banana fruit fly experiment

The Berg Lab

University of washington, genome sciences.

An introduction to fruit flies

This guide is adapted from the University of Arizona Department of Biochemistry and Molecular Biophysics General Biology Program for Science Teachers:  Drosophila Melanogaster and Mendelian Genetics, by Pete Geiger.

An Introduction to Drosophila melanogaster

Drosophila melanogaster is a small, common fly found near unripe and rotted fruit. It has been in use for over a century to study genetics and behavior. Thomas Hunt Morgan was the preeminent biologist studying Drosophila early in the 1900’s. He was the first to discover sex-linkage and genetic recombination, which placed the small fly in the forefront of genetic research. Due to it’s small size, ease of culture and short generation time, geneticists have been using Drosophila ever since.

Fruit flies are easily obtained from the wild and many biological science companies carry a variety of different mutations. In addition these companies sell any equipment needed to culture the flies. Costs are relatively low and most equipment can be used year after year. There are a variety of laboratory exercises one could purchase, although the necessity to do so is questionable.  

Why use Drosophila?

Teachers should use fruit flies for high school genetic studies for several reasons:      1. They are small and easily handled.      2. They can be easily anesthetized and manipulated individually with unsophisticated equipment.      3. They are sexually dimorphic (males and females are different), making it is quite easy to differentiate the sexes.      4. Virgins fruit flies are physically distinctive from mature adults, making it easy to obtain virgin males and females for genetic crosses.      5. Flies have a short generation time (10-12 days) and do well at room temperature.      6. The care and culture of fruit flies requires little equipment, is low in cost and uses little space even for large cultures.

By using Drosophila, students will:      1. Understand Mendelian genetics and inheritance of traits      2. Draw conclusions of heredity patterns from data obtained      3. Construct traps to catch wild populations of D. melanogaster      4. Gain an understanding of the life cycle of D. melanogaster , an insect which exhibits complete metamorphosis      5. Construct crosses of caught and known wild- type and mutated flies      6. Learn techniques to manipulate flies, sex them, and keep concise journal notes      7. Learn culturing techniques to keep the flies healthy      8. Realize many science experiments cannot be conducted and concluded within one or two lab sessions

National standards covered in these lessons: Content:      1. Organisms require a set of instructions for specifying traits (heredity)      2. Hereditary information is located in genes.      3. Combinations of traits can describe the characteristics of an organism.

Students goals:      1. Identify questions and concepts that guide scientific investigations      2. Design and conduct scientific investigations      3. Formulate and revise scientific explanations and models using logic and evidence      4. Communicate and defend a scientific argument

The genetics of Drosophila are well documented and several public-domain web sites feature the complete annotated genome . Therefore, those teachers or students wishing to see where their mutations occur have a ready reference available.

Since Drosophila has been so widely used in genetics, there are many different types of mutations available for purchase. In addition, the attentive student may find mutations within their own wild-caught cultures since, due to a short generation time, mutations are relatively common compared to other animal species.   Classification      Domain: Eukarya      Kingdom: Animalia      Phylum: Arthropoda      Class: Insecta      Order: Diptera      Family: Drosophilidae      Genus: Drosophila (“dew lover”)      Species: melanogaster (“dark gut”)   Life cycle of Drosophila melanogaster Drosophila melanogaster exhibits complete metamorphism, meaning the life cycle includes an egg, larval (worm-like) form, pupa and finally emergence (eclosure) as a flying adult. This is the same as the well-known metamorphosis of butterflies. The larval stage has three instars, or molts.

Females become sexually mature 8-10 hours after eclosion

• The generation time of Drosophila melanogaster varies with temperature. The above cycle is for a temperature of about 22°C (72°F).  Flies raised at lower temperature (to 18°C, or 64°F) will take about twice as long to develop. • Females can lay up to 100 eggs/day. • Virgin females are able to lay eggs; however they will be sterile and few in number.

After the eggs hatch, small larvae should be visible in the growing medium. If your media is white, look for the small black area (the mouth hooks) at the head of the larvae. Some dried premixed media is blue to help identify larvae however this is not a necessity and with a little patience and practice, larvae are easily seen. In addition, as the larvae feed they disrupt the smooth surface of the media and so by looking only at the surface one can tell if larvae are present. However, it is always a good idea to double check using a stereo microscope. After the third instar, larvae will begin to migrate up the culture vial in order to pupate.  

Care, Maintenance and Manipulation of Drosophila

Introduction In order to incorporate fruit flies in the classroom, it will be necessary to maintain cultures of flies for manipulation in crosses and as a backup for any mishaps which may occur. Culturing is very easy and it is recommended to have students maintain their own cultures of flies. In that way, each student or group would be directly responsible for the care and long-term maintenance of the flies, including making large culture populations for their crosses. When directly involved, students gain proficiency and a greater understanding of the flies requirements and behavior. The teacher should remain as coach, not lecturer, assisting students in techniques. The instructor needs to maintain stock cultures of all strains and mutants used by students in case the aforementioned unforeseeable incident occurs and student cultures die out or become intermixed. Losing cultures is the exception rather than the rule, and as long as students re-culture their flies on a regular basis and no mass contamination occurs, flies can be maintained for decades.   Bottles and vials Thomas Hunt Morgan used glass milk bottles for his experiments and, indeed, any container will do, including baby jars and assorted containers. However, for ease of culturing and transferring cultures, uniform bottles and vials are the best approach. Both can be purchased from a biological supply store. Bottles are used mainly for the maintenance of large populations of flies whereas culture vials are useful for maintaining smaller populations and are the preferred container for constructing student crosses. If there is a desire to maintain stock cultures for a long period of time, or to reuse bottles and vials it is important completely clean and sterilize them. This is to prevent outbreaks of pests and diseases.

To clean bottle and vials, first freeze them to kill any flies in them. Remove the food, wash well, then sterilize by autoclaving (for 20 minutes at 121°C and 15 psi; if containers are plastic, be sure they can be autoclaved) or washing in a 10% chlorine bleach solution.

Bottles and vials can be purchased in a variety of sizes and materials. Glass is effective, however if dropped a student could lose 2 weeks of data in a single spill. Autoclaved (sterile) plastic vials are available and are preferable for student use. Vial sizes range from 96 mm by 25 mm to larger sizes, however the smaller size is recommended for making crosses and maintaining small cultures. There are a variety of plugs available from soft cotton to foam plugs. This is a matter of preference and costs, however cotton works fine and can be bought at a local drug store in a pinch.   Where to buy supplies: Carolina Biological Supply Company FlyStuff.com , A division of Genesee Scientific   What they look like:

  Fly food The first step in preparing culture vials is adding food media. There are a variety of types of food available for the flies; some require cooking and others are bought already prepared and dehydrated. The latter can be purchased from a biological supply company. This is, of course, much quicker and easier than preparing cooked media, so much so that students can fill their own vials with media. However, it must be completely rehydrated for best results, since this is the only water source for adults and larvae. Therefore, follow the suggestions below to ensure a completely hydrated media:

Dehydrated media Add dry media to the bottle or vial to about 1/5 to 2/5 volume. Add water until media appears completely moistened. Allow the vial to sit for a few minutes, adding additional water if necessary until the media is completely hydrated. The surface should be moist with a shiny appearance and there should be no spaces in the media. If the media is not completed hydrated, production of vigorous cultures is compromised. Flies may be added minutes after media has been hydrated. Remember to add several grains (but not more) of yeast to the media surface before adding flies.

Cooked media When dispensing cooked media, it should fill the culture vial, bottle or vial 1/5th to 2/5th full. Keep the media out overnight to cure, keeping the vials covered with cloth to keep wild flies from laying eggs in them. The next day, add yeast and plugs. Refrigerate any unused media vials. Cooked media can be stored in a refrigerator for several weeks. Allow media to warm to room temperature before adding flies. Do not allow media to dry out.   Environment The easiest way to grow flies is at room temperature. However, the optimum rearing condition is a temperature of 25°C and 60% humidity. In these conditions generation time is shorter (9-10 days from egg to adult). Unless equipment is readily available this is unnecessary for successful rearing and crossing of flies. It is preferable to keep flies out of drafts and direct sunlight or heat sources. These will rapidly dry the media, necessitating frequent media changes and the potential to dehydrate the flies.   Anesthetizing flies The problem with fruit flies is that they fly! Therefore a variety of methods have been developed to anesthetize flies. Include are ether, commercial brands such as Flynap, carbon dioxide, and cooling. Each has its strengths and weaknesses. Ether is flammable, has a strong odor and will kill flies if they are over-etherized (and can anesthetize younger students!). Flynap, from Carolina Biological, is messy and has an odor that some find offensive. Each of these, however, requires low-cost equipment which can be easily purchased. Carbon dioxide works very well, keeping flies immobile for long periods of time with no side effects, however CO 2 mats (blocks) are expensive and a CO 2 source (usually a bottle) and delivery system (vials and clamps) are necessary, increasing the costs. If resourceful, one can use the CO 2 emitted from Alka-Seltzer tablets to anesthetize flies for short periods of time. Set up a large test tube with a tube and stopper system. Add water in the tube, then the Alka-Seltzer tablet. Carbon dioxide gas will be emitted.

The least harmful to the flies is either carbon dioxide or cooling anesthetizing. Of these two choices, cooling is the simplest, requiring only a freezer, ice and petri dishes. In addition, it is the only method which will not affect fly neurology, therefore behavior studies may begin after the flies have warmed up sufficiently.   Anesthetizing flies by cooling In order to incapacitate the flies, place the culture vial in the freezer until the flies are not moving, generally 8-12 minutes. Dump the flies onto a chilled surface. This can be constructed by using the top of a petri dish, adding crushed ice, then placing the bottom of the petri dish on top. Adding flies to this system will keep them chilled long enough to do each experiment. Simply place the flies back into the culture vial when finished. Flies will “wake up” relatively quickly once off the ice, so keep them cold. There are no long-lasting side effects to this method, although flies left in the refrigerator too long may not recover. Another way to keep flies chilled is adding water to zip-lock type freezer bags, place in the freezer with a petri dish nestled on the bag, and allow to freeze.   Transferring flies from one vial to another Flies should be transferred every 10 to 14 days. Students should maintain a backup culture of their flies and the instructor should maintain backup stock cultures of all fly strains. There are two basic ways to transfer flies when forming new cultures. One requires no anesthetizing but quick hands. A) Place a funnel in the mouth of a fresh culture vial that already has media added. In the old vial (the one with flies in it), gently tap the flies down by softly tamping the vial on a soft surface, such as a mouse pad. The flies will fall to the bottom and remain there for a few seconds (no more than that!), enough time to quickly take the plug off the vial, invert it into the funnel, and gently tamp, together, the two vials to force flies down into the new vial. B) An alternative way is to put the flies in the freezer for about 8 minutes. This will cause the flies to fall into a state of stupor. After placing a funnel on the new vial, invert the vial with motionless flies into the funnel. This is not as much fun but you won’t have any flies flying around the classroom.   Sexing flies It is quite easy to tell males from females and with a little practice students will become confident of their ability to do so. Notice that males are generally smaller and have a darker and more rounded abdomen. The coloration of the abdomen is the easiest to recognize. In addition, males have tarsal sex combs on their first pair of legs. These are black and very distinctive but can only be seen under relatively high magnification. With a little practice, by looking at the abdomen students will become proficient in accurately sexing flies. Sexing flies is critical when making crosses, so be sure student are confident in identifying the difference between the sexes. In order for students to feel comfortable sexing flies, give or have them obtain 25 or more mixed sex flies and allow them to sort the flies into two piles, male and female. Other students in the group and the instructor should verify the sorting. Each member of the group should be able to sex flies.   Pictures of males and females

Note the darker abdomen and more rounded appearance of the male. Females also tend to be larger.   Collecting virgin females While it’s a simple matter of placing virgin females with males, it is important to recognize the time factor involved for obtaining virgins. Females remain virgins for only 8-10 hours after eclosure and must be collected within this time frame. NOTE: Females have the ability to store sperm after a single mating, so if the female for a cross is not a virgin, you will not know the genotype of the male used for your cross. It is strongly suggested that you obtain extra virgins in case a mistake is made in identification or the fly dies before mating and egg lying can occur. In a strong culture, multiple virgin females should be easily obtained. Although females are able to lay eggs as virgins, they will be sterile and no larvae will be produced. Below are three ways to obtain virgins, the ‘removal method’ being most encouraged for beginners.   Removal method Remove all flies 8-10 hours before collecting (generally this is done first thing in the morning). Visually inspect surface of food to ensure complete removal of flies. After 8-10 hours (usually before you leave work) collect all females that are present. All will be virgins. Place in a fresh culture vial and wait 2-3 days look for larvae. Virgin females can lay eggs, but they will be sterile. Since they are photoperiod- sensitive, females tend to eclose early in the morning. Therefore early collections will ensure the greatest number of virgins for experimentation. However, collection is possible later in the day.   Visual method Being able to recognize virgin females removes the necessity of emptying culture vials on a timely basis and allows students to collect their own without the necessity of coming to class at odd times of the day. Note that virgin females are much larger than older females and do not have the dark coloration of mature females. In addition, in the early hours after eclosure, there will be visible a dark greenish spot (the meconium, the remains of their last meal before pupating) on the underside of the abdomen.   Temperature cycling It is possible to maximize the number of virgins in a morning collection by using temperature cycling. When cultures are maintained at a temperature of 18°C, development is slowed so females will not mate until 16 hours after enclosure. By removing flies in the afternoon/evening and placing the vials in an 18°C incubator, 98% of flies obtained in the morning will be virgins. Placing virgins in their own vials for 2-3 days will eliminate those 2% that are non-virgins.   Pictures of virgin males and females:

Crossing flies Once females are deemed virgins, add males. When setting up crosses, a 3:1 ratio of virgin females to males is ideal. Generally, males will mate more efficiently if they have matured 3 days or longer. Be sure to select robust, healthy males; the older the flies, the lower the mating efficiency. Mating occurs quickly and the behavior is interesting to watch, but will not be addressed here. Females begin laying fertile eggs soon after mating. Refer to the life cycle chart for evidence of F1 larvae. Remove adults once it has been established that enough larvae are present (typically 7-8 days after the cross) since you may not be able to distinguish parents from the F1 generation.   Killing Flies: The Morgue This is an unfortunate necessity when using flies. A bottle or beaker with soapy water, or mineral oil is generally used. Dump anesthetized flies directly into the soapy water or mineral oil where they drown. A bottle (beaker, or screw-capped jar) filled with ethanol or isopropanol can also be used as a morgue.   Basic Drosophila Genetics Nomenclature and Definitions

Drosophila melanogaster flies have 4 chromosomes. The genotype is written as:

Chromosome Chromosome or Chromosome / Chromosome

This common nomenclature shows one chromosome on top and its homologue on the bottom, as the chromosomes would appear during meiosis when contributing gametes.

When writing the genotype, in general, chromosomes are separated with a semicolon.

X chromosome; chromosome II; chromosome III; chromosome IV

Wild-type is denoted as “+” or WT

Dominant mutations are written with a capital letter: For example: Bar or B

Recessive mutations are written with a lower case letter: For example: white or w

Mutations are alleles (alternative forms of a gene occupying a given locus on a chromosome) that are inherited with chromosomes.

Homozygote – An individual with the same allele at corresponding loci on the homologous chromosomes.

Heterozygote – An individual with different alleles at corresponding loci on the homologous chromosomes.

Genotype – The genes that an organism possesses.

Phenotype – The observable attributes of an organism.

P1 – Parental generation.

F1 – Filial generation, or offspring generation. F1 is the first offspring generation.

F2 – The second offspring generation.   Other great web resources:

Gerard Manning wrote a simple introduction to Drosophila genetics .

Genetics on the Fly: A Primer on the Drosophila Model System by Karen G. Hales et al (2015).

Taking Stock of the Drosophila Research Ecosystem by David Bilder and Kenneth D. Irvine (2017).

FlyBase is an encyclopedic resource for Drosophila researchers, with detailed information on fly stocks, genes, mutants, researchers, publications and much more.

Choose an Account to Log In

Notifications

Science project, what attracts fruit flies.

Party in the compost! What foods are on a fruit fly’s preferred menu?

In this experiment, you’ll create two compost bins and see if you can determine what foods attract the most fruit flies. This experiment is best done in warm weather, when there are more fruit flies around.

What foods attract fruit flies to your compost bin?

• Four plastic bins with lids
• Safety goggles
• Red wiggler worms
• Vegetable scraps such as lettuce and celery
• Fruit scraps such as banana peels and apple cores
• Shredded newspaper
• Spray bottle
• Get two large plastic bins and get an adult to help you drill holes in the bottom. Make sure that you wear your safety goggles! Drill a quarter-inch hole every 5 inches or so.  Nest the bin with holes inside a bin without holes. Do the same thing with the other two bins. You’ve now created two worm bins – a type of compost bin.
• Shred enough newspaper or used white paper to cover the bottom of the bins, making sure that the bins are about halfway full of shredded paper. It’s best if you use a shredder to do this, but you can also use your hands to shred the paper into thin strips. Spray the paper with water to make it damp.
• Get yourself some worms! Red wiggler worms love to eat compost scraps. A pound of worms per bin is usually enough to start you out. Once you’ve placed the worms into their homes, decide which bin will have fruit and which one will have vegetable scraps. Write Fruit on one container and Vegetables on the other so you don’t get confused.
• Create a hypothesis, your best guess about what is going to happen. Which compost bin will attract and breed more flies: the one with vegetables, or the one with fruit?
• Put each bin in a shady place. These places should be at opposite ends of your house or garden.
• Now, add some food scraps to the bins. Green vegetables like lettuce and celery go into the vegetable bin, while fruit scraps like apple cores and banana peels go into the fruit bin. Weigh all of the food waste before it goes in so that you know you’re giving the worms in each bin the same amount of food. Be careful not to overfeed: Red wigglers can eat half their weight in food every day, so a pound of worms can eat half a pound of food a day.
• Over the next two weeks, visit each bin daily. Track and photograph what’s going on in each bin. What have the worms been eating? What do they seem to like best? Do the vegetable and fruit scraps disappear at the same rate?
• Check out the fruit fly population as well. Are there flies in or around either of the bins? Each day, count or estimate the number of flies that you see above each bin and write it down in your notebook. Does one bin have more flies than the other?

The bin with the fruit will have more fruit flies.

Like their name suggests, fruit flies love fruit. But why? The answer has more to do with the fruit’s sugar and starch than it does with the fruit itself.

Lettuce and celery have very little sugar or starch. This is one of the reasons that these vegetables are so good for you! However, sugars and starches are carbohydrates, one of the basic sources of energy that humans and other animals eat to live.  Most fruit has a lot of carbohydrates, often in the form of sugar.  Some vegetables also have a lot of more complex carbohydrates called starches. A potato is an example of a “starchy” vegetable.

Fruit flies might hang out on overripe fruit, but they’re really there for the yeast that grows on it. When starches in fruit begin to ferment, yeast uses those starches to make alcohol. Fruit flies are attracted to the CO2 produced by this reaction, and come to munch on the yeast. So if you have overripe fruit and starchy vegetables, the fruit flies will fly in for a picnic!

Fruit flies love fermenting fruit for another reason as well. In the wild, fruit flies are threatened by tiny wasps that lay eggs inside baby fruit flies, which can kill the fruit flies. Mother fruit flies lay their eggs in alcohol because alcohol helps kill those wasps. This provides the fruit fly babies with some protection.

How can you avoid ending up with fruit flies in the compost? Try the following techniques: Bury compost under a thin layer of soil or a thick layer of newspaper so that the fruit flies can’t find it. In an outdoor compost, add leaves or newspaper on top of the fruit to hide it and help it compost. Good composting methods will also heat up the compost, making it uncomfortable for the flies.

Related learning resources

Add to collection, create new collection, new collection, new collection>, sign up to start collecting.

Bookmark this to easily find it later. Then send your curated collection to your children, or put together your own custom lesson plan.

You are using an outdated browser. Please upgrade your browser or activate Google Chrome Frame to improve your experience.

• Pest Guides
• Pest Articles
• Pest Quest Episodes
• Bug Battles Episodes
• "Will They Eat It?" Episodes
• Downloadable Activities
• Science Fair Kits
• Report Writing
• Pest Professor Lessons
• Buggy Break
• School Lesson Plans
• Elementary Grade Activities
• Middle School Grade Activities
• Pest Quizzes

Fruit Fly Science Experiment for Middle Schoolers

Science fair project on the life cycle of a fruit fly.

Fruit flies are not only pests but they are also incredibly helpful in scientific research because the reproduce so quickly. For this fruit fly experiment, you will start by observing the fruit fly life cycle. Then, you will follow the scientific method to test how temperature affects fly development during their life cycle.

Difficulty:  This fruit fly science fair project is rated as Easy to Moderate difficulty and, is suggested for grades 4-8.

Time Required:  Two weeks to one month, longer if you choose

Safety:  Because they can carry disease and germs, some fruit flies can on rare occasions make people sick. You should always wash your hands after each observation of the flies. Fruit flies mature and reproduce quickly and in large numbers. Pay attention to this symbol as some steps should be done outdoors to avoid a fruit fly infestation in your house!

Time of Year:   Try spring, summer or fall for easier collection of flies

Material Availability:   Most materials are readily available in your home or grocery store. The fruit flies can be found in your yard hovering near ripening or rotten fruit, fruit that has fallen off a fruit tree onto the ground, and compost piles. You can also order fruit flies from a biological supply or exotic pet food supply companies. These are called Drosophila melanogaster.  

Download Poster

More fun with pests

Open Access is an initiative that aims to make scientific research freely available to all. To date our community has made over 100 million downloads. It’s based on principles of collaboration, unobstructed discovery, and, most importantly, scientific progression. As PhD students, we found it difficult to access the research we needed, so we decided to create a new Open Access publisher that levels the playing field for scientists across the world. How? By making research easy to access, and puts the academic needs of the researchers before the business interests of publishers.

We are a community of more than 103,000 authors and editors from 3,291 institutions spanning 160 countries, including Nobel Prize winners and some of the world’s most-cited researchers. Publishing on IntechOpen allows authors to earn citations and find new collaborators, meaning more people see your work not only from your own field of study, but from other related fields too.

Brief introduction to this section that descibes Open Access especially from an IntechOpen perspective

Want to get in touch? Contact our London head office or media team here

Our team is growing all the time, so we’re always on the lookout for smart people who want to help us reshape the world of scientific publishing.

Home > Books > The Wonders of Diptera - Characteristics, Diversity, and Significance for the World's Ecosystems

Fruit Flies ( Drosophila spp. ) Collection, Handling, and Maintenance: Field to Laboratory

Submitted: 16 October 2020 Reviewed: 04 March 2021 Published: 28 March 2021

DOI: 10.5772/intechopen.97014

Cite this chapter

There are two ways to cite this chapter:

From the Edited Volume

The Wonders of Diptera - Characteristics, Diversity, and Significance for the World's Ecosystems

Edited by Farzana Khan Perveen

To purchase hard copies of this book, please contact the representative in India: CBS Publishers & Distributors Pvt. Ltd. www.cbspd.com | [email protected]

Chapter metrics overview

1,339 Chapter Downloads

Impact of this chapter

Total Chapter Downloads on intechopen.com

Total Chapter Views on intechopen.com

As drosophilids are versatile, low maintenance and non-harming model organisms, they can be easily used in all fields of life sciences like Genetics, Biotechnology, Cancer biology, Genomics, Reproductive biology, Developmental biology, Micro chemical studies, ecology and much more. For using such a model organism, we need to learn capturing, rearing and culturing their progeny along with basic identification and differentiation between males and females. This chapter is being emphasized on techniques of capturing these flies with different and effective techniques. Along with it, most species-specific baits are discussed to catch more yield. Culture food media, a set measurement of different ingredients is used to rear the collected sample. The reasons for using each ingredient are also discussed in this chapter. At last, this chapter highlights the basic clues to identify different species in the field and lab along with learning distinguishing characteristics of males and females easily and effectively.

• attractive baits
• culturing flies
• food preparation
• identification
• sorting out male and female

Author Information

Pragya topal.

• Department of Zoology, HNB Garhwal University, India

Divita Garg

• Division of Biological and Life Sciences, School of Arts and Sciences, Ahmedabad University, India

Rajendra S. Fartyal *

*Address all correspondence to: [email protected]

1. Introduction

In 1911 T.H. Morgan and his students C. Bridges, H. J. Muller and A. H. Sturtevant came across red-eyed insects, fruit-flies. Since then, it has made its way in research labs helping scientists to explore fundamental problems in biological sciences. It has turned out as one of the best metazoan insect model organisms. It is one of the best model organisms for the biological studies ranging from molecular genetics of diseases to the ecosystems and up to the evolutionary scales. Starting from visible mutants and chromosome mapping, today studies of complex genetic networks are possible with the help of multiple genome sequences (the 12 genomes project), systematic gene disruption or knock-down (RNAi stock library), microarray analysis, protein interaction maps and the FlyBase integrated database. High-throughput platform biology and open-source availability are considered as part of modern developments and the Drosophila model has integrated them with success. Recent developments allow the analysis of problems and processes previously inaccessible like complex human diseases caused by developmental, neurological or metabolic defects. Drosophila has clear-cut advantages in this field of research with its sophisticated genetic techniques. Drosophila research also plays an important role in technological transfer to other arthropod models, opening the window to biodiversity resources and macro-evolutionary scale. It has also been successfully integrated in teaching subjects as diverse as genetics, physiology, ecology and evolution.

Drosophilids are ectothermic insects whose body temperature changes with the ambient temperatures. These insects can easily survive between 12–21 degrees Celsius [ 1 ]. Temperature impact on their viability, fertility, developmental period, foraging activity, feeding, and breeding could easily be seen in laboratory stocks and distribution and populations dynamics under field conditions [ 2 , 3 ]. Even Apart from temperature, humidity and rainfall, sunlight also plays a vital role in distribution of drosophilid species.

The family Drosophilidae encompasses 4,450 species distributed in 75 genera with two subfamilies, drosophilidae and steganinae [ 4 ]. The sub family drosophilinae is more diverse, distributed across 47 genera of which genus Drosophila is the largest with 1,213 species. The subfamily steganinae is a smaller one and is distributed across 28 genera of which genus Leucophenga is the largest with 256 species [ 4 ].

In India, more than 347 species are recorded which are spread across 27 genera, of which 58 species belongs to 8 genera of subfamily steganinae and rest 289 species of 19 genera are placed in subfamily drosophilinae (unpublished data).

2. Methods of collection

There are several methods in practice to collect fruit-flies from their natural habitats. Some of the methods are shown below:

2.1 Installing trap-bait

For setting-up traps one could use plastic bottles ranging in the volumes of 250 mL to 1 L. Fruit slices along with a pinch of yeast could be used as a bait. With the use of blade somewhere in the middle of the bottle a section could be carved out for food access. These traps could be hanged on orchard tree at a height 3–4 feet above the ground ( Figure 1 ). The collection area must be damp and moist with minimal human interference. After 1–2 days traps must be recovered to collect flies. Same method could be applied to collect all kinds of Drosophila species. Some flies prefer to breed in the trash and stay close to ground, and some around the trees. The methods could be improved based on little understanding of Drosophila species ecology (authors observations; unpublished).

Trap bait: a) Transparent plastic bottles can be used as traps, a C-shaped window is made in the center of the bottle which serves as an entry point for the flies. Banana pieces placed inside the bottles attracts the flies. b) The bottle can be placed around small bushes/trees and can be collected next morning.

2.2 Net swipe

There are many genera that are not attracted to regular traps and need to be captured from their natural food source (wild rotting fruits). Capturing such drosophilid species nets are more effective. The flies hovering over the rotten fruits or piles of organic wastes could be captured using this method ( Figure 2 ).

Net sweeping: The modified insect nets are used for capturing drosophilids which are not attracted toward baits.

2.3 Use of aspirator

This method is used when fly’s numbers are low and also when investigator is familiar with the species identification and targeting particular species individuals. This method is more appropriate when flies are feeding, mating or resting on petals, leaves, fruits etc. Flies feeding on mushrooms and flowers are mostly collected using this method ( Figure 3 ). Aspirator is also used to transfer flies from bait bottles or insect net to culture vials.

Use of aspirator: The flies performing mating or resting over leaves, mushroom or fruits etc. are captured with help of aspirator.

Some of the common species-specific baits are:

Banana: It is the most commonly used bait. It could be used to collect cosmopolitan Drosophila melanogaster, and other commonly found drosophilid species.

Tomato: Adding yeast to it forms an attractive bait for Drosophila suzukii and Drosophila busckii.

Orange: It is also an attractive bait for Drosophila suzukii and Drosophila busckii.

Grapes: With added yeast grapes act as an attractive bait for species like Drosophila busckii and Drosophila immigrians.

Usage of vinegar in the baits for Drosophila suzukii helps in trapping more flies. Keeping baits in moisture free traps improves the collection yield.

3. Culture handling and maintenance

The culture handling and maintenance is required to be learned for making the collected sample to multiply several times and for making isolines and confirming the species’s level identification.

3.1 Rearing and culturing Drosophila

For rearing and culturing the drosophilids, the collected samples need to be transferred from bait bottles to the narrow culture vials (with culture media) While shifting flies from the bottles to collection tubes the openings of the bottles must be kept under bright light directing the flies toward the light and thus making their transfer easy. For culturing, the flies are kept at room temperature (25 ̊C). If room temperature is not adequate the culture tubes could be kept in the BOD (Biological Oxygen Demand) incubators. This would help the culture media to last longer until the drosophilid colony begins to establish.

3.2 Preparing Drosophila media

The commonly used media for various drosophilid stock maintenance includes Agar, yeast, Maize flour, Brown sugar, Nepagin, and Propionic acid.

To prepare fly media agar is added to the hot water. Following this yeast, maize flour, and sugar is added. After 20–30 minutes of cooking (with 1–2 boil) heat should be turned off. Once media temperature reaches to close to 50 degrees nepagin and propionic acid could be added. Throughout the preparation food needs to be constantly stirred.

Maize powder, brown sugar, and dried yeast are used as food. Yeast holds special nutritional value for drosophilids. While Nepagin is anti-fungal in nature and Propionic acid is a bactericidal and functions as preservative and increases the shelf life of food.

The food could then immediately be transferred to the sterilized vials or bottles. As soon as the media starts hardening the vials or bottles needs to be covered properly with the help of a cheese cloth. The food could be used after a day. The media tubes or bottles could also be stored in a cool place for 1–2 weeks for future use. Other food recipes:

There are almost 7–8 different food recipes for drosophilid species. The following link could be further explored for other recepies [ 5 ].

Cornmeal, sucrose, dextrose, yeast and 2-acid medium

This food recipe was described by Lewis in 1960. This uses the phosphoric acid which allows less propanoic acid usage without affecting the fungicidal effects. This recipe has also been used at Caltech since 1955.

Cornmeal, molasses and yeast medium

This mixture has to be used fresh. This medium was used at Bloomington for several years. It went through certain modifications in order to prevent bacterial contamination.

Cornmeal, dextrose and yeast medium

This recipe was first used by Brent and Oster in 1974. In order to reduce the chances of Leucon Stoc infection of cultures, sucrose was substituted by dextrose [ 5 ].

Identification (taxonomy)

In a biodiversity rich place, one could come across many small insect species. Identification becomes crucial in sorting a particular species from a pool of collection. Over the years, taxonomists came across various methods to identify species. Therefore, some common morphological diagnostic characters were listed out and based on these keys one could identify drosophilid species. However, closely related or sibling species are hard to distinguish with general morphological characteristics.

Drosophilids are usually small flies, ranging from 1.5–7.0 mm in length; yellowish, golden, brownish or blackish in color. They possess a number of the characteristics, such as red eyes and plumose arista. Body is shiny, often with stripes or spots on the thorax. Wings are hyaline or with black patches or marginal areas with dark lines. Abdominal tergite strip patterns (i.e. pigmentation) vary from species to species (dark or light bands or spots in 2–6 tergites). In some species, sexual dimorphism is clear by wing patch or presence of sex-comb on legs.

However, species are identified by their male genital organs (periphallic & phallic) possessing structural compatibility features [ 6 ]. These genital organelle structures are species specific. Example, the two subfamilies viz., drosophilinae and steganinae are differentiated on the basis of the distance between proclinate orbital setae and inner vertical setae from posterior reclinate setae. In subfamily drosophilinae the distance of proclinate orbital seate from posterior reclinate setae is less compared to its the distance of inner vertical setae. While in subfamily Steganinae the distance of proclinate orbital setae from posterior reclinate setae is more than compared to its distance from inner vertical setae.

Flies are too small in size to be observed with naked eyes. Hence magnification is required and this could be achieved with a good hand lens, or a wide-field binocular microscope, or a stereo zoom microscope, or a compound microscope. On a white background pictures emerge better ( Figure 4 ).

Identification: a) after collection of the flies, they are given anesthesia(CO2) which gives 30 minute time to sort the male and female. b) Identification can also be done by dissecting the male genitalia.

For identification, anesthesia could be given to flies. This makes flies unconscious. Ether or Carbon dioxide gas could be used as anesthesia. Flies sensitive to ether or CO2 cold treatment is an option.

Generally, females possess larger body size and have swollen abdomen than males. In some cases of males’ 5th and 6th tergites are pigmented whereas others have wing patches. In a few genera of drosophilids males also possess sex comb on their fore legs. The posterior end of males body is pointed whereas in females it is pointed (due to the presence of their ovipositor and female genital organ). The male genital organs are species-specific and differ from species to species. Dissection of genital organs is used for conformation of species. Dissection of male genitals is usually done with the help of a needle. It requires washing the tissues in 10% KOH approximately at 100 °C for several minutes. This opens the intact reproductive plates and helps investigators to collect additional details. Few drops of glycerol can be added for better resolution.

4. Morphological study of drosophilid

Like in other arthropods, the adult drosophilids body is comprised of three major parts i.e., head, thorax and abdomen. Drosophilids have one pair of wings and halters each. Wings help in flight whereas halters help as balancing organs.

The head of drosophilids has distinguishing parts like ocellar triangle, post-ocellar setae, inner and outer vertical setae, fronto-orbital plates, cibarium and arista. Arista is a distinctive character as it possesses varying dorsal and ventral branches in different genus. The genus Drosophila possesses three elongated orbitals with proclinate setae inserted into the anterior most part. The ratio, length and placement of orbitals, ocellar, vertical and post-ocellar setae is used for the identification of different genera of drosophilids ( Figure 5 ). In genus Chymomyza the anterior reclinate setae is present in front of the proclinate setae whereas in genus Liodrosophila anterior reclinate setae present is small in size. Facial carina is also an important diagnostic characteristic in the species. The chaetotaxy and color of the palps is used to identify the sibling species [ 7 ]. The setae on either side of the face known as vibrissae and sub-vibrissae are important for the identification of species. The number of anterior & posterior sensilla and sensillacampaniformia of the cibarium are important diagnostic characteristics for the identification of the species.

Head: a) D. suzukii, showing major portions of head region. b) Arista and its parts.

The thorax has three main segments: prothorax, mesothorax and metathorax. Mesothorax is significantly enlarged which aids the wings, while prothorax and metathorax are generally reduced. Most of the dorsal surface of the mesothorax is covered with mesonotum. There are numerous regular and irregular rows of acrostichalsetulae and numeral pairs of dorsocentral setae present on the mesonotum. The number and position of the Acrostichalsetulae is taxonomically important for differentiation of the species ( Figure 6 ). Majority of the Drosophila species have six or eight regular acrostichal rows. While most of the Scaptomyza species possess two or four rows. However, other genres are characterized by ten or more irregular rows. In a few species of mycophagous tripunctata and testacea group a set of enlarged acrostichals are located more anteriorly near to the transverse suture known as presutural setae and are also used for identification [ 7 ]. The length and orientation of basal and apical scutellar setae (present on the scutellum) are also taxonomically important for the species identification. Number and length of the katepisternal setae present on katepisternum are also significant for differentiation of the species and species group.

Thorax: D. suzukii a) dorsal b) lateral.

Legs in drosophilids are divided into coxa, trochanter, femur, tibia and tarsus. Tarsus have 5 tarsal segments. Color and arrangements of bristles (chaetotaxy) of male foreleg and relative length of first tarsus is important for distinguishing traits among different species. Their presence and the number of spines on hind leg, tibia and tarsomere (genus Impatiophila) both are considered for the identification of closely related species [ 8 ]. In some species such as Liodrosophila angulata a number of taxonomically important spines are present on the femur of the foreleg.

4.4 Sex Combs

Some species of melanogaster and obscura group are characterized on the basis of sex combs on the male foreleg. The numbers of sex combs are taxonomically important for the distinction of the different species. Drosophila males use their sex combs to grasp the females’ abdomen and genitalia. They also use them to spread their wings prior to copulation.

Wings are attached to the mesothorax of the abdomen. In many drosophilids wings possess spotted patterns. Also,hylanine or fuscous are taxonomically important. Although there is a consistent pattern on wing’s venation but in some taxa additional cross veins could be present eg planitibia species group. Different types of cross-veins (bm-cu, dm-cu, r-m and cuA2), subcostal break, humeral break, wing cells (bm, dm, cup), and wing veins (CuA1, A1, R1, R2 + 3, R4 + 5, M1) occur in the wings ( Figure 7 ). In genus Impatiophila the setae of the middle row on the second costal section is considered important for the differentiation of the species into different species group [ 8 ].

Wing: Abbreviations; h = humeral, hum brk = humeral break, R1 = anterior branch of radium, C1 s = apical seta(e) on 1st costal section, sc brk = subcostal break, r2 + 3 = second + third radial, r4 + 5 = fourth + fifth radial, M1 = 1st posterior (sectorial) branch of media, dm-cu = discal medial-cubital, CuA2 = 2nd anterior branch of cubitus, CuA1 = 1st anterior branch of cubitus, A1 = 1st branch of anal vein, r-m = radial-medial. (source; terminology by professor M.J Toda).

4.6 Abdomen

The dorsolateral portion of abdomen is known as tergite and it is segmented and chitinized. The ventral portion is called sternite. It is generally hairy, chitinized and quadrilateral in shape. The length and width of the sternite is important for species identification ( Figure 8 ).

Abdomen: D. suzukii (a) male abdomen T1 to T6 = Tergite 1 to 6; (b) female abdomen S1 to S6 = Sternite 1 to 6.

4.7 Terminalia

The male terminalia possess many internal and external characters useful for species characterization. In drosophilids the male genitalia exhibit speedy and divergent evolution while female genitalia are thought to evolve slowly among closely-related species. In drosophilids female copulatory structures have been claimed to be mostly invariant compared to male structures. In most animal species with internal fertilization, male external genitalia are the most rapidly evolving organs and they usually are the first organs to diverge morphologically following by speciation. Because of their rapid evolution and species-specificity, their illustration is a common feature of taxonomic literature to discriminate between closely-related species. The morphology of male genitalia can differ dramatically even within very closely related animal species. The male terminalia is further divided into epandrium and hypandrium.

4.8 Epandrium

In males, 9th tergite is known as epandrium and it possesses a number of characteristics such as pair of cerci and surstyli which is present on the posterior and posteroventral of the epandrium respectively. The size, color, shape, morphology and number of setae on the cerci are significant for the distinction of species. Surstyli bears a number of distinct prensistae. The numbers of prensistae vary from species to species and are important for the species identification.

4.9 Hypandrium

The genitalia are the structure linked with 9th sternite or Hypandrium and possesses aedeagus, basal processes of the aedeagus, parameres, and gonopods on it. The posterolateral portion of the hypandrium is known as gonopods. Posterolateral to the gonopods are paraphyses which possess a number of setae. The aedeagus is placed centrally with respect to the rest of hypandrium.

The major distinguishing character at species level are the presence and numbers of sex combs and bristles. Different types of spots present on wings and abdomen, and lining present on thorax of different drosophilids etc. are the characters that provide cues to identify flies ( Figure 9 ).

Species level distinguishing characters.

5. What makes Drosophila a great model organism?

5.1 drosophila as a model organism.

The different characteristics of D. melanogaster make it an ideal model organism, which are following:

Smaller Size and Short lifespan

Shorter life span facilitates large quantities of flies to be produced in a short time.

Minimal culturing requirements

Due to the smaller size and minimal requirements, Drosophila can be cultured and tested in limited resources.

Genetic manipulation

The fly genome has been sequenced and well characterized. It has 100+ years of literature available. Besides this it has four pairs of chromosomes only which makes an ideal system to do genetic crossing and gene editing simpler.

5.2 Basic research

As drosophilids are versatile, low maintenance and non-harming model organisms, they are used in all fields of life sciences like Genetics, Biotechnology, Cancer biology, Genomics, Reproductive biology, Developmental biology, Micro chemical studies, ecology and much more. For more than a century, the low cost, rapid generation time, and excellent genetic tools have made the fly indispensable for basic research. Also, the recent advancements in the field of molecular tools have allowed the organism to be used more efficiently. From human disease modeling to the dissection of cellular morphogenesis and to behavior and aging, the current usage of flies greatly influences fly research. However, this field remains vibrant and exciting, with labs using flies in drug discovery, bioengineering, regenerative biology, and medicine. The future use of fruit flies as a model organism in research is bright.

6. Stock centres

Bloomington stock centre maintains various fly stocks including aberrations, balancers, deficiencies, duplications, clonal, chemically induced mutations, human disease model, mapping, teaching stocks, wild-type lines, transgenes and various other stocks. Further details about the stock centre could be found at site dedicated to stock centre [ 9 ].

Kyoto Stock Centre, Kyoto Institute of Technology, Kyoto, Japan.

Harvard Medical School, Boston, MA, USA.

FlyORF, University of Zurich, Zurich, Switzerland.

NIG-FLY, National Institute of Genetics, Mishima, Japan.

THFC, Tsinghua University, Beijing, China.

Vienna Drosophila Resource Centre (VDRC), Vienna, Austria

Fly maintenance stuff

A detailed information about fly maintenance materials and accessories are valuable at several commercial vendors. The provided links provide access to it [ 10 ].

Important websites for Identification of drosophilids

Taxonomic information database for the world Drosophilidae:

DrosWLD Species (Taxonomic information database for world species of Drosophilidae, maintained by Masanori J. Toda) https://bioinfo.museum.hokudai.ac.jp/

Japan Drosophila Database: JDD http://www.drosophila.jp/jdd/index_en.html

7. Conclusions

This chapter highlights the basic clues to identify different species in the field and lab along with learning distinguishing characteristics of males and females easily and effectively.

8. Recommendations

These protocols will act as baseline data for handing and maintaining drosophilids for young taxonomist. As these processes are easy ones so these could be used at graduation level to let students get familiar with its taxonomy.

Acknowledgments

The authors wish to offer cordial thanks to Dr. Subhash Rajpurohit, Associate Professor, Division of Biological and Life Sciences, School of Arts and Science, Ahmedabad University, Gujrat, India for his careful reading of the manuscript and his valuable suggestions for this study. This work is not financially support from any agency except the publisher.

Conflict of interest

The authors declare no conflict of interest.

Acronyms and abbreviations

humeral break

anterior branch of radium

apical seta(e) on 1st costal section

subcostal break

second + third radial

fourth + fifth radial

1st posterior (sectorial) branch of media

discal medial-cubital

2nd anterior branch of cubitus

1st anterior branch of cubitus

1st branch of anal vein

radial-medial. (Source; Terminology by Professor M.J Toda). Abdomen: T1 to T6 = Tergite 1 to 6; S1 to S6 = Sternite 1 to 6;

• 1. Hoffmann, A. A. Physiological climatic limits in Drosophila: Patterns and implications. Journal of Experimental Biology; (2010). 213(6), 870-880. https://doi.org/10.1242/jeb.037630
• 2. Clemson, A. S., Sgrò, C. M., & Telonis-Scott, M. Thermal plasticity in Drosophila melanogaster populations from eastern Australia: Quantitative traits to transcripts. Journal of Evolutionary Biology; (2016). 29(12), 2447-2463. https://doi.org/10.1111/jeb.12969
• 3. Moghadam, N. N., Ketola, T., Pertoldi, C., Bahrndorff, S., & Kristensen, T. N. Heat hardening capacity in Drosophila melanogaster is life stage-specific and juveniles show the highest plasticity. Biology Letters; (2019). 15(2), 20180628. https://doi.org/10.1098/rsbl.2018.0628
• 4. “Welcome to T a x o D r o s.” https://www.taxodros.uzh.ch/ (accessed Mar. 01, 2021).
• 5. “Bloomington Drosophila Stock Center: Indiana University Bloomington.” https://bdsc.indiana.edu/information/recipes/caltechfood.html (accessed Mar. 01, 2021).
• 6. Bächli, G. Family Drosophilae. In: Papp, L. and Darvas, B. (eds).Contributions to a manual of Palearctic Diptera. III. Higher Brachteera. Science Herald; 1998.Vol −3 pp 503-513.
• 7. Markow, T.A., & O’Grady, P. Drosophila: a guide to species identification and use. Elsevier (2005).
• 8. Fu, Z., Toda, M. J., Li, N.-N., Zhang, Y.-P., & Gao, J.-J. A new genus of anthophilous drosophilids, Impatiophila (Diptera, Drosophilidae): Morphology, DNA barcoding and molecular phylogeny, with descriptions of thirty-nine new species. Zootaxa; (2016). 4120(1), 1-100. https://doi.org/10.11646/zootaxa.4120.1.1
• 9. “Bloomington Drosophila Stock Center,” Bloomington Drosophila Stock Center . https://bdsc.indiana.edu/index.html (accessed Mar. 01, 2021)
• 10. “Flystuff.” https://flystuff.com/ (accessed Mar. 01, 2021).

© 2021 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Continue reading from the same book

The wonders of diptera.

Published: 08 September 2021

By Manoel A. Uchoa, Anderson S. Fernandes and Jimi N....

362 downloads

By Manoel A. Uchoa, Morgana F. Wachter-Serapião and N...

474 downloads

By Zerguine Karima

2061 downloads

IntechOpen Author/Editor? To get your discount, log in .

Discounts available on purchase of multiple copies. View rates

Local taxes (VAT) are calculated in later steps, if applicable.

Support: [email protected]

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: 24 March 2021

Predictive olfactory learning in Drosophila

• Chang Zhao 1   na1 ,
• Yves F. Widmer 2   na1 ,
• Sören Diegelmann 2 ,
• Mihai A. Petrovici 1 ,
• Simon G. Sprecher   ORCID: orcid.org/0000-0001-9060-3750 2 &
• Walter Senn 1  

Scientific Reports volume  11 , Article number:  6795 ( 2021 ) Cite this article

4933 Accesses

18 Citations

8 Altmetric

Metrics details

• Classical conditioning
• Learning algorithms

Olfactory learning and conditioning in the fruit fly is typically modelled by correlation-based associative synaptic plasticity. It was shown that the conditioning of an odor-evoked response by a shock depends on the connections from Kenyon cells (KC) to mushroom body output neurons (MBONs). Although on the behavioral level conditioning is recognized to be predictive, it remains unclear how MBONs form predictions of aversive or appetitive values (valences) of odors on the circuit level. We present behavioral experiments that are not well explained by associative plasticity between conditioned and unconditioned stimuli, and we suggest two alternative models for how predictions can be formed. In error-driven predictive plasticity, dopaminergic neurons (DANs) represent the error between the predictive odor value and the shock strength. In target-driven predictive plasticity, the DANs represent the target for the predictive MBON activity. Predictive plasticity in KC-to-MBON synapses can also explain trace-conditioning, the valence-dependent sign switch in plasticity, and the observed novelty-familiarity representation. The model offers a framework to dissect MBON circuits and interpret DAN activity during olfactory learning.

Similar content being viewed by others

Rapid Bayesian learning in the mammalian olfactory system

Visualization of learning-induced synaptic plasticity in output neurons of the Drosophila mushroom body γ-lobe

Recurrent architecture for adaptive regulation of learning in the insect brain

Introduction.

Predicting the future from sensory input is fundamental for survival. Co-appearing stimuli can be used for improving a prediction, or for predicting important events themselves, as observed in classical conditioning. In fruit fly odor conditioning, an odor that will become the conditioned stimulus (CS), is paired with the unconditioned stimulus (US), here an electroshock, that triggers an avoidance behavior and in internal representation of a negative value (valence). After conditioning, and the negative value representation—although not the full unconditioned response—previously elicited by the electroshock will be reproduced by the odor itself. Classical conditioning theories posit that throughout learning the odor becomes predictive for the electroshock 1 , 2 , 3 . During learning, the prediction error decreases, and learning stops when the predictive odor value matches the strength of the electroshock.

Predicitve olfactory learning in fruit flies is a widely recognised concept in the experimental literature 4 , 5 , and dopaminergic neurons (DANs) in the mushroom body (MB) have been suggested to predict punishment or reward 6 , 7 , 8 . Yet, despite the acknowledgment of its predictive nature, computational models on fruit fly conditioning are mostly guided by the formation of associations, a notion that relates more to memories rather than predictions (for a recent outline of this controversy see 9 ). Similarly, the concept of predictive learning is well recognized for olfactory conditioning in insects in general, but to our knowledge, synaptic plasticity models are not formulated in terms of explicit predictions, but rather in terms associations and correlations, with plasticity being driven by two or three factors, each representing a temporal nonlinear function of the pre- or postsynaptic activities or of a modulatory signal, sometimes combined with homeostatic plasticity. This type of associative models are exist for fruit flies 10 , 11 , locusts 12 , 13 or honey bees 14 , 15 . They differ from target learning, where the unconditioned stimulus sets a target that is learned to be reproduced by the conditioned stimulus. Target learning becomes predictive learning when including a temporal component. It involves a difference operation, and learning stops when the target is reached. The stop-learning feature is difficult to be reproduced by purely correlation-based associative learning, while a purely predictive model intrinsically captures also associative properties.

Associative learning was suggested to be implemented through spike- or stimulus-timing dependent plasticity (STDP) that would underlay conditioning. STDP strengthens or weakens a synapse based on the temporal correlation between the US (electroshock) and the CS (odor), both on the neuronal time scale of 10’s of milliseconds 13 and on the behavioral time scale of 10’s of seconds 7 , 16 , 17 . Whether an association is strengthened by just repeating the pairing until the behavioral saturation is reached 18 , or the association saturates due to a faithful prediction, however, has not been investigated in the fruit fly so far (Fig. 1 ). Here we show that olfactory conditioning in Drosophila is better captured as predictive plasticity that stops when a US-imposed target is reached, rather than by correlation-based plasticity, such as STDP, that does not operate with an explicit error or a target. According to our scheme, it is only the aversive/appetitive value of the US that is predicted by the CS after faithful learning, not the US itself. Based on the common value representation in the mushroom body output neurons (MBONs, Fig. 1 ), the corresponding avoidance/approach reaction as one aspect of the unconditioned response is elicited by the CS alone.

Associative versus error- or target-driven predictive plasticity. (A) Pairing of an odor (CS) with a shock (US) is typically thought to induce correlation-based synaptic strengthening of the synapses mediating the conditioned response, here from the Kenyon cells (KCs) to the mushroom body output neurons (MBONs). Repeated pairing always leads to a stronger association strength. (B) In predictive coding, plasticity stops when the strength of the US is correctly predicted by the CS. (B1) Plasticity of the KC-to-MBON synapses can be driven by the prediction error ‘US-CS’, formed by the dopaminergic neurons (DANs) that calculate the difference between the internal shock representation and the odor-evoked prediction by the MBONs. (B2) Synaptic plasticity can also be driven by a target for the MBON activity, set to the desired aversive value of the odor (i.e. the shock strength) and represented by the DAN activity. To extend the memory life time of the odor value, the DANs may themselves be driven by the recurrent MBON activity, and the MBON-to-DAN synapses may also be learned via target-driven plasticity to predict the shock strength. Our experiments and models exclude A and suggest further experiments to distinguish ( B1 ) and ( B2 ).

The Drosophila olfactory system represents a unique case for studying associative /predictive learning, and the MB is known to be essential in olfactory learning 18 , 19 , 20 , 21 . The Kenyon cells (KCs) receive olfactory input from olfactory projection neurons and form a sparse representation of an odor 14 , 22 , 23 . The parallel axons of the Kenyon cells (KCs) project to the MB lobes 4 , 24 , along which the compartmentalized dendritic arbors of the MB output neurons (MBONs) collect the input from a large number of KCs. Reward or punishment activates specific clusters of DANs PAM and PPL1, respectively which project to corresponding compartments of the MB lobes 25 , 26 , 27 , modulating the activity of the MBONs and the behavioral response 16 , 28 , 29 , 30 . Recently, a detailed mapping of the MB connectome has been accomplished for larvae and of the vertical lobe for the adult Drosophila 31 , 32 . Several studies show that not only the feedforward modulation from DANs to MBONs, but also the feedback from MBONs to DANs play an important role in olfactory learning 33 , 34 .

Previous studies have given insights into the possible cellular and subcellular mechanisms of olfactory conditioning. Yet, the suggested learning rules 10 , 11 , 12 , 13 , 14 , 17 , 35 remain correlation-based and miss the explicit predictive element postulated by the classical conditioning theories 1 , 2 . Here, we present distinctive conditioning experiments showing that olfactory learning is best explained by predictive plasticity (Fig. 1 ). These experiments, in contrast, could not be reproduced by various types of correlation-based associative learning rules. A mathematical model captures the new and previous data on olfactory conditioning, including trace conditioning. The model encompasses the odor/shock encoding and the learning of the aversive odor value with the stochastic response. We further suggest how the predictive plasticity could be implemented in the MB circuit, with MBONs encoding the value (‘valence’) of the odor stimulus, and DANs calculating either the error or the target that drives the KC-to-MBON plasticity. The predictive plasticity rule for the KC-to-MBON synapses is shown to be consistent with the experimental results showing the involvement of these synapses in the novelty-familiarity representation.

Model of the shock representation and the unconditioned response

Aversive odor conditioning is about learning to evoke the avoidance behavior by the conditioned odor alone, as it is evoked by the electroshock. Before describing the acquisition of the conditioned behavior we characterize the unconditioned behavior of the fruit flies.

In the minimal shock detection experiments, fruit flies in a testing chamber had the choice between moving to either of two arms, with one arm being electrified (with voltage strength S ) and the other not. After 30s we counted the number of flies in the electrified and non-electrified control arm, \(N_\text {electr}\) and \(N_\text {nonel}\) , respectively (the few remaining in the testing chamber not being counted). The empirical performance index (PI) for the pure shock application without conditioning is defined as the relative difference, \(\text {PI} = \frac{ N_\text {nonel} - N_\text {electr} }{ N_\text {nonel} + N_\text {electr} }\) . This empirical PI can be approximated by a theoretical PI that is a function of the stimulus strength S . When the stimulus strength is equal to the minimal strength \(S_\circ\) that just elicits a behavioral response, the PI vanishes, \(\text {PI}(S_\circ )=0\) , and for increasing stimulus strength the \(\text {PI}\) asymptotically tends towards 1. We parametrize

with sensitivity parameter \(\alpha\) telling how steeply PI( S ) grows from 0 at \(S = S_\circ\) towards 1 for large S . We experimentally estimated the size of the minimal shock intensity to be \(S_\circ \approx 7 \,V\) (“ Methods ”).

Shock representation

To explain the behavior as emerging from a neuronal representation we map the shock stimulus to hypothetical neuronal activities. Plasticity will then also be described in terms of these internal activities.

We assume an internal representation, s , of the electroshock following Weber-Fechner’s law 36 ,

where \(S_\circ\) and \(\alpha\) are as introduced in Eq. ( 1 ). For \(S<S_\circ\) we set \(s=0\) . Equation ( 2 ) yields a re-interpretation of the behavioral parameters \(S_\circ\) and \(\alpha\) that characterize the PI in terms of sensory ‘perception’: \(S_\circ\) is the just detectable stimulus strength and \(\alpha\) becomes the linear scaling of the sensory activity.

Unconditioned response

To describe the unconditioned response out of the internal representation, we consider the probability \(p_\text {us}(s)\) of escaping from the shock stimulus (the US) given s . We first note that the PI can be expressed in the form \(\text {PI} = 2p_\text {us}-1\) , with \(p_\text {us} = N_\text {nonel} / (N_\text {nonel} + N_\text {electr})\) being the empirical frequency for an individual fruit fly to move to the non-electrified versus the electrified arm. With the avoidance probability of the form \(p_\text {us}(s)=\frac{1}{1+ e^{-s}} \,\) , the PI becomes a function of the internal shock representation s , \(\text {PI}(s) = 2p_\text {us}(s)-1\) . This is consistent with the definition of \(\text {PI}(S)\) from Eq. ( 1 ) as can be checked by substituting the expression for s given in Eq. ( 2 ) (the performance index as a function of s is \(\text {PI}=(1-e^{-s})/(1+e^{-s})\) ).

As an aside, one may also consider other mappings of the shock strength S to the internal representation s as this is not constrained enough by the data. For instance, one may argue that fruit flies perceive electric shocks following Steven’s power law, as it was originally measured in humans 37 . Steven’s law postulates that the internal representation would have the form \(s=(S/S_\circ )^\alpha\) instead of the logarithmic Weber-Fechner law. From this representation the original behavioral response \(p_\text {us}(s)\) is obtained when the readout from the state is of the form \(p_\text {us}(s)=1/(1+s^{-1})\) . The PI is then calculated according to \(\text {PI}(S) = \text {PI}(s) = 2p_\text {us}(s) -1 = (1-s^{-1})/(1+s^{-1})\) .

Odor conditioning depends on the temporal shock distribution

We next turn to the odor conditioning. It was previously investigated how the associative strength of the conditioned odor increases with the strength of the paired electroshock and the number of pairings, while saturating at some level, with respect to both the shock strength and the paring repetition 18 . We asked whether these saturation effects originate from behavioral limitations, or whether they originate from a quick and faithful learning of the intrinsic value of the electroshock strengths.

Temporal sequence conditioning is not fit by Hebbian plasticity. (A) Experimental protocol as explained in the main text, with total shock strength of \(100\,\) V distributed in time across the \(60\,\) s exposure to the conditioning odor (alignment towards the end, A1 , and towards the beginning, A2 ). The duration of an individual voltage shock (a magenta vertical bar) is \(1.5\,\) s. (B) Experimental results showing the LI for the corresponding shock distributions in A1 and A2 , respectively. For shocks at the end, the LI decreases with decreasing individual shock strength. Error bars represent the standard error of the mean (SEM). (C) Fit by the Hebbian plasticity, Eq. ( 6 ). For shocks at the end, a roughly constant LI is produced. For the optimized parameters we extracted \(S_\circ =7\,\) V from the minimal shock detection experiment (Fig. 2-1), \(\tau _o=15\,\) s from trace conditioning experiments 28 , and optimized the product \(\alpha \eta =0.0723\) .

To address this question, we differently packaged the total of 100V into 1*100V, 2*50V, 4*25V and 8*12.5V electroshocks and asked whether the repeated smaller shock strengths (8*12.5V) would lead to premature saturation that would then not be caused by behavioural limitations but rather by some dedicated learning behavior. We distributed these shock packages across the \(60\,\) s odor presentation time (Fig. 2 A, “ Methods ”), and let the fruit flies choose during \(120\,\) s between a conditioned and neutral odor. The learning index (LI) that characterizes the conditioned response is defined analogously to the PI by the relative number of fruit flies that choose the unconditioned control odor ( \(N_\text {CS-}\) , more precisely, the odor that was conditioned with zero shock strength) versus the conditioned odor ( \(N_\text {CS+}\) ), i.e.

The LIs gradually decreased with decreasing electroshock strength if the shocks were applied towards the end of the odor presentation time, and the additional repetitions of the weaker electroshocks could not revert this trend (Fig. 2 B1). Yet, when the same shocks were distributed towards the beginning of the odor presentation time, the LIs remained small, with a tendency to increase with decreasing electroshock strength (Fig. 2 B2).

The avoidance behavior depends in a complicated way on the shock strengths and the shock timings. To explain these behaviors we next formalize the value representation, the decision making, and two different types of plasticity models.

From internal value representation to stochastic responses

The basic observation of conditioning is that, after long enough conditioning time, the conditioned behavior eventually mimics the unconditioned behavior. In our model this implies that the learning index LI converges to the performance index PI (Eq. 1 ). We assume that at any moment in time, a presented odor elicits some activity o in the KCs that reflects the odor intensity. In the experiments, an odor was either present or absent, and hence \(o(t)=1\) or 0. The considered MBON activities are assumed to represent the aversive value ( v ) of the odor. As MBONs are driven by KC, we postulate that the MBON activity takes the form

where w is the synaptic strength (weight) from the KCs to the MBONs.

Model of stochastic action selection

The conditioned response upon odor stimulus appears to be stochastic for an individual fruit fly. It is therefore modeled in terms of the avoidance probability that itself depends on the MBON activity. For simplicity we postulate that this avoidance probability in response to the conditioned stimulus (CS, the odor), has the same form as the one to the unconditioned stimulus introduced above, \(p_\text {us}(s)\) , but with s replaced by v ,

As it is for the PI, the LI can be expressed in terms of this avoidance probability, \(\text {LI}(v) = 2p_\text {cs}(v)-1\) . Remember that fruit flies may remain in the test tube (estimated to be less than \(5\%\) ) and that the LI is calculated based on the fruit flies that effectively moved to one of the two chambers (Eq. 3 ). Hence, the interpretation of \(p_\text {cs}(v)\) on the level of the individual fly is, strictly speaking, the conditional probability that, given the fly ‘decides’ to move, it actually moves away from the conditioned odor.

The model postulates that the decision for each individual fruit fly is a stochastic (Bernoulli) process that only depends on the current MBON activity \(v\!=\!wo\) , and in particular does not depend on previous decisions. In fact, when re-testing the population of fruit flies that escaped from the odor in a first test trial (a fraction \(p_\text {cs}\) of the overall test population), the same fraction \(p_\text {cs}\) of this sub-population escaped again in a second test trial, despite the putative extinction of memory caused by the first test (see the Test-Retest experiment in 38 ). Intriguingly, when waiting \(24\,\) h so that the first conditioning was forgotten, conditioning the successfully escaped and the unsuccessfully non-escaped flies from the first conditioning experiment separately again, the same LI was achieved by both groups. This shows that not only a single response is stochastic, but also the learning (see again 38 , cross-checked by us for a \(8 \times 12.5 V\) stimulation, results not shown). A statistical evaluation of the model with the same number of flies ( \(N_\text {fly}\) ) and trials ( \(N_\text {trial}\) ) as in the experiment gives equal or smaller variance in the LI of the model fruit flies as compared to the experiment (Fig. 3 A1). This implicitly quantifies additional sources of stochasticity in the experimental setup or in the individual fruit fly that have already been described in honeybees 39 , 40 and that go beyond our 1-state stochastic Markov model.

Predictive plasticity captures the sequence conditioning experiments. (A) In contrast to the purely associative learning (Eq. 6 ), the predictive plasticity (Eq. 7 ) fits well the shock-at-end data ( A1 ) and the shock-at-beginning data ( A2 , see Fig. 2 ). Error-bars represent SEM for both data and model. In the model, stochasticity enters through the Bernoulli process according to which each of the model flies chooses to avoid (with probability \(p_\text {cs}(v)\) , Eq. ( 5 )) or approach (with probability \(1-p_\text {cs}(v)\) ) the odor. The same number of flies, \(N_\text {fly}\) , and the same number of trials, \(N_\text {trial}\) , was used in the model as in the experiment. (B) Traces for odor ( o ), shock ( s ), and synaptic strength ( w ) for the odor-to-shock prediction during conditioning, for the shock-at-end ( B1 ) and shock-at-beginning ( B2 ) protocols. Note that between the shocks, w decays towards 0 as the target for \(v=wo\) is \(s=0\) according to the predictive learning rule. The weight does not change if the prediction matches the shock representation, \(w\,o=s\) , e.g. when both are 0. The optimized parameters are: \(S_\circ =6.90\,\) V, \(\alpha =0.79\) , \(\tau _o=14.25\,\) s, \(\Delta \eta = 0.057\,\) , \(\tau _{\eta }=133.48\,\) s, with mean square error \(\text {MSE} = 6.393 \times 10^{-4}\) across all experiments (including the ones below).

The model of stochastic action selection expressed by Eq. 5 assumes that there is only one stimulus type present, either the odor (CS) or the shock (US), and the odor triggers the avoidance reaction with probability \(p_\text {cs}(v)\) , and the shock with probability \(p_\text {us}(s)\) . The experiment may also be setup such that in one arm of the test chamber the CS and in the other the US is present, and the fruit fly can decide whether to move at all or not, for instance, as studied in 41 . In this case the probability of moving in neither of the two arms depends on the difference between the CS- and US-induced value, \(p_\text {cs,us}(v,s) = \frac{1}{1 + e^{-|v-s|}}\) , and this probability may be represented downstream of the MB, as also suggested in 41 . Alternatively, the DAN may represent the US and the MBON may depend on both the CS and US, along the lines of the wiring scheme for the target-driven predictive plasticity outlined below (Discussion with Figs. 1 B, 6 B).

Associative learning models do not fit the conditioning data

Learning is suggested to arise from appropriately modifying the strength w of the KC-to-MBON synapses. The synaptic modification affects the aversive value of the odor following the linear relation \(v=wo\) (Eq. 4 ), and this determines the conditioned response given by the escape probability \(p_\text {cs}(v)\) , see Eq. ( 5 ).

The common conception of conditioning is that the associative strength, w , is changed proportionally to some nonlinear functions of the pre- and postsynaptic activities, possibly modulated by a third factor. To exemplify the essence of associative learning, although this may not do justice to the more complex cited models, we consider a simplified version where the synaptic weight change is proportional to both the strengths of the unconditioned and the conditioned response,

with proportionality factor \(\eta\) defining the learning rate (cf. Fig. 1 A). Here, \(\tilde{o}\) is the low-pass filtered odor o that follows the dynamics \(\,\tau _o \frac{d\tilde{o}}{dt} = - \tilde{o} + o\,\) , with a time constant \(\tau _o\) being on the order of ten seconds. It can be interpreted as a presynaptic eligibility trace that keeps the memory of the presynaptic activity, here the odor o , to be associated with the postsynaptic quantity, here the shock representation s .

This simple Hebbian rule (Eq. 6 ) is not able to fit the sequential conditioning data. In fact, for the shocks at the end, the Hebbian model roughly shows the same LI for the weak and strong stimuli, as it were the total stimulus strength that would count (Fig. 2 C1). The concavity of the logarithmic shock representation by itself would rather favor an increasing LI for the repeated weaker stimuli 8*12.5V as compared to the 1*100V.

We considered a perhaps oversimplified Hebbian learning rule, \(\dot{w} \propto s\tilde{o}\) , as one example of associative plasticity. To consider more sophisticated associative learning rules, we define synaptic weight changes that are functions of the correlation between odor- and shock-induced activity. We also tested these more general forms of associative learning that are based on linear and nonlinear functions of CS-US correlations, such as stimulus-timing dependent synaptic plasticity (STDP) of the form \(\dot{w} = \eta _1 s\tilde{o} - \eta _2 o \tilde{s}\) , with \(\tilde{s}\) being the low-pass filtered s and \(\eta _i\) arbitrary scaling factors. STDP, even after introducing nonlinearities, and also the covariance rule of the form \(\dot{w} = \eta (s - \tilde{s})(o - \tilde{o})\) , did all give roughly a 10 times worse fit (in terms of the \(\text {MSE}\) , “ Methods ” and Fig. 3 -1) than predictive plasticity explained next.

Model of predictive plasticity

The failure of associative learning rules in reproducing our conditioning experiments can be corrected by adding an anti-Hebbian term of the form \(-v\tilde{o}\) to the rule \(\dot{w} \propto s\tilde{o}\) , leading to

We interpret this combined Hebbian/anti-Hebbian plasticity rule as error correcting, with the difference between the shock and odor representation, \(s-v\) , as internal error.

This error-correction learning rule has a long history in the theory of neural networks where it first appeared as Widrow-Hoff rule 42 that was extended to a temporal difference rule 43 , and recently reinterpreted in terms of dendritic prediction of somatic firing 44 , 45 . It also relates to the predictive rule of Rescorla-Wagner 1 previously applied to explain various fruit fly conditioning experiments 46 , although without considering a time-continuous learning scenario and the related temporal aspects. According to this rule, learning stops when the aversive value of the odor, v , predicts the internal representation of the shock stimulus, \(v=s\) . During predictive learning, when the synaptic eligibility trace is active, \(\tilde{o}(t)\!>\!0\) , the synaptic strength w is adapted such that the odor value converges to the internal shock representation, \(v(t) = wo(t) \longrightarrow s(t)\) , with \(s(t)\!>\!0\) when the electroshock-voltage is turned on and \(s(t)\!=\!0\) else. Correspondingly, the conditioned response converges to the unconditioned response, \(p_\text {cs}(v) \longrightarrow p_\text {us}(s)\) . Crucially, during the time when the US is absent, \(s\!=\!0\) (while \(\tilde{o}\!>\!0\) ), a neutral response is learned. On a behavioral level this appears as forgetting the shock prediction, and it also relates to the phenomenon of extinction in classical conditioning 1 .

To fit the conditioning experiments with ongoing electroshock-voltage we need to consider a learning rate that adapts in time. Learning speeds up when the strength of the voltage increases. A stronger voltage triggers initially a higher learning rate that, with ongoing voltage stimulation, decays with a time constant \(\tau _\eta\) on the order of \(2\,\) min. A stepwise increase of s by \(\Delta s\) (as it appears at the onset of an electric shock) leads to a stepwise increase of the initial learning rate \(\eta\) by \(\Delta \eta \, \Delta s\) for an optimized parameter \(\Delta \eta\) (“ Methods ”).

In contrast to the pure associative rule, the predictive rule (Eq. 7 ) qualitatively and quantitatively reproduces the conditioning experiments (Fig. 3 ). With the predictive learning rule, the 1*100V pairing at the end of the odor presentation elicits the strongest conditioned response, while the response is much weaker after the distributed 8*12.5V pairing, as also observed in the experiment. The reason is that the synaptic weight w decreases between the shocks while the odor is still present (green traces in Fig. 3 ). As in the extinction experiments, the presence of the CS alone leads to the prediction that no US is present, and hence to an unlearning of the previously acquired US prediction.

Repetitive and ongoing conditioning reveals its predictive nature

To bolster our hypothesis that olfactory conditioning in the fruit fly is predictive rather than associative, we further tested the model to repetitive and continuously ongoing odor-shock pairings. If the hypothesis is correct, during repeated or extended pairing, learning should in both cases stop when shock strength is correctly predicted by the odor. In particular, the learning performance is expected to saturate at a level below the maximally achievable performance. This is in fact what we observed.

When repeating the previously described block of 4*25V conditioning shocks with 15s inter-shock-intervals, the LI showed a saturation after a single block (Fig. 4 A,B). When conditioning with half of that block, i.e. with only 2*25V conditioning shocks in 15s, roughly \(70\%\) of the saturation level is reached. The same repetition experiment was performed with 4*50V pulses, confirming that also for a stronger US the LI quickly saturated (Fig. 4 C). Again, neither the pure associative rule, nor the covariance rule or the more sophisticated STDP rules, could reproduce this data (Fig. 3-1).

Repetitive and ongoing conditioning is captured by the predictive plasticity. (A) The repeated training consisted of 1, 2, 4 repetitions of a standard training block (dark red bar, as in Fig. 3 A, A1) composed of 4*25V shocks, followed by a break and a control period. Half of a training block was considered with 2*25V shocks towards the end of the 60s odor presentation (yellow bar). (B) The LI saturates after a full block (1 Repetition, gray), as also reproduced by the predictive plasticity model (green). (C) The same protocol with the same number of shocks as in (B), but with 50V instead of 25V shocks. A second training repetition did only slightly increase the LI and for further repetitions it again remains constant. This is reproduced by the predictive plasticity, but not by the various associative plasticity models (see Supplementary Materials). (D) Protocol of ongoing odor-shock pairing, with voltages turned on during the full odor presentation time of 10s, 15s, 30s, 45s, 90s and 120s, both for 25V and 50V. (E) The LI for the time-continuous pairing saturates with a time constant of roughly 20s for the 50V and 30s for the 25V odor-voltage pairings. Predictive plasticity captures this saturation, with \(\text {LI}(v)\) converging towards \(\text {LI}(s)\) (dashed lines, Eq. ( 8 )), for both the 25V and the 50V pairings.

An even more challenging test for the predictive learning rule is an odor-shock pairing where the electric voltage (either 25V or 50V) is turned on throughout the odor presentation time, from 10s up to 120s. After roughly \(1\,\) min of ongoing pairing the LI saturated, both in the data and the model (Fig. 4 D). In the model, learning saturates when the value v of the odor correctly predicts the shock, \(v=s\) , as expressed by a successful predictive learning (i.e. when learning ceases, \(\dot{w} = 0\) , see Eq. ( 7 )). During learning, when the value of the odor converges to the shock representation, \(v \rightarrow s\) , the LI converges to the PI (as defined in Eqs 3 and 1 ),

The equation is obtained from substituting v by s in the expression for \(p_\text {cs}(v)\) , Eq. ( 5 ), and making use of Weber-Fechner’s law translating the shock strength S into the internal representation s (Eq. 2 ). For our simple predictive plasticity model the exposure time to acquire the final performance can be explicitly calculated, and it is shorter for stronger ongoing voltage stimuli (Fig. 4-1).

Trace conditioning is also predictive

Odor conditioning has also been studied in the form of trace conditioning (e.g.  28 , 47 ). A further test of our model is to apply it to these experiments, with the same parameters found to fit our data from Figs. 3 and 4 . In trace conditioning, the electroshock is applied with a variable inter-stimulus-interval (ISI) after the onset of the odor presentation, and this ISI can even extend beyond the presentation time of the odor (Fig. 5 A). We considered the experimental protocol with \(10\,\) s odor presentation and an ISI varying from 5 to 30s, after which 4 conditioning electroshocks of 90V were applied with \(0.2\,\) Hz 28 . The LI gradually decreased with the length of the ISI, with a decay time of roughly \(15\,\) s. The model captures this phenomenon because the odor trace, entering as synaptic eligibility trace ( \(\tilde{o}\) ) in the predictive plasticity rule, is still active for a while after the odor has been cleared up (Fig. 5 B). The identical set of 5 parameters has been used that were extracted from the previous experiments ( \(S_\circ\) , \(\alpha\) , \(\tau _o\) , \(\Delta \eta\) , \(\tau _{\eta }\) , see caption of Fig. 3 ).

Trace conditioning is faithfully reproduced. (A) Experimental protocol of trace conditioning, with variable Inter-Stimulus Intervals (ISIs) from the onset of the odor (green) to the onset of the electroshock train ( \(4*90\) V, 0.2Hz, each 1.25s, red bars). (B) The LI tested immediately after the conditioning with the different ISIs (reproduced from 28 ). The model roughly captures this data (green line) without additional fitting of the parameters.

MB circuits for error- or target-driven predictive plasticity

Based on anatomical connectivity patterns and previous plasticity studies we suggest two forms of how the predictive learning may be implemented in the recurrent MB circuit, via error- and target-driven predictive plasticity (Fig. 1 B). In both versions, learning is mainly a consequence of modifying the KC-to-MBON synaptic boutons 4 , 30 , 33 , 48 , 49 , but the role of the DANs is different. While the KC-to-MBON connections drive the MBONs based on the odor representation in the KCs, the shock information is provided by the DANs and gates the KC-to-MBON plasticity (see also 4 , 7 , 24 , 25 , 27 ). The DANs themselves may either represent the error or the target for KC-to-MBON plasticity.

Suggested implementation of error- and target-driven predictive plasticity. (A1) Mushroom body circuits for olfactory error-driven predictive plasticity. Kenyon cells (KCs) carrying the odor information project to mushroom body output neurons (MBONs) through synapses encoding the aversive value ( v ) of the odor. The input triggered by the electroshock, s , drives the dopaminergic neurons (DANs) that are also inhibited by the MBONs. The DANs represent the prediction error, \(e=s-v\) , and modulate the KC-to-MBON synapses according to \(\dot{w} \propto e\,\tilde{o}\,\) , with \(\tilde{o}\) representing the odor eligibility trace. The conditioned response probability ( \(p_\text {cs}\) , avoidance reaction) is a function of v . (A2) Neuronal activity traces of a DAN ( e , magenta), a KC ( o , light green) and a MBON ( v , dark green), shown at the onset of an odor-shock pairing (‘Pairing onset’, full triangle), \(20\,\) s later (During), and later at the test when only the odor is presented (‘Test’, open triangle, cf. Eq. 9 ). The aversive value v steadily increases (dark green), while the prediction error, e , decreases throughout learning and becomes negative when the odor is presented alone (purple). (B1) Mushroom body circuit for target-driven predictive plasticity. Beside the shock stimulus, the DANs can also indirectly be excited by the MBONs (or directly by the KC, not shown) to form a shock prediction also in the DANs and prevent fast extinction. The shock stimulus ( s ) sets the target for the MBON-to-DAN plasticity, and the DANs ( d ) set the target for the KC-to-MBON plasticity (cf. Eq. 11 ). (B2) As in A2 , but since the DANs now form a prediction of the shock itself based on v , their activity increases throughout learning, and they are also activated during the Test, when the conditioned odor is presented alone (Eq. 10 ). Sketch adapted from 33 and 6 that favor excitatory feedback to the DANs as in version B.

In the first implementation (error-driven predictive plasticity), the DANs themselves represent the prediction error \(e = s - v\,\) . They may extract this error from the excitatory shock input, s , and the inhibitory MBONs feedback providing the aversive value v of the odor ( 33 , see Fig. 6 A1). The modeling captured the effect of learning on the behavioral time scale. To predict specific activity traces in the MB on a fine-grained temporal resolution we introduce the dynamics of the MB neurons. In the case of the DANs as error representation, the firing rates of the MBONs ( v ) and DANs ( e ) is given by

with a neuronal integration time constant \(\tau\) in the order of \(10\,\) ms (Fig. 6 B1). The plasticity of the synapses from the KCs to the MBON is then driven by the DAN-represented prediction error e at any moment in time, \(\dot{w} =\eta \, e\, \tilde{o}\,\) , consistent with the predictive plasticity rule (Eq. 7 ). Note that in the steady state, the DAN activity exactly represents the difference between the shock strength and its odor-induced prediction, \(e=s-v\) . After successful learning, the MBONs accurately match the shock representation and the DAN activity vanishes, \(v=s\) and \(e=0\) .

In the alternative implementation (target-driven predictive plasticity), the DANs provide the learning target to the KC-to-MBON synapses while themselves being driven by the MBONs (Fig. 6 A2). These MBON-to-DAN synapses are also plastic and learn to predict the shock stimulus, just as the KC-to-MBON synapses do. A benefit of this recurrent prediction scheme is that the memory life time of the odor-shock prediction is extended. If after successful learning the odor is presented alone, the target for the KC-to-MBON plasticity is still kept at the original level via MBON-to-DAN feedback, and extinction of the shock memory slows down. The recurrent circuitry between MBONs ( v ) and DANs (with activity d instead of e to indicate that the DANs no longer represent the error but the target for the MBON learning) now becomes

Here, \(w_\text {MK}\) and \(w_\text {DM}\) are the synaptic weights of KC-to-MBON and MBON-to-DAN synapses, respectively, and \(\lambda =0.1\) is the nudging strength of the postsynaptic teaching signal 44 ). Both KC-to-MBON and MBON-to-DAN synapses follow the same form of error-correcting plasticity as in Eq. ( 7 ),

where the DAN activity d now serves as the target for KC-to-MBON synapses, while the shock stimulus s is the target for MBON-to-DAN synapses.

After successful learning, the activity of MBONs and DANs both predict the shock stimulus, \(v=d=s\) (as derived from the steady states of Eq. 11 , see also Fig. 6 B2). If the shock stimulus is absent ( \(s\!=\!0\) ) during the presentation of the conditioned odor o , and the odor was previously conditioned to a shock strength \(s_\circ\) while the DAN activity was fully learned (implying \(w_D=1\) ), the MBON activity, supported by the recurrent DAN activity, becomes \(v = \frac{1-\lambda }{1-\lambda (1-\lambda )} s_\circ \approx 0.99 s_\circ\) (as derived from the steady states of Eq. 10 with \(\lambda =0.1\) , see Fig. 6 B2, column ‘Test’). Hence, the value of the odor faithfully predicts the conditioned shock strength also in this target-driven learning circuitry. Note that in the target-driven plasticity the KC-to-MBON plasticity \(\dot{w}_M\) does not directly relay on the MBON activity since the activity target is imposed by the DAN’s, not by the MBON’s (Eq. 11 ).

Outlook: valence learning and novelty-familiarity representation

The concept of predictive learning can be extended to valence learning where each MBON represents a positive or negative valence, \(v^\pm\) , coding for an appetitive and aversive value of a stimulus, respectively 24 , 30 , 50 , 51 . For each valence, a specific cluster of DANs is involved in the sensory representation, PAM for positive and PPL1 for negative valences 4 , 52 . In the full MB circuit the DANs further receive excitatory drive from the KCs ( 32 , dashed connection in Fig. 6 , here abbreviated by \(w_\text {DK}\) ), and the feedback circuit modulates the plasticity of the KC-to-MBON synapses 48 . The activities of the two valence classes of DANs can be modeled as in Eq. 10 , but with multimodal input from the unconditioned appetitive or aversive stimuli ( \(s^\pm\) ) and the odor representation in the KCs ( \(w_\text {DK} o\) ). Together with the feedback from the corresponding MBONs via weights \(w_\text {DM}\) , and introducing a saturating nonlinear transfer function \(\phi\) , the DAN activities for the two valence clusters become

Plasticity in MBONs is known to be sign flipped when changing the valence of the stimulus 7 , 30 , 52 . This can be captured in the predictive plasticity model by imposing 0 as target when the stimulus and MBON valence do not match. For positive valence MBONs, the target can be set to \(d = d^+(1 - d^-)\) , assuming that the DAN activities are restricted to the range between 0 and 1; for negative valence MBONs the target is \(d^-(1-d^+)\) . When a previously appetitively conditioned odor is now presented ( \(w_\text {MK} o > 0\) for a positive valence MBON), together with a shock ( \(d=0\) ), the postsynaptic error term in the learning rule now becomes negative, \((d - w_\text {MK} o) < 0\) , and the synapses get depressed rather than potentiated as in the first conditioning (Eq. 11 ).

The sign of the KC-to-MBON plasticity can also be changed in other ways. It has been shown that the familiarization to odors can depress MBON responses (in the \(\alpha '3\) compartment), while the response to previously familiarized stimuli is recovered 49 . To explain this phenomenon we extend the predictive plasticity to involve a partial redistribution of the total synaptic strength across the KC-to-MBON synapses, formally expressed by

where we introduced a down-shift in the presynaptic term by the mean odor that exceeds the spontaneous activity level, \(\overline{\tilde{o}} = \frac{1}{n_\text {K}} \sum _{j=1}^{n_\text {K}} \tilde{o}_j - o_\circ\) . Here, the average is across all \(n_\text {K}\) Kenyon cell synapses, and we assume a spontaneous but sparse KC activity \(o_\circ\) such that in average the activity of KC i satisfies \(o_i \ge o_\circ\) 53 . The spontaneous KC maps to the eligibility trace that is strictly positive, \(\tilde{o}_i \ge o_\circ\) , and some spontaneous DAN activity \(d_\circ\) inherited from the KCs, such that \(d \ge d_\circ\) . Because (PPL1 \(-\alpha '3\) ) DAN activity is necessary to observe repetition suppression 49 , we postulate that the learning rate is modulated by the DAN activity, \(\eta = d \eta _\circ\) , for some base learning rate \(\eta _\circ\) .

The various plasticity features of the KC-to-MBON synapses investigated in 49 are consistent with the extended learning rule in Eq. 13 . Repeated odor-evoked KC activation causes synaptic depression, assuming that odors are dominantly activating KCs and MBONs, but less so DANs, \(d < w_\text {MK} o\) (here enters the saturation of the nonlinearity \(\phi\) in Eq. 12 ), leading to the observed repetition suppression ( \(\dot{w}_{\text {MK},i} < 0\) ) and explaining the behavioral familiarization of the flies to odors. The repetition suppression may depress the KC-to-MBON synapses such that in response to spontaneous KC activity ( \(o_\circ )\) the MBON activity is now smaller than the spontaneous DAN activity, \(w_\text {MK} o_\circ < d_\circ\) . In the absence of an odor, the depressed KC-to-MBON synapses will therefore recover due to the spontaneous KC activity (Eq. 13 ), such that eventually the equilibrium is reached again when the spontaneously induces MBON activity matches the spontaneous DAN activity, \(w_\text {MK} o_\circ = d_\circ\) . This explains the ‘passive’ recovery of the MBON responses after odor familiarization 49 .

Further experimental investigations of the KC-to-MBON plasticity shows that optogenetically activating DANs alone potentiates the synapses. In our model this DAN-induced potentiation arises since for the isolated optogenetic DAN activation we have to assume that \(d > w_\text {MK} o\) , and the presynaptic term in the plasticity rule (Eq. 13 ) is positive in average due to the spontaneous KC activity, \((\tilde{o}_i - \overline{\tilde{o}}) = o_\circ >0\) . Next, if we assume that the optogenetic co-activation of MBONs ( \(v>0\) ) and DANs ( \(d>0\) ) applied in 49 is such that \(d < v=w_\text {MK} o\) (but with an increased learning rate \(\eta = d \eta _\circ\) ), then the KC-to-MBON synapses get depressed, as reported from the experiment.

Finally, due to the partial weight redistribution, the repetition suppression during the familiarisation to a new odor implies the potentiation of the other synapses that are not activated, among them most of the previously suppressed synapses that were involved in the representation of a preceding odor. The reason for this heterosynaptic potentiation in our model is that repetition suppression is caused by a negative postsynaptic factor \((d - w_\text {MK} o)<0\) in Eq. 13 as explained above, implying the depression of an active synapses i for which \((\tilde{o}_i - \overline{\tilde{o}}) > 0\) , but also implying the potentiation of not activated synapses, since for those \((\tilde{o}_i - \overline{\tilde{o}}) < 0\) and hence \(\dot{w}_{\text {MK},i} >0\) . This odor-induced potentiation in other synapses explains the ‘active’ recovery from the repetition suppression as seen in the experiment 49 . Technically, \((\tilde{o}_i - \overline{\tilde{o}}) < 0\) holds for not activated synapses only if we assume that the odor-evoked average activity in the KCs is well above the spontaneous activity level such that \(\overline{\tilde{o}} > o_\circ\) .

Predictive, but not correlation-based plasticity, reproduces experimental data

We reconsidered classical odor conditioning in the fruit fly and presented experimental and modeling evidence showing that olfactory learning, also on the synaptic level, is better described as predictive rather than associative. The key observation is that repetitive and time-continuous odor-shock pairing stops strengthening the conditioned response after roughly 1 minute of pairing, even if the shock intensity is below the behavioral saturation level. During conditioning, the odor is learned to predict the co-applied shock stimulus. As a consequence, the odor-evoked avoidance reaction stops strengthening at a level that depends on the shock strength, irrespective of the pairing time beyond \(1\,\) min. We found that associative synaptic plasticity, defined by a possibly nonlinear function of the CS-US correlation strength, as suggested by STDP models, fails to reproduce the early saturation of learning.

We suggest a simple phenomenological model for predictive plasticity according to which synapses change their strength proportionally to the prediction error. This error is expressed as a difference between the internal shock representation and the value representation of the odor. The model encompasses a description of the shock and value representation, the stochastic response behavior of individual flies, and the synaptic dynamics (using a total of 5 parameters). It faithfully reproduces our conditioning experiments (with a total of 28 data points from 3 different types of experiments) as well as previously studied trace conditioning experiments (without need for further fitting). As compared to the associative rules (Hebbian, linear and nonlinear STDP, covariance rule), the predictive plasticity rule obtained the best fits with the least number of parameters. We further compared the model by the Akaike information criterion that considers the number of parameters beside the fitting quality. This criterion yields a likelihood for the predictive plasticity rule to be the best one that is at least 7 orders of magnitude larger as compared to the other four associative rules we considered (see Table 1 , “ Methods ”, and Fig. 3 -2).

Error- versus target-driven predictive plasticity

The same phenomenological model of predictive learning may be implemented in two versions by the recurrent MB circuitry. In both versions the MBONs code for the odor value (‘valence’) that drives the conditioned response. For the error-driven predictive plasticity, the DANs directly represent the shock-prediction error by comparing the shock strength with its MBON estimate, and this prediction error modulates the KC-to-MBON plasticity (Figs. 1 B1 and 6 A). For the target-driven predictive plasticity, the DANs represent the shock stimulus itself that is then provided as a target for the KC-to-MBON plasticity. In this target-driven predictive learning, the DANs may also learn to predict the shock stimulus based on the MBON feedback, preventing a fast extinction of the KC-to-MBON memory (Figs. 1 B2 and 6 B).

Predictive plasticity for both types of implementation has its experimental support. In general, MBON activity is well recognized to encode the aversive or appetitive value of odors and to evoke the corresponding avoidance or approach behavior 4 , 24 , 30 , 54 , 55 , while KC-to-MBON synapses were mostly shown to undergo long-term depression, but also potentiation (see e.g. 50 ). DAN responses are shown to be involved in both the representation of punishment and reward 6 , 7 , 26 , 56 that drive the aversive or appetitive olfactory conditioning 7 . This conditioning further involves the recurrent feedback from MBONs to DANs that may be negative or positive 33 , 50 , see 5 for a recent review. Moreover, the connectome from the larvae and adult fruit fly MBON circuit reveals feedback projections from DANs to the presynaptic side on the KC and the postsynaptic side on the MBONs at the KC-to-MBON synaptic connection 31 , 32 , giving different handles to modulate synaptic plasticity.

With regard to the specific implementations, the error-driven predictive plasticity is consistent with the observation that DAN activity decreases during the conditioning 49 , 51 . The two models have opposite predictions for learning while blocking MBON activity. The error-driven predictive plasticity would yield a higher LI, similarly as observed in 54 , while the target-driven predictive plasticity would yield a lower LI, similarly to 24 . It was also shown that some DANs increased their activity with learning while other DANs, in the same PPL1 cluster that is supposed to represent aversive valences, decreased their activity 51 . In fact, error- and target driven predictive plasticity may both act in concert to enrich and stabilize the representations. As shown in Fig. 6 , DAN activity would decrease in those DANs involved in error-driven and increase in those involved in target-driven predictive plasticity.

While error-driven predictive plasticity offers access to an explicit error representation in DANs, target-driven predictive plasticity has its own merits. If DANs and MBONs code for similar information, they can support a positive feedback-loop to represent a short-term memory beyond the presence of an odor or a shock, as it was observed for aversive valences in PPL1 DANs 6 and for appetitive PAM DANs 33 . A positive feedback-loop between MBONs and DANs is further supported by the persistent firing between these cells after a rejected courtship that may consolidate memory of the rejection, linked to as specific pheromone 8 , 57 .

Distributed learning, memory life-time and novelty-familiarity coding

Target-driven plasticity has further functional advantages in terms of memory retention time. Any odor-related input to the DANs, arising either through a forward hierarchy from KC 48 or a recurrence via MBONs to the DANs 6 , 33 , will extend the memory life-time in a 2-stage prediction process: the unconditioned stimulus ( s ) that drives the DAN activity ( d ) to serve as a target for the value learning in the MBONs via KC-to-MBON synapses ( \(v=w_\text {MK}o\) ), will itself be predicted in the DANs (see Eq. 11 ). Extending the memory life-time through circuit plasticity might be attractive under the light of energy efficiency, showing that long-term memory in a synapse involving de novo protein synthesis can be costly 8 , 58 , while cheaper forms of individual synaptic memories likely have limited retention times. Moreover, distributed memory that includes the learning of an external target representation offers more flexibility, including the regulation of the speed of forgetting 45 .

Target-driven predictive plasticity may also explain the novelty-familiarity representation observed in the recurrent triple of KCs, DANs and MBONs 49 . The distributed representation of valences allows for expressing temporal components of the memories. Spontaneous activity in the KCs and their downstream cells 53 injures a minimal strength of the KC-to-MBON synapses through predictive plasticity. A novel odor that drives KCs will then also drive MBONs and, to a smaller extent (as we assume), also DANs. If the DANs that represent the target for the KC-to-MBON plasticity are only weakly activated by the odor, the KC-to-MBON synapses learn to predict this weaker activity and depress. The depression results in a repetition suppression of MBONs and the corresponding familiarization of the fly to the ongoing odor. However, when the odor is cleared away, the MBON activity induced by spontaneously active KCs via depressed synapses now becomes lower than the spontaneous DAN activity, and predictive plasticity recovers the original synaptic strength. Eventually the spontaneous MBON and DAN activites match again (Eq. 13 ) and the response to the originally novel odor is also recovered, as seen in the experiment 49 .

Olfactory learning is likely distributed across several classes of synapses in the MB. The acquisition of olfactory memories was shown to be independent of transmitter release in KC-to-MBON synapses, although the behavioral recall of these memories required the intact transmission 59 . In fact, learning may also be supported by plasticity upstream of the MBONs such that the effect of blocking KC-to-MBON transmission during learning is behaviorally compensated. Predictive plasticity at the KC-to-MBON synapses requires the summed synaptic transmissions across all synapses in the form of the value \(v=w o\) to be compared with the target d , also during the memory acquisition. This type of plasticity would therefore be impaired by blocking the release.

Distributed learning and absence of blocking

Distributed learning also offers flexibility in acquiring predictions from new cues. While the original Rescorla-Wagner rule would predict blocking 1 , this has not been observed in the fruit fly 46 . Blocking refers to the phenomenon that, if the first odor of a compound-CS is pre-conditioned, the second odor of the compound will not learned to become predictive for the shock. Because our predictive plasticity rules are expressed at the neuronal but not at the phenomenological level, predictions about blocking will depend on the neuronal odor representation. If the two odors activate the same MBONs, blocking would be observed since the MBONs are already driven to the correct value representation by the first odor. If they activate different MBONs, however, blocking would not be observed since the MBONs of the second odor did not yet have the chance to learn the correct value during the first conditioning. Hence, since blocking has not been observed in the fruit fly, we postulate that the odors of the compound-CS in these experiments were represented by different groups of MBONs.

Concentration-specificity and relieve learning

How does our model relate to the concentration-specificity and the timing-specifity of odor conditioning? First, olfactory learning was found to be specific to the odor concentration, with different concentrations changing the subjective odor identity 60 . The response behavior was described to be non-monotonic in the odor intensity, with the strongest response for the specific concentration the flies were conditioned with. It was suggested that this may arise from a non-monotonic odor representation in the KC population as a function of odor intensity 35 , 61 . Given such a presynaptic encoding of odor concentrations, the predictive olfactory learning in the KC-to-MBON connectivity would also inherit the concentration specificity from the odor representation in the KCs. Our predictive plasticity, and also the Rescorla-Wagner model, further predicts that learning with a higher odor concentration (but the same electroshock strength) only speeds up learning, but would not change the asymptotic performance.

Second, olfactory conditioning was also shown to depend on the timing of the shock application before or after the conditioning odor. While a shock application 30s after an odor assigns this odor an aversive valence, an appetitive valence is assigned if the shock application arises 30s before the odor presentation 16 , 17 , 62 , 63 . Modeling the approaching behavior in the context of predictive plasticity would require duplicating our model to also represent appetitive valences, and the action selection would depend on the difference between aversive and appetitive valences. Inverting the timing of CS and US may explain ‘relief learning’ if a stopping electroshock would cause a decrease of the target for aversive MBONs ( \(d^-\) ) and an increase of the target for appetitive MBONs ( \(d^+\) , see Eq. 12 ). An odor presented after the shock would then predict the increased appetitive target and explain the relieve from pain behavior, similarly to the model of relief learning in humans 64 .

Overall, our behavioral experiments and the plasticity model for the KC-to-MBON synapses support the notion of predictive learning in olfactory conditioning, with the DANs representing either the CS-US prediction error or the prediction itself. While predictive coding is recognized as a hierarchical organization principle in the mammalian cortex 65 , 66 , 67 , 68 that explains animal 2 and human behavior 69 it may also offer a framework to investigate the logic of the MB and the multi-layer MBON readout network as studied by various experimental work 24 , 32 .

Materials and methods

We used Drosophila melanogaster of the Canton-S wild-type strain. Flies were reared on standard cornmeal food at \(25^\circ \text {C}\) and exposed to a 12:12 hour light-dark cycle. For the experiments groups of 60-100 flies (1-4 days old) were used.

Behavioral experiments

The apparatus that was used to conduct the behavior experiment is based on 18 and was modified to allow performing four experiments in parallel. Experiments were performed in a climate chamber at \(23-25^\circ \text {C}\) and 70-75 \(\%\) relative humidity. Training procedures were done in dim red light and tests were accomplished in darkness. Two artificial odors, benzaldehyde (Fluka, CAS No. 100-52-7) and limonene (Sigma-Aldrich, CAS No. 5989-27-5), were used for the experiments. \(60\mu l\) of benzaldehyde was filled in plastic containers of 5mm and \(85\mu l\) of limonene was filled in plastic containers of 7mm. Odor containers were attached to the end of the tubes. A vacuum pump was used for odor delivery at a flow rate of \(7\,l\) /min. Tubes lined with an electrifiable copper grid were used to apply electric shock. Shock pulses were \(1.5\,\) s long.

Sequence shock experiments

Groups of flies were loaded in tubes lined with an electrifiable grid. After an initial phase of 90s, one of the odors was presented for 60s. At the same time electric shock pulses were delivered. After 30s of non-odorised airflow, the second odor was presented for 60s, without electric shock. Different electric shock treatments were used (see Fig.  2 ). In half of the cases benzaldehyde was paired with electric shock, while in the other half limonene was the paired odor. Whatever the idendity of the odor is, after pairing with the shock it is called the conditioned stimulus (CS+) while the odor paired with 0 shock strength is called the unconditioned stimulus (CS-). After the training flies were loaded into a sliding compartment and moved to a choice point in the middle of two tubes. Benzaldehyde was attached to one tube and limonene to the other. Flies could choose between the two odors for 120s. Then, the number of flies in each odor tube was counted.

Repeated training experiment

One training block consists of 60s odor, 30s non-odorised air and 60s of the second odor. Four electric shock pulses were delivered after 15, 30, 45 and 60s of the first odor presentation. Flies were exposed to this training block one, two or four times. The time between the training blocks was 90s. For ‘0.5 repetitions’ (as reported in Fig. 4 ) only two pulses were delivered 45 and 60s after onset of the odor and this block was not repeated. Experiments were performed with electric shock pulses of 25 and 50V. After the training, learning performance was tested as in the sequence shock experiment.

Continuous shock experiments

Continuous electric shock was used to train the animals instead of pulses. Electric shock was applied during the entire presentation of the first odor (odor X). odor X and shock duration were 10, 15, 30, 45, 90 or 120s. The second odor (odor Y) was presented for the same duration as odor X and the electric shock. odor Y was always applied 30s after the end of odor X presentation. Experiments were performed with 25 and 50V. The learning test after the training was identical to the sequence shock experiment.

Minimal shock detection

For the electric shock avoidance tests, flies were loaded into a sliding test chamber (compartment). The chamber with the flies was pushed to a choice point between two arms (tubes) with an electrifiable grid at the floor. The grid in one tube was connected to a voltage source (of strength S ), whereas the other was not. Electric shock was delivered continuously for \(30\,\) s and then the number of flies in each tube was counted. For a shock of strength \(S=5, 9\) and \(12.5\,\) V we measure a performance index PI \((S\!=\!5\text {V})=0.006 \pm 0.014\) (mean ± standard error of the mean, SEM), PI \((S\!=\!9\text {V})=0.030 \pm 0.014\) and PI \((S\!=\!12.5\text {V})=0.068 \pm 0.019\) , respectively. For \(S\!=\!7\,\) V we estimated the mean PI to be roughly 0.01, with a SEM to be roughly twice as large, 0.02, see Fig.  2 -1.

Parameter optimization

The parameters are optimized to minimize the least square error between the experimental data and the model simulation. The optimization is done in Matlab (R2014a), using Interior point method with maximum 3000 iterations, 1.0e-06 tolerance. Initial conditions of the parameters are uniformly sampled from a wide interval, and all optimized parameters with similar overall performances were clustered around the ones reported in the caption of Fig.  3 . The same set of parameters for the predictive plasticity (Eq.  7 ) is used throughout. The mean square error ( \(\text {MSE}\) ) between data mean and model mean is calculated by summing the squared error of the means (with the same \(N_\text {fly}\) and \(N_\text {trial}\) ) for all 28 data points across all experiments, divided by 28. The parameters for the predictive learning model are reported in the caption of Fig. 3 , the ones for the other models below.

Adaptable learning rate

The learning rate is assumed to increase with increasing stimulus strength ( \(\dot{s} > 0\) ) and otherwise passively decays. Its dynamics has the form

with optimized parameters \(\tau _\eta\) and \(\Delta \eta\) . We were choosing \(\tau _\eta = 133.48\) s and \(\Delta \eta = 0.057\) in all the experiments using predictive learning rule except for the simulation of the target-driven learning model in Fig. 6 B where we set \(\tau _\eta = 26.7\) s and \(\Delta \eta = 0.74\) . For the discrete time simulations, a step-increase \(\Delta s\) in the shock stimulus triggers a step increase in \(\eta\) by \(\Delta \eta \, \Delta s\) .

Linear STDP, nonlinear STDP, and covariance rule

Given the analogy of the olfactory conditioning to spike-timing (or stimulus-timing) dependent plasticity (STDP) 7 , 13 , 16 , 17 , we considered two different forms of STDP rules. The linear STDP learning rule is

where \(\tilde{s}\) is the low-pass filtered s with filtering time constant \(\tau _s=17.87s\) , \(\tilde{o}\) is the low-pass filtered o with time constant \(\tau _o=7.47s\) , \(\eta _{1}=-0.47\) , \(\eta _{2}=-0.47\) , \(\alpha =0.23\) , and \(S_{\circ }=9.31V\) . For the linear STDP we get \(\text {MSE} = 5.489 \times 10^{-3}\) for the indicated optimized parameters.

The nonlinear STDP rule is of the form

The learning rates \(\eta _{1/2}\) were allowed to be both positive and negative. The optimal parameters are \(\,\eta _{1}=0.01\) , \(\eta _{2}=0.19\) , \(\tau _o=51.20s\) , \(\tau _s=124.12s\) , \(\alpha =9.93\) , \(\alpha _{1}=9.93\) , \(\alpha _{2}=0.44\) , \(S_{\circ }=11.91V\) , with a \(\text {MSE} = 6.550 \times 10^{-3}\) .

The covariance rule has the form

and the optimal parameters are \(\eta =0.12\) , \(\tau _o=300.00s\) , \(\tau _s=19.18s\) , \(\alpha =0.53\) , \(S_{\circ }=9.13V\) , with a \(\text {MSE} = 8.236 \times 10^{-3}\) .

All these 3 rules (Eqs.  15 – 17 ) failed mainly in reproducing the repetitive conditioning experiments (Fig.  4 B), see Supplementary Material. Overall, the \(\text {MSE}\) for all these associative rules is roughly 10 times bigger than the \(\text {MSE}\) for the predictive rule ( \(\text {MSE} =6.393 \times 10^{-4}\) ).

Associative learning rules with adaptive learning rate

We also tested the associative learning rules with the adaptive learning rate from Eq. ( 14 ). Although the \(\text {MSEs}\) get smaller for both linear and nonlinear STDP rules, they remain twice as large as for the predictive learning rule (see Fig.  3 -2). The covariance rule (with optimised parameters \(\alpha =0.12\) , \(S_{\circ }=3.68\) , \(\tau _{o}=37.70s\) , \(\tau _{s}=1498.38s\) , \(\tau _{\eta }=60.54s\) , \(\Delta \eta =1.00\) , with a \(\text {MSE}=1.00 \times 10^{-2}\) ) did not profit from the adaptable learning rate. For the linear STDP learning rule (Eq. 15 ) the optimal parameters with adaptable learning rate are \(\alpha =0.31\) , \(S_{\circ }=3.20\) , \(\tau _{o}=8.29s\) , \(\tau _{s}=171.05s\) , \(\tau _{\eta _{1}}=16.01s\) , \(\tau _{\eta _{2}}=5.78s\) , \(\Delta \eta _{1}=0.39\) , \(\Delta \eta _{2}=5.71\) , with a \(\text {MSE}=1.45 \times 10^{-3}\) .

For the nonlinear STDP learning rule (Eq. 16 ), the optimal parameters are \(\alpha =0.13\) , \(S_{\circ }=3.21V\) , \(\tau _{o}=8.30s\) , \(\tau _{s}=171.10s\) , \(\tau _{\eta _{1}}=16.02s\) , \(\tau _{\eta _{2}}=5.82\) , \(\Delta \eta _{1}=8.96\) , \(\Delta \eta _{2}=5.14\) , \(\alpha _{1}=0.24\) , \(\alpha _{2}=6.05\) , with a \(\text {MSE}=1.46 \times 10^{-3}\) . For the Hebbian additive rule in Eq. ( 6 ) with adaptable learning rate, the optimal parameters are \(\alpha =0.05\) , \(S_{\circ }=5.08\) , \(\tau _{o}=1.50s\) , \(\tau _{\eta }=49.81s\) , \(\Delta \eta =5.46\) , with a \(\text {MSE}=1.24 \times 10^{-2}\) .

Model comparison based on the Akaike information criterion

We compared the various models on the basis of the Akaike information criterion (AIC) that puts the model accuracy on the data set into relation to the number of parameters used to achieve this accuracy 70 , 71 . Assuming that the estimation errors of all n experimental conditions are normally distributed with zero mean, the AIC for a given model M is calculated as a log-likelihood,

where k is the number of parameters in the model, \(n=28\) the number of experimental conditions, and \(C= \frac{n}{2} \, (\ln (2\pi ) +1) + 1 = 40.73\) . The relative likelihood p for model M to be true as compared to the predictive plasticity model \(M_0\) is \(p(M)=\exp {\frac{\text {AIC}(M_0)-\text {AIC}(M)}{2}}\) (Table 1 ).

Data availability

The mathematical model (Matlab) including the experimental data will be available on https://github.com/unibe-cns .

Rescorla, R. A. & Wagner, A. R. A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. Class. Cond. II Curr. Res. Theory 21 , 64–99. https://doi.org/10.1101/gr.110528.110 (1972).

Article   CAS   Google Scholar  

Pearce, J. M. A Model for Pavlovian Learning : Variations in the effectiveness of conditioned but not of unconditioned stimuli. Psych. Rev. 87 , 532–552. https://doi.org/10.1037/0033-295X.87.6.532 (1980).

Rescorla, R. A. Pavlovian conditioning: It’s not what you think it is. Am. Psychol. 43 , 151 (1988).

Aso, Y. et al. The neuronal architecture of the mushroom body provides a logic for associative learning. eLife 3 , e04577. https://doi.org/10.7554/eLife.04577 (2014).

Article   PubMed   PubMed Central   Google Scholar  

Cognigni, P., Felsenberg, J. & Waddell, S. Do the right thing: Neural network mechanisms of memory formation, expression and update in Drosophila. Curr. Opin. Neurobiol. 49 , 51–58. https://doi.org/10.1016/j.conb.2017.12.002 (2018).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Riemensperger, T., Völler, T., Stock, P., Buchner, E. & Fiala, A. Punishment prediction by dopaminergic neurons in Drosophila. Curr. Biol. 15 , 1953–1960. https://doi.org/10.1016/j.cub.2005.09.042 (2005).

Article   CAS   PubMed   Google Scholar  

Aso, Y. & Rubin, G. M. Dopaminergic neurons write and update memories with cell-type-specific rules. eLife 5 , e16135. https://doi.org/10.7554/eLife.16135 (2016).

Zhao, X., Lenek, D., Dag, U., Dickson, B. & Keleman, K. Persistent activity in a recurrent circuit underlies courtship memory in Drosophila. eLife 7 , e31425. https://doi.org/10.7554/eLife.31425 (2018).

Jozefowiez, J. Associative versus predictive processes in Pavlovian conditioning. Behav. Process. https://doi.org/10.1016/j.beproc.2017.12.016 (2017).

Article   Google Scholar  

Smith, D., Wessnitzer, J. & Webb, B. A model of associative learning in the mushroom body. Biol. Cybernet. 99 , 89–103. https://doi.org/10.1007/s00422-008-0241-1 (2008).

Article   MathSciNet   MATH   Google Scholar  

Faghihi, F., Moustafa, A. A., Heinrich, R. & Woergoetter, F. A computational model of conditioning inspired by drosophila olfactory system. Neural Netw. 87 , 96–108. https://doi.org/10.1016/j.neunet.2016.11.002 (2017).

Article   PubMed   Google Scholar  

Cassenaer, S. & Laurent, G. Hebbian STDP in mushroom bodies facilitates the synchronous flow of olfactory information in locusts. Nature 448 , 709–713. https://doi.org/10.1038/nature05973 (2007).

Article   CAS   PubMed   ADS   Google Scholar  

Cassenaer, S. & Laurent, G. Conditional modulation of spike-timing-dependent plasticity for olfactory learning. Nature 482 , 47–52. https://doi.org/10.1038/nature11261 (2012).

Huerta, R., Nowotny, T., Garcia-Sanchez, M., Abarbanel, H. D. I. & Rabinovich, M. I. Learning classification in the olfactory system of insects. Neural Comput. 16 , 1601–1640. https://doi.org/10.1162/089976604774201613 (2004).

Article   PubMed   MATH   Google Scholar  

Strube-Bloss, M. F., Nawrot, M. P. & Menzel, R. Mushroom body output neurons encode odor-reward associations. J. Neurosci. 31 , 3129–3140. https://doi.org/10.1523/JNEUROSCI.2583-10.2011 (2011).

Tanimoto, H., Heisenberg, M. & Gerber, B. Event timing turns punishment to reward. Nature 430 , 983. https://doi.org/10.1038/430983a (2004).

Yarali, A., Nehrkorn, J., Tanimoto, H. & Herz, A. V. M. Event timing in associative learning: from biochemical reaction dynamics to behavioural observations. PloS ONE 7 , 15. https://doi.org/10.1371/journal.pone.0032885 (2012).

Tully, T. & Quinn, W. G. Classical conditioning and retention in normal and mutant Drosophila melanogaster. J. Comp. Physiol. A 157 , 263–277. https://doi.org/10.1007/BF01350033 (1985).

Heisenberg, M., Borst, A., Wagner, S. & Byers, D. Drosophila mushroom body mutants are deficient in olfactory learning. J. Neurogenet. 2 , 1–30. https://doi.org/10.3109/01677068509100140 (1985).

de Belle, J. & Heisenberg, M. Associative odor learning in drosophila abolished by chemical ablation of mushroom bodies. Science 263 , 692–695. https://doi.org/10.1126/science.8303280 (1994).

Article   PubMed   ADS   Google Scholar  

Heisenberg, M. Mushroom body memoir: From maps to models. Nat. Rev. Neurosci. 4 , 266–75. https://doi.org/10.1038/nrn1074 (2003).

Turner, G. C., Bazhenov, M. & Laurent, G. Olfactory representations by drosophila Mushroom body neurons. J. Neurophysiol. 99 , 734–746. https://doi.org/10.1152/jn.01283.2007 (2008).

Litwin-Kumar, A., Harris, K. D., Axel, R., Sompolinsky, H. & Abbott, L. F. Optimal degrees of synaptic connectivity. Neuron 93 , 1153–1164. https://doi.org/10.1016/j.neuron.2017.01.030 (2017).

Aso, Y. et al. Mushroom body output neurons encode valence and guide memory-based action selection in Drosophila. eLife 3 , 004580. https://doi.org/10.7554/eLife.04580 (2014).

Mao, Z. & Davis, R. L. Eight different types of dopaminergic neurons innervate the Drosophila mushroom body neuropil : anatomical and physiological heterogeneity. Front. Neural Circuits 3 , 5. https://doi.org/10.3389/neuro.04.005.2009 (2009).

Burke, C. J. et al. Layered reward signalling through octopamine and dopamine in Drosophila. Nature 492 , 433–437. https://doi.org/10.1038/nature11614 (2012).

Article   CAS   PubMed   PubMed Central   ADS   Google Scholar  

Aso, Y. et al. Three Dopamine pathways induce aversive odor memories with different stability. PLoS Genet. 8 , 1002768. https://doi.org/10.1371/journal.pgen.1002768 (2012).

Galili, D. S., Lüdke, A., Galizia, C. G., Szyszka, P. & Tanimoto, H. Olfactory trace conditioning in Drosophila. J. Neurosci. 31 , 7240–8. https://doi.org/10.1523/JNEUROSCI.6667-10.2011 (2011).

Hige, T. et al. Heterosynaptic plasticity underlies aversive olfactory learning in Drosophila. Neuron 88 , 985–998. https://doi.org/10.1016/j.neuron.2015.11.003 (2015).

Owald, D. et al. Activity of defined mushroom body output neurons underlies learned olfactory behavior in drosophila. Neuron 86 , 417–427. https://doi.org/10.1016/j.neuron.2015.03.025 (2015).

Eichler, K. et al. The complete connectome of a learning and memory centre in an insect brain. Nature 548 , 175–182. https://doi.org/10.1038/nature23455 (2017).

Takemura, S.-Y. et al. A connectome of a learning and memory center in the adult Drosophila brain. eLife 6 , e26975. https://doi.org/10.7554/eLife.26975 (2017).

Ichinose, T. et al. Reward signal in a recurrent circuit drives appetitive long-term memory formation. eLife 4 , 10719. https://doi.org/10.7554/eLife.10719 (2015).

Ueno, K. et al. Coincident postsynaptic activity gates presynaptic dopamine release to induce plasticity in Drosophila mushroom bodies. ELife 6 , e21076. https://doi.org/10.7554/eLife.21076 (2017).

Nehrkorn, J., Tanimoto, H., Herz, A. V. M. & Yarali, A. A model for non-monotonic intensity coding. R. Soc. Open Sci. 2 , 150210. https://doi.org/10.1098/rsos.150120 (2015).

Article   MathSciNet   Google Scholar  

Fechner, G. T., translated by Adler, H. E., edited by Boring, E. G. & Howes, D. H. Elements of Psychophysics (Holt, Rinehart and Winston, INC., 1966).

Stevens, S. S., Carton, A. S. & Shickman, G. M. A scale of apparent intensity of electric shock. J. Exp. Psychol. 56 , 328–334 (1958).

Tully, T., Preat, T., Boynton, S. C. & Del Vecchio, M. Genetic dissection in Drosophila. Cell 79 , 35–47, https://doi.org/10.1016/0092-8674(94)90398-0 (1994).

Pamir, E. et al. Average group behavior does not represent individual behavior in classical conditioning of the honeybee. Learn. Mem. 18 (11), 733–41 (2011).

Pamir, E., Szyszka, P., Scheiner, R. & Nawrot, M. P. Rapid learning dynamics in individual honeybees during classical conditioning. Front. Behav. Neurosci. 8 , 313. https://doi.org/10.3389/fnbeh.2014.00313 (2014).

Diegelmann, S., Klagges, B., Michels, B., Schleyer, M. & Gerber, B. Maggot learning and Synapsin function. J. Exp. Biol. 216 , 939–951. https://doi.org/10.1242/jeb.076208 (2013).

Widrow, B. & Hoff, M. E. Adaptive switching circuits. WESCON Convention. Record 96–104 (1960).

Sutton, R. S. Learning to predict by the methods of temporal differences. Mach. Learn. 3 , 9–44 (1988).

Urbanczik, R. & Senn, W. Learning by the dendritic prediction of somatic spiking. Neuron 81 , 521–528. https://doi.org/10.1016/j.neuron.2013.11.030 (2014).

Brea, J., Gaál, A. T., Urbanczik, R. & Senn, W. Prospective coding by spiking neurons. PLoS Comput. Biol. 12 , 1–25. https://doi.org/10.1371/journal.pcbi.1005003 (2016).

Young, J., Wessnitzer, J., Armstrong, J. & Webb, B. Elemental and non-elemental olfactory learning in drosophila. Neurobiol. Learn. Mem. 96 , 339–352. https://doi.org/10.1016/j.nlm.2011.06.009 (2011).

Article   CAS   PubMed   MATH   Google Scholar  

Tomchik, S. M. & Davis, R. L. Dynamics of learning-related cAMP signaling and stimulus integration in the drosophila olfactory pathway. Neuron 64 , 510–521. https://doi.org/10.1016/j.neuron.2009.09.029 (2009).

Cervantes-Sandoval, I., Phan, A., Chakraborty, M. & Davis, R. L. Reciprocal synapses between mushroom body and dopamine neurons form a positive feedback loop required for learning. eLife 6 , e23789. https://doi.org/10.7554/eLife.23789 (2017).

Hattori, D. et al. Representations of novelty and familiarity in a mushroom body compartment. Cell 169 , 956–969.e17. https://doi.org/10.1016/j.cell.2017.04.028 (2017).

Cohn, R., Morantte, I., Cohn, R., Morantte, I. & Ruta, V. Coordinated and compartmentalized neuromodulation shapes sensory processing in Drosophila. Cell 163 , 1742–1755. https://doi.org/10.1016/j.cell.2015.11.019 (2015).

Felsenberg, J., Barnstedt, O., Cognigni, P., Lin, S. & Waddell, S. Re-evaluation of learned information in Drosophila. Nature https://doi.org/10.1038/nature21716 (2017).

Felsenberg, J. et al. Integration of parallel opposing memories underlies memory extinction. Cell 175 , 709-722.e15. https://doi.org/10.1016/j.cell.2018.08.021 (2018).

Joseph, J., Dunn, F. A. & Stopfer, M. Spontaneous olfactory receptor neuron activity determines follower cell response properties. J. Neurosci. 32 , 2900–2910. https://doi.org/10.1523/JNEUROSCI.4207-11.2012 (2012).

Hige, T., Aso, Y., Rubin, G. M. & Turner, G. C. Plasticity-driven individualization of olfactory coding in mushroom body output neurons. Nature 526 , 258–262. https://doi.org/10.1038/nature15396 (2015).

Perisse, E. et al. Aversive learning and appetitive motivation toggle feed-forward inhibition in the drosophila mushroom body. Neuron 90 , 1086–1099. https://doi.org/10.1016/j.neuron.2016.04.034 (2016).

Liu, C. et al. A subset of dopamine neurons signals reward for odour memory in Drosophila. Nature 488 , 512–516. https://doi.org/10.1038/nature11304 (2012).

Keleman, K. et al. Dopamine neurons modulate pheromone responses in Drosophila courtship learning. Nature 489 , 145–149. https://doi.org/10.1038/nature11345 (2012).

Mery, F. & Kawecki, T. J. Frederic Mery and Tadeusz. J. Kawecki. Sci. 308 , 20–23. https://doi.org/10.1126/science.1111331 (2005).

McGuire, S. E., Le, P. T. & Davis, R. L. The role of Drosophila mushroom body signaling in olfactory memory. Science 293 , 1330–1333. https://doi.org/10.1126/science.1062622 (2001).

Wright, G. A., Thomson, M. G. A. & Smith, B. H. Odour concentration affects odour identity in honeybees. Proc. R. Soc. B 272 , 2417–2422. https://doi.org/10.1098/rspb.2005.3252 (2005).

Luo, S. X., Axel, R. & Abbott, L. F. Generating sparse and selective third-order responses in the olfactory system of the fly. PNAS 107 , 10713–10718. https://doi.org/10.1073/pnas.1005635107 (2010).

Yarali, A. et al. Genetic distortion of the balance between punishment and relief learning in Drosophila. J. Neurogenet. 23 , 235–47. https://doi.org/10.1080/01677060802441372 (2009).

König, C. et al. Reinforcement signaling of punishment versus relief in fruit flies. Learn. Mem. 25 , 247–257. https://doi.org/10.1101/lm.047308 (2018).

Seymour, B. et al. Opponent appetitive-aversive neural processes underlie predictive learning of pain relief. Nat. Neurosci. 8 , 1234–1240. https://doi.org/10.1038/nn1527 (2005).

Rao, R. P. N. & Ballard, D. H. Predictive coding in the visual cortex: A functional interpretation of some extra-classical receptive-field effects. Nat. Neurosci. 2 , 79–87. https://doi.org/10.1038/4580 (1999).

Bastos, A. M. et al. Canonical microcircuits for predictive coding. Neuron 76 , 695–711. https://doi.org/10.1016/j.neuron.2012.10.038 (2012).

Vladimirskiy, B., Urbanczik, R. & Senn, W. Hierarchical novelty-familiarity representation in the visual system by modular predictive coding. PLoS ONE 10 , 1–19. https://doi.org/10.1371/journal.pone.0144636 (2015).

Sacramento, J., Costa, R. P., Bengio, Y. & Senn, W. Dendritic error backpropagation in deep cortical microcircuits. arXiv 1–37 (2017). arXiv:1801.00062v1 .

Lefebvre, G., Lebreton, M., Meyniel, F., Bourgeois-Gironde, S. & Palminteri, S. Behavioural and neural characterization of optimistic reinforcement learning. Nat. Hum. Behav. 1 , 1–9. https://doi.org/10.1038/s41562-017-0067 (2017).

Akaike, H. Information Theory and an Extension of the Maximum Likelihood Principle. In: Parzen E., Tanabe K., Kitagawa G. (eds) Selected Papers of Hirotugu Akaike. Springer Series in Statistics (Perspectives in Statistics) , 199–213 (Springer New York, New York, NY, 1998).

Banks, H. & Joyner, M. L. Aic under the framework of least squares estimation. Appl. Math. Lett. 74 , 33–45. https://doi.org/10.1016/j.aml.2017.05.005 (2017).

Download references

Acknowledgements

We thank members of the Senn and Sprecher lab for helpful support, and in particular Robert Urbanczik \(^\dagger\) , Martin Wiechert for fruitful discussions and Elena Kreutzer, Dominik Spicher, Jakob Jordan, Rui Ponte Costa, Ashok Litwin-Kumar and Emmanuel Perisse for helpful comments on the manuscript. WS is grateful for inspiring conversations with an anonymous reviewer. This research was supported by the SystemsX.ch initiative (for SS and WS, grant 51RT-0-145733, evaluated by the SNSF), the SNSF (grant 310030L-156863, WS), a SNF sinergia grant (CRSII5-180316 led by F. Helmchen) and a grant from the China Scholarship Council (201408080117, CZ). WS and MAP were supported by the European Union’s Horizon 2020 Framework Programme for Research and Innovation under the Specific Grant Agreement No. 720270, 785907, and 945539 (Human Brain Project SGA 1-3, related to predictive coding).

Author information

These authors contributed equally: Chang Zhao and Yves F. Widmer.

Authors and Affiliations

Department of Physiology, University of Bern, Bern, 3012, Switzerland

Chang Zhao, Mihai A. Petrovici & Walter Senn

Department of Biology, University of Fribourg, Fribourg, 1700, Switzerland

Yves F. Widmer, Sören Diegelmann & Simon G. Sprecher

You can also search for this author in PubMed   Google Scholar

Contributions

W.S., S.G.S., C.Z. and Y.F.W. designed the experiments. W.S. and C.Z. developed the mathematical model, Y.W. and S.D. performed the experiments, C.Z. and M.A.P. performed the computer simulations, and C.Z., W.S., Y.W. and S.G.S. wrote the manuscript.

Corresponding authors

Correspondence to Simon G. Sprecher or Walter Senn .

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

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.

Zhao, C., Widmer, Y.F., Diegelmann, S. et al. Predictive olfactory learning in Drosophila . Sci Rep 11 , 6795 (2021). https://doi.org/10.1038/s41598-021-85841-y

Download citation

Received : 09 November 2018

Accepted : 02 March 2021

Published : 24 March 2021

DOI : https://doi.org/10.1038/s41598-021-85841-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

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

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.

• Laboratory Techniques & Protocols

Drosophila: General Information and Methods for Experiments

Introduction.

Whoever thought that the small fly hovering over fruits will become such an essential asset for genetics research!! Fruit flies are one of the first extensively used model organisms in labs. Moreover, it is one of the first model organisms which was brought ‘from nature to labs’ to understand the genetics of the organisms!

N. M. Stevens was the first scientist who chose Drosophila as a model for his studies in 1906, but it gained more popularity after Morgan’s discovery in 1910. Thomas Hunt Morgan’s studies on Drosophila melanogaster led to the discovery of the sex-linked inheritance and linkage. His theories violated the rules of Mendelian inheritance. Followed by his study, A. H. Sturtevant constructed the first genetic map (involving X-linked genes) using Drosophila.

Today, Drosophila is used in various labs to understand the basic cellular and developmental mechanisms of living organisms that mainly coincide with the human system. A few examples of studies that include Drosophila are: mutational studies, developmental studies,  regenerative biology, and drug discovery.

This article presents a brief scenario of the “fly lab”. It unravels some interesting facts and features of Drosophila and introduces you to some widely experimental techniques to study Drosophila in labs.

Why is Drosophila Considered as a Model Organism?

Model organisms are organisms that are used by scientists to perform experimental studies in an effort to understand their biological processes. The choice of a model organism usually depends on the type of research the scientist intends to conduct.

Drosophila has a number of qualities that make it desirable for scientific studies. Some of its features include:

• Small size (Adults-3mm and eggs-0.5mm in length).
• Easy to handle.
• Sexual dimorphism (different males and females)
• One female can lay about 100 eggs.
• Short generation time (9-10 days)
• 4 pairs of chromosomes and the whole genome is sequenced.
• Low culture and maintenance cost (requires maize food, cultured in small bottles, and re lesser lab space).

Morphology of Drosophila melanogaster

Life Cycle of Drosophila

Drosophila requires 9 to 10 days (at 25 ℃) to complete its life cycle; to develop from egg to adult―the development is also temperature-dependent as it takes 19 days to complete the life cycle at 18  ℃.

The process of embryogenesis takes 24 hrs followed by 4-5 days in the larval stage―1st, 2nd, and 3rd instar larval stage. After the 5th day, the larvae enter into the pupal stage which lasts around 6-9 days of the development.

At the pupal stage, the tissue progenitor cells start transforming into adult characters (for example eyes, legs, and wings)―the ecdysone hormone is also involved in the transitional development of flies, from the pupal stage to adult.

Then, the adult fly emerges from the pupal case―the process called eclosion―and enters the process of maturation, which takes around 8-12 hours.

Sex determination in Drosophila

Drosophila  has an XY system of sex determination like humans. Males are heteromorphic (XY), and females are homomorphic (XX).  However, in  Drosophila,  sex is determined by the ratio of the X-chromosome to an autosome.

When the X:A ratio is ≥ 1.0 (1X:1A, 2X:2A, 3X:3A, and 3A:2A) then the fly will develop into a female. On the other hand, when the ratio of X:A is ≤ 0.5, then a male will be developed. To understand this, have a look at the table below:

What is the genetic mechanism that operates behind this sex-determining mechanism?

The ratio of X:A, regulates the expression of the Sxl gene. If the ratio is ≥ 1.0, then transcription of the Sxl (Sex-lethal) gene occurs, followed by female-specific splicing-removal of exons required for the development of male sex.

The splicing leads to the formation of a functional Sxl protein. The Sxl protein further activates the Tra (Transformer) gene which encodes Tra protein. The Tra protein binds to the Dsx  (Doublesex) gene, which leads to the formation of female-specific Dsx protein with 430 aa residues [2] .

When the X:A ratio is >1, then there won’t be an expression of the Sxl gene and the absence of a function Tra protein. This condition leads to the activation of male-specific splicing, which further leads to the formation of male-specific Dsx protein with 550 amino acid residues [2] .

The Dsx (Doublesex) gene is essential for the sexual differentiation of both the sexes (male and female). The absence of the Dsx gene can turn flies (of both XX and XY arrangements) into intersex [2] .

Genetics of Drosophila

Drosophila  has 4 pairs of chromosomes and the size of its genome is estimated to be approx 180 Mb―the euchromatin (transcriptionally active region ) is about 120 Mb. Out of 4 pairs of chromosomes, one is sex chromosome (acrocentric X chromosomes and metacentric Y chromosomes) and three are autosomes.

Other than mitotic chromosomes,  Drosophila  has “ Polytene chromosomes -giant chromosomes.” This chromosome is formed by the process of “ endoreduplication -multiple round of DNA replication, without cell division.” The polytene chromosomes are commonly found in salivary glands and most commonly isolated from larval salivary glands.

How does the genetics of Drosophila aid in the experiments?

A lot of information has been decoded about the Drosophila genome. Some genetic characteristics of drosophila are known as “ genetic tools ”, which provide essential aid in various researches including genetics, biotechnology, drug discovery, etc.

1. Classical genetics

It has been almost 100 years since the use of Drosophila in various researches. These meticulous studies have contributed a lot to the understanding of heredity and gene activity, construction of recombination maps, and the relation between sex linkage and inheritance of sex chromosomes.

Mutagenesis studies on Drosophila led to the construction of genetic maps, which was also facilitated by the studies of banding patterns of polytene chromosomes found in salivary glands of  Drosophila  larva.

2. Transposable elements

The P elements carry the gene for transposase activity and are responsible for hybrid dysgenesis―series of mutations as a result of crossing male with P element and a female without P element which leads to sterility and chromosomal aberration―in Drosophila. Its use in introducing cloned DNA in Drosophila was discovered way back in 1982.

The P elements are now an essential tool in genetics for gene tagging, gene disruption, chromosome engineering, and inducible gene expression. Moreover, the mutation, by the introduction of P elements into a gene, allows the molecular identification of the affected gene.

The controlled activity of transposase aids scientists to insert P elements at the desired location and generate a large number of mutations required for their studies.

3. The Drosophila genome

The whole genome of Drosophila was sequenced in 2000. Researchers have found that Drosophila has approximately 70% of orthologous genes associated with human diseases.

The availability of a chunk of information about Drosophila genomes aid researchers (as a model organism) for human disease research and drug discovery and it also avoids the labor of performing a number of molecular manipulations of Drosophila DNA.

Experimental methods to study Drosophila in labs

The study of Drosophila in labs serves many purposes which include the understanding of neurons, developmental processes, mutational studies, gene regulation, and signaling. A number of methods are used to study the different processes in different experiments. In this section, you will get a brief of cytogenetic methods used to study different parts/processes of Drosophila and its culturing in labs.

Other than the mentioned cytogenetic experiments in this section, the two other, most commonly used bioengineered, techniques used to study the Drosophila are:

• The GAL4/UAS system: Used to study gene expression and function in organisms.
• The FLP/ FRT system: Site-directed recombination technology.

1. Culturing Drosophila in labs

The  Drosophila  culturing does not require standard equipment. The great scientist T. H. Morgan used glass milk bottles to culture  Drosophila  for his experiments! Now, laboratories use bottles and vials to culture  Drosophila .

Bottles are used to maintain a large population and culturing vials are used to maintain a small population and make crosses. Generally, glass bottles are preferred but autoclaved plastic bottles can also work well. Moreover, the size of small vials ranges from  96 mm by 25 mm to larger vials. To cover the bottles, plugs are used which can be cotton or foam plugs. However, cotton plugs are mostly preferred.

The preparation of media involves two ways: cooking the media or using ready-made and dehydrated media. The latter is most preferred, which avoids the labor of cooking the media, and is easier and quicker than the former. However, rehydration is required to use readymade media. To ensure the complete rehydration and culture process, follow the given procedure:

Rehydration Procedure of ready-made media:

• Add ⅕ to ⅖ volume of dry media to the bottle/vial.
• Add water to completely moisten the media.
• Let the vial rest for a few minutes and add water until it seems hydrated.
• The surface of completely hydrated media looks shiny without any spaces in the media.
• Allow the media to warm to room temperature. The optimum temperature to grow flies is at 25 ℃ with 60 % humidity.
• Add several grains of yeast on the surface of the media after the media is hydrated.
• Transfer the flies in the vial/bottles and plug it.
• Flies should be transferred to different clean vials/bottles to maintain the culture.

NOTE: Change the food always after or between 10-14 days.

2. Orcein staining to identify Polytene chromosomes

• Select 3rd instar larva from the culture.
• Place them in a container having PBS and 0.8% saline.
• Put the larva on the slide containing a drop of 45% acetic acid and dissect the larva.
• Remove the salivary gland and allow it to sit in 45% acetic acid for 2-3 minutes.
• Stain the gland with the lactic–acetic-orcein stain for 5 minutes.
• Take a clean slide and put a drop of 1: 2 : 3 fixative. Transfer the gland to this slide.
• Put a coverslip over the gland and by using bibulous paper score a zigzag pattern over the coverslip to shear the chromosomes into fragments.
• Seal the edges of the coverslip and observe it in the microscope.
• To identify the chromosome, resolve the banding patterns of the telomeric region and look for some landmarks such as puffs, constriction, and banded regions.

3. In-situ hybridization to polytene chromosomes

• Dissect the salivary glands of the Drosophila in a drop of 0.7% NaCl and then transfer to 45 % acetic acid.
• Transfer the glands to the fixative for 3 minutes.
• Spread chromosomes by tapping the coverslip with a pencil in a circular motion.
• Cover the slide in blotting paper and leave overnight for better fixation.
• Freeze the slide in liquid nitrogen and flip off the coverslip.
• Place the slide in 70% ethanol for 5 min and air dry. The chromosomes can be stored desiccated at room temperature.
• Two methods of denaturation are available: Alkali denaturation (involves the use of freshly prepared NaOH)and Heat denaturation (involves the use of Tris HCl, pH 7.5, and then transfer to 70 % and 96% ethanol).
• Use the denatured slide on the same day.
• Prepare probes using DNA amplified by the polymerase chain reaction (PCR) in the presence of biotinylated nucleotides.
• Pipette probe onto the chromosomes and cover it with a siliconized coverslip.
• Overnight incubate the slide in a box lined with moist tissue at 58 ℃.
• Remove the slide from the box and put it in 2X SSC. Wash the slide at 58 ℃ for 1 hour, with three changes of the solution.
• Wash the slides in PBS and PBS-TX for 2 minutes.
• Make a 1:250 dilution of streptavidin–horseradish peroxidase conjugate in the buffer and add to the chromosomes.
• Place the slide in a humid box and incubate at 37°C for 30 minutes.
• Wash the slide in PBS and PBS-TX.
• Place the DAB solution onto the chromosomes.
• Incubate at room temperature for 10-15 minutes.
• Rinse the slides in PBS.
• Observe the slide under the Phase Contrast microscope.

4. Immunostaining to study Larval brains

• Transfer the 3rd instar larva in a few drops of physiological solution (0.7% NaCl in distilled water) and dissect out its brain.
• Transfer the brain to a few drops of hypotonic solution (0.5% Trisodium citrate dihydrate in distilled water) for 8 minutes.
• Transfer the brain to a fixative solution (2% formaldehyde, 45% acetic acid in distilled water) for 1-8 minutes.
• Transfer four fixed brains to 4 drops of fixative solution and cover it with a siliconized coverslip.
• Put another clean slide on the coverslip, invert the sandwich, and squash for 1 minute between blotting paper.
• Freeze in liquid nitrogen, flip off the coverslip, and immerse the slide in 1X PBS at room temperature.
• Put the slide in 1X PBS with dried nonfat milk and incubate in it for 30 minutes.
• Clean the slides in 1X PBS for 3 minutes.
• Put the primary antibody solution on the mitotic preparation and incubate for 1 hour at room temperature.
• Wash the slides in PBS for 5 minutes (three times).
• Put the secondary antibody solution on the mitotic preparation and incubate it for 1 hour at room temperature.
• Wash the slides in 1X PBS (three times) for 5 min in the dark.
• Stain the slides in the DAP1 staining solution at room temperature for 4 minutes.
• Wash the slides in 1X PBS for 30 seconds.
• Mount the slide in a drop of antifade solution.
• Analyze the slides using an epifluorescence microscope (The slide can also be used to perform the FISH technique).

Note: To grasp the principle and theory of Cytogenetic techniques you can refer to “ Cytogenetics: An advanced technique to color Chromosomes  and  Molecular Cytogenetics: in situ Hybridization-based technology ”.

Drosophila studies have unraveled many mysteries of biological processes and today its use in biomedical research can not be ignored.

A number of proven methods are available to study any characteristics/features of Drosophila and understand different processes (neural functions, signaling, development, etc.). The availability of huge information on Drosophila in combination with an array of genetic tools allows researchers to tackle any problems in biology.

Today, it is extensively used in applied and translational research towards human health and it is certainly expected by scientists that it can make a major breakthrough in regenerative medicine.

References:

• Daryl S. Henderson (2004). Drosophila Cytogenetics Protocols. Methods in Molecular Biology, vol. 247.  Human press  inc., Totowa, NJ. DOI https://doi.org/10.1385/1592596657.
• Hake, L. & O’Connor, C. (2008) Genetic mechanisms of sex determination.  Nature Education 1(1) , 25.
• Jennings, B. H. (2011). Drosophila – a versatile model in biology & medicine . Materials Today , 14(5), 190–195. DOI:10.1016/s1369-7021(11)70113-4.
• Karen G. Hales, Christopher A. Korey, Amanda M. Larracuente, and David M. Roberts (2015). Genetics on the Fly: A Primer on the Drosophila Model System.  Genetics , 201, 815–842. DOI: 10.1534/genetics.115.183392.
• Mariateresa Allocca, Sheri Zola, and Paola Bellosta (2018). The Fruit Fly, Drosophila melanogaster: The Making of a Model (Part I), Drosophila melanogaster – Model for Recent Advances in Genetics and Therapeutics, Farzana Khan Perveen.  IntechOpen . DOI: 10.5772/intechopen.72832
• O’Kane, C. J. (2001). Drosophila melanogaster.  Encyclopedia of Genetics , 584–585. DOI:10.1006/rwgn.2001.1698.
• https://depts.washington.edu/cberglab/wordpress/outreach/an-introduction-to-fruit-flies/
• https://www.nasa.gov/sites/default/files/atoms/files/fruit_fly-iss-mini-book-tagged.pdf
• https://www.differencebetween.com/difference-between-male-and-female-drosophila-melanogaster/ .

Our Location

• 5250 Old Orchard Rd Suite 300 Skokie, IL 60077
• +1 847 983 3672

Monday – Friday 9 AM – 5 PM EST

Conduct Science

• Become a Partner
• Social Media
• Career/Academia
• Privacy Policy
• Shipping & Returns
• Request a quote

Customer service

• Account Details
• Lost Password

DISCLAIMER: ConductScience and affiliate products are NOT designed for human consumption, testing, or clinical utilization. They are designed for pre-clinical utilization only. Customers purchasing apparatus for the purposes of scientific research or veterinary care affirm adherence to applicable regulatory bodies for the country in which their research or care is conducted.

• Fruit Fly Behavior

Explore fruit flies in the classroom

Drosophila melanogaster is one of many species of vinegar fly, and is found around the world living in close proximity to humans. D. melanogaster flies have been used in labs for over a century, because they’re simple to care for, they don’t take up much space, they’re cheap, and they produce offspring rapidly. These characteristics also make flies ideal for classroom use. Maintaining fly crosses in the classroom looks very similar to what scientists do in the lab – just on a smaller scale. Think of fly bottles as a study system – a starting point from which many open-ended projects could develop.

What you'll need

Carolina Biological supply website: https://www.carolina.com   (a good source for supplies and even starter fly cultures)

Bloomington Drosophila Stock Center: bdsc.indiana.edu (researcher’s go-to source for transgenic flies)

Fly Morgue and Vinegar Trap

Be prepared to dispose of dead flies and to capture escaped flies. The following materials are helpful to have on hand:

• 70% ethanol
• Apple cider vinegar
• A bottle or jar, possibly with a funnel on top
• Plastic wrap
• Rubber bands

Fly Culture Bottles

You’ll want clear containers containing about 1.5 cm of food media and plugged with something porous (cotton is easiest).

Here are some container options:

• Commercial plastic vials and/or bottles designed for fly culture to cover with
• Reuse spice jars/baby food/takeout containers that have been thoroughly cleaned and sterilized

Here are some covering options:

• Custom plugs here / here
• Rip off bits of cotton wool (buy rolls). Even large cotton balls work! If your container is wide, you can cover it with cheesecloth or a similar porous fabric and secure it with a rubber band.

There are various options for fly food, choose the best fit for your situation:

• Homemade with corner store ingredients, but more likely to mold over time: Banana Food
• Homemade with more robust ingredients: Agar/Cornmeal Food
• Commercially pre-mixed food for purchase

Protocols for the homemade foods are available for download in the Save & Share menu.

"Fly Pushing" Supplies

To make fly crosses for propagating your fly population and/or making unique genetic offspring, you will want the following:

• A dissecting microscope
• A simple paintbrush or other fine instrument for “pushing” the flies
• CO 2 (such as from mixing baking soda and vinegar in a squeeze bottle), FlyNap , or other fly anesthetic

Protocols for fly pushing and anesthesia are available for download in the Save & Share menu.

Drosophila husbandry explained

Once you have the supplies ready for your flies, plan out how you will maintain your fly colony over time. Flies reproduce rapidly, making them easy to prepare and maintain within timeframes reasonable for the classroom or student research timeline. Below we show a simple proposed timeline for setting crosses. When set this way you’ll have a new batch of flies for study emerging every other week. Because you can keep a batch for up to two weeks, you can have a continuous supply of flies or leave yourself some time to analyze data and plan for the next experiments on weeks 1, 3, etc.

A timeline of fly crosses to maintain live flies over time

Setting up a cross is very simple! Stick a small rectangle of tissue paper into a new container of food. Put 5-10 females and 5-10 males in the new container, and leave them be for 10 days at room temperature.

Check out this guide book (which can also be downloaded as a PDF from the Save & Share menu) to get an overview of setting up and maintaining flies in your own learning environment.

• D4P: People and Medicine
• D4P: Social Behavior
• Everyday Emulsions
• What is a Virus, Anyway?
• Fermentation of Foodstuffs
• Learn about the Styles of Scientific Reasoning
• RockEDU Mentorship Handbook
• Continued Learning for Mentors
• Guide to LGBTQIA+ Allyship for Mentors
• In-Reach Infrastructure
• Perspectives
• Student Spotlight
• BIOME Blurbs
• Call for Contributions
• About The Incubator
• For Teachers
• For Students
• For Scientists
• For Outreach Professionals
• Our Online Resources
• Partners and Funders for RockEDU Online
• Our Programs in NYC
• Who are we?

Update your browser to view this website correctly. Update my browser now

• Faculty Resource Center
• Biochemistry
• Bioengineering
• Cancer Research
• Developmental Biology
• Engineering
• Environment
• Immunology and Infection
• Neuroscience
• JoVE Journal
• JoVE Encyclopedia of Experiments
• JoVE Chrome Extension
• Environmental Sciences
• Pharmacology
• JoVE Science Education
• JoVE Lab Manual
• JoVE Business
• Videos Mapped to your Course
• High Schools
• Videos Mapped to Your Course

Encyclopedia Of Experiments

Drosophila melanogaster  (fruit fly).

How fruit flies control the brain's 'steering wheel'

When we walk down the street, we have an internal sense of which way we are heading, from looking at street signals and physical landmarks, and also a sense of where we'd like to go. But how does the brain coordinate between these directions, doing the mental math that tells us which way to turn?

Now, new research describes such a neural process in fruit flies, providing insight into how an animal's brain steers it in the right direction. The study, published in Nature , shows how neurons that signal the direction in which a fly is currently oriented work together with neurons that signal the direction in which way the fly wishes to be oriented -- its goal direction -- to form a circuit that guides the animal.

"The fundamental question is how brains enable navigation," says Rockefeller's Gaby Maimon. "In this study, we describe neurons that provide goal-direction signals, alongside a brain circuit that uses these signals to direct steering."

Navigational goals

The cells responsible for signaling which way a fly is oriented in the world (known as "compass" neurons) were first discovered in 2015. A few years later, work from the Maimon lab and others demonstrated that flies with defective compass neurons are unable to navigate in a straight line along any arbitrary goal direction. Building on that discovery, Peter Mussells Pires, a student in Maimon's lab and the lead author on this paper, set out to discover the cells responsible for keeping track of a fly's goal angle.

Pires and colleagues used two-photon microscopy to monitor flies' neurons while the insects walked on an air-levitated ball in a virtual environment. Whenever the researchers rotated the virtual environment, the activity of the fly's compass neurons rotated in the brain as well. Interestingly, however, a population of cells, identified as FC2 neurons, remained unmoved and focused on the original heading.

"Imagine walking uptown in Manhattan, and someone pulls your shoulder and turns you east. Something in your brain continues to track which way is north, so that you can return to your original heading," Maimon explains. "In flies, these are FC2 neurons."

To confirm the role of FC2 neurons in tracking a goal, the team used optogenetics -- a technique that uses light to control the activity of neurons. By manipulating the activity of FC2 cells, the researchers were able to change the fly's navigation direction in predictable ways. "This was the experiment that really convinced us that these cells can actually determine the fly's goal," Pires says.

Mental math

With heading neurons and goal neurons identified, the team shifted its focus to the brain circuit responsible for combining the two signals. Recent work fleshing out the fly brain connectome -- a map detailing the connections between different neurons -- helped the researchers zero in on the circuit in question. The connectome made clear that a set of cells, called PFL3 cells, receive inputs from both the compass and goal neurons.

A series of experiments confirmed that PFL3 neurons tell the fly's body which way to turn by influencing the brain's motor system. They do so by comparing internal heading and goal inputs, functioning a bit like the steering wheel of the fly's navigation system. Larry Abbott, a theorist at Columbia University, collaborated with the team to develop a mathematical understanding of the system. Abbott's model captured how compass and goal signals, which are represented in world or map coordinates (for example, north/east/south/west), are converted into motor-related signals in the body's coordinate system -- that is, left and right turns. Complementary results on PFL neurons, closely tied to the present study, are detailed in a parallel Nature paper.

Future work from the Maimon lab will focus on how flies build and store longer term spatial memories and goals to guide behavior; the goal signal characterized in this study only explains what the flies will do in the next few seconds. Maimon is also curious to learn whether these new findings might catalyze the discovery of similar brain circuits in mammals and ultimately humans.

"By studying the fly brain," he says, "we have provided an initial glimpse into how a simple 'thought' is converted into an action. Hopefully, these findings will allow us to understanding more complex forms of this process in mammals down the road."

• Animal Learning and Intelligence
• Insects (including Butterflies)
• Behavioral Science
• Developmental Biology
• Biochemistry Research
• Brain tumor
• Chemical synapse
• Neurobiology
• Artificial neural network

Story Source:

Materials provided by Rockefeller University . Note: Content may be edited for style and length.

Journal Reference :

• Peter Mussells Pires, Lingwei Zhang, Victoria Parache, L. F. Abbott, Gaby Maimon. Converting an allocentric goal into an egocentric steering signal . Nature , 2024; DOI: 10.1038/s41586-023-07006-3

Cite This Page :

Explore More

• US Droughts, Rainy Extremes Becoming More Severe
• Clean Up Mistakes: Gene Expression in Cells
• Aging Immune System Fuels Cancer Growth: ...
• Bacteria Can Anticipate the Seasons
• Removing Long-Lasting Chemicals from Water
• Some Pterosaurs Would Flap, Others Would Soar
• Astrophysics: AI Shines New Light On Exoplanets
• One-Dimensional Gas Created out of Light
• 100x Improvement in Sight After Gene Therapy
• Milky Ways Black Hole: Massive Cosmic Merger

Trending Topics

Strange & offbeat.

• News/Events
• Arts and Sciences
• Design and the Arts
• Engineering
• Global Futures
• Health Solutions
• Nursing and Health Innovation
• Public Service and Community Solutions
• University College
• Thunderbird School of Global Management
• Polytechnic
• Downtown Phoenix
• Online and Extended
• Lake Havasu
• Research Park
• Washington D.C.
• Biology Bits
• Bird Finder
• Coloring Pages
• Experiments and Activities
• Games and Simulations
• Quizzes in Other Languages
• Virtual Reality (VR)
• World of Biology
• Meet Our Biologists

Listen and Watch

• PLOSable Biology
• All About Autism
• Xs and Ys: How Our Sex Is Decided
• When Blood Types Shouldn’t Mix: Rh and Pregnancy
• What Is the Menstrual Cycle?
• Understanding Intersex
• The Mysterious Case of the Missing Periods
• Summarizing Sex Traits
• Shedding Light on Endometriosis
• Periods: What Should You Expect?
• Menstruation Matters
• Investigating In Vitro Fertilization
• Introducing the IUD
• How Fast Do Embryos Grow?
• Helpful Sex Hormones
• Getting to Know the Germ Layers
• Gender versus Biological Sex: What’s the Difference?
• Gender Identities and Expression
• Focusing on Female Infertility
• Fetal Alcohol Syndrome and Pregnancy
• Ectopic Pregnancy: An Unexpected Path
• Creating Chimeras
• Confronting Human Chimerism
• Cells, Frozen in Time
• EvMed Edits
• Stories in Other Languages
• Virtual Reality
• Zoom Gallery
• Ugly Bug Galleries
• Ask a Question
• Top Questions
• Question Guidelines
• Permissions
• Information Collected
• Author and Artist Notes
• Share Ask A Biologist
• Articles & News
• Our Volunteers
• Teacher Toolbox

show/hide words to know

Extract: to draw or pull something out (especially from another material).

Gene: a region of DNA that instructs the cell on how to build protein(s). As a human, you usually get a set of instructions from your mom and another set from your dad... more

Molecule: a chemical structure that has two or more atoms held together by a chemical bond. Water is a molecule of two hydrogen atoms and one oxygen atom (H2O)...  more

Extracting DNA from a Banana and Other Fruits

Introduction

All living things, bananas and people included, pass on information from one generation to the next using the same basic material, DNA . Within every living organism, most cells contain a complete set of DNA instructions. The information in DNA tells our bodies how to develop, grow, and work. It also controls many of the features that make an organism unique.

DNA or deoxyribonucleic acid is found in all living things. Its natural shape is called a double helix and when seen under extremely high-powered microscopes, it looks kind of like a ladder twisted into a spiral shape.

These instructions are in segments of DNA called genes. Genes, along with other parts of our DNA that turn genes on and off, hold information for how our body develops and functions. They produce molecules called proteins that do most of the work in the body. Variants of genes, called alleles, are responsible for differences in hair color, eye color, and earlobe shape.

All of these instructions fit within tiny packages within our tiny cells, so that is all way too tiny for anyone to ever really see or touch, right? Well, not entirely. Because DNA is in every cell, there is a lot of it in an organism. If you took all of the DNA out of some middle-sized organism (or part of an organism, like a piece of fruit), you could see and even touch DNA. We will use common household products to break apart the cells in a banana and extract out the DNA. While you may know of the double-helix structure of DNA, you can't see that structure with the naked eye. So when seeing it without a high-powered microscope...what does DNA look like?

Materials     

• 1/2 peeled ripe banana (you can also use strawberries or other fruit)
• 1/2 cup hot water
• 1/2 tsp liquid dishwashing soap
• resealable zip-top bag (quart size)
• very cold rubbing alcohol (isopropyl alcohol) placed in freezer ahead of time
• coffee filter
• narrow glass
• wooden stirrer

Watch biologist Melissa Wilson Sayres as she shows you step-by-step how to extract DNA from a banana.

Extracting DNA in 10 Easy Steps

• Mush the banana in the resealable bag for about a minute until all the lumps are gone and it almost looks like pudding.
• Fill a cup with the hot water and salt.
• Pour the saltwater mix into the bag. Close the bag and very gently squeeze and move the saltwater and banana mush together. Do this for 30 to 45 seconds.
• Add the dishwashing soap into the bag and gently mix the contents. Try to avoid making too much foam.
• Place the coffee filter in a clear glass cup, securing the top of the filter around the lip of the cup.
• Pour the mix into the filter and let it sit until all of the liquid drips down into the cup.
• Remove and discard the used coffee filter.
• Tilt the glass and slowly add cold alcohol down the side of the cup. You want the alcohol to form a layer on top of the banana mix, staying separated, so be careful not to pour it too fast. Make a layer of alcohol that is 2.5-5cm (1-2in) thick.
• After the alcohol layer is set up, wait for eight minutes. You may see some bubbles and cloudy material moving around in the alcohol. This is the DNA pieces clumping together.
• Use the wooden stirrer to start poking the cloudy stuff in the alcohol layer. Spin the stirrer it in place to start gathering the cloudy stuff. When you are done, take a closer look at the stuff on the stirrer. You are looking at DNA!

(Teacher & student packet is available.)

What Happened?

You may understand that mashing a banana can break cells apart and help break apart cell walls, but why was all that other stuff added? And how did we get inside the cells and get the DNA to stick together?

Let's think of three of the main items we added to the bananas.

• Saltwater - The bananas were mashed with saltwater before anything else was added. But this was a special step preparing for the addition of the dish soap. Once the dish soap helps release the DNA, this salt will help the DNA strands to stick to each other in clumps large enough for you to see.
• Dish soap - Dish soap can help split apart the membranes (the outer "skin") that hold cells together. These membranes are made of a type of molecule called lipids. When you think of lipids, think of fats and oils. Dish soap "cuts through grease" because it actually separates those greasy molecules from each other. Now, the molecules that make the membranes around cells and the nucleus (which holds DNA) are lipids. So when dish soap is added, the cell membrane and the nuclei are broken apart, releasing the DNA.
• Alcohol - The DNA clumps are soluble (can be dissolved) in some liquids, but not in alcohol. So adding alcohol helps the clumps of DNA to form. 

Banana and Strawberry image by Ralph Daily via Wikimedia Commons.

Read more about: Seeing DNA

View citation, bibliographic details:.

• Article: Seeing DNA
• Author(s): Melissa Wilson Sayres
• Publisher: Arizona State University School of Life Sciences Ask A Biologist
• Site name: ASU - Ask A Biologist
• Date published: April 19, 2016
• Date accessed: September 4, 2024
• Link: https://askabiologist.asu.edu/activities/banana-dna

Melissa Wilson Sayres. (2016, April 19). Seeing DNA. ASU - Ask A Biologist. Retrieved September 4, 2024 from https://askabiologist.asu.edu/activities/banana-dna

Chicago Manual of Style

Melissa Wilson Sayres. "Seeing DNA". ASU - Ask A Biologist. 19 April, 2016. https://askabiologist.asu.edu/activities/banana-dna

MLA 2017 Style

Melissa Wilson Sayres. "Seeing DNA". ASU - Ask A Biologist. 19 Apr 2016. ASU - Ask A Biologist, Web. 4 Sep 2024. https://askabiologist.asu.edu/activities/banana-dna

 DNA is in every living thing (and it's only in living things).

To extract DNA for this activity, it is best to use mushy fruit. Bananas and strawberries are great choices.

Download the - Banana DNA activity (PDF)

Learn more about Melissa Wilson Sayres' work with Monster DNA .

Monster DNA

Be Part of Ask A Biologist

By volunteering, or simply sending us feedback on the site. Scientists, teachers, writers, illustrators, and translators are all important to the program. If you are interested in helping with the website we have a Volunteers page to get the process started.

Share to Google Classroom

Fruit flies hit the gym: Tiny treadmills push flies to set 50 mm/s speed record

The biological insights may shed light on how animals navigate complex environments..

Mrigakshi Dixit

In their treadmill experiments, UW Medicine researchers obtained the fastest walking speed (50 mm/second) ever reported for fruit flies.

Alice C. Gray  

Researchers have turned fruit flies into tiny athletes to unlock the secrets of locomotion.

The University of Washington researchers have been working to understand the neural mechanisms behind movement.

For this, they created small-scale treadmill-like machines for fruit flies. They conducted this study to understand how fruit flies’ nervous system controls movement.

Interestingly, the treadmill experiments also led to a new revelation: the “fastest walking speed” ever recorded in fruit flies.

“They were able to surpass an instantaneous walking speed of 50 millimeters per second,” the researchers noted. 

Fruit fly’s proprioception

The nervous systems of insects and humans play an important role in movement. It helps maintain balance by detecting unexpected changes and adjusting the body’s motions accordingly. This is known as proprioception, which is the body’s awareness of its own position and movement.

Animals would struggle to navigate their environment without this neurological control, putting them at risk of falling and being injured. 

However, neuroscientists have struggled to understand the underlying mechanisms of how the nervous system recognizes and responds to unexpected changes. This is why researchers turned to fruit flies.  

“These tiny creatures are a good model system to study the neural control of locomotion because they have a compact, fully mapped nervous system,” the researchers noted in the press release. 

The team used a “linear treadmill” to train flies to walk and track their movements in 3D over a longer duration. Researchers compared walking rates in flies with and without impaired proprioception.

Decoding hidden neural control

The flies on the treadmill exhibited a burst-like walking pattern, sprinting forward and then coasting backward on the belt. They spent about half their time walking and increased their pace as the treadmill speed climbed.

The study found that fruit flies can adjust their walking patterns to maintain balance and stability, even when faced with unexpected challenges.

The key highlight came when certain neurons associated with proprioception were genetically silenced. When deprived of sensory feedback, the flies took “fewer but larger steps.”

The team noticed that the flies’ leg coordination was surprisingly unaffected by the lack of sensory feedback. This suggests that either other proprioceptive neurons are more critical for walking coordination or the nervous system can adapt to compensate.

They also compared the movement pattern using the split-belt treadmill, which has two separate belts that can move at different speeds.

“The scientists found that the split-belt treadmill had little effect on the coordination between legs. However, flies substantially changed the step distances of their middle legs when the two belts moved at different speeds,” the team noted. 

“The middle legs are ideally positioned to stably pivot the body of the fly about its center of mass, like rowing a boat from its center,” the researchers added in the press release .

The study proposes that flies adjust their steps to maintain a straight path when faced with rotational disturbances. Through the split-belt treadmill experiments, researchers gained valuable insights into how fruit flies adapt to environmental changes and coordinate their leg movements for stability and locomotion.

RECOMMENDED ARTICLES

Treadmills have a long history of being used to study the neural control of locomotion in various animals, including invertebrates like cockroaches and stick insects, as well as vertebrates like rodents, cats, and humans. 

The findings were published in the Cell Press journal.

The Blueprint Daily

Stay up-to-date on engineering, tech, space, and science news with The Blueprint.

By clicking sign up, you confirm that you accept this site's Terms of Use and Privacy Policy

ABOUT THE EDITOR

Mrigakshi Dixit Mrigakshi is a science journalist who enjoys writing about space exploration, biology, and technological innovations. Her work has been featured in well-known publications including Nature India, Supercluster, The Weather Channel and Astronomy magazine. If you have pitches in mind, please do not hesitate to email her.

POPULAR ARTICLES

Roman military camp at 7,000 feet height found near historic alps battlefield, lifestraw and veinviewer, the best of ie this week, us nuclear reactor breakthrough unravels plutonium oxide’s secrets at 3000 k, does the iphone have a future, related articles.

Foreign fighter jet carrying nuclear bomb under fuselage spotted over US

• transportation

Chinese EV travels 554 miles in single charge with 150-kWh semi-solid state battery

Boeing’s Starliner capsule returns to Earth without astronauts, new issues found

Planet Nine: Astronomers could discover elusive world within months