model were modest ( Figure 11—figure supplement 1C ).

Pooling across symmetric nodes in the maze

Another attempt to increase the counts for each history involved pooling counts over multiple T-junctions in the maze that are closely related by symmetry. For example, all the T-junctions at the same level of the binary tree look locally similar, in that they all have corridors of identical length leading from the junction. If one supposes that the animal acts the same way at each of those junctions, one would be justified in pooling across these nodes, leading to a better estimate of the action probabilities, and perhaps less over-fitting. This particular procedure was unsuccessful, in that it produced higher cross-entropy than without pooling.

However, one may want to distinguish two types of junctions within a given level: L-nodes are reached by a left branch from their parent junction one level lower in the tree, R-nodes by a right branch. For example, in Figure 3—figure supplement 1 , node 1 is L-type and node 2 is R-type. When we pooled histories over all the L-nodes at a given level and separately over all the R-nodes the cross-entropy indeed dropped, by about 5% on average. This pooling greatly reduced the amount of over-fitting ( Figure 11—figure supplement 1B ), which allowed the use of longer histories, which in turn improved the predictions on test data. The benefit of distinguishing L- and R-nodes probably relates to the animal’s tendency to alternate left and right turns.

All the Markov model results we report are obtained using pooling over L-nodes and R-nodes at each maze level.

The behavioral data and code that produced the figures are available in a public Github repository cited in the article https://github.com/markusmeister/Rosenberg-2021-Repository (copy archived at https://archive.softwareheritage.org/swh:1:rev:224141473e53d6e8963a77fbe625f570b0903ef1 ). We also prepared a permanent institutional repository at https://data.caltech.edu/badge/latestdoi/329740227 .

• Rosenberg M
• van der Meij J
• Google Scholar
• Apollodorus
• Behrens TEJ
• Whittington JCR
• Stachenfeld KL
• Kurth-Nelson Z
• Bitterman ME
• Bourtchuladze R
• Frenguelli B
• Steinmetz NA
• Zatka-Haas P
• Bai Reddy C
• Carandini M
• Churchland AK
• Fanselow MS
• Benjamini Y
• Gallistel CR
• Fairhurst S
• Grobéty M-C
• Krechevsky I
• McNamara CG
• Tejero-Cantero Á
• Campo-Urriza N
• Horicsányi A
• Vásárhelyi G
• Chartarifsky-Lynn L
• Rondi-Reig L
• Fujishita C
• Yamagishi A
• Shinohara S
• Tchernichovski O
• Peterson BK
• Hoekstra HE

Author details

Contribution, contributed equally with, competing interests.

For correspondence

Simons foundation (543015), simons foundation (543025), national science foundation (1564330).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Funding: This work was supported by the Simons Collaboration on the Global Brain (grant 543015 to MM and 543025 to PP), by NSF award 1564330 to PP, and by a gift from Google to PP.

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to animal protocol 1656 approved by the institutional animal care and use committee (IACUC) at Caltech.

© 2021, Rosenberg et al.

This article is distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use and redistribution provided that the original author and source are credited.

• 20,575 views
• 1,529 downloads
• 55 citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

Downloads (link to download the article as pdf).

• Article PDF
• Figures PDF

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools), categories and tags.

• few-shot learning
• decision-making
• predictive models
• cognitive map

Research organism

Further reading, neural population dynamics underlying evidence accumulation in multiple rat brain regions.

Accumulating evidence to make decisions is a core cognitive function. Previous studies have tended to estimate accumulation using either neural or behavioral data alone. Here we develop a unified framework for modeling stimulus-driven behavior and multi-neuron activity simultaneously. We applied our method to choices and neural recordings from three rat brain regions - the posterior parietal cortex (PPC), the frontal orienting fields (FOF), and the anterior-dorsal striatum (ADS) - while subjects performed a pulse-based accumulation task. Each region was best described by a distinct accumulation model, which all differed from the model that best described the animal's choices. FOF activity was consistent with an accumulator where early evidence was favored while the ADS reflected near perfect accumulation. Neural responses within an accumulation framework unveiled a distinct association between each brain region and choice. Choices were better predicted from all regions using a comprehensive, accumulation-based framework and different brain regions were found to differentially reflect choice-related accumulation signals: FOF and ADS both reflected choice but ADS showed more instances of decision vacillation. Previous studies relating neural data to behaviorally-inferred accumulation dynamics have implicitly assumed that individual brain regions reflect the whole-animal level accumulator. Our results suggest that different brain regions represent accumulated evidence in dramatically different ways and that accumulation at the whole-animal level may be constructed from a variety of neural-level accumulators.

A rapid phylogeny-based method for accurate community profiling of large-scale metabarcoding datasets

Environmental DNA (eDNA) is becoming an increasingly important tool in diverse scientific fields from ecological biomonitoring to wastewater surveillance of viruses. The fundamental challenge in eDNA analyses has been the bioinformatical assignment of reads to taxonomic groups. It has long been known that full probabilistic methods for phylogenetic assignment are preferable, but unfortunately, such methods are computationally intensive and are typically inapplicable to modern Next-Generation Sequencing data. We here present a fast approximate likelihood method for phylogenetic assignment of DNA sequences. Applying the new method to several mock communities and simulated datasets, we show that it identifies more reads at both high and low taxonomic levels more accurately than other leading methods. The advantage of the method is particularly apparent in the presence of polymorphisms and/or sequencing errors and when the true species is not represented in the reference database.

• Physics of Living Systems

Combining mutation and recombination statistics to infer clonal families in antibody repertoires

B-cell repertoires are characterized by a diverse set of receptors of distinct specificities generated through two processes of somatic diversification: V(D)J recombination and somatic hypermutations. B cell clonal families stem from the same V(D)J recombination event, but differ in their hypermutations. Clonal families identification is key to understanding B-cell repertoire function, evolution and dynamics. We present HILARy (High-precision Inference of Lineages in Antibody Repertoires), an efficient, fast and precise method to identify clonal families from single- or paired-chain repertoire sequencing datasets. HILARy combines probabilistic models that capture the receptor generation and selection statistics with adapted clustering methods to achieve consistently high inference accuracy. It automatically leverages the phylogenetic signal of shared mutations in difficult repertoire subsets. Exploiting the high sensitivity of the method, we find the statistics of evolutionary properties such as the site frequency spectrum and 𝑑𝑁∕𝑑𝑆 ratio do not depend on the junction length. We also identify a broad range of selection pressures spanning two orders of magnitude.

Be the first to read new articles from eLife

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Procedures for Behavioral Experiments in Head-Fixed Mice

Affiliation Janelia Farm Research Campus, Howard Hughes Medical Institute, Ashburn, Virginia, United States of America

Present address: Section of Neurobiology and Department of Biological Sciences, University of Southern California, Los Angeles, California, United States of America

Present address: The Solomon H. Snyder Department of Neuroscience and Brain Science Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

Present address: Section of Neurobiology and Department of Neurosciences, University of California, San Diego, California, United States of America

Affiliations Janelia Farm Research Campus, Howard Hughes Medical Institute, Ashburn, Virginia, United States of America, Department of Basic Neurosciences, University of Geneva, Geneva, Switzerland

Affiliation Department of Basic Neurosciences, University of Geneva, Geneva, Switzerland

Present address: Institute of Neuroscience, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China

* E-mail: [email protected]

• Zengcai V. Guo, 
• S. Andrew Hires, 
• Nuo Li, 
• Daniel H. O'Connor, 
• Takaki Komiyama, 
• Eran Ophir, 
• Daniel Huber, 
• Claudia Bonardi, 
• Karin Morandell, 

• Published: February 10, 2014
• https://doi.org/10.1371/journal.pone.0088678
• Reader Comments

26 Jun 2014: The PLOS ONE Staff (2014) Correction: Procedures for Behavioral Experiments in Head-Fixed Mice. PLOS ONE 9(6): e101397. https://doi.org/10.1371/journal.pone.0101397 View correction

The mouse is an increasingly prominent model for the analysis of mammalian neuronal circuits. Neural circuits ultimately have to be probed during behaviors that engage the circuits. Linking circuit dynamics to behavior requires precise control of sensory stimuli and measurement of body movements. Head-fixation has been used for behavioral research, particularly in non-human primates, to facilitate precise stimulus control, behavioral monitoring and neural recording. However, choice-based, perceptual decision tasks by head-fixed mice have only recently been introduced. Training mice relies on motivating mice using water restriction. Here we describe procedures for head-fixation, water restriction and behavioral training for head-fixed mice, with a focus on active, whisker-based tactile behaviors. In these experiments mice had restricted access to water (typically 1 ml/day). After ten days of water restriction, body weight stabilized at approximately 80% of initial weight. At that point mice were trained to discriminate sensory stimuli using operant conditioning. Head-fixed mice reported stimuli by licking in go/no-go tasks and also using a forced choice paradigm using a dual lickport. In some cases mice learned to discriminate sensory stimuli in a few trials within the first behavioral session. Delay epochs lasting a second or more were used to separate sensation (e.g. tactile exploration) and action (i.e. licking). Mice performed a variety of perceptual decision tasks with high performance for hundreds of trials per behavioral session. Up to four months of continuous water restriction showed no adverse health effects. Behavioral performance correlated with the degree of water restriction, supporting the importance of controlling access to water. These behavioral paradigms can be combined with cellular resolution imaging, random access photostimulation, and whole cell recordings.

Citation: Guo ZV, Hires SA, Li N, O'Connor DH, Komiyama T, Ophir E, et al. (2014) Procedures for Behavioral Experiments in Head-Fixed Mice. PLoS ONE 9(2): e88678. https://doi.org/10.1371/journal.pone.0088678

Editor: Sidney Arthur Simon, Duke University Medical Center, United States of America

Received: October 31, 2013; Accepted: December 14, 2013; Published: February 10, 2014

Copyright: © 2014 Guo et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: Funding provided by Howard Hughes Medical Institute. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Neural circuits are composed of defined neuronal populations that are connected in a highly specific manner. A central goal of modern neuroscience is to link the dynamics of these neural circuits to behavior [1] . Deciphering the logic of neural circuits thus requires cell-type specific neurophysiology and manipulation [2] . Because of the wide availability of transgenic mice that allow cell-type specific targeting, the mouse is a leading model system for mammalian circuit neuroscience [3] .

Over the last fifty years, experiments in behaving primates have led the way in separating causation from correlation in neurophysiological experiments. Head-fixation and body restraint have been critical because they facilitate stimulus control and measurement of movement. Non-human primates can be trained in sophisticated tasks that isolate specific brain functions. Repeated trials, often many hundreds per day, unleash powerful statistical methods to relate behavior and neurophysiological measurements. Although head-fixed monkeys have been the ‘gold standard’ system in relating the dynamics of individual neurons to behavior, cell-type-specific measurements [4] , [5] and manipulation remain exceptional in non-human primates.

In contrast, in the mouse brain, cell-type-specific neurobiology is becoming routine. Transgenes can be targeted to specific types of neurons, which are nodes of the circuit diagram [2] . These transgenes can be used to identify cell-types during recordings and to manipulate circuit nodes during behavior. Mice also have a rich behavioral repertoire involving many basic sensory, cognitive and motor functions. Mice are relatively cheap, promising high-throughput approaches to neurophysiology. The microcircuit organization of the brain, as far as it is known, is similar in mice and other higher mammals. Finally, the lissencephalic macrostructure of the mouse brain allows unobstructed access to a large fraction of the brain for neurophysiology and imaging [6] , [7] .

Over the last decade, inspired by experiments on behaving primates, increasingly sophisticated procedures for quantitative head-fixed behaviors have been developed for mice (for a review of the literature on head-fixed behaving rats see [8] ). For example, learning in the vestibulo-ocular reflex, long studied in monkeys, has been successfully probed in mice [9] . Head-fixation is critical because precise control of head motion with respect to visual stimuli is essential, as is measurement of eye position. Beyond reflexive behavior, mice have also been trained in choice-based tasks using operant conditioning. Head-fixed mice have been trained to discriminate odors [10] , [11] , auditory stimuli [12] , visual stimuli [13] – [16] , and tactile cues [7] , [17] – [25] . Head-fixed mice can navigate simple mazes in a visual virtual reality environment [26] . As in most primate studies, in these types of experiment mice are motivated by thirst.

In this paper, we describe procedures for water restriction and behavioral training. We illustrate the procedures with detailed training protocols for head-fixed mice performing whisker-based tactile behaviors. Rodents use their whiskers to detect and locate objects when moving through an environment [27] , [28] . The measurement of the locations of object features is a critical aspect of object identification and navigation. Inspired by previous work in freely moving rats [29] , we have trained head-fixed mice to locate an object (a vertical pole) near their heads with their whiskers [7] , [17] – [23] . This is by construction an active sensation behavior: mice have to move their whiskers in an intelligent manner to collect information about the world. High-speed imaging of whisker position, facilitated by head-fixation, reveals the whisker movements underlying discrimination [30] . Changes in whisker shape, caused by contact between whisker and object, report the mechanical inputs to the somatosensory system. The object-localization task is ideally suited to probing the neural basis of tactile spatial perception and sensorimotor integration [31] .

Procedures and Results

We describe our current best practice for head-fixation, water restriction and behavioral training for head-fixed mice performing tactile behaviors. The procedures are introduced in roughly the order in which they are performed in the laboratory. We first outline the surgery and apparatus for head-fixation. We then introduce water restriction, which is critical to motivate the mice for behavioral experiments [32] . Mice are then briefly acclimatized to handling by the experimenter and to head-fixation, followed by operant conditioning. The apparatus [7] , [10] , [17] , [23] and software ( http://brodylab.princeton.edu/bcontrol ) for behavior, whisker tracking ( https://openwiki.janelia.org/wiki/display/MyersLab/Whisker+Tracking ) [30] , electrophysiology (ephus.org) [7] , [18] , [22] , and imaging ( https://openwiki.janelia.org/wiki/display/shareddesigns/Shared+Two-photon+Microscope+Designs ) (scanimage.org) [19] – [21] have been described elsewhere.

1. Surgery and head-fixation

Head bar surgery..

All procedures were in accordance with protocols approved by the Janelia Farm Institutional Animal Care and Use Committee. All surgeries used standard aseptic procedures. Mice (∼2–6 months old, typically males) were deeply anesthetized with 2% isoflurane (by volume in O2; SurgiVet; Smiths Medical) and mounted in a stereotaxic apparatus (Kopf Instruments). Mice were kept on a thermal blanket (Harvard Apparatus) and their eyes were covered with a thin layer of petroleum jelly. During the surgery, the anesthesia levels were adjusted to 1–1.5% to achieve ∼1/second breathing rate in mice. The scalp was cleaned with 70% ethanol and betadine. Marcaine (50 µl 0.5% solution) was injected under the scalp for topical anesthesia. Ketofen (non-steroidal anti-inflammatory drug, 5 mg/kg) was injected subcutaneously and buprenorphine (opiod analgesic, 0.05 mg/kg) was injected into the intraperitoneal cavity. A flap of skin, approximately 1 cm 2 , was removed from the dorsal skull with a single cut. The remaining gelatinous periostium was removed with small scissors. The skull was cleaned and dried with sterile cotton swabs. The bone was scraped with a scalpel or slowly turning dental drill for better bonding with the glue. The exposed skull was covered with a thin layer of cyanoacrylic glue. The head bar was positioned directly onto the wet glue. Dental acrylic (Jet Repair Acrylic) was added to cover the glue and cement the head bar in position. The head bar links the skull rigidly to the behavioral apparatus.

For experiments requiring maximal mechanical stability, we typically use an extended head bar, with a plate that is fitted in three dimensions to the shape of the dorsal mouse skull ( Figure 1A ). When cemented to the skull this plate bonds with all skull plates over large surface areas and thereby links the skull plates and rigidifies the skull. With the head-plate clamped to the head-plate holder, all remaining brain motion is caused by movement of the brain within the skull (data not shown). For experiments requiring access to large areas of the brain we use a minimal head bar (22.3×3.2 mm) [7] .

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

A. Left, two types of titanium head plates. Right, stainless steel head bar holder and clamp (only one of two sides is shown). The head plate is inserted into notches in the holder and fastened with the clamp (right, top) and a thumbscrew (not shown). The simple head bar (left, top) is used when access to large parts of the brain is necessary. The larger head plate (left, middle) provides better stability. The simple head bar was cemented to the skull of the mouse (left, bottom). The head of the mouse (top view) was pointing downward. The skull was outfitted with a clear skull cap [7] . The head bar was aligned at the lambda sutures. The red dot indicates the location of bregma. B. Plexiglass body tube used for head-fixed mice. Mice rest their front paws on the front ledge. The bottom of the tube is coated with aluminum foil to produce electrical contact for electric lickports. The aluminum foil is connected to the red banana socket which will be connected to electric lickports for detecting licking events. C. Example caddy used in training apparatus, assembled from standard optomechanical components (Thorlabs). The head bar holder is mounted towards the left. D. A head-fixed mouse in the caddy.

https://doi.org/10.1371/journal.pone.0088678.g001

Optional viral gene transfer.

In some cases viral reagents, typically adeno-associated virus (AAV) were introduced during the head bar surgery [18] – [21] . Using a dental drill with an FG 1/4 drill bit, a small hole was drilled into the skull. The virus was introduced using a fine glass injection pipette (tip diameter approximately 15–20 µm) beveled to a sharp tip (outer diameter, 20–30 µm). Beveling is critical since it allows the pipette to penetrate the dura without dimpling the cortex, greatly reducing tissue damage. The pipette containing virus was lowered into the brain region of interest. Viral suspension is injected slowly into the parenchyma (10 nL per minute). Approximately 30 nL of AAV (approximately 10 12 titer) is sufficient to transduce neurons in a 500 µm diameter column of the neocortex [33] . Following the surgery, buprenorphine (0.1 mg/kg) was administered once. Ketoprofen (5 mg/kg) was administered once a day for two days as an analgesic to reduce inflammation. Animals were examined once a day for three days for signs of infection, lethargy, and grooming.

In other cases it may be necessary to introduce viruses during training. As viral transduction efficiency can be low in water restricted mice, water should be supplemented for 2 days prior surgery (3–4 ml water per day) [26] .

Head-fixation and lickport.

For head-fixation, the wings of the head bar are seated into notches in a stainless steel holder and fixed with a pair of clamps and thumbscrews ( Figure 1A ). The mouse body is inserted into an acrylic ‘body tube’ (1⅛ inch i.d.; McMaster; P/N 8486K433) ( Figure 1B ), with the mouse head extending out and the front paws gripping the tube edge or a ledge after head-fixation. The holder and body tube in turn are attached to a caddy ( Figure 1C ). Typically, the head bar is about 30 mm above the bottom of the body tube. The caddy is fixed to the behavior box using magnetic kinematic bases (e.g. Thorlabs, KB3X3). These mounts allow the experimenter to conveniently head-fix mice outside of the apparatus in the caddy. The caddy with mouse can then be placed into the apparatus rapidly and consistently. A head-fixed mouse should crouch in a natural position in the body tube, with its paws resting on a tube edge or a ledge ( Figure 1D ).

Water rewards are provided by different types of custom-made lickports that sense the movement of the tongue. Electrical lickports are activated by the tongue making contact with the steel nozzle of the lickport [34] . Optical lickports are activated by interruptions in the light path between an LED and a phototransistor [23] . Optical lickports require regular cleaning to ensure that the optical path remains unobstructed. Electrical lickports are more robust, but can introduce artifacts in electrophysiological measurements.

The lickport position relative to the mouse is a critical parameter during training. If the lickport is too close to the mouth, the mouse might lick compulsively. If the lickport is too far, the mouse might miss rewards and become discouraged. We typically start with the lickport 0.5 mm below the lower lip, and 5 mm posterior to the tip of the nose. During training the lickport typically is moved away from the mouth to discourage compulsive licking (see Discussion ).

2. Water restriction

How can we motivate experimental subjects to cooperate in behavioral experiments? In the case of human subjects, this is typically achieved by the subjects' willingness to participate in scientific experiments, or by providing subjects with economic rewards. For non-human subjects, experimenters can restrict the animal subjects' access to basic needs such as food and water [35] – [39] , and use them as rewards during behavioral experiments. Rodents generally cope better with water restriction than food restriction [40] . In an attempt to use food restriction (2–3 grams of solid food per day with free access to water) some mice developed significant health problems (high health scores) before reaching 15% weight loss. Here we describe procedures for motivating mice by limiting their access to water, based on pioneering work by Slotnick and colleagues in the context of freely moving olfactory behavior in mice [32] , [41] . Although most tested mice were male, females showed similar weight loss and behavioral performance after water restriction. On days when behavioral experiments were carried out, mice typically obtained all of their water during performance in the behavior apparatus (approximately 1 ml water per day). On other days, including weekends and holidays, mice received 1 ml water per day.

Water restriction was started after mice recovered from surgery (at least three days after surgery). Mice were housed singly in cages containing tunnels and bedding material, in a reverse light cycle room. Housing in small groups of siblings is also possible. Training and behavioral testing occurred mainly during the dark phase. Relative humidity critically affects the animals' need of water [42] and was kept at 40–50%, with little seasonal variations. Following full and complete recovery from a previous surgery (at least three days post surgery), mice were placed on a water restriction schedule in preparation for behavioral conditioning. Dry food was continuously available (Rodent diet 5053). One ml of water was dispensed manually into bowls which were attached to the inside walls of individual cages, at consistent times of day. Mice consumed this water within minutes. This corresponds to approximately 35% of ad libitum water consumption for C57BL/6J mice (Mouse Phenome Database from the Jackson Laboratory: http://www.jax.org/phenome ).

All mice undergoing water restriction were monitored daily for hydration, weight, ruffled fur, and movement ( Figure 2 ). The pre-restriction body weight is typically in the range 23–30 g for 2–6 months old males. If mice drop below 70% of pre-restriction weight, or if mice show signs of dehydration or pain, their health is assessed in more detail. The health assessment is summarized in a health score ( Figure 3 ). Health scores in the range of 1–2 typically reflect slightly reduced activity and ruffled fur around the margins of the head bar surgery. If the health score is above three, mice receive supplemental water ( Figure 2 ). After stabilization of body weight, typically after seven to ten days of water restriction, the training procedure began ( Figure 4 ). The body weight tends to increase with long periods of restriction after the initial dip ( Figure 4B,F ). With shorter periods of water restriction, mice will not be sufficiently motivated to overcome fear-related reflexes, triggered by new environments that are invariably part of initial stages of training. Without strong motivation, mice often stop working after a few trials and may learn undesired behaviors. Trained mice often receive all of their water (1 ml, sometimes more; Figure 4G ) during performance in the behavioral apparatus. After behavioral sessions in which mice consumed little water (<0.5 ml) a water supplement (0.2–0.5 ml) was typically provided to a total water consumption of 0.6 ml per day or more.

https://doi.org/10.1371/journal.pone.0088678.g002

Activity levels, grooming, and indicators of eating and drinking are scored daily in a health assessment sheet. The total aggregate health score determines if mice are supplied with additional water (see flowchart in Figure 2 ).

https://doi.org/10.1371/journal.pone.0088678.g003

All mice were trained in a lick/no-lick object location discrimination task using a single whisker (same mice as in Figures 2 & 3 of [18] ). Rewards consisted of approximately 8 µl of water per trial. A. Experimental time-course for one example mouse, from the beginning of water restriction to the end of electrophysiological recordings. An 85 day old mouse (25.4 g) was put on water restriction for eight days, followed by training (starting on day 9) and recording (starting on day 28). B. Body weight as a function of time. Same mouse as in A. The dashed line indicates 30% weight loss. C. Water consumed per day. After start of training mice mostly received their water during the training session. A larger number of correct trials will lead to more consumed water. Same mouse as in A. D. Health score as a function of time. A health score larger than 3 (dashed line) triggers more detailed evaluation and possibly water supplements. Same mouse as in A. E. Experimental time-course for a group of 5 mice. Same format as A. F. Average body weight of 5 mice (black line) and 2 mice with free access to water (grey line). Shading indicates standard deviation. Experimental time-course for all mice was similar, but not identical to A. G. Average water consumed. H. Average health score.

https://doi.org/10.1371/journal.pone.0088678.g004

At steady state, mice typically lose 20% of body weight compared to age-matched controls ( Figure 4B, F ) while consuming 1 ml of water per day. Our experience has shown that mice must lose at least 15% of body weight to be motivated to perform challenging behavioral tasks for large numbers of trials. During early stages of training the number of trials performed per session, as well as the fraction of correct trials, correlate with weight loss ( Figure 5A, B ). This indicates that water restriction determines the mouse's motivation and drives learning and performance. Consistent water restriction, including weekends, is critical. This is because even one day of free access to water causes substantial weight gain ( Figure 6 ) and loss of motivation for several days.

A. Performance as a function of normalized body weight. Each circle corresponds to one behavioral session. Different colors correspond to different mice (7–8 sessions per mouse). The sessions included are the first seven to eight sessions of discrimination training (corresponding to the training phase shown by open symbols in Figure 3a of [23] . Multiple factors can compromise performance in behavioral experiments. In this experiment mice were trained in serial with individualized attention to reduce variability due to uncontrolled factors. The correlation coefficient is R 2  = 0.52 ( p <0.001). B. Number of trials as a function of normalized body weight. Mice usually perform less trials in the first few sessions of training. Same sessions as in (A). The correlation coefficient is R 2  = 0.24 ( p <0.001).

https://doi.org/10.1371/journal.pone.0088678.g005

https://doi.org/10.1371/journal.pone.0088678.g006

Under our conditions health scores remain in a normal range (<3) for four months of continuous water restriction (see Training the lick-left/lick-right task with a delay epoch ). Higher scores are typically related to other factors, such as stressful surgeries, large head-implants, or infection. We performed a histological analysis for 6 male C57Bl/6J mice after one month of water restriction. Most organ weights, including heart, spleen, kidneys, adrenal glands, and testes, were indistinguishable from control mice (6 male mice; ad libitum water consumption). The brain (94±1% of control, mean ± SD, p<0.001, t-test; all tests with Bonferroni correction) and spleen (54.6±6.7%, p<0.001) were smaller in the water deprived mice. Water restricted rodents tend to have lower organ weights [43] . The reason for the pronounced reduction of spleen size is unknown.

Blood samples were further extracted to analyze the physiological state of water restricted mice. The concentrations of most solutes were in the normal range, including sodium, potassium, chloride, aspartate aminotransferase (AST), alanine aminotransferase (ALT), blood urea nitrogen (BUN), CO 2 , total protein, albumin, tibili and creatinine. Glucose (55±16%; p<0.01) and alkaline phosphatase (ALP) (67±18%; p<0.05) were reduced in the water-deprived mice. Mice eat when water is available. The reduced glucose and ALP likely reflect that the mice were euthanized long after eating.

3. Handling and head-fixation

Four days prior to instrumental training (at least three days after starting water restriction) mice should be handled so that they become habituated to the training environment, including the experimenter's hands, body tube, head-fixation, rig, sounds in the experimental room, and other factors. As a result mice will be less stressed and learn faster. Here we describe our current procedures, but procedures with less extensive habituation have also been successful [23] .

Handling proceeds in three steps, typically on successive days.

Day 1 . The mouse is acclimatized to the experimenter's hands. We typically start by placing two sunflower seeds into the mouse's holding cage for 10–15 minutes, while removing any objects that the mouse can hide in (tubes, running wheel, cotton nests, etc). After the agitated mouse has settled down, we corner it with our hands with deliberate and gentle movements and allow the mouse to climb on our hand. We hold the mouse in our hands for 5–10 minutes until it calms down, as evidenced by grooming behaviors, and offer the mouse water using a syringe (approximately 0.2 ml). Drinking is a sign of relaxation.

We then let the mouse explore the body tube until he enters it. If the mouse enters the body tube we repeat the procedure 4–5 times without forcing the mouse. Otherwise we try again on Day 2.

Day 2 . The mouse is further acclimatized to the experimenter's hands and the apparatus. We hold the mouse and have it nibble at a sunflower seed ( Figure 7A ). The mouse will eat only if he feels comfortable. The mouse then explores the body tube again. A water reward (0.1–0.2 ml) is given after the mouse has entered the tube ( Figure 7B, C ). At this point the mouse is head-fixed rapidly (<10 s), with its body in the holding tube. Additional water (0.2 ml/5 minutes) is provided during head-fixation (10–15 minutes).

A. Mouse eating a sunflower seed on the experimenter's hand. The pins emanating from the top of the mouse head correspond to ground and reference electrodes for extracellular recordings. B. Mouse being familiarized with the body tube. C. Mouse receiving a water reward in the body tube.

https://doi.org/10.1371/journal.pone.0088678.g007

Day 3 . The mouse is acclimatized to the apparatus. The mouse is head-fixed and the caddy is placed into the behavioral apparatus for 30 minutes. Water rewards (0.2 ml) are provided every few minutes, for a total of 1 ml.

Day 4 . The procedures from the third day are repeated, but extended to 45 minutes. In addition, the mouse is introduced to a lickport as a source of water.

4. Training the lick/no-lick object location discrimination task

In this section we describe training of one version of a lick/no-lick (go/no-go) object location discrimination task in the dark (corresponding to the data in Figures 4 , 8 , 9 ). The goal is to train mice to use a single whisker (typically C2) to locate a vertical pole for a water reward. Single whisker tasks greatly simplify linking sensory stimuli to behavior and neurophysiology [18] .

A. Block-diagram of the possible events in a single trial. B. Schematic representation of event timing during a single lick trial. C. Schematic representation of the behavioral contingency. Mice had to lick for a water reward when the pole was in a posterior position and hold their tongue when the pole was in an anterior position. In some experiments, the contingency of the pole positions was reversed. D. Behavioral data from one session. The abscissa shows the time from trial start. Lick and no-lick trials are randomly interleaved. The pink ticks indicate licks. The red ticks indicate the first licks after the grace period. The blue bars correspond to the open times of the reward water valve. The horizontal green and red bars indicate whether each trial is correct or incorrect, respectively. The dark gray shading indicates that the pole is fully descended and in reach of the whiskers.

https://doi.org/10.1371/journal.pone.0088678.g008

A. Time-course of experiments. B. Learning curves showing the discriminability index, d'. Thin lines correspond to individual mice. Thick lines, average. Red, recording sessions. C. Learning curves showing the fraction of correct trials. D. Water consumed. E. Health score. A health score larger than 3 (dashed line) triggers more detailed evaluation and possibly water supplements.

https://doi.org/10.1371/journal.pone.0088678.g009

During each trial the object, a vertical pole (0.5–1 mm in diameter), was presented at one of several possible positions on one side of the face ( Figure 8C, D ). The no-lick position was a single anterior pole location. The lick position was one, or optionally multiple [19] , [20] , relatively posterior pole locations. In some experiments the contingency was reversed. The distance of the posterior pole location to the whisker pad was 5–8 mm. The final distance between the no-lick and the most anterior lick position was 4.29 mm along the anterior-posterior axis. Water was available through a single lickport centered on the midline. Movement of the pole took 0.5 s, after which the animal was given 2.5 s to search for the object with its whisker and indicate object location by licking or withholding licking ( Figure 8A, B ). To encourage multiple whisker-object contacts before signaling a response, the animal was given a grace period (0.5–1.5 s) from onset of pole movement where licking did not signal the response outcome. Following the grace period, a lick in the remaining pole availability time (answer lick) was scored as a hit if the pole was in a lick position or a false alarm if the pole was in the no-lick position. Hits triggered opening of a water valve to deliver approximately 8 uL of water. Two seconds after the answer lick, the pole retracted and the intertrial period began. On false alarm trials the mouse was given a timeout, typically 2–5 s, which retriggered on any additional licks during the timeout. If no lick occurred during the response window, the trial was scored as a miss (lick trial) or a correct rejection (no-lick trial). On both misses and correct rejections the intertrial period began immediately following the end of the response window. The intertrial period typically lasted two seconds, during which the pole first moved to the midpoint of the two pole positions and then to the position of the next trial.

Training proceeded through multiple stages. Mice were trained once a day for sessions lasting 45 to 90 minutes. The first day of training began with association between the presence of the pole and water availability. The pole was moved into the center of the whisker field (to ensure whisker-pole contacts) and any licking triggered a water reward. After three lick-triggered rewards the protocol was paused and the pole was moved out of reach of the whiskers. After a 10 s delay, the process was repeated, until mice licked concurrently with touch between whiskers and pole. If the mouse failed to lick after one minute, the lickport was manually seeded with a water droplet by briefly opening the valve using the behavioral control software. Mice often lick when smelling the water emerging from the lickport. If the mouse still refused to lick, the lickport was moved closer such that the droplet touched the fur. This always caused the animal to lick.

Mice were then exposed to the timing of the trials. The pole was moved to a single ‘lick’ position on repeated trials. Mice received rewards when licking 1–2 s after the pole came within reach and were not punished for excessive licking. Once the mouse received rewards on five consecutive trials, the pole was introduced in the no-lick position on 20% of trials. The initial no-lick position was far anterior, out of reach of the whiskers. This specifically links detection of the pole within the whisker field, rather than other cues such as sound and vibration, to availability of reward. Once the mouse licked on >75% of lick trials the probability of the no-lick position was increased to 50%, with a maximum of three consecutive trials of a single type. In cases of five or more consecutive misses, the no-lick probability was reduced to 0% until the animal began responding. About one half of the mice progressed to the 50% no-lick probability stage by the end of the first day of training, whereas others had difficulty moving beyond the initial association of pole presence and water availability.

Prior to the second day training session all whiskers except C2 were trimmed to 3 mm in length (i.e. too short to contact the pole). The lick (go) location was positioned 2 mm anterior to the resting position of the C2 whisker for each mouse, whereas the no-lick (no-go) position was out of reach. The pole was placed randomly in lick and no-lick positions with 50% probability, with a maximum of 3 consecutive trials of a single type. Whisking and licking were examined to identify possible training failure modes for each mouse. In case of high miss rates on trials where the whisker touched the pole, the lickport position was adjusted to ensure it was triggered properly on each attempted lick. If the animal had a high miss rate and the whisker did not strike the pole, the pole location was moved closer to the resting position of the whisker. If the animal was licking compulsively on lick and no-lick trials, the lickport was moved further from the animal's mouth and/or the no-lick probability was increased to 80% until several correct rejection trials occurred. If the animal was licking cautiously at least once on both trial types to probe for water rewards the timeout punishment was increased to 5 s. As the performance of the mouse increased during or across sessions, the no-lick position was progressively moved toward the lick position, within easy reach of a vigorous whisk of the C2 whisker, making this an object location discrimination task. The final distance between the lick position and the no-lick position was 4.29 mm. Sessions were terminated when mice missed 10 lick trials in a row (even after adjusting the lickport position for the early training sessions).

Individual mice learn at a variety of rates. After one week of training, the best mice achieved peak performance of >90/100 consecutive trials correct, with total session performance of >80% correct (discriminability index, d' >2), whereas other mice required up to 3 weeks to achieve similar performance levels ( Figure 9A–C ). In our experience, object localization with single whiskers is challenging for mice, and the training time might reflect the inherent difficulty of the task. With one row of intact whiskers training times are much shorter: mice typically learn the lick/no-lick pole detection task in 1–3 days [21] . Even faster learning can be achieved in lick/no-lick olfactory discrimination behaviors. We have found that mice routinely learn to report two different odors within one session [10] ( Figure 10 ).

A. Schematic representation of the behavioral contingency. Mice had to lick for a water reward when odor B was presented and hold their tongue when odor A was presented. B. Performance in the first session of the odor discrimination task (data from [10] ). Colored lines correspond to individual mice ( n  = 5).

https://doi.org/10.1371/journal.pone.0088678.g010

We have also observed that the distance of the pole from the whisker pad has a large impact on performance. The whisker is linearly tapered and its bending stiffness decreases gradually with distance from the whisker pad over five orders of magnitude [17] , [44] . Forces exerted by the pole on the whisker are usually larger when the pole is closer to the whisker pad, leading to faster learning in mice. In our experiments the distance of the pole to the whisker pad was 5–8 mm. Future innovations in shaping mouse behavior will no doubt shorten training times.

On days with behavioral sessions, mice generally obtained all water for the day during the session and were allowed to perform until sated. Mice typically performed 300 trials and received 0.6–1.2 mL of water. The amount of water consumed was determined by weighing the mouse before and after the session (including any excrement). If the mouse consumed an unusally small volume of water (<0.5 ml) a small water supplement (0.2–0.5 ml) was provided a few hours after training. Mice maintained body weight with health scores in the normal range (<3; Figure 9D, E ).

5. Training the lick-left/lick-right task with a delay epoch

The lick/no-lick object location discrimination task described above has several disadvantages for the study of decision making. First, animals are biased towards licking. Second, sensation and action (i.e. the answer lick) happen nearly simultaneously. For numerous experiments it is of interest to separate “sensation” and “action” in time. We therefore designed a task in which both pole positions are rewarded, with a delay epoch that separates sensation and action. The temporal structure of the task was modeled after behavioral paradigms widely used in psychophysics [45] .

Mice were trained to perform a symmetric response lick-left/lick-right object location discrimination task with a short-term memory component ( Figure 11 ) [7] . The behavioral apparatus and training procedures have been described [7] . In short, mice need to use their whiskers to locate a vertical pole (0.9 mm in diameter), presented at one of two possible positions on the right side of the face. The posterior pole position was placed 5 mm from the whisker pad. The two pole positions were spaced 4.29 mm apart along the anterior-posterior axis (40 degrees of whisking angle) and were held constant from session to session. Water was available through two lickports, spaced 4.5 mm apart. Mice were trained to indicate the posterior pole position with licking right, and the anterior pole position with licking left ( Figure 11C ); in some experiments the contingency was reversed. The pole was only available to the animals during the sample epoch and the animals need to hold their response for a brief delay epoch ( Figure 11B ). The delay epoch thus separated “sensation” and “action” in time. At the beginning of each trial, the vertical pole quickly moved within reach of the C2 whisker (0.2 s travel time). The pole remained within reach for 1 s, after which it was retracted. The retraction time was 0.2 s, of which the pole remained within reach in the first 0.1 s. The delay epoch lasted for another 1.2 s after the completion of pole retraction (delay epoch, 1.3 s total, Figure 11B ). At the end of the delay epoch, an auditory “response” cue (pure tone, 3.4 kHz, 0.1 s) was issued.

A. Block-diagram showing the possible events in a single trial. Licking during the sample or delay epochs leads to a brief timeout (1–1.2 s) and were not shown for clarity. B. Schematic of event timing during a single trial. Same as Figure 1C of [7] . C. Schematic representation of the behavioral contingency. Mice had to touch a left lickport for a water reward for an anterior pole location and a right lickport for a posterior pole location. In some experiments the contingency of the pole positions was reversed. D. Behavioral data from one session. Trials with the licking response before the response cue were excluded for clarity (25% of total trials). The abscissa shows the time from trial start. Lick-left and lick-right trials are randomly interleaved. The blue and light blue ticks indicate the onset time of the first and subsequent contacts respectively. The red and pink ticks indicate the first and subsequent licks respectively. The horizontal green and red bars indicate whether each trial is correct or incorrect respectively. The dark gray shading indicates the sample epoch during which the pole is within reach of the whiskers. The black vertical lines delineate the sample, delay and response epochs.

https://doi.org/10.1371/journal.pone.0088678.g011

Training was carried out in daily behavioral sessions that lasted 1–1.5 hours [7] . In the first behavioral session, mice received liquid rewards simply by licking either lickport. The auditory “response” cue was played immediately before water delivery; this contingency was kept constant throughout training. In the following sessions, the vertical pole was presented to indicate the rewarded lickport (e.g. the pole presented to the posterior position indicated that the right-side lickport was rewarded, see Figure 11C ). The rewarded lickport alternated between the two lickports after three rewards. Occasionally, water delivery by manually clicking a computer -controlled valve was necessary to prompt the mice to lick the other lickport. This phase of training lasted for 1–3 sessions. Presentation of the pole allowed the mice to gradually associate a pole position with licking the correct lickport. Presentation of the pole at the posterior position always touched some of the whiskers, whereas presentation of the pole at the anterior position made fewer contacts. Often, mice would start to associate the pole with licking the correct lickport. Signs of this could be gauged by the observation that mice quickly switched to lick the right-side lickport when the pole was presented at the posterior position (which typically contacted their whiskers). Once such signs were observed, mice were subjected to the object location discrimination task with no delay epoch, in which the presentation of the pole position was randomized. The mice were free to lick the correct lickport immediately after the pole was presented. Licking before the “response” cue was not punished. Licking the incorrect lickport after the “response” cue led to no liquid reward and a brief timeout (2–5 s). Typical mice learned this step quickly (5 sessions, Figure 12 ). After mice reached criterion performance with full whisker fields (typically >75% correct), the delay epoch was introduced. First, mice were trained to lick only after the “response” cue. Licking before the “response cue” was punished by a loud “alarm” sound (siren buzzer, 0.05 s duration, 2–4.5 KHz, 102 dB without shielding, RadioShack, 273-079), followed by a brief timeout (1–1.2 s). Continued licking triggered additional timeouts. The trial was allowed to resume once the timeout was complete, but these trials were excluded from the analyses (“lick early” trials, Figure 12E ). Mice gradually learned to suppress their licking before the “response” cue. Once mice were successfully conditioned to lick following the “response” cue, the pole was removed at the end of the sample epoch and the delay epoch was added in incremental steps (typical steps of 0.2–0.4 s added once per session).

A. Schematic of time-course of experiments. B. Learning curves showing the performance. Thin lines correspond to individual mice. Thick lines, average. Colors correspond to whisker trimming. Vertical dashed line indicates when the delay epoch was introduced. The four mice were from the same litter (2 males and 2 females). Same as Figure S1B in [7] . C. Learning curves showing the discriminability index, d'. D. Bias: performance of lick-right trials minus performance of lick-left trials. Same as Figure S1C [7] . E. The fraction of trials with licking responses during the sample or delay epoch. Same as Figure S1D [7] . F. Water consumed. G. Trials per session. H. Health score. A health score larger than 3 (dashed line) triggers more detailed evaluation and possibly water supplements. I. Health score for four mice that were under water restriction for four months. A health score larger than 3 (dashed line) triggers more detailed evaluation and possibly water supplements.

https://doi.org/10.1371/journal.pone.0088678.g012

After mice achieved criterion performance (>70%) on the object location discrimination task with a delay epoch, their whiskers were progressively trimmed (full whiskers→C row→C2, see Figure 12 ). The total training time for the full task is 3–4 weeks ( Figure 12A–D ). Trials in which mice did not lick within a 1.5 second window after the “response” cue were counted as “ignore” and excluded from the analyses. These “ignore” trials were rare and typically occurred at the end of a session, signaling that the mouse was sated or tired. Sessions were terminated when signs of fatigue were observed (e.g. reduced whisking, occurrence of “ignore” trials). Typically, the last 20 trials within each session were excluded from analyses. In a typical experimental session, fully trained mice performed 400 behavioral trials ( Figure 12G ). Under our conditions animals typically receive 0.8–1 ml water per day during training ( Figure 12F ). The health scores remain in a normal range (0–3) for up to four months of continuous water restriction ( Figure 12H, I ).

6. Modifications of the lick-left/lick-right task

The lick-left/lick-right object location discrimination task described above has a delay epoch to separate sensation and action, enabling study of perceptual decision. It usually takes 3–6 weeks to train mice to perform this task using a single (C2) whisker. Higher performance and shorter training times can be achieved if either the delay epoch is removed or mice are allowed to perform the task with multiple whiskers [46] . We often use a modified lick-left/lick-right object location discrimination task without delay (data in Figure 13 ). This task does not have a delay epoch, and mice perform object location discrimination with a row of whiskers. In addition, there were eight possible pole positions (evenly spaced at 1 mm) on the right side of the face (5 mm lateral to the whisker pad). The pole positions were held constant from session to session. Mice were trained to indicate the four posterior pole positions with licking right, and the four anterior pole positions with licking left.

A. Example experiment, with water (black circles) and sucrose (red circles) rewards provided on alternating sessions. B. The number of trials is 23% larger with sucrose (p<0.001 in two mice; n.s. in the third). C. The number of rewards per session is larger (p<0.001 in two mice; n.s. in the third). D. The discriminability index is unchanged.

https://doi.org/10.1371/journal.pone.0088678.g013

The lick-left/lick-right task with a delay epoch was also trained using an alternative strategy that used a motorized lickport. The left and right lickports were mounted on a stepper motor (Zaber Technologies, P/N NA08B30) which was controlled by a computer (i.e. the motorized lickport). The lickport was positioned so that it was centered along the animal's medial-lateral axis, but rested approximately 5 mm out of reach of the tongue. Immediately before the response epoch, the lickport was quickly moved within reach of the tongue (0.25 s) and mice initiated licking. Upon reward collection, or immediately after an incorrect response, the lickport was withdrawn. Most mice learn to withhold licking until the lickport moves into reach. This version of the task does not have a punitive stimulus (sound or timeout) to train a delay.

7. Sucrose rewards

To motivate mice to consume more water and thus perform more trials, we supplemented sucrose in water at 0.1 g/ml concentration (50 g sucrose and 1.7 g cool-aid black cherry mixed with water to 500 ml final volume). We trained three mice to perform the modified lick-left/lick-right object location discrimination task. Water or sucrose water was used on alternating sessions ( Figure 13A ). The reward liquid drop size was kept constant at 4 µL. The behavioral session was terminated when the mouse showed signs of being sated (e.g. reduced whisking, occurrence of trials without licking response). Mice were supplemented to 1 ml if they drank less than that amount in any behavioral session. This is to prevent mice from being thirstier on the subsequent session. Mice performed a significantly higher number of trials and obtained more rewards in sucrose water sessions ( Figure 13B, C ). The performance using sucrose water was not increased ( Figure 13D ). To assess potential adaption to sucrose reward, after one month of interleaved testing, we tested sucrose reward for an additional 15 consecutive sessions. Mice consistently consumed more sucrose reward compared with water. The caloric intake from sucrose is about 5% of total caloric intake in a normal mouse ( http://www.jax.org/phenome ). We did not observe obesity in mice trained on sucrose water for up to four months. Thus sucrose water boosted the number of trials per session without compromising the animals' performance and health.

We describe procedures for training head-fixed mice to perform robust perceptual behaviors. In each trial mice were exposed to one of several sensory stimuli and had to choose one of two responses based on the sensory stimuli. The behavioral choice was signaled by mice touching a water port with their tongue. Mice were water restricted, and thus motivated by thirst. Mice performed many hundreds of behavioral trials per session for water rewards. Weight loss associated with water restriction was positively correlated with the animals' behavioral performance and the number of correct trials ( Figure 5A, B ). Trained mice consumed 1 ml water per day during behavioral sessions. Mice maintained good health for four months of continuous water restriction ( Figure 12 ).

The water restriction procedure was developed for C57BL/6J mice and worked for all inbred laboratory strains we have used (C57BL/6Crl, PV-IRES-Cre, Six3-Cre, Scnn1a-Tg3-Cre, VGAT-ChR2-EYFP) [7] , [17] – [23] . Water restriction has to be adjusted depending on the relative humidity. Many species of mice survive, and even maintain their weight, without access to water at moderate levels of humidity [42] . Mice can derive their entire fluid intake from moist food. Laboratory mouse strains can vary with respect to their water consumption by several-fold ( http://www.jax.org/phenome ). The water schedule may also have to be adjusted according to mouse strain and sex. Furthermore, water restriction schedules also have to take activity in the home cage into account. Mice housed in enriched environments with access to treadmills need more water.

Our studies have focused on active tactile sensation in the sense that mice have to move their whiskers to accumulate information about tactile stimuli. Although it has long been appreciated that natural sensation is active [27] , [47] – [50] , neurophysiological studies of perception usually probe situations in which stimuli are applied passively (i.e. in fixating or immobilized non-human primates) [51] , [52] . In our behaviors mice controlled the position of the whiskers (but not their head) and thus the sensory input. Head-fixation was critical for these experiments because it facilitates precise measurements of the dynamics of whiskers and their interactions with objects [17] , [18] , [30] .

Mice were trained on either a lick/no-lick (go/no-go) or a lick-left/lick-right object location discrimination task. The lick/no-lick task has been successfully used to study neuronal correlates of perception [18] , [22] , sensorimotor integration and learning [19] – [21] . The lick/no-lick task has some disadvantages for the study of perceptual decisions. First, mice are intrinsically biased towards licking; that is, animals usually prefer licking to get water reward in “go” than withholding licking to avoid timeouts in “no-go” trials. This complicates the interpretation of psychometric curves and perturbation experiments [8] , [18] . Second, after a few touches with the pole, mice initiate licking within 100's of ms.. Thus the sensation of touch and action (i.e. licking) happen nearly simultaneously. To delineate “sensation” and “action” in time, we developed the lick-left/lick-right object location discrimination task with a delay epoch [7] . Mice accumulated tactile information during the sample epoch and maintained a memory of pole location or motor choice during the delay epoch. Though the lick-left/lick-right task has the advantage of separating behavioral events (e.g. whisker touch and licking) in time, it typically requires additional training time. In addition, the lick/no-lick task has trials without reward and licking, which can be helpful to isolate neural activity related to specific behavioral variables. We have also noticed differences in whisking strategies across the two types of behavioral tasks [7] , [18] .

The lickport position plays a crucial role in training. In the lick/no-lick task, if the lickport is too close mice tend to lick compulsively irrespective of trial type. If the lickport is too far, mice will tend to miss rewards and become discouraged. Adjusting the lickport position for individual mice is critical in behavioral shaping. In the lick-left/lick-right task, the left and right lickports are usually placed symmetrically along the midline of the animal's mouth. However, some mice have intrinsic licking bias and prefer to lick to one side over the other. This intrinsic bias can be countered by moving the preferred lickport laterally away from the animal's mouth. We ensured that the lickport positions are unchanged between experimental sessions, with occasional modifications to counteract animals' bias.

Although we focus our description on training active tactile behaviors, the core components of the methods can be used to train mice on other perceptual tasks. Training was divided into multiple stages (e.g. Figure 12 ). These stages can be grouped as follows: learning the mechanics of water rewards; learning trial and reward timing; associating reward with a stimulus (sometimes this stage was combined with the previous stage); when appropriate, learning about delays between stimulus and reward; learning perceptually more difficulty discriminations; reversal of stimulus – reward contingency (not discussed here). Mice were advanced from easier tasks to the next level when they performed at 70% correct. Mice were advanced promptly to avoid habit formation.

Acknowledgments

We thank Luciana Walendy for photography and help with experiments, Tanya Tabachnik for help with machining, Alison Vollmer for sucrose water recipe, Christopher D. Harvey, Michael Hausser, Dara Sosulski, Adam Packer, Beverley Clark and Martine Groen for comments on the manuscript.

Author Contributions

Conceived and designed the experiments: ZVG SAH NL DHO TK EO DH CB KM DG SP NX JC KS. Performed the experiments: ZVG SAH NL DHO TK EO DH CB KM DG SP NX JC KS. Analyzed the data: ZVG SAH NL DHO TK EO DH CB KM DG SP NX JC KS. Wrote the paper: ZVG KS.

• View Article
• Google Scholar
• 50. Yarbus AL (1967) Eye Movements and Vision. New York: Plenum Press.

• Climate Change
• Public Science Common

Using Ice to Boil Water: Science Experiment

Did you know that you can boil water using ice? In this video, Alex Dainis describes the science behind this unique trick! Water will boil at lower temperatures at higher altitudes, because the atmospheric pressure there is lower. Thus, if you can create a low pressure atmosphere in a jar, you can cause water to boil at a temperature lower than 100 degrees Celsisus. To execute this experiment, boil water in a heat proof jar, then screw on the lid. The water will stop boiling, but when you place ice on top of the lid, the water vapor and gasses inside the jar will cool down, creating a lower pressure atmosphere that allows the water to start boiling again!

• Women in STEM

• Research article
• Open access
• Published: 23 August 2024

Caspase-11 signaling promotes damage to hippocampal CA3 to enhance cognitive dysfunction in infection

• Ni Liang 2   na1 ,
• Yi Li 2   na1 ,
• Chuang Yuan 3 ,
• Xiaoli Zhong 4 ,
• Yanliang Yang 1 ,
• Fang Liang 2 ,
• Kai Zhao 2 ,
• Fangfang Yuan 2 ,
• Jian Shi 2 ,
• Erhua Wang 2 ,
• Yanjun Zhong 4 ,
• Guixiang Tian 5 ,
• Ben Lu 2 , 6 , 7 &
• Yiting Tang   ORCID: orcid.org/0000-0002-6878-4775 1  

Molecular Medicine volume  30 , Article number:  127 ( 2024 ) Cite this article

Metrics details

Cognitive dysfunction caused by infection frequently emerges as a complication in sepsis survivor patients. However, a comprehensive understanding of its pathogenesis remains elusive.

In our in vivo experiments, an animal model of endotoxemia was employed, utilizing the Novel Object Recognition Test and Morris Water Maze Test to assess cognitive function. Various techniques, including immunofluorescent staining, Western blotting, blood‒brain barrier permeability assessment, Limulus Amebocyte Lysate (LAL) assay, and Proximity-ligation assay, were employed to identify brain pathological injury and neuroinflammation. To discern the role of Caspase-11 (Casp11) in hematopoietic or non-hematopoietic cells in endotoxemia-induced cognitive decline, bone marrow chimeras were generated through bone marrow transplantation (BMT) using wild-type (WT) and Casp11-deficient mice. In vitro studies involved treating BV2 cells with E. coli-derived outer membrane vesicles to mimic in vivo conditions.

Our findings indicate that the deficiency of Casp11-GSDMD signaling pathways reverses infection-induced cognitive dysfunction. Moreover, cognitive dysfunction can be ameliorated by blocking the IL-1 effect. Mechanistically, the absence of Casp11 signaling significantly mitigated blood‒brain barrier leakage, microglial activation, and synaptic damage in the hippocampal CA3 region, ultimately leading to improved cognitive function.

This study unveils the crucial contribution of Casp11 and GSDMD to cognitive impairments and spatial memory loss in a murine sepsis model. Targeting Casp11 signaling emerges as a promising strategy for preventing or treating cognitive dysfunction in patients with severe infections.

Severe infection leads to sepsis, which has a high mortality rate and complex pathogenesis (Singer et al. 2016 ). Patients are often admitted to the intensive care unit due to severe infection, and nearly half of them exhibit encephalopathy symptoms such as altered consciousness, which can further increase their mortality. Even if some patients survived from survive sepsis, they will also have persistent cognitive impairment (Iwashyna et al. 2010 ; Barichello et al. 2019 ). Cognitive impairment resulting from infection represents a significant clinical challenge requiring resolution. However, the complete understanding of its pathogenesis remains elusive.

Endotoxin, or lipopolysaccharide (LPS), serves as the primary pathogenic component in the cell wall of gram-negative bacteria. In 1998, Toll-like receptor 4 (TLR4) was identified as the cell surface receptor for LPS (Poltorak et al. 1998 ). In recent years, researchers have discovered the receptor of intracellular LPS in mice, namely Caspase-11 (Casp11) (Kayagaki et al. 2011 , 2013 ; Shi et al. 2014 ). Gasdermin D (GSDMD) is a protein that is activated by Casp11 and then undergoes self-shear to become the N-terminal and C-terminal regions. The N-terminus undergoes oligomerization to form holes in the cell membrane (Shi et al. 2015 ; Ding et al. 2016 ). These GSDMD pores not only changes the internal and external osmotic pressure of cells, leading to cell swelling and lysis, but also releases a large number of inflammatory mediators, such as IL-1α and IL-1β, which cause severe inflammatory effects (Broz and Dixit 2016 ; Galluzzi et al. 2018 ; Rathinam et al. 2019 ). Recently, it was reported that living cells can release inflammatory mediators such as IL-1 through GSDMD pores without pyroptosis (Evavold et al. 2018 ).

Activated microglia have been recognized as pivotal contributors to the onset of cognitive impairment induced by severe infection (Zrzavy et al. 2019 ). Furthermore, MRI examinations in sepsis patients have revealed vasogenic edema and white matter hyperintensities associated with blood‒brain barrier (BBB) damage (Ehler et al. 2017 ). Notably, the hippocampus, intricately linked to learning and memory functions, stands out as a focal area of interest (Barbizet 1963 ; Kimble and Pribram 1963 ). The hippocampal CA3 region is important for the rapid encoding of memory that arises from different spatial functions (O'Keefe and Dostrovsky 1971 ). However, how they are linked to infection-induced cognitive dysfunction remains unclear.

Addressing the significant clinical issue of infection-induced cognitive impairment, we employed a mouse model of endotoxemia. This study aims to uncover the role and mechanism of Casp11-GSDMD signaling specifically in infection-related cognitive impairment.

Methods and materials

Animal model.

The mice of B6.SJL and C57BL/6 were purchased from Hunan SJA Company. Gsdmd-KO mice were donated by Academician Jiahuai Han (School of Life Sciences, Xiamen University) (He et al. 2015 ). Il-1r-KO mice were purchased from the Jackson Laboratory, while Casp11-KO mice were generously provided by Professor Timothy R. Billiar (University of Pittsburgh Medical Center) (Deng et al. 2018 ). The average body weight of the mice was 25 to 30 g and aged 8–12 weeks unless specified otherwise. The mice were raised in a specific pathogen-free conditions in the Department of Laboratory Animals of Central South University. Standard conditions included a 12 h light–dark cycle and a temperature range of 22–25 °C. The animal model uses intraperitoneal injection of 8 mg/kg LPS (from Escherichia coli (O111:B4), Sigma L2630) to induce cognitive disorder.

Novel object recognition test

On Day 5 post-LPS challenge, we conducted a Novel Object Recognition (NOR) test in a 40 × 40 cm field arena to evaluate the mice's ability to discriminate a novel object. The test comprised two main phases: training and testing (Antunes and Biala 2012 ). In the training phase, we used two identical objects and symmetrically placed on the diagonal of the arena. Each mouse, positioned in one corner facing the wall, was given 10 min of free exploration while being recorded with the Smart system. To prevent olfactory cues, the arena was thoroughly cleaned with 75% alcohol between each test. After a 24 h interval, the testing phase commenced. We replaced the previous object with a new object and the mice's preference for the new object reflected the mice's desire to explore. Mouse movement recording and data analysis were using the VisuTrack system (Shanghai XinRuan Information Technology Co., Ltd). The discrimination index is calculated as the ratio of time spent with the new novel object to the total time with both objects, served as the metric for evaluating discrimination ability. All behavioral experiment operations and data analyses were conducted in a blinded, randomized fashion.

Morris water maze test

The Morris Water Maze test, conducted from Day 7 to Day 11 post-LPS challenge, assessed spatial learning and memory in mice (Morris 1984 ). Mouse movements recording and data analysis were using the Smart3.0 system (Panlab Harvard Apparatus). The test employed a circular water tank, 120 cm in diameter, divided into four quadrants, each marked by distinct visual cues on the pool wall. An escape platform, hide in 1 cm below the water’s surface and concealed with edible titanium powder, was placed in one quadrant. The temperature of water was maintained at approximately 20 °C. Prior to the formal experiments, mice were acclimated to the water environment by spending 30 s on the platform. Subsequently, mice were introduced into the water from different entry points, facing the pool wall. Over four days, mice underwent training sessions three times a day, with a 1 min time limit to locate the platform. Mice failing to reach the platform within the allotted time were guided to it and allowed a 10 s stay; the latency to the platform was recorded. On the fifth day of the test period, we removed the platform and released the mice from a quadrant opposite the platform location into the water for 60 s. Spatial cognitive ability and memory were assessed through indicators such as the frequency of platform crossings, the latency to first reach the platform, and the proportion of time spent by mice in the target quadrant.

Western blot analysis

The separated proteins were using 10% or 12% sodium dodecyl sulfate‒polyacrylamide gel electrophoresis and subsequently transferred onto 0.22 μm PVDF membranes (Millipore). Following a 1 h room temperature blockade with 10% nonfat dry milk in TBST buffer, PVDF membranes were incubated with primary antibodies overnight at 4 °C. Detailed antibody information is as follows: IBA1 (Wako#019–19741, 1:1000), Casp11 (Novusbio, NB120-10454, 1:200), GSDMD (Abcam, ab209845, 1:1000), β-actin (Cell Signaling Technology, 1:5000). The second antibody (Jackson, 1:5000) was incubated at room temperature for 1 h. Western Bright ECL Spray was employed for band visualization.

Immunofluorescent staining

All animals underwent perfusion from the left ventricle with 20 ml PBS at 4 °C followed by 4% PFA (Paraformaldehyde). The entire brains, preserved in 4% PFA overnight, were subsequently removed. Dehydration was carried out in incrementing sucrose concentrations (15%, 25%, 35%, dissolved with PBS), and the brains were embedded in OCT, snap-frozen at -80 °C, and 20μm thick coronal sections were prepared at −20 °C low temperature thermostat. Following a 1 h 25 ℃ blocked with 0.1% Triton X-100 and 5% BSA, sliced brain was incubated with primary antibodies overnight at 4 °C: Ionized calcium-binding adaptor molecule 1 (Wako#019–19741, IBA1, 1:500), PSD-95 (Thermo Fisher Scientific#51–6900, 1:200,), Synaptophysin (Sigma#S5768, 1:200). Subsequently, sections underwent a 2 h incubation with secondary antibodies (Thermo Fisher Scientific#A10521 and F-2765, 1:500). Washed with PBST (0.01 M PBS containing 0.1% Triton X-100) for three times. Visualization and quantification by using Nikon Ni-U microscope and Image-J software. Confocal images were obtained using the Zeiss LSM800 confocal microscope.

Blood brain barrier permeability

The experiment of blood brain barrier permeability was assessed using Evans blue extravasation (Sigma E2129). A 2% solution of Evans blue (diluted in saline, 80 ml/kg) was intravenously injected through the medial canthus vein immediately after the LPS challenge, and circulation was allowed for 2 h. Then, the brains were perfused with 20 ml PBS and removed on a background of paper to take gross photographs. Brain samples were weighed to determine the wet weight and fixed using dimethylformamide (dimethylformamide: brain wet weight = 1000 μl: 300 mg). Following homogenization, incubation at 37 °C for 24 h, and subsequent centrifugation, fluorescence levels were measured with a microplate reader (emission at 680 nm and excitation at 620 nm). The Evans blue in tissue content was quantified using a linear standard curve derived from known dye amounts. For fluorescent panoramic scanning, images of 30 μm thick coronal sections post-Evans blue challenge were captured using a Keyence BZ-X microscope.

Purification of bacterial OMVs

We purified Outer Membrane Vesicles (OMVs) with E. coli BL21, the method following a previously described method with some modifications (Chutkan et al. 2013 ). Briefly, E. coli BL21 were cultured in an appropriate amount of LB solution until reaching an OD600 of ~ 0.5. To obtain the sterile supernatant, we first centrifuge the bacterial solution from the ice bath at 10,000 × g for 10 min at 4 °C. Then, the supernatant underwent further filtration through a 0.45 µm filter (Millipore), followed centrifugation at 10,000 × g for 10 min at 4 °C for the second time and filtration through a 0.22 µm filter (Millipore). OMVs were then obtained by ultracentrifugation at ~ 100,000 × g for 2 h at 4 °C using a Beckman Ti70 rotor. We removed the supernatant, the OMVs were resuspended in 500 µl sterile Dulbecco's Phosphate-Buffered Saline (DPBS) (Gibco) and filtered through a 0.22 µm filter. Purified OMVs underwent agar plating to confirm the absence of bacterial contamination, using a BCA protein assay kit (Thermo Scientific) to determine the protein content of OMV.

Cell culture and stimulations

1 × 10^6 BV2 cells were treated with OMVs for 18h at indicated doses or left untreated in a 6-well plate. Then the supernatants were collected 18h post-stimulation to test cell death. For in vitro siRNA silencing of Casp11, BV2 cells were cultured in 6-well plates at a density of 0.2 million cells/well for transfection. We utilized Lipofectamine RNAi MAX Transfection Reagent (Invitrogen; 13778150) for siRNA transfection. Cells were stimulated with OMVs after siRNA transfection 48 h. The siRNA target sequences were CCUGAAGAGUUCACAAGGCUUTT (mCasp11) and UUCUCCGAACGUGUCACGUTT (control). The knock down efficiency were assessed by western blot.

Limulus amebocyte lysate (LAL) assay

The Limulus Amebocyte Lysate (LAL) assay (Xiamen Bioendo Technology; EC64405) was employed to quantify LPS following the manufacturer's instructions. We first injected the mice with LPS, and then all the blood was sucked out and discarded through cardiac blood collection. We then injected a sufficient amount of endotoxin-free DPBS from the hearts of the mice to flush the remaining blood in the vascular circulation as clean as possible. At the end of perfusion, the brain tissue of mice was removed and mixed with endotoxin-free DPBS for grinding. Then brain homogenate extract was collected for LAL assay. At the same time, the last perfusion effluents were drawn from the outlet (right atrial appendage) for LAL assay. All LAL assay reagent and endotoxin-free consumables were supplied by kit.

ELISA and LDH assay

Plasma samples from adult mice and cell-free supernatant were examined using TNF-α (ThermoFisher; 88–7324), IL-1α (ThermoFisher; 88–5019), IL-6 (ThermoFisher; 88–701364), or IL-1β (ThermoFisher; 88–7013), ELISA kits. Cell death was assessed via the LDH Cytotoxicity Assay kit (Beyotime Biotechnology).

Proximity-ligation assay (PLA)

The experiment of Proximity Ligation Assay kit (Sigma DUO92008) was utilized to reveal the interaction between Casp11 protein and LPS in sections of mouse brain tissue. This unique method allows for the visualization of subcellular localization and protein–protein interactions in situ. Mice were exposed to LPS for 11 h, and 20 μm thick frozen sections of brain tissue were prepared as described earlier. Following fixation with 4% formaldehyde and penetration with TritonX-100, sections were incubated overnight with different kinds of primary antibody pairs for Casp11 (rat monoclonal 17D9, Novus NB120-10454) and LPS (mouse monoclonal 2D7/1, Abcam ab35654). After the primary antibody incubation, the sections were treated with the corresponding PLA probe—coupled oligonucleotide secondary antibody combination (rat MINUS and mouse PLUS for Casp11 and LPS interactions). The Proximity Ligation Assay was executed according to the manufacturer’s instructions, and images were captured by using a Nikon Ni-U microscope.

Bone marrow transplantation (BMT)

BMT mice, aged between 5 and 9 weeks, were employed, ensuring the use of sex-matched donor-recipient pairs (all BMT groups represented as donor → recipient). WT donor mice were on the B6.SJL background, and knockouts were on the C57BL/6 background, while recipient mice were on the C57BL/6 background, encompassing both WT and knockout. Bone marrow transplant recipients received 0.5 million Bone marrow cells from donors following irradiation with 12 Gy (administered in split doses of 2 × 6.0 Gy spaced 4 h apart at a rate of 0.5 Gy/min). The recipients were then raised aseptically in an air-laminar flow cabinet and given antibiotic water for 2 weeks. Hematopoietic reconstitution of chimeric mice was detected 4–5 weeks after BMT. Briefly, PBMC were obtained from peripheral blood of recipient mice, and the expressions of CD45.1 and CD45.2 were detected by flow cytometry. Subsequently, the chimeric mice were injected with LPS and the NOR experiment was performed several days after injection.

Statistical analysis

We used the Version 9.0 Prism GraphPad and IBM SPSS Statistics 21 for statistical analyses. Data underwent analysis used by One-way or Two-way ANOVA and Repeated Measures ANOVA. p < 0.05 was applied for statistical significance. All experiments were replicated at least three times independently, and all values are expressed as the mean ± SEM.

Casp11 promotes cognitive dysfunction after LPS challenge.

Cognitive dysfunction caused by infection is an important clinical problem. To elucidate the impact of Casp11 on cognition during infection, we employed an LPS-induced endotoxemia mouse model (Savi et al. 2021 ). We designed a behavioral experiment to assess the cognitive function of mice (Fig.  1 A). The NOR experiment was used to assess the object recognition memory of the mice (a novel object was placed into the red circle). Discrimination ratio presents the proportion of time spent exploring one of the two identical objects. During the training phase of the NOR test, we found that the discrimination ratio of the mice was approximately 0.5, with no significant differences between the groups, meaning that the results would not be biased by the experimental conditions and setup (Fig.  1 C). The discrimination ratio of the mice in the WT-LPS group was also approximately 0.5 in the test phase, while the mice in the Casp11 −/− -LPS and WT-saline groups had discrimination ratios over 0.6 with statistically significant differences (Fig.  1 D). The trajectory density visually illustrates the mice's inclination to explore objects (Fig.  1 B). This suggests impaired recognition memory for the old object in the WT-LPS group, where mice struggled to recognize the new object. By contrast, akin to WT saline mice, those in the Casp11 −/− -LPS group exhibited restored cognitive function. In the Morris water maze, LPS-treated mice displayed prolonged escape latency compared to the WT-normal saline group during hidden platform water maze training (Fig.  1 F). Notably, the Casp11 −/− -LPS group exhibited performance akin to the WT-saline group. The LPS-treated mice showed diminished target platform crossings, spent reduced percentages of time in the target quadrant, and experienced increased difficulty locating the hidden platforms. Additionally, the Casp11 −/− -LPS group closely resembled the WT-saline group (Fig.  1 G–J). These findings showed that Casp11 signaling is involved in cognitive impairment of mice after LPS challenge.

Casp11 promotes cognitive dysfunction after LPS challenge. A The strategy of the experiment. Systemic treatments consisted of an intraperitoneal injection of 8 mg/kg LPS or saline. B – D NOR experiment. WT-Saline (n = 21), Casp11 −/− -Saline (n = 10), WT-LPS (n = 17), Casp11 −/− -LPS (n = 27), from four independent experiments. B Trajectory tracking of the representative mice. C Discrimination ratio in the training period. D Discrimination ratio in the testing period. E – J MWM experiment. WT-Saline (n = 12), Casp11 −/− -Saline (n = 10), WT-LPS (n = 16), Casp11 −/− -LPS (n = 14), from three independent experiments. E Diagram of the water maze partition. F Escape latency in the MWM training task. G Trajectory tracking of the representative mice in testing task by the Smart v3.0-Panlab Harvard Apparatus, from three independent experiments with similar results. H The number of platform area crossings of mice in testing task. I The percentage of time spent in the target quadrant by mice in testing task. J The time of first occurrence to the platform area (latency of the 1st entrance to the target) of mice in testing task. The data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

GSDMD enhances cognitive dysfunction in endotoxemic mice.

GSDMD can be cleaved by activated Casp11 (Shi et al. 2015 ; He et al. 2015 ; Kayagaki et al. 2015 ). To determine whether GSDMD was also involved in the formation of cognitive impairment in endotoxemic mice, behavioral tests were performed. During the NOR task, deletion of GSDMD significantly ameliorated the novel object recognition performance in LPS-challenged mice (Fig.  2 B, C). Traces of movements in the NOR task showed similar trends (Fig.  2 A). Moreover, GSDMD-deficient mice exhibited enhanced performance in the water maze task (Fig.  2 D, E), demonstrating more target platform crossings, higher percentages of time spent in the target quadrant, and a faster location finding compared to WT mice (Fig.  2 F–H). These findings showed that GSDMD also enhances cognitive dysfunction.

GSDMD enhances cognitive dysfunction in endotoxemic mice. (A-C) NOR experiment. WT-Saline (n = 17), Gsdmd −/− -Saline (n = 11), WT-LPS (n = 23), Gsdmd −/− -LPS (n = 18), from three independent experiments. A Trajectory tracking of representative mice. B Discrimination ratio in the training period. C Discrimination ratio in the testing period. D – H MWM experiment. WT-Saline (n = 14), Gsdmd −/− -Saline (n = 11), WT-LPS (n = 12), Gsdmd −/− -LPS (n = 19), from three independent experiments. D Escape latency in the MWM training task. E Trajectory tracking of representative mice in testing task, from three independent experiments with similar results. F The number of platform area crossings of mice in testing task. G The percentage of time spent in the target quadrant by mice in testing task. H The time of first occurrence to the platform area of mice in testing task. The data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

The effect of the IL-1 cytokine aggravates cognitive dysfunction.

Casp11 and GSDMD signal may mediate the release of inflammatory factors, including IL-1α and IL-1β (Shi et al. 2015 ). To determine the influence of inflammatory cytokines in more detail, levels of inflammatory cytokines were measured in endotoxemic mice. Casp11 or GSDMD deficiency significantly reduced the release of the inflammatory cytokines IL-1α and IL-1β in the plasma (Fig.  3 A). Next, NOR experiment results showed that IL-1R knockout mice performed better than wild-type mice (Fig.  3 B–D). Furthermore, the deletion of IL-1r ameliorated the Morris water maze performance in LPS-challenged mice, with shorter escape latency during hidden platform water maze training (Fig.  3 E). In the testing period, IL-1r-deficient mice showed more target platform crossings and greater percentages of time spent in the target quadrant and were faster at finding the location than WT mice (Fig.  3 F–I). Altogether, these results reveal that inflammatory cytokines make an important contribution to cognitive dysfunction.

The effect of the IL-1 cytokine aggravates cognitive dysfunction. A IL-1α and IL-1β levels in the plasma of mice intraperitoneally challenged with LPS for 16h, n = 3–4 per group. B – D NOR experiment. WT-Saline (n = 9), Il-1r −/− -Saline (n = 10), WT-LPS (n = 12), Il-1r −/− -LPS (n = 9), from three independent experiments. B Trajectory tracking of representative mice. B Discrimination ratio in the training period C Discrimination ratio in the testing period. E - I MWM experiment. WT-Saline (n = 15), Il-1r −/− -Saline (n = 10), WT-LPS (n = 22), Il-1r −/− -LPS (n = 13), from three independent experiments. E Escape latency in the MWM training task. F Trajectory tracking of representative mice in testing task, from three independent experiments with similar results. G The number of platform area crossings of mice in testing task. H The percentage of time spent in the target quadrant by mice in testing task. I The time of first occurrence to the platform area of mice in testing task. The data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Casp11-GSDMD signaling promotes damage to the hippocampal CA3 region.

The hippocampus, vital for spatial memory, plays a pivotal role in understanding how Casp11 and GSDMD contribute to cognitive dysfunction in sepsis. Our investigation into microglial activation revealed a significant increase in microglial activity in the hippocampal CA3 region of LPS-challenged mice, a response that was restored in Casp11 or GSDMD-deficient mice (Fig.  4 A, B). Subsequently, western blot analysis confirmed a notable elevation in IBA1 protein levels in the hippocampus of LPS-exposed groups compared to saline controls, a rise that was mitigated in Casp11 or GSDMD-deficient mice (Fig.  4 C). Synapse loss was evident in the hippocampal CA3 regions (Fig.  4 D, E). The noteworthy impact of Casp11 and GSDMD warrants attention. In vivo, bacteria-released OMVs were found capable of delivering LPS into mouse macrophages' cytosol, activating Casp11 (Vanaja et al. 2016 ). Furthermore, to elucidate the effect of Casp11-GSDMD on microglia, immunoblot showed that BV2 microglia exhibited Casp11 activation and GSDMD cleavage, and LDH assay or ELISA showed the release of lactate dehydrogenase and IL1 when treated BV2 cells with E.coli-derived OMVs (Fig.  4 F, G). These results suggest that Casp11-GSDMD signaling may play an important role in associated with overactivation of microglia.

Casp11 and GSDMD promotes damage to the hippocampal CA3 region. A Immunofluorescent staining in the hippocampal CA3 region. (IBA1: green, the marker of microglia). Scale bar represents 100μm, n = 3 per group. B The fluorescence intensity of microglial activation was quantified by ImageJ software. C Immunoblot indicating IBA1 in the hippocampus of mice. D Loss of synapses in CA3 as determined by quantification of colocalized pre- and postsynaptic puncta on confocal images. Scale bar represents 20μm, n = 4–6 mice per group. E ImageJ was used to determine the number of synaptic colocalized. F Immunoblot to detect Casp11, GSDMD in supernatants (SN) or cell lysates (Cell) in BV2 microglia transfected with scrambled siRNA or Casp11-specifific siRNA upon OMV 25μg/mL for 18h. G LDH assay for cytotoxicity and ELISA for total IL-6, IL-1α, IL-1β, and TNF-α in cell culture medium. The data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Infection increased BBB permeability and neuroinflammation via Casp11 signaling.

There is a special structure between the peripheral circulation and the internal brain environment, named the BBB (Gao and Hernandes 2021 ). To further determine the role of Casp11-GSDMD signaling, BBB permeability was measured with Evans blue dye (Gibson and Evans 1937 ). As seen in the gross photographs, the degree of blue staining in Casp11, GSDMD or IL-1r deficient mice was significantly less than that in the WT mice after LPS challenge and similar to that in the saline group (Fig.  5 A). Meanwhile, Evans blue permeation quantitative experiments showed the same results as the gross photographs (Fig.  5 B). Panoramic scanning images of coronal brain sections of mice showed the lower degree of fluorescence intensity in Casp11, GSDMD or IL-1r deficient mice after LPS challenge (Fig.  5 C, D). In addition, elevated LPS levels were detected in brain homogenates when brain barrier leakage occurs after LPS challenge, which was significantly reduced by genetic deletion of Casp11, Gsdmd or IL-1r (Fig.  5 E). Meanwhile, the levels of IL1α, IL1β and IL6 in brain homogenates of Casp11 or Gsdmd deficient mice were significantly lower than those of WT mice, it appears that damage to the blood–brain barrier may be associated with (Fig.  5 F–H). These results suggest a key link between blood–brain barrier leakage, neuroinflammation and cognitive decline. To investigate the entry of LPS into brain cells and its activation of Casp11 in vivo, we conducted a proximity-ligation assay (PLA), providing a visual representation of the LPS- Casp11 interaction in endotoxemia tissues. PLA results revealed the induction of LPS- Casp11 interaction in the brain tissue of WT mice but not Casp11-deficient mice (I-J), suggesting the entry of LPS into the brain and subsequent Casp11 activation. To discern whether Casp11 in hematopoietic or non-hematopoietic cells contributes to endotoxemia-induced cognitive decline, we generated bone marrow chimeras through bone marrow transplantation (BMT) using WT and Casp11-deficient mice (L-M). NOR test results demonstrated that the deletion of Casp11 in the non-hematopoietic compartment, not the hematopoietic compartment, prevented endotoxemia-induced cognitive decline (M). Thus, non-hematopoietic Casp11 emerges as a crucial player in endotoxemia-induced cognitive decline.

Infection increased BBB permeability and neuroinflammation via Casp11 signaling. A - C Mice were treated with LPS and 2% Evans blue (2 h, i.v.). A Representative photographs of the gross brains of mice, from three independent experiments with similar results. B The concentration of Evans blue was expressed as μg Evans blue per mL dimethylformamide. n = 3–6 per group. C Graphs of frozen coronal brain sections were acquired from three independent experiments with similar results, scale bar represents 500μm. D The fluorescence intensity of Graphs of frozen coronal brain sections were quantified by ImageJ. E Limulus Amebocyte Lysate (LAL) assay to detect LPS in brain homogenate after perfusion. F – H IL-1α, IL1β and IL6 levels in brain homogenate of mice intraperitoneally challenged with LPS, n = 3 per group. I - J The interaction between Casp11 and LPS were visualized as the red spots by PLA in mouse frozen sections of brain tissue. Scale bar represents 50 μm. K The expressions of CD45.1 and CD45.2 in chimeric mice. L , M NOR experiment of chimeric mice, WT-Saline (n = 9); WT → WT-Saline (n = 6); WT → Casp11 −/− -Saline (n = 7); Casp11 −/−  → WT-Saline (n = 6); Casp11 −/−  → Casp11 −/− -Saline (n = 4); WT-LPS (n = 8); WT → WT-LPS (n = 9); WT → Casp11 −/− -LPS (n = 9); Casp11 −/−  → WT-LPS (n = 10); Casp11 −/−  → Casp11 −/− -LPS (n = 11), from three independent experiments. The data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

This study unveils a significant contribution of Casp11 and GSDMD to cognitive impairments and spatial memory loss in a murine sepsis model. In Gram-negative sepsis, circulating endotoxin can infiltrate host cell cytoplasm through high-mobility group box 1 protein (HMGB1) or bacterial outer membrane vesicles. This intrusion triggers the activation of cytosolic Casp11, the endotoxin's intracellular receptor (Deng et al. 2018 ; Vanaja et al. 2016 ). Activated Casp11 then enzymatically cleaves its downstream substrate GSDMD, producing membrane-pore-forming peptides that aggregate on the cell membrane. This cascade results in the release of numerous inflammatory mediators, including interleukin-1α (IL-1α) and interleukin-1β (IL-1β), along with the induction of lytic cell death.

The Casp11 and GSDMD-dependent functional abnormalities after endotoxin challenge are associated with the pathological changes in the CA3 region of the hippocampus, including cellular structural disruptions and synaptic loss. During the course of studying the underlying mechanisms, we found that Casp11 and GSDMD were required for endotoxin-increased permeability of blood–brain barrier (BBB), which is a vital structure that importantly regulates the exchange of molecules between the brain and the blood stream and provides protection against substance such as certain cytokines that are harmful for the central nerve system. Previous studies have reported changes in BBB permeability in sepsis (Gao and Hernandes 2021 ; Chung et al. 2020 ). We show that the endotoxin-induced BBB leakage is mediated by the activation of the Casp11 pathway. Casp11 is highly expressed in endothelial and myeloid cells. As endothelial cells are the major components that constitute BBB, it is conceivable that Casp11 induced endothelial damage or dysfunction might directly increase the permeability of BBB. Previous studies also implicate that immune cell-derived proinflammatory cytokines could also increase the permeability of BBB. Because activation of GSDMD in myeloid cells is associated with the release of interleukin-1 family cytokines, myeloid Casp11 might also contribute to the BBB leakage.

Understanding the mechanisms underlying infection-induced cognitive impairments is of great importance. Our findings shed light on the role of Casp11 and GSDMD in the disruption of the BBB and the subsequent neuroinflammation that contributes to synaptic loss and cognitive dysfunction. Further investigations should focus on elucidating whether and how Casp11 dependent BBB permeability causes neuroinflammation and cognitive decline in sepsis. In conclusion, data in current study demonstrates the involvement of the Casp11 and GSDMD in the breakdown of the blood–brain barrier, the enhance of neuroinflammation, and the loss of synaptic connections in the hippocampus, all of which are associated with the cognitive decline. These findings not only provide scientific evidence and a mechanistic basis for infection-induced persistent cognitive impairments, but also offer potential therapeutic targets for the prevention and treatment of cognitive impairment caused by infection.

Availability of data and material

The datasets used and/or analysed during the current study are available from the corresponding author or on reasonable request.

Antunes M, Biala G. The novel object recognition memory: neurobiology, test procedure, and its modifications. Cogn Process. 2012;13(2):93–110.

Article   CAS   PubMed   Google Scholar  

Barbizet J. Defect of memorizing of hippocampal-mammillary origin: a review. J Neurol Neurosurg Psychiatry. 1963;26(2):127–35.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Barichello T, Sayana P, Giridharan VV, et al. Long-term cognitive outcomes after sepsis: a translational systematic review. Mol Neurobiol. 2019;56(1):186–251.

Broz P, Dixit VM. Inflammasomes: mechanism of assembly, regulation and signalling. Nat Rev Immunol. 2016;16(7):407–20.

Chung HY, Wickel J, Brunkhorst FM, et al. Sepsis-associated encephalopathy from delirium to dementia? J Clin Med. 2020;9:3.

Article   Google Scholar  

Chutkan H, Macdonald I, Manning A, et al. Quantitative and qualitative preparations of bacterial outer membrane vesicles. Methods Mol Biol. 2013;966:259–72.

Deng M, Tang Y, Li W, et al. The endotoxin delivery protein HMGB1 mediates caspase-11-dependent lethality in sepsis. Immunity. 2018;49(4):740-53.e7.

Ding J, Wang K, Liu W, et al. Pore-forming activity and structural autoinhibition of the gasdermin family. Nature. 2016;535(7610):111–6.

Ehler J, Barrett LK, Taylor V, et al. Translational evidence for two distinct patterns of neuroaxonal injury in sepsis: a longitudinal, prospective translational study. Crit Care. 2017;21(1):262.

Article   PubMed   PubMed Central   Google Scholar  

Evavold CL, Ruan J, Tan Y, et al. The pore-forming protein gasdermin D regulates interleukin-1 secretion from living macrophages. Immunity. 2018;48(1):35-44.e6.

Galluzzi L, Vitale I, Aaronson SA, et al. Molecular mechanisms of cell death: recommendations of the Nomenclature Committee on Cell Death 2018. Cell Death Differ. 2018;25(3):486–541.

Gao Q, Hernandes MS. Sepsis-associated encephalopathy and blood-brain barrier dysfunction. Inflammation. 2021;44(6):2143–50.

Article   PubMed   Google Scholar  

Gibson JG, Evans WA. Clinical studies of the blood volume I Clinical application of a method employing the azo dye “evans blue” and the spectrophotometer. J Clin Invest. 1937;16(3):301–16.

He WT, Wan H, Hu L, et al. Gasdermin D is an executor of pyroptosis and required for interleukin-1β secretion. Cell Res. 2015;25(12):1285–98.

Iwashyna TJ, Ely EW, Smith DM, et al. Long-term cognitive impairment and functional disability among survivors of severe sepsis. JAMA. 2010;304(16):1787–94.

Kayagaki N, Warming S, Lamkanfi M, et al. Non-canonical inflammasome activation targets caspase-11. Nature. 2011;479(7371):117–21.

Kayagaki N, Wong MT, Stowe IB, et al. Noncanonical inflammasome activation by intracellular LPS independent of TLR4. Science. 2013;341(6151):1246–9.

Kayagaki N, Stowe IB, Lee BL, et al. Caspase-11 cleaves gasdermin D for non-canonical inflammasome signalling. Nature. 2015;526(7575):666–71.

Kimble DP, Pribram KH. Hippocampectomy and behavior sequences. Science. 1963;139(3557):824–5.

Morris R. Developments of a water-maze procedure for studying spatial learning in the rat [J]. J Neurosci Methods. 1984;11(1):47–60.

O’keefe J, Dostrovsky J. The hippocampus as a spatial map Preliminary evidence from unit activity in the freely-moving rat. Brain Res. 1971;34(1):171–5.

Poltorak A, He X, Smirnova I, et al. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice: mutations in Tlr4 gene. Science. 1998;282(5396):2085–8.

Rathinam VAK, Zhao Y, Shao F. Innate immunity to intracellular LPS. Nat Immunol. 2019;20(5):527–33.

Savi FF, De Oliveira A, De Medeiros GF, et al. What animal models can tell us about long-term cognitive dysfunction following sepsis: A systematic review. Neurosci Biobehav Rev. 2021;124:386–404.

Shi J, Zhao Y, Wang Y, et al. Inflammatory caspases are innate immune receptors for intracellular LPS. Nature. 2014;514(7521):187–92.

Shi J, Zhao Y, Wang K, et al. Cleavage of GSDMD by inflammatory caspases determines pyroptotic cell death. Nature. 2015;526(7575):660–5.

Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. 2016;315(8):801–10.

Vanaja SK, Russo AJ, Behl B, et al. Bacterial outer membrane vesicles mediate cytosolic localization of LPS and caspase-11 activation. Cell. 2016;165(5):1106–19.

Zrzavy T, Hoftberger R, Berger T, et al. Pro-inflammatory activation of microglia in the brain of patients with sepsis. Neuropathol Appl Neurobiol. 2019;45(3):278–90.

Download references

Acknowledgements

The authors thank Qianqian Xue, Ling Li, Weixiao Qin, Mingjie Shao, Feiyi Yang and Wenhao Yang for managing mouse colonies and research assistance.

This work was supported by the Hunan Provincial Outstanding Youth Science Fund Project (2022JJ10088), the Key Scientific Project of Hunan Province (2022SK2056), the National Natural Science Foundation of China (No.82272218 & No.81930059), the National Outstanding Youth Science Fund Project of the National Natural Science Foundation (No. 82025021) and the Scientific Research Program of FuRong Laboratory (No.2023SK2086). 

Author information

Ni Liang and Yi Li These authors contributed equally.

Authors and Affiliations

Department of Physiology, School of Basic Medical Science, Central South University, Changsha, 410000, People’s Republic of China

Yanliang Yang & Yiting Tang

Department of Critical Care Medicine and Hematology, The 3Rd Xiangya Hospital, Central South University, Changsha, 410000, People’s Republic of China

Ni Liang, Yi Li, Fang Liang, Kai Zhao, Fangfang Yuan, Jian Shi, Erhua Wang & Ben Lu

Xiangya School of Pharmaceutical Sciences, Central South University, Changsha, 410000, People’s Republic of China

Chuang Yuan

ICU Center, The Second Xiangya Hospital, Central South University, Changsha, 410000, People’s Republic of China

Xiaoli Zhong & Yanjun Zhong

Department of Ultrasound, The Second Xiangya Hospital of Central South University, Changsha, 410000, People’s Republic of China

Guixiang Tian

Hunan Key Laboratory of Organ Fibrosis, Xiangya Hospital, Central South University, Changsha, 410000, People’s Republic of China

Key Laboratory of Sepsis Translational Medicine of Hunan, Central South University, Changsha, Hunan Province, 410000, People’s Republic of China

You can also search for this author in PubMed   Google Scholar

Contributions

Yiting Tang and Ben Lu designed the experiments and revised the manuscript. Ni Liang and Yi Li collected and analyzed the data and wrote the manuscript. Xiaoli Zhong, Yanliang Yang participated in the design of the study and performed the statistical analysis. Chuang Yuan, Yanjun Zhong, Guixiang Tian conceived of the study, and participated in its design and coordination and helped to draft the manuscript. Fang Liang, Kai Zhao, Jian Shi, Fangfang Yuan and Erhua Wang provided technical assistance. Yiting Tang, Ben Lu, and Jian Shi made revisions to the final version of the manuscript. The strategy of the experiment graph was created with BioRender.com.

Corresponding author

Correspondence to Yiting Tang .

Ethics declarations

Ethics approval and consent to participate.

The study was approved by the Institutional Animal Protection and Utilization Committee of Central South University (No.2019sydw0325).

Consent for publication

Not application.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Liang, N., Li, Y., Yuan, C. et al. Caspase-11 signaling promotes damage to hippocampal CA3 to enhance cognitive dysfunction in infection. Mol Med 30 , 127 (2024). https://doi.org/10.1186/s10020-024-00891-y

Download citation

Received : 05 January 2024

Accepted : 05 August 2024

Published : 23 August 2024

DOI : https://doi.org/10.1186/s10020-024-00891-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Endotoxemia
• Cognitive dysfunction
• Neuroinflammation
• Blood‒brain barrier

Molecular Medicine

ISSN: 1528-3658

• General enquiries: [email protected]

• Architecture and Design
• Asian and Pacific Studies
• Business and Economics
• Classical and Ancient Near Eastern Studies
• Computer Sciences
• Cultural Studies
• Engineering
• General Interest
• Geosciences
• Industrial Chemistry
• Islamic and Middle Eastern Studies
• Jewish Studies
• Library and Information Science, Book Studies
• Life Sciences
• Linguistics and Semiotics
• Literary Studies
• Materials Sciences
• Mathematics
• Social Sciences
• Sports and Recreation
• Theology and Religion
• Publish your article
• The role of authors
• Promoting your article
• Abstracting & indexing
• Publishing Ethics
• Why publish with De Gruyter
• How to publish with De Gruyter
• Our book series
• Our subject areas
• Your digital product at De Gruyter
• Contribute to our reference works
• Product information
• Tools & resources
• Product Information
• Promotional Materials
• Orders and Inquiries
• FAQ for Library Suppliers and Book Sellers
• Repository Policy
• Free access policy
• Open Access agreements
• Database portals
• For Authors
• Customer service
• People + Culture
• Journal Management
• How to join us
• Working at De Gruyter
• Mission & Vision
• De Gruyter Foundation
• De Gruyter Ebound
• Our Responsibility
• Partner publishers

Your purchase has been completed. Your documents are now available to view.

Mice can recognise water depths and will avoid entering deep water

Rodents are averse to bodies of water, and this aversion has been exploited in experiments designed to study stress in mice. However, a few studies have elucidated the characteristics of murine water aversion. In this study, we investigated how mice behave in and around areas filled with water. Using variants of the open field test that contained pools of water at corners or sides of the field, we recorded the movements of mice throughout the field under various conditions. When the water was 8 mm deep, the mice explored the water pool regardless of whether an object was placed within it, but when the water was 20 mm deep, the mice were less willing to enter it. When the mice were placed on a dry area surrounded by 3 mm-deep water, they explored the water, but when they were surrounded by 8 mm-deep water, they stayed within the dry area. Our results indicate that mice exhibit exploratory behaviours around water, they can recognise water depths and avoid unacceptably deep water, and their willingness to enter water may be reduced by situational anxiety. Our experimental method could be used to investigate water-related anxiety-like behaviours in mice.

1 Background

In the wild, mice exhibit a tendency to avoid water as much as possible [ 1 ], and when mice are placed in water, they fall and immediately stiffen. Although it is difficult to determine the exact feelings of mice with regard to water, these observations clearly indicate that they do not like water. The aversion of mice to water has been exploited in the design of various behavioural tests, including the forced swim test, the water T-maze test, and the Morris water maze test [ 2 ]. Furthermore, this aversion is frequently used to induce chronic stress in mice through repeated forced swimming sessions and the placement of wet rugs in breeding cages [ 3 , 4 , 5 ].

The irrational fear of water in humans is known as aquaphobia, and aquaphobia is among the common simple phobias. Phobias, which are defined as abnormal psychological and physiological fears for a specific thing [ 6 ], are classified as anxiety disorders in the tenth revision of the International Statistical Classification of Diseases and Related Health Problems and are common comorbidities in patients with other anxiety disorders [ 7 , 8 ]. A past investigation [ 9 ] reported an aquaphobia prevalence of 1.8%, or approximately 1 in 50 people, in the general Icelandic population, and the symptoms of aquaphobia, which can include headache, suffocation, panic attacks, and decreased water intake [ 10 ], can adversely affect productivity, confidence levels, and overall health. However, only 9% of patients with general phobias report having consulted a physician about their conditions [ 9 ].

The causes of phobias remain largely undetermined. Researchers have speculated that aquaphobia may arise from a combination of genetics and experiential factors (e.g. swimming ability and instances of needing to be rescued from water) [ 11 ]. It is often thought that experiential factors are the most important contributors to aquaphobia in adults, but Poulton et al. found no association between swimming experience during the first 9 years of life and aquaphobia at the age of 18 years [ 12 ]. In accordance with Darwin’s non-associative model of fear acquisition, aquaphobia may constitute a type of innate fear that can manifest without any history of distressing experiences. Such innate fears may diminish over time due to repeated safe exposures to fear-inducing stimuli [ 13 ], and people can indeed learn to overcome or manage aquaphobia.

The neural mechanisms underlying phobias and innate fears are unknown, and this lack of knowledge has prevented the development of any mechanistic therapies or diagnostic markers for aquaphobia [ 14 ]. The advancement of therapeutic strategies for aquaphobia thus depends on the acquisition of experimental evidence identifying the neurochemical and neuroanatomical pathways underlying phobias [ 15 ], and given their aversion to water, mice are a tempting model organism for investigations into aquaphobia. However, it is unclear whether the aversion of mice to water truly reflects a fear of water, and certain key parameters of murine water aversion that could clarify the matter remain unexplored. For example, researchers have not fully determined the degree to which mice will modify their behaviours to avoid water, and it remains unknown whether a mouse’s behaviour around a body of water depends on the water’s depth.

To elucidate the behavioural parameters of murine water aversion and facilitate the development of new behavioural tests that could aid the identification of relevant neural circuits, we investigated the behaviours of mice when placed in proximity to water. We examined whether and under what conditions mice placed in a box with an area of shallow water would approach and enter the water.

2.1 Tests with an object in the water

To determine whether a mouse’s interest in exploring objects could tempt it to explore a water pool containing an object, we compared the behaviours of mice in an enriched environment (i.e. one with objects present both within and outside the water pool) with their behaviours in an empty environment (i.e. one with objects only present within the water pool) ( Figures 1a and b ). We observed no significant between-condition differences in the total distance travelled ( Figures 1c and d ), the number of entries into the zone surrounding the water pool ( Figure 1e ), or the total time spent in that zone ( Figure 1f ). However, we observed that the number of entries into the water pool and the total time spent in the water were greater under the empty environment condition than under the enriched environment condition ( Figures 1g and h ; Supplementary Videos 1 and 2), which suggests that the mice were more motivated to explore the water pool when the water pool was the only area that contained an object.

Behavioural tests featuring an object in the water pool. (a) Schematics of the experimental environments. (b) Outline of the experimental protocol. (c) Sample traces of a mouse’s movements through the enriched and empty environments. (d–h) Boxplots showing the total distance travelled (d), the number of entries into the zone surrounding the water pool (e), the total time spent in that zone (f), the number of entries into the water (g), and the total time spent in the water (h) under each experimental condition. Statistical significance was defined as * p < 0.05.

2.2 Murine interest in the water pool

To determine whether mice would exhibit any interest in exploring a water pool by itself, we compared the behaviours of mice in an enriched environment (i.e. one with objects present outside the water pool) with their behaviours in an empty environment (i.e. one without any objects) ( Figure 2a ). We observed no significant between-condition difference in the total distance travelled ( Figures 2b and c ), but we observed that the number of entries into the zone surrounding the water pool and the total time spent in that zone were greater under the empty environment condition than under the enriched environment condition ( Figures 2d and e ). We also observed more entries into the water pool under the empty environment condition than under the enriched environment condition ( Figure 2f ; Supplementary Videos 3 and 4), but we observed no significant difference in the total time spent in the water ( Figure 2g ). These results indicate that the willingness of a mouse to enter the water is not dependent on the presence of objects in the water pool. However, the fact that the mice spent more time walking around the water pool than in it suggests that the mice were still hesitant to enter the water.

Behavioural tests without an object in the water pool. (a) Schematics of the experimental environments. (b) Sample traces of a mouse’s movements through the enriched and empty environments. (c–g) Boxplots showing the total distance travelled (c), the number of entries into the zone surrounding the water pool (d), the total time spent in that zone (e), the number of entries into the water (f), and the total time spent in the water (g) under each experimental condition. Statistical significance was defined as * p < 0.05.

2.3 Effects of variable water depths on mouse behaviours

To determine the effects of water depths on a mouse’s willingness to enter the water, we compared the behaviours of mice in the presence of an 8 mm-deep water pool with those in the presence of a 20 mm-deep water pool ( Figures 3a and b ). We observed no significant between-condition difference in the total distance travelled ( Figures 3c and d ), but we observed that the number of entries into the zone surrounding the water pool and the time spent within that zone were significantly greater under the 20 mm depth condition than under the 8 mm depth condition ( Figures 3e and f ). The number of entries into the water pool and the total time spent in the water were both lower under the 20 mm depth condition than under the 8 mm depth condition, with most mice not entering the water at all under the 20 mm depth condition ( Figures 3g and h ; Supplementary Video 5). These results suggest that mice can determine the depth of a pool of water.

Behavioural tests with variable water depths. (a) Schematic of the experimental environments. (b) Schematic of the difference in water depths. (c) Sample traces of a mouse’s movements under the 8 and 20 mm depth conditions. (d–h) Boxplots showing the total distance travelled (d), the number of entries into the zone surrounding the water pool (e), the total time spent in that zone (f), the number of entries into the water (g), and the total time spent in the water (h) under each experimental condition. Statistical significance was defined as * p < 0.05.

2.4 Behaviours of mice surrounded by water

To determine how a stressful situation affects a mouse’s willingness to enter the water, we compared the behaviours of mice surrounded by 3 mm-deep water with those of mice surrounded by 8 mm-deep water ( Figures 4a and b ). We observed no significant between-condition difference in the total distance travelled ( Figure 4d ), but we observed that the mice only entered the water under the 3 mm depth condition ( Figures 4c, e and f ; Supplementary Video 6). These results constitute further evidence that mice can determine the depth of a pool of water. They also indicate that being surrounded by water reduces a mouse’s willingness to enter the water.

Behavioural tests with the mice surrounded by water. (a) Schematic of the experimental environments. (b) Schematic of the difference in water depths. (c) Sample traces of a mouse’s movements under the 3 mm and 8 mm depth conditions. (d–f) Boxplots showing the total distance travelled (d), the number of entries into the water (e), and the total time spent in the water (f) under each experimental condition. Statistical significance was defined as * p < 0.05.

3 Discussion

In this study, we have shown that a strain of mice widely used in experiments can recognise water depths, they are unwilling to enter unacceptably deep bodies of water, and the distressing situation of being surrounded by water reduces their depth tolerance.

Although the mice frequently exhibited some apparent hesitancy about entering the water, with protracted periods in which they restricted themselves to exploring the zones surrounding the water, they usually exhibited an eventual willingness to enter and explore the water pools regardless of whether objects were present within the open field device. These behaviours differ markedly from those observed in situations where mice are presumably simply attempting to avoid potential predators, such as the tendency of mice in a boxed environment to focus on exploration near the wall while avoiding the less protected open areas [ 16 , 17 ]. Indeed, the water-entry behaviours that we observed are more reminiscent of the risk-accepting behaviours after initial periods of risk-avoidance observed in various behavioural tests [ 18 ]. For example, in the open field test, light/dark transition test, and elevated plus maze test, mice initially avoid open areas, highly illuminated areas, and heights, but they also exhibit an eventual willingness to explore new and potentially risky spaces [ 19 – 21 ]. The behaviours of mice when confronted with novelty are thus determined by a conflict between the willingness to explore unknown areas and objects and the motivation to avoid potential danger. The willingness of mice in our experiments to approach and enter bodies of water probably reflects an innate motivation to explore new environments and is thus analogous to the willingness of mice to enter the open arms in the elevated plus maze test, the central area in the open field test, and the illuminated area in the light/dark transition test.

Avoidance behaviours depend on an animal’s senses and are influenced by its motor activity, motivational factors, and search strategies [ 22 ]. The elevated plus maze test, light/dark transition test, and open field test are all designed to evaluate anxiety-like behaviours by taking advantage of known avoidance behaviours [ 23 ], but the results obtained from the different tests are sometimes inconsistent [ 20 , 24 – 26 ]. Discrepancies may arise from the fact that these tests are based on distinct anxiety-like behaviours [ 23 ]: avoidance of illuminated spaces in the light/dark transition test, avoidance of open spaces in the open field test, and avoidance of heights in the elevated plus maze test [ 27 ]. The observed discrepancies suggest that different forms of anxiety may involve distinct mechanisms, and this, in turn, implies that it is important for researchers to have access to diverse tests with which to assess different forms of anxiety [ 28 , 29 ]. Our findings concerning the characteristics of murine water avoidance may be used to develop a new test of anxiety-like behaviours based on water avoidance.

In this study, mice exhibited a markedly reduced willingness to enter the water when the depth was increased to 20 mm, even though they could still walk through the water without needing to swim at that depth. Mice can swim but they will normally avoid water as much as possible [ 1 ]. Our results clearly show that mice can recognise the depth of a body of water and can choose to avoid entering deep water, just as experiments with elevated plus mazes have shown that mice can recognise heights and can choose to avoid them [ 30 ]. To the best of our knowledge, our study is the first to provide empirical evidence that mice are more willing to enter shallow water than to enter deep water.

Interestingly, we found that when mice were placed in a central area surrounded by water, they exhibited an increased aversion to deep water, with no entries into the water with an 8 mm depth that had been acceptable for mice that were not surrounded by water. This suggests that being surrounded by water prompted increased feelings of anxiety in the mice and reduced their willingness to engage in exploratory behaviours, a finding that is consistent with past investigations showing that anxiety suppresses exploratory behaviours [ 31 ]. Other factors, such as situation complexity, novelty, and the animal’s baseline emotional state, can also reduce a mouse’s willingness to explore a new environment [ 32 ].

Anxiety and fear are normal emotions that are selected for in the evolutionary process because they aid an organism in avoiding dangerous situations. For example, humans can experience fear around water because of the risk of drowning. However, individuals with aquaphobia experience abnormal symptoms around water such as headaches, feelings of suffocation, panic attacks, and decreased water intake [ 10 ]. Fear occurs in response to threats, but the physiological mechanisms underlying anxious behaviours remain unclear [ 33 ]. For small rodents, entering small spaces, holes, or tunnels is an important behaviour, but mice that lack leucine-rich repeat transmembrane neuronal proteins exhibit claustrophobia-like phenotypes that involve avoidance of small enclosures [ 34 ]. Interestingly, mice with hippocampal lesions also exhibit an unwillingness to enter small holes and tunnels [ 35 ]. Investigations into social phobias have found that such phobias may be related to interactions between the noradrenalinergic and serotonergic systems and the hypothalamic–pituitary–adrenal system [ 15 , 36 ]. Collectively, these findings indicate the existence of neural mechanisms underlying innate fears and offer clues as to how therapeutic strategies for phobias and anxiety could be developed. Our results add to the existing knowledge concerning phobias and may aid efforts to elucidate the mechanisms underlying water avoidance in mice and aquaphobia in humans.

4 Conclusion

Our results clearly indicate that mice exhibit exploratory behaviours in the context of entering shallow water. Furthermore, mice can recognise water depths and can choose not to enter the water if it is too deep. We speculate that the extent of a mouse’s exploratory behaviour in the presence of bodies of water is partially determined by anxieties related to water, and we propose that the dependence of a mouse’s exploratory behaviours on water depths could be used to design new tests of anxiety-like behaviours that could aid research into aquaphobia in humans.

5 Materials and methods

5.1 animals.

All efforts were made to minimise the number of animals used and to prevent unavoidable discomfort. Male C57BL/6N mice (age: 10 weeks) were purchased from Charles River Laboratories Japan (Kanagawa, Japan) and were housed five to a cage with food and water provided ad libitum under a 12 h light/dark cycle at 23–26°C.

Ethical approval: The research related to animals’ use has complied with all the relevant national regulations and institutional policies for the care and use of animals. All animal experiments were performed in accordance with the U.S. National Institutes of Health (NIH) – Guide for the Care and Use of Laboratory Animals (NIH Publication No. 80-23, revised in 1996) and approved by the Committee for Animal Experiments at the Kawasaki Medical School Advanced Research Center.

5.2 Behavioural tests

All behavioural tests were conducted in behavioural testing rooms between 09:00 h and 16:00 h during the light phase of the light/dark cycle. After the tests, the equipment was cleaned with 70% ethanol and super hypochlorous water to eliminate olfactory cues. Hypochlorous acid is an effective odour removal agent with a weak intrinsic odour [ 30 , 37 ]. The behavioural testing rooms were illuminated at a 100 lux intensity.

For the behavioural tests, we used an open field test apparatus that consisted of a 45 cm × 45 cm square area surrounded by 40 cm-high walls. The tests involved various arrangements of water pools and objects. Prior to object placement, each mouse was placed in the box for a 10 min free exploration period to produce habituation to the environment, after which the mouse was briefly returned to its home cage while object placement occurred. Unless otherwise noted, the water pools were filled during the habituation period. During the behavioural tests, data were video-recorded.

5.3 Tests with an object in the water pool

In this experiment, the open field included a 13.0 cm × 13.0 cm pool of 3 mm-deep water that was positioned at the centre of one wall ( Figure 1a ). A tower model was placed in the centre of the water pool, and a 4.0 cm × 4.0 cm cotton square, a 4.0 cm × 4.0 cm × 1.0 cm polystyrene rectangular prism, and a 4.0 cm × 8.0 cm × 4.0 cm wire cage were placed on the side of the open field opposite the water pool.

Ten mice were used in this experiment. In test 1, all objects were placed in the box ( Figure 1a ; enriched environment), and the mouse was placed in a corner before being allowed to move freely around the box for 12 min ( Figure 1b ). The mouse was then returned to its home cage for 5 min. In test 2, all objects except the tower model in the water pool were removed ( Figure 1a ; empty environment), and the mouse was again placed in a corner before being allowed to move freely around the box for 12 min ( Figure 1b ). The same mice were used in test 1 and test 2.

5.4 Tests of murine interest in the water

In this experiment, the open field included a 13.0 cm × 13.0 cm pool of 3 mm-deep water that was positioned in a corner ( Figure 2a ). A 4.0 cm × 4.0 cm cotton square, a 4.0 cm × 4.0 cm × 1.0 cm polystyrene rectangular prism, a 4.0 cm × 7.0 cm × 4.0 cm wire cage, and a 50 mL tube without a lid were placed in the areas away from the water pool.

Ten mice were used in this experiment. In test 1, all objects were placed in the box ( Figure 2a ; enriched environment), and the mouse was placed in a corner before being allowed to move freely around the box for 12 min. The mouse was then returned to its home cage for 5 min. In test 2, all objects were removed ( Figure 2a ; empty environment), and the mouse was again placed in a corner before being allowed to move freely around the box for 12 min. The same mice were used in test 1 and test 2.

5.5 Tests with variable water depths

In this experiment, the open field included a 14.0 cm × 20.0 cm water pool with a depth of 8 mm or 20 mm positioned in a corner ( Figures 3a and b ). The same container was used for all experiments to ensure equal container heights. A 4.0 cm × 4.0 cm cotton square, a 50 mL tube without a lid, and a third object (miniature-home) were placed in areas away from the water pool. Separate sets of 10 mice were used for tests involving the 8 mm and 20 mm depths. In this test, the mouse was placed in the corner before being allowed to move freely around the box for 12 min. In this experiment, each mouse was used in only one experiment.

5.6 Tests with water surrounding a dry zone

In this experiment, the open field consisted of a 13.0 cm × 13.0 cm dry central area that was surrounded by a water pool with a depth of 3 mm or 8 mm ( Figures 4a and b ). Separate sets of 10 mice were used for tests involving the 3 mm and 8 mm depths. In contrast to the other experiments, water was not added to the box until after the 10 min habituation period. In this test, the mouse was placed on the dry central area before being allowed to move freely around the box for 12 min. In this experiment, each mouse was used in only one experiment.

5.7 Data analyses

The video-recorded data were analysed with video-tracking software (ANY-MAZE; Stoelting, Wood Dale, IL, USA). For each 12 min test period, we determined the total distance travelled, the number of entries into the zone surrounding the water pool, the amount of time spent in that zone, the number of entries into the water pool, and the amount of time spent in the water pool. For comparing two groups, Student’s t -test was used for normally distributed data, and Mann–Whitney U test was used for not normally distributed data. In addition, one-way repeated-measures analysis of variance was used for normally distributed data, and the Friedman test was used for not normally distributed data. p < 0.05 was used as the definition of statistical significance. Statistical analyses were performed with SPSS software (IBM, Armonk, NY, USA).

Acknowledgements

We thank the Kawasaki Medical School Central Research Institute for providing the instruments used to perform this study. We also thank Editage ( www.editage.jp ) for English-language editing.

Funding information: This work was supported by a Grant Aid for the Sanyo Broadcasting Foundation and the Okayama Medical Foundation. The funding source had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; and in the decision to submit the article for publication.

Author contributions: All authors had complete access to all study data and assume complete responsibility for the integrity of the data and accuracy of the data analysis. Study concept and design: H.U., Y.T., M.O., and T.I. Acquisition of data: H.U., Y.T., S.S., and Y.T. Analysis and interpretation of data: H.U., S.S., and Y.T. Drafting of the manuscript: H.U. and M.O. Critical revision of the manuscript for important intellectual content: S.M., N.K., K.W., Y.M., and T.I. Study supervision: M.O. and T.I.

Conflict of interest : The authors state no conflict of interest.

Data availability statement: The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

[1] Ellenbroek B, Youn J. Rodent models in neuroscience research: is it a rat race? Dis Model Mech. 2016;9:1079–87. 10.1242/dmm.026120 Search in Google Scholar PubMed PubMed Central

[2] Wahlsten D. Mouse Behavioral Testing. 1st edn. Amsterdam, The Netherlands: Academic Press; 2010. 10.1016/B978-0-12-375674-9.10001-1 Search in Google Scholar

[3] Kim KS, Kwon HJ, Baek IS, Han PL. Repeated Short-term (2h × 14d) Emotional Stress Induces Lasting Depression-like Behavior in Mice. Exp Neurobiol. 2012;21:16–22. 10.5607/en.2012.21.1.16 Search in Google Scholar PubMed PubMed Central

[4] Mul JD, Zheng J, Goodyear LJ. Validity assessment of 5 day repeated forced-swim stress to model human depression in young-adult C57BL/6J and BALB/cJ mice. eNeuro. 2016;29:3. 10.1523/ENEURO.0201-16.2016 Search in Google Scholar PubMed PubMed Central

[5] Karagiannides I, Golovatscka V, Bakirtzi K, Sideri A, Salas M, Stavrakis D, et al. Chronic unpredictable stress regulates visceral adipocyte-mediated glucose metabolism and inflammatory circuits in male rats. Physiol Rep. 2014;2:e00284. 10.14814/phy2.284 Search in Google Scholar PubMed PubMed Central

[6] Choy Y, Fyer AJ, Lipsitz JD. Treatment of specific phobia in adults. Clin Psychol Rev. 2007;27:266–86. 10.1016/j.cpr.2006.10.002 Search in Google Scholar PubMed

[7] LeBeau RT, Glenn D, Liao B, Wittchen HU, Beesdo-Baum K, Ollendick T, et al. Specific phobia: a review of DSM-IV specific phobia and preliminary recommendations for DSM-V. Depress Anxiety. 2010;27:148–67. 10.1002/da.20655 Search in Google Scholar PubMed

[8] Wittchen HU, Jacobi F, Rehm J, Gustavsson A, Svensson M, Jönsson B, et al. The size and burden of mental disorders and other disorders of the brain in Europe 2010. Eur Neuropsychopharmacol. 2011;21:655–79. 10.1016/j.euroneuro.2011.07.018 Search in Google Scholar PubMed

[9] Líndal E, Stefánsson JG. The lifetime prevalence of anxiety disorders in Iceland as estimated by the US National Institute of Mental Health Diagnostic Interview Schedule. Acta Psychiatr Scand. 1993;88:29–34. 10.1111/j.1600-0447.1993.tb03410.x Search in Google Scholar PubMed

[10] Ajinkya SA. Cognitive hypnotherapy for panic disorder with Aquaphobia. Sleep Hypn. 2015;17:1–2. Search in Google Scholar

[11] Hall LL. Fighting Phobias, the Things That Go Bump in the Mind. FDA Consum Mag. 1997;31:2. 10.1037/e543742006-002 Search in Google Scholar

[12] Poulton R, Menzies RG, Craske MG, Langley JD, Silva PA. Water trauma and swimming experiences up to age 9 and fear of water at age 18: a longitudinal study. Behav Res Ther. 1999;37:39–48. 10.1016/S0005-7967(98)00103-X Search in Google Scholar

[13] Clarke JC, Jackson JA. Hypnosis and behavior therapy: the treatment of anxiety and phobias. New York: Springer; 1983. Search in Google Scholar

[14] Hyde J, Ryan KM, Waters AM. Psychophysiological markers of fear and anxiety. Curr Psychiatry Rep. 2019;21:56. 10.1007/s11920-019-1036-x Search in Google Scholar

[15] Marcin MS, Nemeroff CB. The neurobiology of social anxiety disorder: the relevance of fear and anxiety. Acta Psychiatr Scand Suppl. 2003;417:51–64. 10.1034/j.1600-0447.108.s417.4.x Search in Google Scholar

[16] Hwa LS, Neira S, Pina MM, Pati D, Calloway R, Kash TL. Predator odor increases avoidance and glutamatergic synaptic transmission in the prelimbic cortex via corticotropin-releasing factor receptor 1 signaling. Neuropsychopharmacology. 2019;44:766–75. 10.1038/s41386-018-0279-2 Search in Google Scholar

[17] Harris AZ, Atsak P, Bretton ZH, Holt ES, Alam R, Morton MP, et al. A novel method for chronic social defeat stress in female mice. Neuropsychopharmacology. 2018;43:1276–83. 10.1038/npp.2017.259 Search in Google Scholar

[18] Belzung C, Le Pape G. Comparison of different behavioral test situations used in psychopharmacology for measurement of anxiety. Physiol Behav. 1994;56:623–8. 10.1016/0031-9384(94)90311-5 Search in Google Scholar

[19] Teegarden S. Behavioral phenotyping in rats and mice. Mater Methods. 2012;2:122. 10.13070/mm.en.2.122 Search in Google Scholar

[20] Belovicova K, Bogi E, Csatlosova K, Dubovicky M. Animal tests for anxiety-like and depression-like behavior in rats. Interdiscip Toxicol. 2017;10:40–3. 10.1515/intox-2017-0006 Search in Google Scholar

[21] Crawley J, Goodwin FK. Preliminary report of a simple animal behavior model for the anxiolytic effects of benzodiazepines. Pharmacol Biochem Behav. 1980;13:167–70. 10.1016/0091-3057(80)90067-2 Search in Google Scholar

[22] Ohl F, Roedel A, Binder E, Holsboer F. Impact of high and low anxiety on cognitive performance in a modified hole board test in C57BL/6 and DBA/2 mice. Eur J Neurosci. 2003;17:128–36. 10.1046/j.1460-9568.2003.02436.x Search in Google Scholar PubMed

[23] Andersen ML, Tufik S. Rodent Model as Tools in Ethical Biomedical Research. Cham: Springer; 2016. 10.1007/978-3-319-11578-8 Search in Google Scholar

[24] Holmes A, Parmigiani S, Ferrari PF, Palanza P, Rodgers RJ. Behavioral profile of wild mice in the elevated plus-maze test for anxiety. Physiol Behav. 2000;71:509–16. 10.1016/S0031-9384(00)00373-5 Search in Google Scholar

[25] Tsujimura A, Matsuki M, Takao K, Yamanishi K, Miyakawa T, Hashimoto-Gotoh T. Mice lacking the kf-1 gene exhibit increased anxiety- but not despair-like behavior. Front Behav Neurosci. 2008;2:4. 10.3389/neuro.08.004.2008 Search in Google Scholar

[26] Nakajima R, Takao K, Huang SM, Takano J, Iwata N, Miyakawa T, et al. Comprehensive behavioral phenotyping of calpastatin-knockout mice. Mol Brain. 2008;1:7. 10.1186/1756-6606-1-7 Search in Google Scholar

[27] Takao K, Miyakawa T. Light/dark transition test for mice. J Vis Exp. 2006;1:104. 10.3791/104 Search in Google Scholar

[28] Lezak KR, Missig G, Carlezon Jr. WA. Behavioral methods to study anxiety in rodents. Dialogues Clin Neurosci. 2017;19:181–91. 10.31887/DCNS.2017.19.2/wcarlezon Search in Google Scholar

[29] Shoji H, Miyakawa T. Effects of test experience, closed-arm wall color, and illumination level on behavior and plasma corticosterone response in an elevated plus maze in male C57BL/6J mice: a challenge against conventional interpretation of the test. Mol Brain. 2021;14:34. 10.1186/s13041-020-00721-2 Search in Google Scholar

[30] Komada M, Takao K, Miyakawa T. Elevated plus maze for mice. J Vis Exp. 2008;22:22. 10.3791/1088 Search in Google Scholar

[31] Bertoglio LJ, Carobrez AP. Previous maze experience required to increase open arms avoidance in rats submitted to the elevated plus-maze model of anxiety. Behav Brain Res. 2000;108:197–203. 10.1016/S0166-4328(99)00148-5 Search in Google Scholar

[32] Lister RG. Ethologically-based animal models of anxiety disorders. Pharmacol Ther. 1990;46:321–40. 10.1016/B978-0-08-040698-5.50012-0 Search in Google Scholar

[33] Graeff FG, Guimarães FS. Fundamentos de psicofarmacologia. 2nd edn. São Paulo: Ateneu; 2012. Search in Google Scholar

[34] Voikar V, Kulesskaya N, Laakso T, Lauren J, Strittmatter SM, Airaksinen MS. LRRTM1-deficient mice show a rare phenotype of avoiding small enclosures--a tentative mouse model for claustrophobia-like behaviour. Behav Brain Res. 2013;238:69–78. 10.1016/j.bbr.2012.10.013 Search in Google Scholar PubMed PubMed Central

[35] Deacon RM, Rawlins JN. Hippocampal lesions, species-typical behaviours and anxiety in mice. Behav Brain Res. 2005;156:241–9. 10.1016/j.bbr.2004.05.027 Search in Google Scholar PubMed

[36] Bell CJ, Malizia AL, Nutt DJ. The neurobiology of social phobia. Eur Arch Psychiatry Clin Neurosci. 1999;249:S11–8. 10.1007/PL00014162 Search in Google Scholar PubMed

[37] Leo LM, Pamplona FA. Elevated plus maze test to assess anxiety-like behavior in the mouse. Bio-protocol. 2014;4:e1211. 10.21769/BioProtoc.1211 Search in Google Scholar

© 2022 Hiroshi Ueno et al ., published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

• X / Twitter

Supplementary Materials

• Supplementary video 1
• Supplementary video 2
• Supplementary video 3
• Supplementary video 4
• Supplementary video 5
• Supplementary video 6
• Supplementary caption

Please login or register with De Gruyter to order this product.

Journal and Issue

Articles in the same issue.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 21 August 2024

Transmission dynamics of MERS-CoV in a transgenic human DPP4 mouse model

• Neeltje van Doremalen 1 ,
• Trenton Bushmaker 1 ,
• Robert J. Fischer 1 ,
• Atsushi Okumura 2 ,
• Dania M. Figueroa Acosta 1   nAff4 ,
• Rebekah J. McMinn 1 ,
• Michael Letko 1 ,
• Dana Scott 3 ,
• Greg Saturday 3 &
• Vincent J. Munster 1  

npj Viruses volume  2 , Article number:  36 ( 2024 ) Cite this article

88 Accesses

3 Altmetric

Metrics details

• Viral pathogenesis
• Viral transmission

Since 2002, three novel coronavirus outbreaks have occurred: severe acute respiratory syndrome coronavirus (SARS-CoV-1), Middle East respiratory syndrome coronavirus (MERS-CoV), and SARS-CoV-2. A better understanding of the transmission potential of coronaviruses will result in adequate infection control precautions and an early halt of transmission within the human population. Experiments on the stability of coronaviruses in the environment, as well as transmission models, are thus pertinent.

Here, we show that transgenic mice expressing human DPP4 can be infected with MERS-CoV via the aerosol route. Exposure to 5 × 106 TCID50 and 5 × 104 TCID50 MERS-CoV per cage via fomites resulted in transmission in 15 out of 20 and 11 out of 18 animals, respectively. Exposure of sentinel mice to donor mice one day post inoculation with 105 TCID50 MERS-CoV resulted in transmission in 1 out of 38 mice via direct contact and 4 out of 54 mice via airborne contact. Exposure to donor mice inoculated with 104 TCID50 MERS-CoV resulted in transmission in 0 out of 20 pairs via direct contact and 0 out of 5 pairs via the airborne route. Our model shows limited transmission of MERS-CoV via the fomite, direct contact, and airborne routes. The hDPP4 mouse model will allow assessment of the ongoing evolution of MERS-CoV in the context of acquiring enhanced human-to-human transmission kinetics and will inform the development of other transmission models.

Similar content being viewed by others

SARS-CoV-2 disease severity and transmission efficiency is increased for airborne compared to fomite exposure in Syrian hamsters

Increased small particle aerosol transmission of B.1.1.7 compared with SARS-CoV-2 lineage A in vivo

A pangolin-origin SARS-CoV-2-related coronavirus: infectivity, pathogenicity, and cross-protection by preexisting immunity

Introduction.

In the last two decades, three novel coronaviruses have caused outbreaks in the human population. Severe acute respiratory syndrome coronavirus (SARS-CoV-1) was first identified in 2003 after it caused human cases in the Guangdong province, China, in November 2002. From Guangdong, the virus spread to 37 countries, resulting in >8000 infected people with a case fatality rate of 9.5% 1 . Middle East respiratory syndrome coronavirus (MERS-CoV) was first detected in 2012 and still circulates in the dromedary camel population, from which it infects the human population. Since 2012, more than 2600 cases with a case fatality rate of 36% 2 have been reported. The largest pandemic with a coronavirus to date started in December 2019 and was caused by SARS-CoV-2. To date, more than 770 million cases have been reported, resulting in nearly 7 million deaths 3 .

A better understanding of the transmission potential of coronavirus is crucial when devising personal protection equipment for healthcare staff and quarantine measurements. The stability of MERS-CoV, SARS-CoV-1, and SARS-CoV-2 has been investigated under several different environmental conditions in both fomites and aerosols 4 , 5 , 6 . Experimental transmission models have been developed for SARS-CoV-2 and focus mainly on hamsters and ferrets 7 , 8 , 9 , 10 , 11 , 12 , 13 , whereas SARS-CoV-1 transmission has been shown in ferrets and cats 13 , 14 . However, transmission of MERS-CoV in animal models has not yet been reported.

A review of 681 MERS cases in the Kingdom of Saudi Arabia (KSA) estimated that 12% of cases were infected via direct exposure to dromedary camels, and 88% resulted from human-to-human transmission 15 . Analysis of transmission dynamics showed that the number of subsequent generations is limited. The risk of a human-to-human transmission event differs per generation: the initial zoonotic transmission risk is 22.7%. Then it drops to 10.5% for the second generation, 6.1% for the third generation, and 3.9% for the fourth generation 16 . This shows that although human-to-human transmission contributes significantly to the number of human MERS-CoV cases, transmission is not sustained. Human-to-human transmission of MERS-CoV can be divided into household-associated healthcare-associated (nosocomial transmission). Epidemiological modeling of MERS-CoV transmission estimates nosocomial transmission to be ten times higher than household transmission 17 .

Virus transmission can occur via different routes, including fomites, direct contact, and aerosols. Knowledge on the transmission routes of emerging coronaviruses is essential in designing broad preemptive countermeasures against zoonotic and human-to-human transmission. More specifically, it will improve the ability of hospitals to reduce the likelihood of human-to-human transmission by implementing appropriate personal protective equipment and hospital hygiene procedures.

The best transmission models for SARS-CoV-2 are the hamster and ferret models 7 , 8 , 9 , 10 , 11 , 12 , 13 . However, these animals are not naturally susceptible to MERS-CoV 18 , 19 and likely will require expression of hDPP4. Although transmission in mice is likely more limited than in hamsters or ferrets, SARS-CoV-2 transmission has been shown in mice 20 , 21 . Therefore, we utilize a transgenic mouse model expressing human DPP4 (hDPP4 mice) to investigate whether MERS-CoV can transmit via fomites, direct contact, or the airborne route.

Intranasal or aerosol inoculation of hDPP4 mice with MERS-CoV

In our hDPP4 mouse model, expression of hDPP4 can be found in the nasal turbinates, trachea, lungs, and the kidney (Fig. 1a−d ). Intranasal inoculation with 10 5 TCID 50 MERS-CoV strain HCoV-EMC/2012 resulted in >20% weight loss and 100% lethality (Fig. 1e, f ). Viral gRNA could be detected in oropharyngeal swabs on all six days during the experiment (Fig. 1g ). Infectious virus was detected in lung tissue at three days post-inoculation (dpi) and at endpoint, as well as in brain tissue at end point, but not in the kidney (Fig. 1h ). In contrast, littermates of the hDPP4 mice which were negative for hDPP4, were not susceptible to MERS-CoV infection. We then investigated whether inoculation via aerosols would result in productive infection of hDPP4 mice. We inoculated ten mice via aerosols with an estimated 3.8 × 10 2 TCID 50 MERS-CoV per mouse. All mice reached the endpoint criteria (Fig. 2a ). Shedding, as measured by gRNA presence in oropharyngeal swabs, could be detected on all days (Fig. 2b ). Viral RNA was detected in lung tissue (four out of four) and brain tissue (one out of four) on three dpi and all tissue on the day of endpoint criteria (Fig. 2c ). Using immunohistochemistry, we compared the cellular tropism of MERS-CoV replication in hDPP4 mice between intranasal (inoculation dose = 10 3 TCID 50 ) and aerosol inoculation (3.8 × 10 2 TCID 50 ) in the upper and lower respiratory tract at three dpi (Fig. 3 ). For both groups, viral replication was detected in lung tissue: type I and type II pneumocytes were positive for MERS-CoV antigen. However, even though we showed that cells lining the trachea, as well as the nasal turbinates, of hDPP4 mice express hDPP4 (Fig. 1a−c ), MERS-CoV antigen staining was negative in the trachea. No staining in the nasal turbinates was detected for the aerosol-inoculated group, whereas staining in the nasal turbinates of the intranasally inoculated group was very limited (Fig. 3 ). This MERS-CoV respiratory tropism is similar to what has been observed in humans 22 ; MERS-CoV predominantly targets the cells in the lower respiratory tract.

Detection of hDPP4 expression in hDPP4 mice using immunohistochemistry in ( a ) nasal mucosa; ( b ) trachea; and ( c ) type I and II pneumocytes, bronchiolar and endothelial cells in lung tissue. d Comparison of hDPP4 expression in lung and kidney tissue obtained from wildtype and hDPP4 mice using flow cytometry. N  = 3, bars represent median. e Survival curves of mice after inoculation with 5 × 10 5 TCID 50 MERS-CoV. N  = 4 (Wildtype) or 5 (hDPP4). ** =  p  < 0.01. f Relative weight loss in mice after MERS-CoV inoculation. The lines represent median±range. Mice were euthanized upon reaching >20% of body weight loss (dotted line). g Viral load (gRNA) in oropharyngeal swabs obtained from mice after inoculation with 5 × 10 5 TCID 50 MERS-CoV. h Infectious MERS-CoV titers in lung, brain, and kidney tissue of hDPP4 mice. Dotted line detection limit.

a Survival curves of mice inoculated with 3.8 × 10 2 TCID 50 MERS-CoV via aerosols. N  = 6. b Viral load (gRNA) in oropharyngeal swabs obtained from mice inoculated with 3.8 × 10 2 TCID 50 MERS-CoV via aerosols. Bars represent median. b Viral load (gRNA) and c (infectious virus) in lung and brain tissue of hDPP4 mice inoculated with 3.8 × 10 2 TCID 50 MERS-CoV via aerosols.

a , b Nasal turbinates; ( c , d ) Trachea; ( e , f ) Lungs; ( a , c , e ) Intranasal inoculation; ( b , d , f ) Aerosol inoculation. Tissues were stained with an in-house produced rabbit polyclonal antiserum against HCoV-EMC/2012 for the detection of viral antigen. Immunohistochemistry staining reveals MERS-CoV antigen in type I and type II pneumocytes and limited staining in nasal turbinates (image represents only staining found). Inserts highlight affected cells. Nasal turbinates (200x), trachea (200x) and lung tissue (100 and 400x insert) are shown.

Transmission of MERS-CoV in hDPP4 mice

Next, we investigated the transmission of MERS-CoV within our mouse model via three different routes: fomite, direct contact, and the airborne route.

Fomite transmission

MERS-CoV-containing media was pipetted on different objects in the cage, including on two plastic and two metal washers, after which two mice were introduced per cage (Fig. S1 ). Upon exposure to 5 × 10 4 TCID 50 /cage, 11 out of 18 mice reached endpoint criteria (Fig. 4a ). Brain and lung tissue were harvested from non-survivors, and viral gRNA was detected in brain tissue of eleven mice and lung tissue of seven mice (Fig. 4b ). sgRNA was detected in brain tissue of three mice, and lung tissue of one mouse (Fig. 4c ). Infectious virus was recovered from brain tissue, but not lung tissue, of three mice (Fig. 4d ). The remaining seven mice were euthanized at 28 dpe; six animals were seropositive for MERS-CoV S1 (Fig. 4e ). We then exposed twenty mice to 5 × 10 6 TCID 50 /cage, again two mice per cage. Fifteen out of 20 mice were euthanized (Fig. 4a ). Viral gRNA was detected in lung and brain tissue of all the non-survivors (Fig. 4b ). Viral sgRNA was detected in brain tissue of all mice and lung tissue of six mice (Fig. 4c ). Infectious virus was isolated from brain tissue, but not lung tissue, of eight mice (Fig. 4d ). Four out of five surviving animals were found to be seropositive for MERS-CoV S1 protein (Fig. 4e ).

hDPP4 mice were exposed to 5 × 10 4 TCID 50 MERS-CoV ( N  = 18) or 5 × 10 6 TCID 50 MERS-CoV ( N  = 20). a Survival curves of hDPP4 mice exposed to fomites containing MERS-CoV. Viral gRNA ( b ) or sgRNA ( c ) in lung and brain tissue of hDPP4 mice that reached endpoint criteria. d Infectious MERS-CoV detected in lung and brain tissue of hDPP4 mice that reached endpoint criteria. e Serology titers in sera of survivors obtained 28 dpe. ELISA assays were performed using MERS-CoV S1 protein. Dotted line = limit of detection. ** =  p  < 0.01.

Contact transmission

Twenty mice in two separate experiments were inoculated intranasally with 10 4 TCID 50 MERS-CoV, and single-housed. One day later, one sentinel mouse per cage was introduced. No sentinel mice reached endpoint criteria (Fig. 5a ). One mouse seroconverted with a titer of 100 (Fig. 5e ). Subsequently, 38 mice in three separate experiments were inoculated with 10 5 TCID 50 MERS-CoV, and sentinel mice were introduced one day later. One sentinel mouse was euthanized five days post-exposure (dpe) (Fig. 5a ). A low amount of viral gRNA was detected in both lung and brain tissue (Fig. 5b ), but no sgNA was detected (Fig. 5c ). Likewise, no infectious virus was recovered from tissue (Fig. 5d ). All 37 remaining sentinels survived exposure. Antibodies against S1 were detected in one surviving animal (Fig. 5e ).

hDPP4 mice were inoculated intranasally with 10 4 TCID 50 MERS-CoV ( N  = 20) or 10 5 TCID 50 MERS-CoV ( N  = 38). a Survival curves of mice directly exposed to donor mice infected with MERS-CoV. Viral gRNA ( b ) or sgRNA ( c ) in lung and brain tissue of hDPP4 mice that reached endpoint criteria. d Infectious MERS-CoV detected in lung and brain tissue of hDPP4 mice that reached endpoint criteria. e Serology titers in sera of survivors obtained 28 dpe. ELISA assays were performed using MERS-CoV S1 protein. Dotted line = limit of detection.

Airborne transmission

Five hDPP4 mice were inoculated intranasally with 10 4 TCID 50 MERS-CoV. Sentinel mice were introduced in the same cage as the donor animal at one dpi. Animals were separated by a perforated divider, which did not allow direct contact but allowed airflow from the donor to the sentinel animal (Fig. S1 , also described in 7 ). No animals reached endpoint criteria (Fig. 6a ). Three surviving animals were found to be seropositive for S1 at very low titers (Fig. 6e ). In eight different experiments, 54 hDPP4 mice were inoculated with 10 5 TCID 50 MERS-CoV. Sentinel animals were introduced one dpi. Four animals reached endpoint criteria (Fig. 6a ). Viral gRNA could be detected in all brain tissues and three out of four lung tissues (Fig. 6b ). Viral sgRNA was detected in brain and lung tissue of two mice, and infectious virus was found in the brain, but not lung tissue, of one mouse (Fig. 6c, d ). All 50 remaining sentinels survived exposure. Seven of the surviving animals were found to be seropositive for S1 (Fig. 5e ).

hDPP4 mice were inoculated intranasally with 10 4 TCID 50 MERS-CoV ( N  = 5) or 10 5 TCID 50 MERS-CoV ( N  = 54). a Survival curves of mice exposed to donor mice infected with MERS-CoV. Viral gRNA ( b ) or sgRNA ( c ) in lung and brain tissue of hDPP4 mice that reached endpoint criteria. d Infectious MERS-CoV detected in lung and brain tissue of hDPP4 mice that reached endpoint criteria. e Serology titers in sera of survivors obtained 28 dpe. ELISA assays were performed using MERS-CoV S1 protein. Dotted line = limit of detection.

The continued circulation of MERS-CoV in the dromedary camel population highlights the need for a better understanding of the transmission potential of MERS-CoV. Like MERS-CoV, SARS-CoV-1 and SARS-CoV-2 are thought to use a combination of different transmission routes between humans, including fomite, direct contact, and airborne transmission. Which one of these routes is most important is difficult to ascertain. However, it is clear that human-to-human transmission of MERS-CoV is restricted compared to SARS-CoV-1 and, in particular, SARS-CoV-2, and is primarily nosocomial 17 .

Within hospitals, MERS-CoV viral RNA has been detected on various surfaces up to five days after viral RNA was detected in patient samples. In addition, infectious virus was isolated from different hospital surfaces 23 and air samples 24 . Likewise, SARS-CoV-1 and SARS-CoV-2 RNA could be detected in air and surface samples 25 , 26 . Experimentally, infectious MERS-CoV at 20 °C and 40% relative humidity could be recovered from plastic and steel surfaces for up to 48 h 4 , and using a similar setup, SARS-CoV-1 and SARS-CoV-2 likewise retained viability for 48 h 5 . Superspreader events have been documented for all three viruses 27 , 28 , 29 , and in some scenarios, airborne transmission of SARS-CoV-1 and SARS-CoV-2 appears to be the most likely scenario for specific human-to-human transmission clusters 27 , 30 , 31 , 32 .

Animal models are crucial for experimental transmission studies, as transmission involves several factors: shedding of virus from an infected host, survival of the virus in aerosols or on surfaces, and infection of the sentinel host. Our data suggest that MERS-CoV can utilize a variety of different transmission routes, although the fomite route was much more efficient than both the direct contact and airborne routes.

In our airborne transmission setup the cage divider prevents direct contact, but allows the movement of larger droplets and aerosols from the donor cage to the sentinel cage. Therefore, we cannot distinguish between transmission events by aerosols (droplets < 100 µm), larger droplets (>100 µm), or a combination of these two 33 . An experimental setup exclusively allowing transmission of aerosols as designed for SARS-CoV-2 34 would be able to distinguish between aerosol and droplet transmission.

Our overall data agree with limited human-to-human transmission of MERS-CoV in the general population. Given the tropism of MERS-CoV for the lower respiratory tract and minimal evidence of infection of the upper respiratory tract 22 , 35 , 36 , 37 , there is likely little to no natural generation of infectious aerosols in patients. We hypothesize that the propensity of MERS-CoV to transmit relatively efficiently in hospital settings is linked to performing aerosol-generating procedures, such as intubation and bronchoscopy, on infected patients 38 , 39 , 40 rather than the natural generation of aerosols containing MERS-CoV from the respiratory tract of the patient. Combined with a potentially more susceptible hospital population 41 , this could lead to a relatively high nosocomial transmission compared to household transmission.

In this study, we showed MERS-CoV transmission via the fomite and in limited numbers via direct contact and airborne routes. The hDPP4 transmission model will be invaluable in assessing the transmission potential of novel MERS-CoV strains without prior adaptation to the mouse host in light of the continuing virus evolution during human outbreaks and the camel population 42 , 43 . In addition, this work will allow the development of effective countermeasures to block human-to-human transmission of MERS-CoV.

Ethics statement

Approval of animal experiments was obtained from the Institutional Animal Care and Use Committee of the Rocky Mountain Laboratories. Performance of experiments was done following the guidelines and basic principles in the United States Public Health Service Policy on Humane Care and Use of Laboratory Animals and the Guide for the Care and Use of Laboratory Animals. Work with infectious MERS-CoV strains under BSL3 conditions was approved by the Institutional Biosafety Committee (IBC). Inactivation and removal of samples from high containment was performed per IBC-approved standard operating procedures.

Development of hDPP4 mice

hDPP4 mice were developed by ingenious Targeting Laboratory. A ROSA26 knock-in vector containing a 3’ splice acceptor, LoxP -flanked neomycin stop cassette, Kozak sequence, human DPP4 cDNA sequence and bovine growth hormone poly-A tail was injected into balb/c embryonic stem (ES) cells via electroporation. ES cells were injected into balb/c blastocysts. Resulting chimeric mice were bred with wildtype balb/c mice. Heterozygous offspring were bred with BALB/c-Tg(CMV-cre)1Cgn/J mice (Jackson Laboratory) to produce mice ubiquitously expressing hDPP4. Deletion of the LoxP -flanked neomycin stop cassette occurs in all tissues, including germ cells, in BALB/c-Tg(CMV-cre)1Cgn/J mice. Polymerase chain reaction (PCR) was performed to genotype each mouse using a three primer set-up; forward primer (FP) AGCACTTGCTCTCCCAAAGTC, reverse primer 1 (RP1) GACAACGCCCACACACCAGGTTAG and reverse primer 2 (RP2) TCTTCTGTAATCAGCTGCCTTTTA.

Virus and cells

HCoV-EMC/2012 was kindly provided by Erasmus Medical Center, Rotterdam, The Netherlands. Virus propagation was performed in VeroE6 cells in DMEM (Sigma) supplemented with 2% fetal calf serum (Logan), 1 mM L-glutamine (Lonza), 50 U/ml penicillin and 50 μg/ml streptomycin (Gibco) (2% DMEM). VeroE6 cells were maintained in DMEM supplemented with 10% fetal calf serum, 1 mM L glutamine, 50 U/ml penicillin and 50 μg/ml streptomycin. Virus was titrated by inoculating VeroE6 cells with tenfold serial dilutions of virus in 2% DMEM. Five days after inoculation, cytopathic effect (CPE) was scored and TCID 50 was calculated from four replicates by the method of Spearman-Karber.

DPP4 expression

The expression of DPP4 in mouse tissue was examined using flow cytometry. Lung and kidney tissues were collected from animals and a single cell suspension was made using the mouse lung dissociation kit (Miltenyi Biotec). Cells were washed two times in PBS containing 2% FBS, and incubated with anti-DPP4 (AF1180, R&D systems, 8 µg/mL in PBS with 2% FBS). After 1 h at 4 °C, cells were washed as above and incubated with donkey anti-goat AF488 (A-11055, Thermo Fisher, 1:500 in PBS with 2% FBS). Hereafter, cells were washed, fixed with 2% formalin and analyzed on a BD FACSymphony A5 (BD Biosciences) flow cytometer using a high-throughput sampler. DPP4 expression was analyzed using FlowJo 8 (BD Biosciences).

Inoculation experiments

Animal numbers were determined using statistical methods before study start. hDPP4 mice (4−10 weeks, male and female) were separated by sex and then randomly divided between the experimental group, ensuring sex distribution was even. Animals were acclimatized for at least 3 days. Hereafter, animals were anesthetized with inhalation isoflurane and inoculated intranasally with MERS-CoV isolate HCoV-EMC/2012 in a total volume of 30 µl. Aerosol inoculation using the AeroMP aerosol management platform (Biaera technologies, USA) was performed as described previously 18 . Briefly, mice that were awake were exposed to a single 10-minute aerosol exposure whilst contained in a stainless-steel wire mesh cage (5 mice per cage, 2 cages per run, mo anesthesia). Aerosol particles were generated by a 3-jet collison nebulizer (Biaera technologies, USA) and ranged from 1 to 5 µm in size. Respiratory minute volume rates of the animals were determined using Alexander et al. 44 . Weights of the animals were averaged and the estimate inhaled dose was calculated using the simplified formula D = R x C aero  x T exp 45 , where D is the inhaled dose, R is the respiratory minute volume (L/min), C aero is the aerosol concentration (TCID 50 /L), and T exp is duration of the exposure (min). After inoculation, animals were observed daily for signs of disease. Euthanasia was indicated at >20% loss of initial body weight, if severe respiratory distress was observed, or if neurological signs were observed. Oropharyngeal and nasal swabs were collected daily in 1 ml DMEM with 100 U/ml penicillin and 100 µg/ml streptomycin. Researchers were not blinded to study groups.

Transmission experiments

All transmission studies were done at 21−23 °C and 40−45% relative humidity. Fomite transmission was examined by contaminating cages containing two metal and two plastic discs [3] with 0.5 mL of MERS-CoV isolate HCoV-EMC/2012 (total dose: 5 × 10 6 or 5 × 10 4 TCID 50 per cage) in the following manner: 50 µl of virus was placed on the water bottle, 50 µl of virus was placed on the food, 50 µl of virus was placed per disc, and 4 × 50 µl of virus was placed on the flooring of the cage. Mice (4−10 weeks, male and female) were placed in the cages 10 min post contamination and followed as described above. Contact and airborne transmission were examined by intranasal inoculation of donors with 10 4 or 10 5 TCID 50 of MERS-CoV. At 1 dpi, sentinel animals (4−10 weeks, male and female) were placed in the same cage (contact transmission) or in the same cage on the other side of a divider (airborne transmission) (Fig. S1 ). This divider prevented direct contact between the donor and sentinel mouse and the movement of bedding material. Only one transmission pair was housed per cage. Hereafter, mice were followed as described above.

Histopathology and immunohistochemistry

Murine tissues were evaluated for pathology and the presence of viral antigen as described previously 46 . Briefly, tissues were fixed in 10% neutral-buffered formalin for 7 days and paraffin-embedded. Tissue sections were stained with hematoxylin and eosin (H&E). An in-house produced rabbit polyclonal antiserum against HCoV-EMC/2012 (1:1000) was used as a primary antibody for the detection of viral antigen. A commercial antibody (AF1180, R&D resources) was used for the detection of DPP4.

Viral RNA detection

Tissues were homogenized in RLT buffer and RNA was extracted using the RNeasy method on the QIAxtractor (Qiagen) according to the manufacturer’s instructions. RNA was extracted from swab samples using the QiaAmp Viral RNA kit on the QIAxtractor. For one-step real-time qPCR, 5 μl RNA was used in the Rotor-GeneTM probe kit (Qiagen) according to instructions of the manufacturer. Standard dilutions of a virus stock with known titer were run in parallel in each run, to calculate TCID 50 equivalents in the samples. Initial detection of viral RNA was targeted upstream of the envelope gene sequence (UpE) 47 , confirmation was targeted at the ORF1A 48 . MERS-CoV M-gene mRNA was detected as described by Coleman et al. 49 . Values with Ct-values > 38 were excluded. For digital droplet PCR (ddPCR, Biorad), 8 µl RNA was added to supermix for probes and assay was run according to instructions of the manufacturer using UpE primers and probe allowing absolute quantification of target RNA.

Enzyme-linked immunosorbent assay (ELISA) was performed as described previously 50 . Briefly, spike S1 antigen Sino Biological Inc., 0.5 µg/mL in 50 mM bicarb binding buffer (4.41 g KHCO 3 and 0.75 g Na 2 CO 3 in 1 L water) was bound to MaxiSorb plates (Nunc) and then blocked with 5% non-fat dried milk in PBS-0.1% Tween (5MPT). Serum samples were diluted in 5MPT. Detection of MERS-specific antibodies was performed with HRP-conjugated IgG (H + L) secondary antibody and developing solution (KPL) followed by measurement at 405 nm. Sera was termed seropositive if the optical density value was higher than the average + 3x standard deviation of sera obtained from randomly selected mice before MERS-CoV inoculation.

Statistical analysis

Statistical analysis was performed by Log-rank (Mantel-Cox) test to compare survival curves, and by Kruskall-Wallis test followed by Mann-Whitney test. P  < 0.05 were considered significant.

WHO. Severe Acute Respiratory Syndrome (SARS). https://www.who.int/health-topics/severe-acute-respiratory-syndrome#tab=tab_1 .

WHO. MERS situation update, September 2023. https://www.emro.who.int/health-topics/mers-cov/mers-outbreaks.html#:~:text=From%20April%202012%20till%20September,(CFR)%20of%2036%25 .

WHO. WHO Coronavirus (COVID-19) Dashboard. https://covid19.who.int/ .

van Doremalen, N., Bushmaker, T. & Munster, V. J. Stability of Middle East respiratory syndrome coronavirus (MERS-CoV) under different environmental conditions. Eurosurveillence 18 , 20590 (2013).

Google Scholar  

van Doremalen, N. et al. Aerosol and surface stability of SARS-CoV-2 as compared with SARS-CoV-1. N Engl. J. Med. 382 , 1564–1567 (2020).

Article   PubMed   Google Scholar  

van Doremalen, N. et al. Surface‒aerosol stability and pathogenicity of diverse middle east respiratory syndrome coronavirus strains, 2012‒2018. Emerg Infect. Dis. 27 , 3052–3062 (2021).

Article   PubMed   PubMed Central   Google Scholar  

Port, J. R. et al. SARS-CoV-2 disease severity and transmission efficiency is increased for airborne but not fomite exposure in Syrian hamsters. https://doi.org/10.1101/2020.12.28.424565 (2020).

Sia, S. F. et al. Pathogenesis and transmission of SARS-CoV-2 in golden hamsters. Nature 583 , 834–838 (2020).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Chan, J. F.-W. et al. Surgical mask partition reduces the risk of noncontact transmission in a golden syrian hamster model for Coronavirus Disease 2019 (COVID-19). Clin. Infect. Dis. 71 , 2139–2149 (2020).

Article   CAS   PubMed   Google Scholar  

Kim, Y.-I. et al. Infection and rapid transmission of SARS-CoV-2 in Ferrets. Cell Host Microbe 27 , 704–709.e2 (2020).

Schlottau, K. et al. SARS-CoV-2 in fruit bats, ferrets, pigs, and chickens: an experimental transmission study. Lancet Microbe 1 , e218–e225 (2020).

Richard, M. et al. SARS-CoV-2 is transmitted via contact and via the air between ferrets. Nat. Commun. 11 , 3496 (2020).

Kutter, J. S. et al. SARS-CoV and SARS-CoV-2 are transmitted through the air between ferrets over more than one meter distance. Nat. Commun. 12 , 1653 (2021).

Martina, B. E. E. et al. Virology: SARS virus infection of cats and ferrets. Nature 425 , 915 (2003).

Cauchemez, S. et al. Unraveling the drivers of MERS-CoV transmission. Proc. Natl Acad. Sci. USA 113 , 9081–9086 (2016).

Nishiura, H., Miyamatsu, Y., Chowell, G. & Saitoh, M. Assessing the risk of observing multiple generations of Middle East respiratory syndrome (MERS) cases given an imported case. Eurosurveillence 20 , 21181 (2015).

Poletto, C., Boëlle, P.-Y. & Colizza, V. Risk of MERS importation and onward transmission: a systematic review and analysis of cases reported to WHO. BMC Infect. Dis. 16 , 448 (2016).

de Wit, E. et al. The Middle East respiratory syndrome coronavirus (MERS-CoV) does not replicate in Syrian hamsters. PLoS ONE 8 , e69127 (2013).

Raj, V. S. et al. Adenosine deaminase acts as a natural antagonist for dipeptidyl peptidase 4-mediated entry of the Middle East respiratory syndrome coronavirus. J. Virol. 88 , 1834–1838 (2014).

Pan, T. et al. Infection of wild-type mice by SARS-CoV-2 B.1.351 variant indicates a possible novel cross-species transmission route. Sig. Transduct. Target Ther. 6 , 1–12 (2021).

Article   CAS   Google Scholar  

Bao, L. et al. Transmission of severe acute respiratory syndrome Coronavirus 2 via Close contact and respiratory droplets among human angiotensin-converting enzyme 2 mice. J. Infect. Dis. 222 , 551–555 (2020).

Ng, D. L. et al. Clinicopathologic, immunohistochemical, and ultrastructural findings of a fatal case of middle east respiratory syndrome coronavirus infection in the United Arab Emirates, April 2014. Am. J. Pathol. 186 , 652–658 (2016).

Bin, S. Y. et al. Environmental contamination and viral shedding in MERS patients during MERS-CoV outbreak in South Korea. Clin. Infect. Dis. 62 , 755–760 (2016).

Kim, S.-H. et al. Extensive viable Middle East Respiratory Syndrome (MERS) Coronavirus contamination in air and surrounding environment in MERS isolation wards. Clin. Infect. Dis. 63 , 363–369 (2016).

Booth, T. F. et al. Detection of airborne severe acute respiratory syndrome (SARS) coronavirus and environmental contamination in SARS outbreak units. J. Infect. Dis. 191 , 1472–1477 (2005).

Tan, K. S. et al. Detection of hospital environmental contamination during SARS-CoV-2 Omicron predominance using a highly sensitive air sampling device. Front. Public Health 10 , 1067575 (2022).

Cho, S. Y. et al. MERS-CoV outbreak following a single patient exposure in an emergency room in South Korea: an epidemiological outbreak study. Lancet 388 , 994–1001 (2016).

Wang, S. X. et al. The SARS outbreak in a general hospital in Tianjin, China – the case of super-spreader. Epidemiol. Infect. 134 , 786–791 (2006).

Majra, D., Benson, J., Pitts, J. & Stebbing, J. SARS-CoV-2 (COVID-19) superspreader events. J. Infect. 82 , 36–40 (2021).

Yu, I. T. S. et al. Evidence of airborne transmission of the severe acute respiratory syndrome virus. N Engl J. Med. 350 , 1731–1739 (2004).

Yu, I. T. S., Wong, T. W., Chiu, Y. L., Lee, N. & Li, Y. Temporal-spatial analysis of severe acute respiratory syndrome among hospital inpatients. Clin. Infect. Dis. 40 , 1237–1243 (2005).

Duval, D. et al. Long distance airborne transmission of SARS-CoV-2: rapid systematic review. BMJ 377 , e068743 (2022).

Prather, K. A. et al. Airborne transmission of SARS-CoV-2. Science 370 , 303–304 (2020).

Port, J. R. et al. Increased small particle aerosol transmission of B.1.1.7 compared with SARS-CoV-2 lineage A in vivo. Nat. Microbiol. 7 , 213–223 (2022).

Alsaad, K. O. et al. Histopathology of Middle East respiratory syndrome coronovirus (MERS-CoV) infection - clinicopathological and ultrastructural study. Histopathology 72 , 516–524 (2018).

Memish, Z. A. et al. Respiratory tract samples, viral load, and genome fraction yield in patients with Middle East respiratory syndrome. J. Infect. Dis. 210 , 1590–1594 (2014).

Guery, B. et al. Clinical features and viral diagnosis of two cases of infection with Middle East Respiratory Syndrome coronavirus: a report of nosocomial transmission. Lancet 381 , 2265–2272 (2013).

Fowler, R. A. et al. Transmission of severe acute respiratory syndrome during intubation and mechanical ventilation. Am. J. Respir. Crit. Care Med. 169 , 1198–1202 (2004).

Christian, M. D. et al. Possible SARS coronavirus transmission during cardiopulmonary resuscitation. Emerg. Infect. Dis. 10 , 287–293 (2004).

Tran, K., Cimon, K., Severn, M., Pessoa-Silva, C. L. & Conly, J. Aerosol generating procedures and risk of transmission of acute respiratory infections to healthcare workers: a systematic review. PLoS ONE 7 , e35797 (2012).

Judson, S. D. & Munster, V. J. Nosocomial transmission of emerging viruses via aerosol-generating medical procedures. Viruses 11 , 940 (2019).

Kim, Y. et al. Spread of mutant middle east respiratory syndrome coronavirus with reduced affinity to human CD26 during the South Korean Outbreak. mBio 7 , e00019 (2016).

Sabir, J. S. M. et al. Co-circulation of three camel coronavirus species and recombination of MERS-CoVs in Saudi Arabia. Science 351 , 81–84 (2016).

Alexander, D. J. et al. Association of Inhalation Toxicologists (AIT) working party recommendation for standard delivered dose calculation and expression in non-clinical aerosol inhalation toxicology studies with pharmaceuticals. Inhal. Toxicol. 20 , 1179–1189 (2008).

Hartings, J. M. & Roy, C. J. The automated bioaerosol exposure system: preclinical platform development and a respiratory dosimetry application with nonhuman primates. J. Pharmacol. Toxicol. Methods 49 , 39–55 (2004).

de Wit, E. et al. Middle East respiratory syndrome coronavirus (MERS-CoV) causes transient lower respiratory tract infection in rhesus macaques. Proc. Natl Acad. Sci. USA 110 , 16598–16603 (2013).

Corman, V. M. et al. Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR. Eurosurveillance 25 , 2000045 (2020).

Corman, V. M. et al. Assays for laboratory confirmation of novel human coronavirus (hCoV-EMC) infections. Eurosurveillence 17 , 20334 (2012).

Coleman, C. M. & Frieman, M. B. Growth and Quantification of MERS-CoV Infection. Curr. Protoc. Microbiol. 37 , 15E.2.1–9 (2015).

van Doremalen, N. et al. High prevalence of Middle East respiratory coronavirus in young dromedary camels in Jordan. Vector Borne Zoonotic Dis. 17 , 155–159 (2017).

Download references

Acknowledgements

The authors would like to thank Laura Tally and colleagues for assistance with mouse breeding, the Rocky Mountain Veterinary branch for assistance with high containment husbandry and cage design, Tina Thomas, Dan Long and Rebecca Rosenke for assistance with pathology and Anita Mora and Ryan Kissinger for assistance with the figures. This research was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH).

Open access funding provided by the National Institutes of Health.

Author information

Dania M. Figueroa Acosta

Present address: Division of Infectious Diseases, Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA

Authors and Affiliations

Division of Intramural Research, Laboratory of Virology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Hamilton, MT, USA

Neeltje van Doremalen, Trenton Bushmaker, Robert J. Fischer, Dania M. Figueroa Acosta, Rebekah J. McMinn, Michael Letko & Vincent J. Munster

Paul G. Allen School for Global Health, Washington State University, Pullman, WA, USA

Atsushi Okumura

Division of Intramural Research, Rocky Mountain Veterinary Branch, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Hamilton, MT, USA

Dana Scott & Greg Saturday

You can also search for this author in PubMed   Google Scholar

Contributions

N.v.D. and V.J.M. wrote the main manuscript text. N.v.D., R.F., T.B., A.O., R.M., M.L., D.S., G.S., D.F.A., and V.J.M. performed the research. All authors reviewed the manuscript.

Corresponding author

Correspondence to Vincent J. Munster .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary data, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

van Doremalen, N., Bushmaker, T., Fischer, R.J. et al. Transmission dynamics of MERS-CoV in a transgenic human DPP4 mouse model. npj Viruses 2 , 36 (2024). https://doi.org/10.1038/s44298-024-00048-y

Download citation

Received : 19 January 2024

Accepted : 20 July 2024

Published : 21 August 2024

DOI : https://doi.org/10.1038/s44298-024-00048-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.