aim of yeast fermentation experiment

The fermentation of sugars using yeast: A discovery experiment

Charles Pepin (student) and Charles Marzzacco (retired), Melbourne, FL

Introduction

Enzyme catalysis 1  is an important topic which is often neglected in introductory chemistry courses. In this paper, we present a simple experiment involving the yeast-catalyzed fermentation of sugars. The experiment is easy to carry out, does not require expensive equipment and is suitable for introductory chemistry courses.

The sugars used in this study are sucrose and lactose (disaccharides), and glucose, fructose and galactose (monosaccharides). Lactose, glucose and fructose were obtained from a health food store and the galactose from Carolina Science Supply Company. The sucrose was obtained at the grocery store as white sugar. The question that we wanted to answer was “Do all sugars undergo yeast fermentation at the same rate?”

Sugar fermentation results in the production of ethanol and carbon dioxide. In the case of sucrose, the fermentation reaction is:

\[C_{12}H_{22}O_{11}(aq)+H_2 O\overset{Yeast\:Enzymes}{\longrightarrow}4C_{2}H_{5}OH(aq) + 4CO_{2}(g)\]

Lactose is also C 12 H 22 O 11  but the atoms are arranged differently. Before the disaccharides sucrose and lactose can undergo fermentation, they have to be broken down into monosaccharides by the hydrolysis reaction shown below:

\[C_{12}H_{22}O_{11} + H_{2}O \longrightarrow 2C_{6}H_{12}O_{6}\]

The hydrolysis of sucrose results in the formation of glucose and fructose, while lactose produces glucose and galactose.

sucrose + water \(\longrightarrow\) glucose + fructose

lactose + water \(\longrightarrow\) glucose + galactose

The enzymes sucrase and lactase are capable of catalyzing the hydrolysis of sucrose and lactose, respectively.

The monosaccharides glucose, fructose and galactose all have the molecular formula C 6 H 12 O 6  and ferment as follows:

\[C_{6}H_{12}O_{6}(aq)\overset{Yeast Enzymes}{\longrightarrow}2C_{2}H_{5}OH(aq) + 2CO_{2}(g)\]

In our experiments 20.0 g of the sugar was dissolved in 100 mL of tap water. Next 7.0 g of Red Star ®  Quick-Rise Yeast was added to the solution and the mixture was microwaved for 15 seconds at full power in order to fully activate the yeast. (The microwave power is 1.65 kW.) This resulted in a temperature of about 110  o F (43  o C) which is in the recommended temperature range for activation. The cap was loosened to allow the carbon dioxide to escape. The mass of the reaction mixture was measured as a function of time. The reaction mixture was kept at ambient temperature, and no attempt at temperature control was used. Each package of Red Star Quick-Rise Yeast has a mass of 7.0 g so this amount was selected for convenience. Other brands of baker’s yeast could have been used.

This method of studying chemical reactions has been reported by Lugemwa and Duffy et al. 2,3  We used a balance good to 0.1 g to do the measurements. Although fermentation is an anaerobic process, it is not necessary to exclude oxygen to do these experiments. Lactose and galactose dissolve slowly. Mild heat using a microwave greatly speeds up the process. When using these sugars, allow the sugar solutions to cool to room temperature before adding the yeast and microwaving for an additional 15 seconds.

Fermentation rate of sucrose, lactose alone, and lactose with lactase

Fig. 1 shows plots of mass loss vs time for sucrose, lactose alone and lactose with a dietary supplement lactase tablet added 1.5 hours before starting the experiment. All samples had 20.0 g of the respective sugar and 7.0 g of Red Star Quick-Rise Yeast. Initially the mass loss was recorded every 30 minutes. We continued taking readings until the mass leveled off which was about 600 minutes. If one wanted to speed up the reaction, a larger amount of yeast could be used. The results show that while sucrose readily undergoes mass loss and thus fermentation, lactose does not. Clearly the enzymes in the yeast are unable to cause the lactose to ferment. However, when lactase is present significant fermentation occurs. Lactase causes lactose to split into glucose and galactose. A comparison of the sucrose fermentation curve with the lactose containing lactase curve shows that initially they both ferment at the same rate.

Fig. 1. Comparison of the mass of CO 2 released vs time for the fermentation of sucrose, lactose alone, and lactose with a lactase tablet. Each 20.0 g sample was dissolved in 100 mL of tap water and then 7.0 g of Red Star Quick-Rise Yeast was added.

However, when the reactions go to completion, the lactose, lactase and yeast mixture gives off only about half as much CO 2  as the sucrose and yeast mixture. This suggests that one of the two sugars that result when lactose undergoes hydrolysis does not undergo yeast fermentation. In order to verify this, we compared the rates of fermentation of glucose and galactose using yeast and found that in the presence of yeast glucose readily undergoes fermentation while no fermentation occurs in galactose.

Fig. 2. Comparison of the mass of CO 2 released vs time for the fermentation of sucrose, glucose and fructose. Each 20 g sugar sample was dissolved in 100 mL of water and then 7.0 g of yeast was added.

Fermentation rate of sucrose, glucose and fructose

Next we decided to compare the rate of fermentation of sucrose with that glucose and fructose, the two compounds that make up sucrose. We hypothesized that the disaccharide would ferment more slowly because it would first have to undergo hydrolysis. In fact, though, Fig. 2 shows that the three sugars give off CO 2  at about the same rate. Our hypothesis was wrong. Although there is some divergence of the three curves at longer times, the sucrose curve is always as high as or higher than the glucose and fructose curves. The observation that the total amount of CO 2  released at the end is not the same for the three sugars may be due to the purity of the fructose and glucose samples not being as high as that of the sucrose.

Fermentation rate and sugar concentration

Next, we decided to investigate how the rate of fermentation depends on the concentration of the sugar. Fig. 3 shows the yeast fermentation curves for 10.0 g and 20.0 g of glucose. It can be seen that the initial rate of CO 2  mass loss is the same for the 10.0 and 20.0 g samples. Of course the total amount of CO 2  given off by the 20.0 g sample is twice as much as that for the 10.0 g sample as is expected. Later, we repeated this experiment using sucrose in place of glucose and obtained the same result.

Fig. 3. Comparison of the mass of CO 2  released vs time for the fermentation of 20.0 g of glucose and 10.0 g of glucose. Each sugar sample was dissolved in 100 mL of water and then 7.0 g of yeast was added.

Fermentation rate and yeast concentration

After seeing that the rate of yeast fermentation does not depend on the concentration of sugar under the conditions of our experiments, we decided to see if it depends on the concentration of the yeast. We took two 20.0 g samples of glucose and added 7.0 g of yeast to one and 3.5 g to the other. The results are shown in Fig. 4. It can clearly be seen that the rate of CO 2  release does depend on the concentration of the yeast. The slope of the sample with 7.0 g of yeast is about twice as large as that with 3.5 g of yeast. We repeated the experiment with sucrose and fructose in place of glucose and obtained similar results.

Fig. 4. Comparison of the mass of CO 2 released vs time for the fermentation of two 20.0 g samples of glucose dissolved in 100 mL of water. One had 7.0 g of yeast and the other had 3.5 g of yeast.

In hindsight, the observation that the rate of fermentation is dependent on the concentration of yeast but independent of the concentration of sugar is not surprising. Enzyme saturation can be explained to students in very simple terms. A molecule such as glucose is rather small compared to a typical enzyme. Enzymes are proteins with large molar masses that are typically greater than 100,000 g/mol. 1  Clearly, there are many more glucose molecules in the reaction mixture than enzyme molecules. The large molecular ratio of sugar to enzyme clearly means that every enzyme site is occupied by a sugar molecule. Thus, doubling or halving the sugar concentration cannot make a significant difference in the initial rate of the reaction. On the other hand, doubling the concentration of the enzyme should double the rate of reaction since you are doubling the number of enzyme sites.

The experiments described here are easy to perform and require only a balance good to 0.1 g and a timer. The results of these experiments can be discussed at various levels of sophistication and are consistent with enzyme kinetics as described by the Michaelis-Menten model. 1  The experiments can be extended to look at the effect of temperature on the rate of reaction. For enzyme reactions such as this, the reaction does not take place if the temperature is too high because the enzymes get denatured. The effect of pH and salt concentration can also be investigated.

• Jeremy M. Berg, John L. Tymoczko and Lubert Stryer,  Biochemistry , 6th edition, W.H. Freeman and Company, 2007, pages 205-237.
• Fugentius Lugemwa, Decomposition of Hydrogen Peroxide,  Chemical Educator , April 2013, pages 85-87.
• Daniel Q. Duffy, Stephanie A. Shaw, William D. Bare, Kenneth A. Goldsby, More Chemistry in a Soda Bottle, A Conservation of Mass Activity,  Journal of Chemical Education , August 1995, pages 734-736.
• Jessica L Epstein, Matthew Vieira, Binod Aryal, Nicolas Vera and Melissa Solis, Developing Biofuel in the Teaching Laboratory: Ethanol from Various Sources,  Journal of Chemical Education , April 2010, pages 708–710.

More about April 2015

Department of Chemistry 200 University Ave. W Waterloo, Ontario, Canada N2L 3G1

[email protected]

Choose an Account to Log In

Notifications

Science project, growing yeast: sugar fermentation.

Yeast is most commonly used in the kitchen to make dough rise. Have you ever watched pizza crust or a loaf of bread swell in the oven? Yeast makes the dough expand. But what is yeast exactly and how does it work? Yeast strains are actually made up of living eukaryotic microbes, meaning that they contain cells with nuclei. Being classified as fungi (the same kingdom as mushrooms), yeast is more closely related to you than plants! In this experiment we will be watching yeast come to life as it breaks down sugar, also known as sucrose , through a process called fermentation . Let’s explore how this happens and why!

What is sugar’s effect on yeast?

• 3 Clear glass cups
• 2 Teaspoons sugar
• Water (warm and cold)
• 3 Small dishes
• Permanent marker

• Fill all three dishes with about 2 inches of cold water
• Place your clear glasses in each dish and label them 1, 2, and 3.
• In glass 1, mix one teaspoon of yeast, ¼ cup of warm water, and 2 teaspoons of sugar.
• In glass 2, mix one teaspoon of yeast with ¼ cup of warm water.
• In glass 3, place one teaspoon of yeast in the glass.
• Observe each cups reaction. Why do you think the reactions in each glass differed from one another? Try using more of your senses to evaluate your three glasses; sight, touch, hearing and smell especially!

The warm water and sugar in glass 1 caused foaming due to fermentation. 

Fermentation is a chemical process of breaking down a particular substance by bacteria, microorganisms, or in this case, yeast. The yeast in glass 1 was activated by adding warm water and sugar. The foaming results from the yeast eating the sucrose. Did glass 1 smell different? Typically, the sugar fermentation process gives off heat and/or gas as a waste product. In this experiment glass 1 gave off carbon dioxide as its waste.

Yeast microbes react different in varying environments. Had you tried to mix yeast with sugar and cold water, you would not have had the same results. The environment matters, and if the water were too hot, it would kill the yeast microorganisms. The yeast alone does not react until sugar and warm water are added and mixed to create the fermentation process. To further investigate how carbon dioxide works in this process, you can mix yeast, warm water and sugar in a bottle while attaching a balloon to the open mouth. The balloon will expand as the gas from the yeast fermentation rises.

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.

• The Open University
• Accessibility hub
• Guest user / Sign out
• Study with The Open University

My OpenLearn Profile

Personalise your OpenLearn profile, save your favourite content and get recognition for your learning

About this free course

Become an ou student, download this course, share this free course.

Start this free course now. Just create an account and sign in. Enrol and complete the course for a free statement of participation or digital badge if available.

3.1.3 Yeast experiment explained

You’ve seen the results of the yeast experiment, but what do these results mean?

Yeasts are microscopic, single-celled organisms, and are a type of fungus that is found all around us, in water, soil, on plants, on animals and in the air. Like all organisms, when yeasts are put in the right type of environment they will thrive; growing and reproducing.

Your experiments were designed to help you identify which environment promotes the most yeast growth. The first three glasses in your experiment contained different temperature environments (cold water, hot water and body temperature water). At very low temperatures the yeast simply does not grow but it is still alive – if the environment were to warm up a bit, it would gradually begin to grow. At very high temperatures the cells within the yeast become damaged beyond repair and even if the temperature of that environment cooled, the yeast would still be unable to grow. At optimum temperatures the yeast thrives.

Your third and fourth glasses both contained environments at optimum temperature (body temperature) for yeast growth, the difference being, the fourth glass was sealed. The variable between these two experiments was the amount of available oxygen. You may have been surprised by your results here, thinking that a living organism in an environment without oxygen cannot survive? However, you should have found that yeast grew pretty well in both experiments.

To understand why yeast was able to thrive in both conditions we need to understand the chemical process occurring in each glass during the experiment. In the three open glasses, oxygen is readily available, and from the moment you added the yeast to the sugar solution it began to chemically convert the sugar in the water and the oxygen in the air into energy, water, and carbon dioxide in a process called aerobic respiration.

Yeast is a slightly unusual organism – it is a ‘facultative anaerobe’. This means that in oxygen-free environments they can still survive. The yeast simply switches from aerobic respiration (requiring oxygen) to anaerobic respiration (not requiring oxygen) and converts its food without oxygen in a process known as fermentation. Due to the absence of oxygen, the waste products of this chemical reaction are different and this fermentation process results in carbon dioxide and ethanol.

Depending on how long you monitored your experiment for and how much space your yeast had to grow you may have noticed that, with time, the experiment sealed with cling film slowed down. This is for two reasons; firstly because less energy is produced by anaerobic respiration than by aerobic respiration and, secondly, because the ethanol produced is actually toxic to the yeast. As the ethanol concentration in the environment increases, the yeast cells begin to get damaged, slowing their growth.

The ethanol produced is a type of alcohol, so it is this process that allows us to use it to make beer and wine. When used in bread making, the yeast begins by respiring aerobically, the carbon dioxide from which makes the bread rise. Eventually the available oxygen is used up, and the yeast switches to anaerobic respiration producing alcohol and carbon dioxide instead. Do not worry though; this alcohol evaporates during the baking process, so you won’t get drunk at lunchtime from eating your sandwiches.

Sciencing_Icons_Science SCIENCE

Sciencing_icons_biology biology, sciencing_icons_cells cells, sciencing_icons_molecular molecular, sciencing_icons_microorganisms microorganisms, sciencing_icons_genetics genetics, sciencing_icons_human body human body, sciencing_icons_ecology ecology, sciencing_icons_chemistry chemistry, sciencing_icons_atomic & molecular structure atomic & molecular structure, sciencing_icons_bonds bonds, sciencing_icons_reactions reactions, sciencing_icons_stoichiometry stoichiometry, sciencing_icons_solutions solutions, sciencing_icons_acids & bases acids & bases, sciencing_icons_thermodynamics thermodynamics, sciencing_icons_organic chemistry organic chemistry, sciencing_icons_physics physics, sciencing_icons_fundamentals-physics fundamentals, sciencing_icons_electronics electronics, sciencing_icons_waves waves, sciencing_icons_energy energy, sciencing_icons_fluid fluid, sciencing_icons_astronomy astronomy, sciencing_icons_geology geology, sciencing_icons_fundamentals-geology fundamentals, sciencing_icons_minerals & rocks minerals & rocks, sciencing_icons_earth scructure earth structure, sciencing_icons_fossils fossils, sciencing_icons_natural disasters natural disasters, sciencing_icons_nature nature, sciencing_icons_ecosystems ecosystems, sciencing_icons_environment environment, sciencing_icons_insects insects, sciencing_icons_plants & mushrooms plants & mushrooms, sciencing_icons_animals animals, sciencing_icons_math math, sciencing_icons_arithmetic arithmetic, sciencing_icons_addition & subtraction addition & subtraction, sciencing_icons_multiplication & division multiplication & division, sciencing_icons_decimals decimals, sciencing_icons_fractions fractions, sciencing_icons_conversions conversions, sciencing_icons_algebra algebra, sciencing_icons_working with units working with units, sciencing_icons_equations & expressions equations & expressions, sciencing_icons_ratios & proportions ratios & proportions, sciencing_icons_inequalities inequalities, sciencing_icons_exponents & logarithms exponents & logarithms, sciencing_icons_factorization factorization, sciencing_icons_functions functions, sciencing_icons_linear equations linear equations, sciencing_icons_graphs graphs, sciencing_icons_quadratics quadratics, sciencing_icons_polynomials polynomials, sciencing_icons_geometry geometry, sciencing_icons_fundamentals-geometry fundamentals, sciencing_icons_cartesian cartesian, sciencing_icons_circles circles, sciencing_icons_solids solids, sciencing_icons_trigonometry trigonometry, sciencing_icons_probability-statistics probability & statistics, sciencing_icons_mean-median-mode mean/median/mode, sciencing_icons_independent-dependent variables independent/dependent variables, sciencing_icons_deviation deviation, sciencing_icons_correlation correlation, sciencing_icons_sampling sampling, sciencing_icons_distributions distributions, sciencing_icons_probability probability, sciencing_icons_calculus calculus, sciencing_icons_differentiation-integration differentiation/integration, sciencing_icons_application application, sciencing_icons_projects projects, sciencing_icons_news news.

• Share Tweet Email Print
• Home ⋅
• Science Fair Project Ideas for Kids, Middle & High School Students ⋅

Biology Experiments on the Fermentation of Yeast

What Are Some Common Uses of Yeast?

Yeast is a fungal microorganism that man has usedsince before he had a written word. Even to this day, it remains a common component of modern beer and bread manufacture. Because it is a simple organism capable of rapid reproduction and even faster metabolism, yeast is an ideal candidate for simple biology science experiments that involve the study of fermentation.

What is Fermentation?

Fermentation is the biological process by which yeast consumes simple sugars and releases alcohol and carbon dioxide. For the most part, fermentation requires a mostly aquatic environment to occur. Different yeasts respond differently to changes in environment, making some better for baking and others for brewing. Bakers use fermentation to add CO2 bubbles to bread dough. During baking, these bubbles make the bread light and fluffy while the alcohol boils away. Brewers take care to preserve the alcohol of fermentation and use the CO2 to help build a frothy head for their potent beverages.

Indirect Life Test Experiments

The first experiment that should come to mind when examining yeast is determining whether or not yeast is a living organism. While it would be easy to rely on foreknowledge about the nature of yeast, more is learned by application of scientific method. If yeast is alive, it should consume food, respire and reproduce. Indirect tests look for clues that these processes are taking place. For such experiments, you should measure the amount of CO2 released by yeast that are digesting sugar water in test tubes with balloons attached. Use Benedict's solution to test for the presence of sugar in the final product.

Salinity Experiments

Fermentation is a delicate process that relies on ideal conditions to occur. Experiments that study how it responds to salinity are of particular interest to science and industry alike. Your project can either take a single type of yeast and vary the amount of salt in the solution to see if there is an ideal salinity, or alternately, use various yeasts to see how they respond to the same level of salt. In the latter experiment, make sure to use yeasts from many industries, since most baker's yeasts fare poorly in saline conditions.

Sugar Experiments

While it's clear that yeast requires sugar for fermentation, there are many different sugars that yeast could use for fuel. You can perform a number of experiments to determine which ones promote the highest level of yeast growth. In one, you can add yeast to various beverages, such as fruit juices and non-carbonated sports drinks to see which environment produces the most CO2. Another can use various sweeteners such as granulated sugars, syrups and nectars (such as agave) placed in weak solutions. You can measure CO2 production with balloons placed over the reacting test tubes, or simply observe the bubbles produced and make a relative comparison.

Related Articles

How to make a vitamin c indicator, what are the independent variables for a moldy bread..., science fair project on the effect of soda on the body, cell respiration experiments, 5 uses of fermentation, girly science fair project ideas, teeth science projects, science project for the effects that beverages have..., how-to science experiments for kids with iodine and..., science projects for cut flowers, simple & easy science fair projects, the effect of temperature on pea respiration, experiments on which mouthwash kills bacteria, chemistry science fair projects, scientific experiment ideas for ap biology, definitions of control, constant, independent and dependent..., cheese mold experiments, how to make a 1% sucrose solution, what happens when there is no oxygen available at the....

• The Science of Cooking: Yeast Air Balloons

About the Author

Andy Klaus started his writing career contributing science and fiction articles to Dickinson High School's newsletters back in 1984. Since then, he has authored novels and written technical books for health-care companies such as VersaSuite. He has covered topics varying from aerospace to zoology and received an associate degree in science from College of the Mainland.

Find Your Next Great Science Fair Project! GO

• Fermentation of Food

Testing Substrate Specificity in Yeast Fermentation

As Buchner discovered at the turn of the 20th century, the process of fermentation is a multistep, enzyme-catalyzed reaction. Inherent to maximal enzyme action is a defined set of optimal conditions and substrates. Therefore, the enzymes responsible for glycolysis and subsequent fermentation reactions will exhibit optimal reaction rates in an environment that mimics physiological conditions. To demonstrate this point, we will use “active dry” yeast packets from the supermarket (such as Fleischmann’s or Red Star brands). This common product is a freeze dried collection of Saccharomyces cerevisiae , also known as “baker’s yeast.”

Naturally found on ripe fruits, like grapes, as well as on and in the human body, S. cerevisiae is a facultative anaerobe, and its biodiversity and carbon utilization is dictated by the carbon and energy sources available in its specific habitat. This demonstrates the incredible metabolic flexibility housed in these tiny eukaryotes, however, some substrates are more efficiently metabolized than others. In the absence of oxygen, S. cerevisiae will switch on its fermentation pathway as a mechanism to maintain a favorable cellular redox status (fermentation regenerates NAD + , which is essential for glycolysis), generating ethanol and carbon dioxide as byproducts. 

To understand how different sugar substrates are utilized by S. Cerevisiae , we can measure the amount of CO 2 produced. If you recall the stoichiometry for fermentation, for every mole of glucose, yeast cells will produce two moles of CO 2 , which makes a quantification of sugar metabolism fairly straightforward. While scientists have invented a number of devices to quantifiably measure the rate CO 2 production resulting from fermentation in yeast, these devices are not practical for classroom settings. Here we will use a basic 15ml conical tube (such as Falcon or Corning brands) with gradation markings as a device to measure CO 2 production in response to a variety of carbon sources.

Corner Store: Grocery Items

• Dextrose (glucose), Galactose, Lactose, Maltose, Sucrose (table sugar)
• Baker’s Yeast Packets or a jar

Common Items

Laboratory equipment.

• Water bath, hot plate, or pan & stove
• Beakers or flat-bottomed bowl/baking dish
• Micropipettes & tips, sterile transfer pipettes (or straws or medicine syringe)
• 15mL Conical Tubes and a needle OR test tubes and glass slides to cover

Preparation

• Prepare sugar stock solutions at 40% w/v (40g per 100ml of H 2 O)
• For a recommended starting concentration of 0.5%, v/v, you’ll need 0.5 ml stock in 10 ml final volume
• Heat water baths and/or hot plates to 4 0 ° C
• Warm all solutions to be used in the 40 °C bath
• Prepare a gas-collection setup (see below)
• Immediately before the experiment, prepare the 7% yeast solution in H 2 O

In the example at right, plastic tubes were used with a few holes poked into each plastic cap using a needle. The tubes were filled with yeast solution and food source (e.g. sucrose), covered, and inverted quickly into the water bath. The samples should be held inside the tubes and displaced into the water bath as gas is produced.

A similar setup can be used with test tubes and stoppers with holes (cover with your finger when inverting) or using a glass slide to invert each tube into the bath before sliding off the slide and allowing the fermentation gas produced to displace the yeast solution. You may also be able to use a typical gas collection apparatus from a chemistry laboratory at your school.

Finally, volumes of gas can be determined by emptying a marked tube and filling it with water exactly to the mark. Then that water can be poured into a graduated cylinder of appropriate size to determine the volume of gas produced.

• Fill 15ml conical tube with 8ml of a sugar solution
• Mix the 7% yeast solution to be a uniform suspension. Fill the remainder of the tube (~7ml) with yeast solution such that the meniscus rises above the lip of the tube. 
• NOTE: make sure that there are no sizeable air bubbles in the tube
• NOTE: you can also test the effect of temperature on fermentation by adjusting temperature of water bath or hot plate
• Immediately mark the bottom of the CO 2 bubble (if there is one). Mark this point at 5 minute intervals for 30 minutes.
• At the end of the experiment, record the level of CO 2 produced at each time interval by emptying the tube, filling with water to the mark, and pouring the water into a graduated cylinder or onto a balance for accurate measurement

What conclusions can you draw about the metabolism efficiency of different substrates by  S. cerevisiae ?

Related resources

Biochemical deconstruction of an egg sandwich, the chemistry that holds us together.

• Everyday Emulsions

Escape the Cell

Science is lit, science saturday special: ants, beyond dirt: the ecosystem beneath our feet, virus hunters.

• Fruit Fly Behavior
• D4P: People and Medicine
• D4P: Social Behavior
• 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

• Science & Math
• Sociology & Philosophy
• Law & Politics

Lab Explained: Production of Yeast Fermentation

• Lab Explained: Production of Yeast…

Introduction

Yeast is commonly known as a baking material for making bread and beer. However, under biological jargon it is a group of eukaryotic, single-celled microorganisms that makes up almost 1% of all fungal species. 1 They are found in soils and on plant surfaces, like flower nectar and fruits, and they reproduce asexually through budding. This is where a small daughter cell grows on the parent cell whose nucleus then duplicates and then separates with the daughter cell. This reproduction can happen at a rate of up to once every 90 minutes. 2 However, some yeast is an exception and reproduces by binary fission, a process in which the DNA duplicates and the cytoplasm splits evenly into two identical daughter cells.

Most importantly, yeast undergoes metabolism without the presence of oxygen, thus respiring anaerobically. The carbon dioxide that is produced is what makes bread rise when using the specific type; Saccharomyces cerevisiae , also known as baker’s yeast.

The breakdown of glucose by yeast:

C6H12O6 → 2C2H5OH + 2CO2

Glucose → ethanol + carbon dioxide

The glucose is broken down by glycolysis. However, there is more to the process than this simple arrow. When respiring anaerobically, yeast performs glycolysis where it converts pyruvate further into ethanol and carbon dioxide. Although the process requires 2 ATP, it produces 4; a net gain of 2 ATP molecules that can be used for energy. Fermentation rate is affected by factors like temperature, saccharide concentration, and type of sugar solution.

Monosaccharides are the monomers 3 of carbohydrates, consisting of glucose, fructose, and galactose. They provide quick, accessible energy that is easily broken down. Disaccharides are made of two monosaccharides, consisting of lactose, maltose, and sucrose. They are held together by a covalent bond 4 through a condensation reaction. 5 Thus, their more complex structures make them harder to break down compared to monosaccharides. These are the sugars and the monomers each sugar is made up of.

Investigations

2.1 hypothesis.

Glucose will produce the largest amount of carbon dioxide, followed, by fructose, sucrose, and lactose. Although they all have similar chemical formulas, they differ in structure and

stereochemistry. 6 The monosaccharides will perform better because they require less energy to be broken down and thus create the product of carbon dioxide at a faster and higher rate.

H : The carbon dioxide produced will be the same for each sugar type.

H 1 : The addition of monosaccharides will produce more carbon dioxide than disaccharides.

2.2 Variables

Independent Variable: Type of saccharide

Dependent Variable: Time of each observation (minutes)

Controlled Variables:

• Beaker size: 250ml
• Keeping this consistent would allow the same space for each reaction to take place.
• Amount of yeast: 2.0g
• Type of yeast: Alnatura Backhefe
• Ingredients: Yeast dried from organic farming
• This was important to monitor because different yeast that has different origins could have different reactions.
• Amount of distilled water: 50ml
• Maintaining a consistent amount of ingredients ensured that the chemical reactions would have fair, equal environments.
• Amount of saccharide: 5.0g

Uncontrolled Variables:

• Room temperature: 21°C
• This was monitored using a thermometer. This was important to keep consistent because increased temperatures speed up reactions. However, the lab is quite large, so it was impossible to control this variable.

2.3 Preliminary Experiment

A preliminary experiment was carried out a day before our designated internal assessment time. I thought it was appropriate to prepare for my actual experiment to ensure that all the materials were working, and my procedure demonstrates efficacy. My method was identical to that of my real one, however, I did make one change. Originally, my saccharide to distilled water ratio was 5g:100ml. This was not fitting for my time frame of 30 minutes. Thus, I changed my measurements and made the solution much more concentrated; 5g:50ml. This increased the speed of the reaction so I could take more readings within a smaller span of time, allowing me to also improve my reliability. 7

In addition to adapting my method, I also conducted a control test. I mixed 2g of yeast with 50ml of distilled water and no carbon dioxide was released. Now that I had a method that worked, I could safely move on to my final internal assessment procedure.

3.1 Apparatus

• Gas Syringe – to measure produced carbon dioxide (±0.05ml)
• Mixer – to create and mix the solutions of a saccharide and distilled water
• Stir Rod (or Flea) – magnet used to mix the solutions
• Stands – to hold apparatus in place
• Scale – to measure necessary materials (± 0.05g)
• Stopper – to prevent any carbon dioxide loss throughout the structure
• Clamps – to hold flask and gas syringe in place
• Graduated cylinder – to measure the amount of each ingredient (±0.1ml)
• Flask – to contain the site of the reaction
• Rubber Tube – to connect the glass pipes
• Baking Yeast (Backhafer)
• Glucose, Fructose, Sucrose, Lactose
• Distilled Water – to create solution that is added to yeast

3.2 Methodology

• Pour 2g of yeast into the flask
• Add 50ml distilled water
• Put mixer on 500 rpm
• Record results at both 15 and 30
• The mixture should not release any carbon dioxide, as there is no respiration.

Saccharide Experiment:

• Mix 5g of the saccharide with 50ml distilled water in a beaker until clear
• Make sure the valve is open so the carbon dioxide can flow through to the gas syringe
• Pour the saccharide and distilled water solution into the flask
• Immediately close the flask with the stopper
• Immediately start timer
• Record results at 15 minutes (from gas syringe)
• Stop timer at 30 minutes and record results
• Clean apparatus between uses
• Repeat 4 times per sugar for each sugar

3.3 Justification

All of my measurements for my materials were done with a digital scale where I could control how much of each ingredient I was using down to the nearest milligram. I additionally used two of them, in case one were to malfunction or give a different result. Therefore, I know that I did not have any false readings or amounts or yeast, sugars, or distilled water. Unless, of course, both digital apparatuses malfunctioned.

I repeated my method four times with each sugar to increase reliability and make sure I was getting consistent results. Additionally, when a trial went wrong in any way, I would clean up and re-do it. For example, one time I realized that I had labelled my flask incorrectly and lost track of which sugar I had put in the solution. I properly disposed of it, cleaned the materials, and started over to ensure that I was being accurate, and no leftover waste or excess ingredients would affect the results i.e. each experiment would be independent from one another.

3.4 Risk Assessment

Safety issues: None of the materials I used were harmful or put me at risk in any way.

Ethical issues: Nothing in my internal assessment could be deemed unethical in any way. Environmental issues: The disposal of yeast in a drainage or water system can result in adecrease of oxygen for anything that is further down the pipes. This oxygen shortage can throw off the balance of any ecosystems associated with the drain in question. Thus, I diluted the yeast mixture with water and then disposed of it in the waste container.

Raw and Processed Data

4.1 qualitative observations.

• The saccharide mixture combined with the yeast started to create a layer of froth/bubbles at the top.
• The mixture previously mentioned also left condensation all over the inside of the flask.
• Some sugars were harder to dissolve in the distilled water than others, namely the fructose. This is most likely due to the more complex structure that makes it harder to mingle with the H2O. The more monomers and bonds within a compound, the more difficult it is for any other molecule, in this case water, to interfere and dissolve it.

4.2 Raw Data

4.3 Processed Data

From this graph, we can very well see that some sugars simply performed better, specifically glucose and fructose. The disaccharides – fructose and lactose – had a much lower fermentation rate. Thus, my hypothesis is being supported.

Had I not discarded the anomalies present in my raw data for sucrose, the standard deviations would have been over 5x as large. Because standard deviation is the value describing how much the data differs from the mean, it is very evident that such a large change in the value all due to one anomaly was unnecessary in my final conclusions and analysis of data. It would have negatively impacted the scope of my investigation because it is very obviously due to an error in my method (see weaknesses).

5.1 Conclusion

Looking at table 3 and graph 1, the hierarchy of efficacy in producing carbon dioxide reads as follows; glucose, fructose, sucrose, and lactose. When speaking in terms of standard deviation, sucrose and lactose had the smallest. Table 4 further supports this. Thus, sucrose and lactose were the most consistent and differed the least from their mean. This could be due to the substances being stored correctly or never having been disrupted since their extraction from the lab they were bought from. From table 4, we can further infer that since the F value is greater than the F crit value, the values are not at all equal, rejecting my null hypothesis.

A higher standard deviation, as seen in fructose, could be a result of improper care of the sugars themselves. For example, placing them in warm temperatures or letting them be vulnerable to sunlight could cause them to melt or change in structure and thus alter from molecule to molecule.

Regarding other academic investigations of yeast, my investigation gave results that were expected by other biologists. To quote from the Journal of Undergraduate Biology Laboratory Investigations, “The monosaccharides we used in our experiments (glucose and honey)produced a higher rate of CO2 than the disaccharides (refined sucrose and lactose) did.” This directly aligns with my alternative hypothesis and experiment results.

From my results, it can be concluded that yeast is affected differently depending on the type of saccharide it encounters. Thus, it supports my hypothesis: Glucose will produce the largest

amount of carbon dioxide, followed, by fructose, sucrose, and lactose. As mentioned before, their chemical formulas being almost identical has little to do with it. It is all in the structure. The monosaccharides require less energy to be broken down and thus create the product of carbon dioxide. This is proven by the fact that every cell has the ability to break down glucose, whereas only the liver can break down fructose. 9 Even then, when breaking down fructose, it is first converted into glucose where then glycolysis can take place.

5.2 Improvements

The apparatus I used was not optimal. For example, my mixer did not have a digital display of how fast my rpm was. Thus, I could have been mixing the yeast and sugar solutions at different speeds for each trial. The speed could have affected the rate of the reaction due to how well the molecules were interacting and colliding with each other, and thus affecting the production of carbon dioxide.

In addition, I used quite a large flask compared to the 50ml of solution I had to use. It was made for 250ml, so it is imaginable how much extra space was within the container. This may have decreased the accuracy of the measurements I took from the gas syringe, because there was also gas trapped in the flask itself. Perhaps I could have additionally measured the volume of the container that did not have the solution (the gap between the liquid surface and the gas syringe) and added that to my volumetric values of carbon dioxide to increase my accuracy.

Finally, the yeast packages I used were all from the same store, however it is possible that they were sourced from different farms. Therefore, the structure of the yeast molecules used could have differed due to different methods used while they were farmed. Albeit, this would be a very difficult, arbitrary thing to track, and is a stretch in affecting the results of my assessment.

What my investigation did do extremely well was independence. It is extremely unlikely that and ingredients unintentionally mixed or met due to my rigorous cleaning and replacing of the apparatus. In addition, my method of using the mixer was helpful because it ensured the solutions were extremely even and constant throughout (e.g., no leftover sugar at the bottom).

5.3 Extensions

To extend the current scope of my study, I would be enthusiastic to try and see how different types of yeast would affect my results. In this internal assessment, I only used brewers/baker’s yeast that is simply dried from a farm. However, there are hundreds of other kinds found in a variety of different places. I would measure how these saccharides, and possibly also more, e.g. maltose, affect the fermentation rate of yeasts like torula (used to create paper) and fission yeast (alternative to brewing yeast) or fresh yeast.

Related Posts

• Glucose & Sucrose Fermentation: Carbon Dioxide Production Lab Answers
• Clam Dissection Lab: Explained
• Percent Composition Lab: Explained
• The Energy of Phase Changes Lab Explained
• TLC of ASPIRIN: Lab Explained

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

Save my name, email, and website in this browser for the next time I comment.

Post comment

An official website of the United States government

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

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

• Publications
• Account settings
• My Bibliography
• Collections
• Citation manager

Save citation to file

Email citation, add to collections.

• Create a new collection
• Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

• Search in PubMed
• Search in NLM Catalog
• Add to Search

Mechanistic insights into rumen function promotion through yeast culture ( Saccharomyces cerevisiae ) metabolites using in vitro and in vivo models

Affiliations.

• 1 Key Laboratory of Animal Nutrition and Feed Science of Jilin Province, Key Laboratory of Animal Production Product Quality and Security Ministry of Education, JLAU-Borui Dairy Science and Technology R&D Center, College of Animal Science and Technology, Jilin Agricultural University, Changchun, China.
• 2 Postdoctoral Scientific Research Workstation, Feed Engineering Technology Research Center of Jilin Province, Changchun Borui Science and Technology Co., Ltd., Changchun, China.
• 3 College of Life Sciences, Engineering Research Center of Bioreactor and Pharmaceutical Development, Ministry of Education, Jilin Agricultural University, Changchun, China.
• PMID: 39081884
• PMCID: PMC11287897
• DOI: 10.3389/fmicb.2024.1407024

Introduction: Yeast culture (YC) enhances ruminant performance, but its functional mechanism remains unclear because of the complex composition of YC and the uncertain substances affecting rumen fermentation. The objective of this study was to determine the composition of effective metabolites in YC by exploring its effects on rumen fermentation in vitro , growth and slaughter performance, serum index, rumen fermentation parameters, rumen microorganisms, and metabolites in lambs.

Methods: In Trial 1, various YCs were successfully produced, providing raw materials for identifying effective metabolites. The experiment was divided into 5 treatment groups with 5 replicates in each group: the control group (basal diet without additives) and YC groups were supplemented with 0.625‰ of four different yeast cultures, respectively (groups A, B, C, and D). Rumen fermentation parameters were determined at 3, 6, 12, and 24 h in vitro. A univariate regression model multiple factor associative effects index (MFAEI; y) was established to correlate the most influential factors on in vitro rumen fermentation with YC metabolites (x). This identified the metabolites promoting rumen fermentation and optimal YC substance levels. In Trial 2, metabolites in YC not positively correlated with MFAEI were excluded, and effective substances were combined with pure chemicals (M group). This experiment validated the effectiveness of YC metabolites in lamb production based on their impact on growth, slaughter performance, serum indices, rumen parameters, microorganisms, and metabolites. Thirty cross-generation rams (Small tail Han-yang ♀ × Australian white sheep ♂) with good body condition and similar body weight were divided into three treatment groups with 10 replicates in each group: control group, YC group, pure chemicals combination group (M group).

Results: Growth performance and serum index were measured on days 30 and 60, and slaughter performance, rumen fermentation parameters, microorganisms, and metabolites were measured on day 60. The M group significantly increased the dressing percentage, and significantly decreased the GR values of lambs ( p < 0.05). The concentration of growth hormone (GH), Cortisol, insulin (INS), and rumen VFA in the M group significantly increased ( p < 0.05).

Discussion: These experiments confirmed that YC or its screened effective metabolites positively impact lamb slaughter performance, rumen fermentation, and microbial metabolism.

Keywords: growth performance; lambs; microorganism; rumen; slaughter performance; yeast culture metabolites.

Copyright © 2024 Chen, Xiao, Zhao, Li, Zhao, Zhang, Xin, Han, Wang, Aschalew, Zhang, Wang, Qin, Sun and Zhen.

PubMed Disclaimer

Conflict of interest statement

XC, JX, WeZ, WGZ, LX, ZH, LW, TW, ZS, and YZ were employed by Changchun Borui Science and Technology Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Venn diagram of the metabolites…

Venn diagram of the metabolites in products A, B, C, and D.

Effect of different yeast culture…

Effect of different yeast culture (YC) on (A–K) rumen fermentation parameters, (L) single-factor…

Correlation analysis between different yeast…

Correlation analysis between different yeast cultures and multiple factor associative effects index (MFAEI).…

Effects of combined yeast culture…

Effects of combined yeast culture (YC) and pure chemicals (M) on the serum…

Effects of combined yeast culture (YC) and pure chemicals (M) on rumen fermentation…

Effects of combined yeast culture (YC) and pure chemicals (M) on ruminal bacterial…

Effect of combined yeast culture…

Effect of combined yeast culture (YC) and pure chemicals (M) on rumen metabolites.…

Mechanistic insights into the rumen…

Mechanistic insights into the rumen function promotion by yeast metabolites using in vitro…

• Abu El-Kassim M., Abdou S., Hassan E., Abdullah M. (2021). Effect of macroalgae and yeast culture on body performance, blood metabolites, ruminal fermentation and digestibility coefficients of Ossimi lambs. Arch. Agric. Sci. J. 4, 156–167. doi: 10.21608/AASJ.2021.77678.1065 - DOI
• Alzahal O., Dionissopoulos L., Laarman A. H., Walker N., Mcbride B. W. (2014). Active dry Saccharomyces cerevisiae can alleviate the effect of subacute ruminal acidosis in lactating dairy cows. J. Dairy Sci. 97, 7751–7763. doi: 10.3168/jds.2014-8212, PMID: - DOI - PubMed
• Amin M. R., Onodera R. (1997). Synthesis of phenylalanine and production of other related compounds from phenylpyruvic acid and phenylacetic acid by ruminal bacteria, protozoa, and their mixture in vitro. J. Gen. Appl. Microbiol. 43, 9–15. doi: 10.2323/jgam.43.9, PMID: - DOI - PubMed
• Arfuso F., Minuti A., Liotta L., Giannetto C., Trevisi E., Piccione G., et al. . (2023). Stress and inflammatory response of cows and their calves during peripartum and early neonatal period. Theriogenology 196, 157–166. doi: 10.1016/j.theriogenology.2022.11.019, PMID: - DOI - PubMed
• Bäckhed F., Ding H., Wang T., Hooper L. V., Koh G. Y., Nagy A., et al. . (2004). The gut microbiota as an environmental factor that regulates fat storage. Proc. Natl. Acad. Sci. USA 101, 15718–15723. doi: 10.1073/pnas.0407076101, PMID: - DOI - PMC - PubMed

Related information

Grants and funding, linkout - more resources, full text sources.

• Frontiers Media SA
• PubMed Central

• Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.

Remember Me

Shop Experiment Sugar Fermentation by Yeast Experiments​

Sugar fermentation by yeast.

Experiment #24 from Investigating Chemistry through Inquiry

Introduction

Yeast can metabolize sugar in two ways, aerobically , with the aid of oxygen, or anaerobically , without oxygen. When yeast metabolizes a sugar under anaerobic conditions, ethanol (CH 3 CH 2 OH) and carbon dioxide (CO 2 ) gas are produced. An equation for the fermentation of the simple sugar glucose (C 6 H 12 O 6 ) is:

The metabolic activity of yeast can be determined by the measurement of gas pressure inside the fermentation vessel.

In the Preliminary Activity, you will use a Gas Pressure Sensor to monitor the pressure inside a test tube as yeast metabolizes glucose anaerobically. When data collection is complete, you will perform a linear fit on the resultant graph to determine the fermentation rate.

After completing the Preliminary Activity, you will first use reference sources to find out more about sugar fermentation by yeast before you choose and investigate a researchable question dealing with fermentation.

Sensors and Equipment

This experiment features the following sensors and equipment. Additional equipment may be required.

Ready to Experiment?

Ask an expert.

Get answers to your questions about how to teach this experiment with our support team.

• Call toll-free: 888-837-6437
• Chat with Us
• Email [email protected]

Purchase the Lab Book

This experiment is #24 of Investigating Chemistry through Inquiry . The experiment in the book includes student instructions as well as instructor information for set up, helpful hints, and sample graphs and data.

share this!

August 1, 2024

This article has been reviewed according to Science X's editorial process and policies . Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

peer-reviewed publication

trusted source

When it comes to DNA replication, humans and baker's yeast are more alike than different, scientists discover

by Van Andel Research Institute

Humans and baker's yeast have more in common than meets the eye, including an important mechanism that helps ensure DNA is copied correctly, reports a pair of studies published in the journals Science and Proceedings of the National Academy of Sciences .

The findings visualize for the first time a molecular complex—called CTF18-RFC in humans and Ctf18-RFC in yeast —that loads a "clamp" onto DNA to keep parts of the replication machinery from falling off the DNA strand.

It is the latest discovery from longtime collaborators Huilin Li, Ph.D., of Van Andel Institute, and Michael O'Donnell, Ph.D., of The Rockefeller University, to shed light on the intricate mechanisms that enable the faithful passage of genetic information from generation to generation of cells.

"The accurate copying of DNA is fundamental to the propagation of life," Li said. "Our findings add key pieces to the puzzle of DNA replication and could improve understanding of DNA replication-related health conditions."

DNA replication is a tightly controlled process that copies the genetic code , allowing its instructions to be conveyed from one generation of cells to the next. In diseases like cancer, these mechanisms can fail, leading to uncontrolled or faulty replication with devastating consequences.

To date, at least 40 diseases, including many cancers and rare disorders, have been linked to problems with DNA replication.

The process begins by unzipping DNA's ladder-like structure, resulting in two strands called the leading and lagging strands. A molecular construction crew then assembles the missing halves of the strands, turning a single DNA helix into two. Much of this work falls to enzymes called polymerases, which assemble the building blocks of DNA.

On their own, however, polymerases aren't good at staying on the DNA strand. They require CTF18-RFC in humans and Ctf18-RFC in yeast to thread a ring-shaped clamp onto the DNA leading strand, and another clamp loader called RFC in both human and yeast to thread the clamp onto the lagging strand. The clamp then closes and signals to the polymerases that they can begin replicating DNA.

Using high-powered cryo-electron microscopes, Li, O'Donnell and their teams revealed previously unknown facets of the leading strand clamp loaders' structures, including a "hook" that forces the leading strand polymerase to let go of the new DNA strand so it can be recognized by the clamp loader.

This distinction represents a key difference between the functions of the leading strand clamp loader (CTF18-RFC) and the lagging strand clamp loader (RFC) and illuminates an important aspect of varying DNA duplication mechanisms on the leading and lagging strands.

Lastly, the study identified shared features between the yeast and human leading strand clamp loaders, which demonstrate an evolutionary link between the two. This finding underscores the value of yeast as powerful yet simple models for studying genetics.

Qing He et al, Cryo-EM reveals a nearly complete PCNA loading process and unique features of the human alternative clamp loader CTF18-RFC, Proceedings of the National Academy of Sciences (2024). DOI: 10.1073/pnas.2319727121

Journal information: Science , Proceedings of the National Academy of Sciences

Provided by Van Andel Research Institute

Explore further

Feedback to editors

Saturday Citations: Warp drive disasters; cancer prospects across generations; a large COVID vaccination study

Aug 3, 2024

Study yields new insights into the link between global warming and rising sea levels

Aug 2, 2024

Coinfecting viruses obstruct each other's cell invasion

Scientists discuss why we might not spot solar panel technosignatures

BNP-Track algorithm offers a clearer picture of biomolecules in motion

New compound found to be effective against 'flesh-eating' bacteria

Not the day after tomorrow: Why we can't predict the timing of climate tipping points

Scientists pin down the origins of the moon's tenuous atmosphere

Solving the doping problem: Enhancing performance in organic semiconductors

Genetic signatures of domestication identified in pigs and chickens

Relevant physicsforums posts, contradictory statements made by two different professors about iq scores, new and interesting publications relevant to the origin of life, the cass report (uk).

Jul 30, 2024

The predictive brain (Stimulus-Specific Error Prediction Neurons)

Jul 21, 2024

Understanding COVID Quarantine Guidance

Jul 19, 2024

Innovative ideas and technologies to help folks with disabilities

Jul 18, 2024

More from Biology and Medical

Related Stories

Combining unique methods, researchers discover new protein functions, explore physical space of proteins

Feb 9, 2024

Cryo-EM reveals how '911' molecule helps fix damaged DNA

Mar 21, 2022

Double trouble at chromosome ends: New study alters our understanding of telomere biology

Feb 28, 2024

Scientists solve mystery of DNA damage detection and repair caused by sunlight, alcohol, and pollution

Jul 31, 2024

Cancer genome research analyzes molecular evolution of tumors after chemical exposure

Jun 13, 2024

Researchers identify how the bacterial replicative helicase opens to start DNA replication process

Feb 26, 2019

Recommended for you

Genetic mutation prompts 'deadbeat dad' fish to start raising their offspring

Scientists 'cautiously optimistic' about AI's role in drug discovery

Let us know if there is a problem with our content.

Use this form if you have come across a typo, inaccuracy or would like to send an edit request for the content on this page. For general inquiries, please use our contact form . For general feedback, use the public comments section below (please adhere to guidelines ).

Please select the most appropriate category to facilitate processing of your request

Thank you for taking time to provide your feedback to the editors.

Your feedback is important to us. However, we do not guarantee individual replies due to the high volume of messages.

E-mail the story

Your email address is used only to let the recipient know who sent the email. Neither your address nor the recipient's address will be used for any other purpose. The information you enter will appear in your e-mail message and is not retained by Phys.org in any form.

Newsletter sign up

Get weekly and/or daily updates delivered to your inbox. You can unsubscribe at any time and we'll never share your details to third parties.

More information Privacy policy

Donate and enjoy an ad-free experience

We keep our content available to everyone. Consider supporting Science X's mission by getting a premium account.

E-mail newsletter

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Article Menu

• Subscribe SciFeed
• Recommended Articles
• Google Scholar
• on Google Scholar
• Table of Contents

Find support for a specific problem in the support section of our website.

Please let us know what you think of our products and services.

Visit our dedicated information section to learn more about MDPI.

JSmol Viewer

Hybrid modeling for on-line fermentation optimization and scale-up: a review.

1. Introduction

2. mechanistic modeling, 3. data-driven modeling.

4. Hybrid Modeling

Applications in fermentation processes, 5. model aided scale-up, 5.1. use of cfd-coupled kinetic models, 5.2. use of hybrid modelling in scale-up, 6. conclusions.

• Both mechanistic and data-driven models continue to be relevant strategies in the development of fermentation processes, both with different specific use cases. Data-driven models are particularly relevant for online process models and are frequently used in the development of soft-sensors. On the other hand, the interpretability and extrapolation capabilities of mechanistic models make them suitable for process optimization and understanding the impact of different parameters on the cell’s metabolic responses.
• Hybrid modeling is a rapidly evolving field and offers substantial benefits in the context of fermentation processes. It enables the exploitation of the strengths of both types of aforementioned models while combatting their weaknesses, ideally leading to a more agile development process.
• The level of mechanistic knowledge included in hybrid models must be carefully selected to avoid overparametrizing or biasing the model. If performed adequately, the result will be a more accurate and extrapolative model, with lower data requirements than a data-driven counterpart.
• Most use cases still focus on the prediction and monitoring of relevant process variables, but they present great potential for model predictive control applications. Furthermore, it appears to be an interesting tool for aiding in process upscaling due to good extrapolation capabilities across scales.
• The technology readiness level of hybrid modeling is still considered low. Some challenges, like the expansion of models as more data becomes available or the complexity in parameter estimation, need to be overcome for their successful implementation as relevant tools for industrial bioprocesses.

Author Contributions

Acknowledgments, conflicts of interest, abbreviations.

• Behera, S.S.; Ray, R.C.; Das, U.; Panda, S.K.; Saranraj, P. Microorganisms in Fermentation. In Essentials in Fermentation Technology ; Berenjian, A., Ed.; Springer International Publishing: Cham, Switzerland, 2019; pp. 1–39. [ Google Scholar ] [ CrossRef ]
• González-Figueredo, C.; Flores-Estrella, R.A.; Rojas-Rejón, O.A. Fermentation: Metabolism, kinetic models, and bioprocessing. In Current Topics in Biochemical Engineering ; IntechOpen: Rijeka, Croatia, 2018; Volume 1. [ Google Scholar ]
• Nadal-Rey, G.; McClure, D.D.; Kavanagh, J.M.; Cornelissen, S.; Fletcher, D.F.; Gernaey, K.V. Understanding gradients in industrial bioreactors. Biotechnol. Adv. 2021 , 46 , 107660. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Pigou, M.; Morchain, J. Investigating the interactions between physical and biological heterogeneities in bioreactors using compartment, population balance and metabolic models. Chem. Eng. Sci. 2015 , 126 , 267–282. [ Google Scholar ] [ CrossRef ]
• Gargalo, C.L.; de las Heras, S.C.; Jones, M.N.; Udugama, I.; Mansouri, S.S.; Krühne, U.; Gernaey, K.V. Towards the Development of Digital Twins for the Bio-manufacturing Industry. In Digital Twins: Tools and Concepts for Smart Biomanufacturing ; Herwig, C., Pörtner, R., Möller, J., Eds.; Springer International Publishing: Cham, Switzerland, 2021; pp. 1–34. [ Google Scholar ] [ CrossRef ]
• Becker, T.; Enders, T.; Delgado, A. Dynamic neural networks as a tool for the online optimization of industrial fermentation. Bioprocess Biosyst. Eng. 2002 , 24 , 347–354. [ Google Scholar ] [ CrossRef ]
• Lourenço, N.D.; Lopes, J.A.; Almeida, C.F.; Sarraguça, M.C.; Pinheiro, H.M. Bioreactor monitoring with spectroscopy and Chemometrics: A Review. Anal. Bioanal. Chem. 2012 , 404 , 1211–1237. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Janoska, A.; Buijs, J.; van Gulik, W.M. Predicting the influence of combined oxygen and glucose gradients based on scale-down and modelling approaches for the scale-up of penicillin fermentations. Process Biochem. 2023 , 124 , 100–112. [ Google Scholar ] [ CrossRef ]
• Mears, L.; Stocks, S.M.; Albaek, M.O.; Sin, G.; Gernaey, K.V. Mechanistic fermentation models for process design, monitoring, and Control. Trends Biotechnol. 2017 , 35 , 914–924. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Gernaey, K.V.; Lantz, A.E.; Tufvesson, P.; Woodley, J.M.; Sin, G. Application of mechanistic models to fermentation and biocatalysis for next-generation processes. Trends Biotechnol. 2010 , 28 , 346–354. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Tsopanoglou, A.; Jiménez del Val, I. Moving towards an era of hybrid modelling: Advantages and challenges of coupling mechanistic and data-driven models for upstream pharmaceutical bioprocesses. Curr. Opin. Chem. Eng. 2021 , 32 , 100691. [ Google Scholar ] [ CrossRef ]
• von Stosch, M.; Oliveira, R.; Peres, J.; Feyo de Azevedo, S. Hybrid semi-parametric modeling in process systems engineering: Past, present and future. Comput. Chem. Eng. 2014 , 60 , 86–101. [ Google Scholar ] [ CrossRef ]
• Narayanan, H.; Luna, M.; Sokolov, M.; Butté, A.; Morbidelli, M. Hybrid Models Based on Machine Learning and an Increasing Degree of Process Knowledge: Application to Cell Culture Processes. Ind. Eng. Chem. Res. 2022 , 61 , 8658–8672. [ Google Scholar ] [ CrossRef ]
• Rogers, A.W.; Song, Z.; Ramon, F.V.; Jing, K.; Zhang, D. Investigating ‘greyness’ of hybrid model for bioprocess predictive modelling. Biochem. Eng. J. 2023 , 190 , 108761. [ Google Scholar ] [ CrossRef ]
• Shah, P.; Sheriff, M.Z.; Bangi, M.S.F.; Kravaris, C.; Kwon, J.S.I.; Botre, C.; Hirota, J. Deep neural network-based hybrid modeling and experimental validation for an industry-scale fermentation process: Identification of time-varying dependencies among parameters. Chem. Eng. J. 2022 , 441 , 135643. [ Google Scholar ] [ CrossRef ]
• Bangi, M.S.F.; Kao, K.; Kwon, J.S.I. Physics-informed neural networks for hybrid modeling of lab-scale batch fermentation for β -carotene production using Saccharomyces cerevisiae. Chem. Eng. Res. Des. 2022 , 179 , 415–423. [ Google Scholar ] [ CrossRef ]
• Moser, A.; Appl, C.; Brüning, S.; Hass, V.C. Mechanistic Mathematical Models as a Basis for Digital Twins. In Digital Twins: Tools and Concepts for Smart Biomanufacturing ; Springer: Berlin/Heidelberg, Germany, 2021. [ Google Scholar ] [ CrossRef ]
• Waldherr, S. Estimation methods for heterogeneous cell population models in systems biology. J. R. Soc. Interface 2018 , 15 , 20180530. [ Google Scholar ] [ CrossRef ]
• Anane, E.; López C, D.C.; Neubauer, P.; Cruz Bournazou, M.N. Modelling overflow metabolism in Escherichia coli by acetate cycling. Biochem. Eng. J. 2017 , 125 , 23–30. [ Google Scholar ] [ CrossRef ]
• Albaek, M.O.; Gernaey, K.V.; Hansen, M.S.; Stocks, S.M. Modeling enzyme production with Aspergillus oryzae in pilot scale vessels with different agitation, aeration, and agitator types. Biotechnol. Bioeng. 2011 , 108 , 1828–1840. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Grisales Díaz, V.H.; Willis, M.J. Ethanol production using Zymomonas mobilis: Development of a kinetic model describing glucose and xylose co-fermentation. Biomass Bioenergy 2019 , 123 , 41–50. [ Google Scholar ] [ CrossRef ]
• Du, Y.H.; Wang, M.Y.; Yang, L.H.; Tong, L.L.; Guo, D.S.; Ji, X.J. Optimization and Scale-Up of Fermentation Processes Driven by Models. Bioengineering 2022 , 9 , 473. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Tang, W.; Deshmukh, A.T.; Haringa, C.; Wang, G.; van Gulik, W.; van Winden, W.; Reuss, M.; Heijnen, J.J.; Xia, J.; Chu, J.; et al. A 9-pool metabolic structured kinetic model describing days to seconds dynamics of growth and product formation by Penicillium chrysogenum. Biotechnol. Bioeng. 2017 , 114 , 1733–1743. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Jahan, N.; Maeda, K.; Matsuoka, Y.; Sugimoto, Y.; Kurata, H. Development of an accurate kinetic model for the central carbon metabolism of Escherichia coli . Microb. Cell Factories 2016 , 15 , 112. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Çelik, E.; Çalık, P.; Oliver, S.G. A structured kinetic model for recombinant protein production by Mut+ strain of Pichia pastoris. Chem. Eng. Sci. 2009 , 64 , 5028–5035. [ Google Scholar ] [ CrossRef ]
• Haringa, C.; Tang, W.; Wang, G.; Deshmukh, A.T.; van Winden, W.A.; Chu, J.; van Gulik, W.M.; Heijnen, J.J.; Mudde, R.F.; Noorman, H.J. Computational fluid dynamics simulation of an industrial P. chrysogenum fermentation with a coupled 9-pool metabolic model: Towards rational scale-down and design optimization. Chem. Eng. Sci. 2018 , 175 , 12–24. [ Google Scholar ] [ CrossRef ]
• Pan, S.; Chen, G.; Zeng, J.; Cao, X.; Zheng, X.; Zeng, W.; Liang, Z. Fibrinolytic enzyme production from low-cost substrates by marine Bacillus subtilis: Process optimization and kinetic modeling. Biochem. Eng. J. 2019 , 141 , 268–277. [ Google Scholar ] [ CrossRef ]
• Xu, J.; Tang, P.; Yongky, A.; Drew, B.; Borys, M.C.; Liu, S.; Li, Z.J. Systematic development of temperature shift strategies for Chinese hamster ovary cells based on short duration cultures and kinetic modeling. mAbs 2019 , 11 , 191–204. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Goldrick, S.; Ştefan, A.; Lovett, D.; Montague, G.; Lennox, B. The development of an industrial-scale fed-batch fermentation simulation. J. Biotechnol. 2015 , 193 , 70–82. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Barrigon, J.M.; Valero, F.; Montesinos, J.L. A macrokinetic model-based comparative meta-analysis of recombinant protein production by Pichia pastoris under AOX1 promoter. Biotechnol. Bioeng. 2015 , 112 , 1132–1145. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Germec, M.; Karhan, M.; Demirci, A.; Turhan, I. Kinetic modeling, sensitivity analysis, and techno-economic feasibility of ethanol fermentation from non-sterile carob extract-based media in Saccharomyces cerevisiae biofilm reactor under a repeated-batch fermentation process. Fuel 2022 , 324 , 124729. [ Google Scholar ] [ CrossRef ]
• Mears, L.; Stocks, S.M.; Albaek, M.O.; Sin, G.; Gernaey, K.V. Application of a mechanistic model as a tool for on-line monitoring of pilot scale filamentous fungal fermentation processes—The importance of evaporation effects. Biotechnol. Bioeng. 2017 , 114 , 589–599. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Alloghani, M.; Al-Jumeily, D.; Mustafina, J.; Hussain, A.; Aljaaf, A.J. A Systematic Review on Supervised and Unsupervised Machine Learning Algorithms for Data Science. In Supervised and Unsupervised Learning for Data Science ; Springer: Berlin/Heidelberg, Germany, 2020. [ Google Scholar ] [ CrossRef ]
• Wiering, M.A.; Van Otterlo, M. Reinforcement learning. Adapt. Learn. Optim. 2012 , 12 , 739. [ Google Scholar ]
• Tavasoli, T.; Arjmand, S.; Ranaei Siadat, S.O.; Shojaosadati, S.A.; Sahebghadam Lotfi, A. A robust feeding control strategy adjusted and optimized by a neural network for enhancing of alpha 1-antitrypsin production in Pichia pastoris. Biochem. Eng. J. 2019 , 144 , 18–27. [ Google Scholar ] [ CrossRef ]
• Nagy, Z.K. Model based control of a yeast fermentation bioreactor using optimally designed artificial neural networks. Chem. Eng. J. 2007 , 127 , 95–109. [ Google Scholar ] [ CrossRef ]
• Andersen, S.W.; Runger, G.C. Partitioned partial least squares regression with application to a batch fermentation process. J. Chemom. 2011 , 25 , 159–168. [ Google Scholar ] [ CrossRef ]
• Barton, M.; Duran-Villalobos, C.A.; Lennox, B. Multivariate batch to batch optimisation of fermentation processes to improve productivity. J. Process. Control 2021 , 108 , 148–156. [ Google Scholar ] [ CrossRef ]
• Nucci, E.R.; Cruz, A.J.; Giordano, R.C. Monitoring bioreactors using Principal Component Analysis: Production of penicillin G acylase as a case study. Bioprocess Biosyst. Eng. 2009 , 33 , 557–564. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Dewasme, L.; Cote, F.; Filee, P.; Hantson, A.L.; Vande Wouwer, A. Dynamic modeling of hybridoma cell cultures using maximum likelihood principal component analysis. IFAC-PapersOnLine 2017 , 50 , 12143–12148. [ Google Scholar ] [ CrossRef ]
• Pimentel, G.A.; Dewasme, L.; Wouwer, A.V. Data-driven Linear Predictor based on Maximum Likelihood Nonnegative Matrix Decomposition for Batch Cultures of Hybridoma Cells. IFAC-PapersOnLine 2022 , 55 , 903–908. [ Google Scholar ] [ CrossRef ]
• Kaelbling, L.P.; Littman, M.L.; Moore, A.W. Reinforcement learning: A survey. J. Artif. Intell. Res. 1996 , 4 , 237–285. [ Google Scholar ] [ CrossRef ]
• Treloar, N.J.; Fedorec, A.J.H.; Ingalls, B.; Barnes, C.P. Deep reinforcement learning for the control of microbial co-cultures in bioreactors. PLoS Comput. Biol. 2020 , 16 , e1007783. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Kim, J.W.; Park, B.J.; Oh, T.H.; Lee, J.M. Model-based reinforcement learning and predictive control for two-stage optimal control of fed-batch bioreactor. Comput. Chem. Eng. 2021 , 154 , 107465. [ Google Scholar ] [ CrossRef ]
• Oh, T.H.; Park, H.M.; Kim, J.W.; Lee, J.M. Integration of reinforcement learning and model predictive control to optimize semi-batch bioreactor. AIChE J. 2022 , 68 , e17658. [ Google Scholar ] [ CrossRef ]
• Paquet-Durand, O.; Assawarajuwan, S.; Hitzmann, B. Artificial neural network for bioprocess monitoring based on fluorescence measurements: Training without offline measurements. Eng. Life Sci. 2017 , 17 , 874–880. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Lopes, J.; Menezes, J.; Westerhuis, J.; Smilde, A. Multiblock PLS analysis of an industrial pharmaceutical process. Biotechnol. Bioeng. 2002 , 80 , 419–427. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Schweidtmann, A.M.; Zhang, D.; von Stosch, M. A review and perspective on hybrid modeling methodologies. Digit. Chem. Eng. 2024 , 10 , 100136. [ Google Scholar ] [ CrossRef ]
• von Stosch, M.; Hamelink, J.M.; Oliveira, R. Hybrid modeling as a QBD/PAT tool in process development: An industrial E. coli case study. Bioprocess Biosyst. Eng. 2016 , 39 , 773–784. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Bayer, B.; Duerkop, M.; Striedner, G.; Sissolak, B. Model Transferability and Reduced Experimental Burden in Cell Culture Process Development Facilitated by Hybrid Modeling and Intensified Design of Experiments. Front. Bioeng. Biotechnol. 2021 , 9 , 740215. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Brunner, V.; Siegl, M.; Geier, D.; Becker, T. Biomass soft sensor for a Pichia pastoris fed-batch process based on phase detection and hybrid modeling. Biotechnol. Bioeng. 2020 , 117 , 2749–2759. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• von Stosch, M.; Davy, S.; Francois, K.; Galvanauskas, V.; Hamelink, J.M.; Luebbert, A.; Mayer, M.; Oliveira, R.; O’Kennedy, R.; Rice, P.; et al. Hybrid modeling for quality by design and PAT-benefits and challenges of applications in biopharmaceutical industry. Biotechnol. J. 2014 , 9 , 719–726. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• von Stosch, M.; Oliveira, R.; Peres, J.; Feyo de Azevedo, S. A general hybrid semi-parametric process control framework. J. Process. Control. 2012 , 22 , 1171–1181. [ Google Scholar ] [ CrossRef ]
• Psichogios, D.C.; Ungar, L.H. A hybrid neural network-first principles approach to process modeling. AIChE J. 1992 , 38 , 1499–1511. [ Google Scholar ] [ CrossRef ]
• Cruz-Bournazou, M.N.; Narayanan, H.; Fagnani, A.; Butté, A. Hybrid Gaussian Process Models for continuous time series in bolus fed-batch cultures. IFAC-PapersOnLine 2022 , 55 , 204–209. [ Google Scholar ] [ CrossRef ]
• Vega-Ramon, F.; Zhu, X.; Savage, T.R.; Petsagkourakis, P.; Jing, K.; Zhang, D. Kinetic and hybrid modeling for yeast astaxanthin production under uncertainty. Biotechnol. Bioeng. 2021 , 118 , 4854–4866. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Read, J.; Bifet, A.; Pfahringer, B.; Holmes, G. Batch-Incremental versus Instance-Incremental Learning in Dynamic and Evolving Data. In Proceedings of the Advances in Intelligent Data Analysis XI, Helsinki, Finland, 25–27 October 2012; Springer: Berlin/Heidelberg, Germany, 2012; pp. 313–323. [ Google Scholar ]
• Rogers, A.W.; Cardenas, I.O.S.; Del Rio-Chanona, E.A.; Zhang, D. Investigating physics-informed neural networks for bioprocess hybrid model construction. Comput. Aided Chem. Eng. 2023 , 52 , 83–88. [ Google Scholar ] [ CrossRef ]
• Rydal, T.; Frandsen, J.; Nadal-Rey, G.; Albæk, M.O.; Ramin, P. Bringing a scalable adaptive hybrid modeling framework closer to industrial use: Application on a multiscale fungal fermentation. Biotechnol. Bioeng. 2024 , 121 , 1609–1625. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• von Stosch, M.; Oliveria, R.; Peres, J.; de Azevedo, S.F. Hybrid modeling framework for process analytical technology: Application to Bordetella pertussis cultures. Biotechnol. Prog. 2012 , 28 , 284–291. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Boareto, Á.J.M.; De Souza, M.B., Jr.; Valero, F.; Valdman, B. A hybrid neural model (HNM) for the on-line monitoring of lipase production by Candida rugosa. J. Chem. Technol. Biotechnol. 2007 , 82 , 319–327. [ Google Scholar ] [ CrossRef ]
• Jenzsch, M.; Gnoth, S.; Kleinschmidt, M.; Simutis, R.; Lübbert, A. Improving the batch-to-batch reproducibility of microbial cultures during recombinant protein production by regulation of the total carbon dioxide production. J. Biotechnol. 2007 , 128 , 858–867. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Dors, M.; Simutis, R.; Lübbert, A. Hybrid Process Modeling for Advanced Process State Estimation, Prediction, and Control Exemplified in a Production-Scale Mammalian Cell Culture. In Biosensor and Chemical Sensor Technology ; American Chemical Society: Washington, DC, USA, 1996. [ Google Scholar ] [ CrossRef ]
• Pinto, J.; Mestre, M.; Ramos, J.; Costa, R.S.; Striedner, G.; Oliveira, R. A general deep hybrid model for bioreactor systems: Combining first principles with deep neural networks. Comput. Chem. Eng. 2022 , 165 , 107952. [ Google Scholar ] [ CrossRef ]
• Cabaneros Lopez, P.; Udugama, I.A.; Thomsen, S.T.; Roslander, C.; Junicke, H.; Iglesias, M.M.; Gernaey, K.V. Transforming data to information: A parallel hybrid model for real-time state estimation in lignocellulosic ethanol fermentation. Biotechnol. Bioeng. 2021 , 118 , 579–591. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Schmidt, F.R. Optimization and scale up of industrial fermentation processes. Appl. Microbiol. Biotechnol. 2005 , 68 , 425–435. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Formenti, L.R.; Nørregaard, A.; Bolic, A.; Hernandez, D.Q.; Hagemann, T.; Heins, A.L.; Larsson, H.; Mears, L.; Mauricio-Iglesias, M.; Krühne, U.; et al. Challenges in industrial fermentation technology research. Biotechnol. J. 2014 , 9 , 727–738. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Funke, M.; Buchenauer, A.; Schnakenberg, U.; Mokwa, W.; Diederichs, S.; Mertens, A.; Müller, C.; Kensy, F.; Büchs, J. Microfluidic biolector—microfluidic bioprocess control in microtiter plates. Biotechnol. Bioeng. 2010 , 107 , 497–505. [ Google Scholar ] [ CrossRef ]
• Tajsoleiman, T.; Mears, L.; Krühne, U.; Gernaey, K.V.; Cornelissen, S. An industrial perspective on scale-down challenges using miniaturized bioreactors. Trends Biotechnol. 2019 , 37 , 697–706. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Junne, S.; Klingner, A.; Kabisch, J.; Schweder, T.; Neubauer, P. A two-compartment bioreactor system made of commercial parts for bioprocess scale-down studies: Impact of oscillations on Bacillus subtilis fed-batch cultivations. Biotechnol. J. 2011 , 6 , 1009–1017. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Wang, G.; Haringa, C.; Noorman, H.; Chu, J.; Zhuang, Y. Developing a Computational Framework To Advance Bioprocess Scale-Up. Trends Biotechnol. 2020 , 38 , 846–856. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Siebler, F.; Lapin, A.; Hermann, M.; Takors, R. The impact of CO gradients on C. ljungdahlii in a 125 m 3 bubble column: Mass transfer, circulation time and lifeline analysis. Chem. Eng. Sci. 2019 , 207 , 410–423. [ Google Scholar ] [ CrossRef ]
• Haringa, C.; Tang, W.; Deshmukh, A.T.; Xia, J.; Reuss, M.; Heijnen, J.J.; Mudde, R.F.; Noorman, H.J. Euler-Lagrange computational fluid dynamics for (bio)reactor scale down: An analysis of organism lifelines. Eng. Life Sci. 2016 , 16 , 652–663. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Kuschel, M.; Siebler, F.; Takors, R. Lagrangian Trajectories to Predict the Formation of Population Heterogeneity in Large-Scale Bioreactors. Bioengineering 2017 , 4 , 27. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Lapin, A.; Müller, D.; Reuss, M. Dynamic behavior of microbial populations in stirred bioreactors simulated with Euler-Lagrange methods: Traveling along the lifelines of single cells. Ind. Eng. Chem. Res. 2004 , 43 , 4647–4656. [ Google Scholar ] [ CrossRef ]
• Simon, L.L.; Fischer, U.; Hungerbühler, K. Modeling of a Three-Phase Industrial Batch Reactor Using a Hybrid First-Principles Neural-Network Model. Ind. Eng. Chem. Res. 2006 , 45 , 7336–7343. [ Google Scholar ] [ CrossRef ]

Click here to enlarge figure

Share and Cite

Albino, M.; Gargalo, C.L.; Nadal-Rey, G.; Albæk, M.O.; Krühne, U.; Gernaey, K.V. Hybrid Modeling for On-Line Fermentation Optimization and Scale-Up: A Review. Processes 2024 , 12 , 1635. https://doi.org/10.3390/pr12081635

Albino M, Gargalo CL, Nadal-Rey G, Albæk MO, Krühne U, Gernaey KV. Hybrid Modeling for On-Line Fermentation Optimization and Scale-Up: A Review. Processes . 2024; 12(8):1635. https://doi.org/10.3390/pr12081635

Albino, Mariana, Carina L. Gargalo, Gisela Nadal-Rey, Mads O. Albæk, Ulrich Krühne, and Krist V. Gernaey. 2024. "Hybrid Modeling for On-Line Fermentation Optimization and Scale-Up: A Review" Processes 12, no. 8: 1635. https://doi.org/10.3390/pr12081635

Article Metrics

Article access statistics, further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

An official website of the United States government

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

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

• Publications
• Account settings

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

• Advanced Search
• Journal List
• PMC11297284

Phenotypic heterogeneity follows a growth-viability tradeoff in response to amino acid identity

Kiyan shabestary.

1 Department of Bioengineering and Imperial College Centre for Synthetic Biology, Imperial College London, London, SW7 2AZ UK

Cinzia Klemm

Benedict carling.

2 London Biofoundry, Imperial College Translation & Innovation Hub, London, UK

James Marshall

Juline savigny, marko storch.

3 Department of Infectious Disease, Imperial College London, London, SW7 2AZ, UK

Rodrigo Ledesma-Amaro

Associated data.

Data for Fig.  1c was from ref. 23 obtained from NCBI with accession number GSE125162. Raw sequencing data of subpopulation RNA sequencing generated in this study was deposited in NCBI GO with accession number GSE235239 . All data generated or analysed during this study are included in this article and in the  Supplementary Information .  Source data are provided with this paper.

All scripts used for data analysis and plotting are available on GitHub ( https://github.com/KiyanShabestary/2023-NLIM-heterogeneity for general analysis and plotting and https://github.com/Benedict-Carling/YeaZ-Output-Analysis for high-throughput microscopy).

In their natural environments, microorganisms mainly operate at suboptimal growth conditions with fluctuations in nutrient abundance. The resulting cellular adaptation is subject to conflicting tasks: growth or survival maximisation. Here, we study this adaptation by systematically measuring the impact of a nitrogen downshift to 24 nitrogen sources on cellular metabolism at the single-cell level. Saccharomyces lineages grown in rich media and exposed to a nitrogen downshift gradually differentiate to form two subpopulations of different cell sizes where one favours growth while the other favours viability with an extended chronological lifespan. This differentiation is asymmetrical with daughter cells representing the new differentiated state with increased viability. We characterise the metabolic response of the subpopulations using RNA sequencing, metabolic biosensors and a transcription factor-tagged GFP library coupled to high-throughput microscopy, imaging more than 800,000 cells. We find that the subpopulation with increased viability is associated with a dormant quiescent state displaying differences in MAPK signalling. Depending on the identity of the nitrogen source present, differentiation into the quiescent state can be actively maintained, attenuated, or aborted. These results establish amino acids as important signalling molecules for the formation of genetically identical subpopulations, involved in chronological lifespan and growth rate determination.

Microbes frequently encounter suboptimal conditions. Here, Shabestary et al. show that phenotypic heterogeneity is an important feature of Saccharomyces species adaptation where amino acid identity serves as an environmental cue driving this adaptive process.

Introduction

Most microorganisms spend most of their lifetime in a non-growing, quiescent state 1 – 3 . Upon occasional exposure to nutrients, they exit this state and resume growth. In microbes such as yeast, quiescence and proliferative growth are fundamentally opposite cellular states with very distinct gene expression profiles and metabolic signatures 4 – 8 . While the metabolism of growing cells is dominated by anabolic reactions, quiescent cells rely on catabolism for survival and typically undergo important metabolic rewiring associated with an upregulation of the stress response, recycling of internal macromolecules and an overall reduced metabolic activity 4 , 6 , 9 , 10 . Physiologically, quiescent cells are smaller and possess a thicker cell wall that provides resistance to a wide variety of stresses 4 and different quiescence states can be accessed depending on the environmental insult experienced 9 .

Understanding how microorganisms regulate their cell size, growth rate and survivability in response to environmental signals including starvation has been a major challenge in quantitative cellular physiology 11 – 14 . Studies at the population level have drawn empirical relationships between cell growth, cell size and nutrient availability 15 – 17 . Yet, population-averaged observations often mask single-cell behaviours due to phenotypic variations across cells 14 , 18 , 19 . In particular, cell-to-cell heterogeneity is often found in a population of genetically identical (isogenic) cells, even when growing under steady-state assumptions, due to differences in stochasticity, cell ageing or cell cycle progression 14 , 20 – 24 . Recent advances in single-cell phenotyping such as cell segmentation and tracking in microscopy 25 – 27 or single-cell RNA sequencing 22 , 23 , 28 – 31 have given new insights into the emergence of phenotypic heterogeneity in microorganisms. While the underlying differentiation processes are still poorly understood, the fitness benefits are clear. Phenotypic heterogeneity, or population multimodality 32 , denoting the presence of two or more distinct isogenic subpopulations (Fig.  1a ), can reduce the risks of investing all resources towards a specific phenotype as in bet-hedging 20 , 33 – 35 or can lead to a more efficient proteome partitioning through cellular differentiation or division of labour as in multicellular organisms 36 – 38 .

a Conditions tested throughout this study. Cells were grown in rich (YPD) media until they reached exponential phase, washed twice with PBS and resuspended in one of the media. Different modalities (unimodal or bimodal) were observed across conditions. Flow cytometer and microscope schematics were made with BioRender.com released under a CC-BY-NC-ND 4.0 International license. b Pipeline used throughout this study to monitor single-cell differentiation. c scRNAseq datasets were obtained from ref. 23 and describe cells shifted to 0.8 mM proline (NLIM-PRO) or glutamine (NLIM-GLN) as well as the control (YPD). Plots represent dimensionality-reduced data using UMAP, where each point represents a single cell. Cells in the top UMAP plot are coloured by growth scores, calculated from a regression model 39 trained on bulk RNAseq data. Histogram above represents the density of growth scores (GS) for each condition. d Flow cytometry data for cells exposed to a nitrogen downshift. GFP fluorescence (BL1-H channel) and cell size (FSC-H forward scatter channel) were used to measure single-cell heterogeneity. Arbitrary units are shown (abbreviated a.u.).

Here, we study the single-cell response to suboptimal growth conditions in the model eukaryote Saccharomyces cerevisiae . Using nitrogen downshift as a case study, we report the presence of both isogenic quiescent and growing subpopulations displaying differences in cell size, chronological lifespan and growth resumption capability. Based on previously published single-cell RNA sequencing datasets 23 , we identify subpopulation markers that allow high-throughput interrogation of cellular fate at the onset of a nitrogen downshift and study differentiation across 24 different nitrogen sources present in limited or replete amounts (Fig.  1b ). We perform a multi-omics analysis of the differentiation process using subpopulation RNA sequencing and analyse the single-cell metabolic response using a prototrophic GFP-tagged transcription factor library and metabolic biosensors (Fig.  1a , Supplementary Fig.  1 ). Our results reveal the presence of two distinct subpopulations reflecting a global population-wide strategy where isogenic subpopulations are metabolically specialised in either growth or viability depending on the nitrogen source present. Results obtained in wild and laboratory Saccharomyces strains suggest a previously unknown amino acid-dependent and conserved behaviour shaping population dynamics.

Nitrogen shift leads to phenotypic heterogeneity

To study population adaptation following a nitrogen shift, we took advantage of recently available yeast single-cell RNA sequencing datasets (scRNAseq). Jackson et al. 23 observed that diploid prototrophic S. cerevisiae cells, generated from FY4 and FY5 laboratory strains, exposed to nitrogen downshifts display two subpopulation clusters with distinct transcriptome profiles. The differentiation process was also impacted by the quality of the nitrogen source where growth on the non-preferred amino acid proline gave a stronger differentiation than for the preferred amino acid glutamine (Fig.  1c ). We further leveraged this data to find subpopulation markers that could be used to study the emergence of this differentiation. Remarkably, in both conditions, one of the subpopulation clusters (cluster P2 for proline and G2 for glutamine) showed a decrease in expression of ribosomal genes while having a higher stress signature (Fig.  1c ). We investigated whether this difference in transcriptome could translate into a difference in growth. Using a previous regression model predicting growth rate based on bulk transcriptomic data obtained from nutrient-limited chemostats 39 , we found that these clusters (P1/P2 and G1/G2 for proline and glutamine, respectively) had distinct predicted growth profiles 4 h into nitrogen limitation according to scRNAseq data (Fig.  1c ). We identified a subpopulation marker RPL28 (YGL103W) whose transcript levels were high and significantly different between G1 and G2, as well as between P1 and P2 (Supplementary Fig.  2 and Fig.  1c ). Flow cytometry of isogenic BY4741 laboratory cells containing super-folded GFP (sfGFP) under the control of pRPL28 promoter (700 bp upstream of start codon) was first used to reproduce the observed heterogeneity at 4 h post-shift. As suggested from scRNAseq data, bimodality could be detected in nitrogen-limited media with proline or glutamine as nitrogen source but not in rich Yeast extract-Peptone-Dextrose (YPD) media (Fig.  1d ). For each of these nitrogen-limited conditions, two subpopulations of cells could be detected for both pRPL28-sfGFP intensity and cell size (Fig.  1d ). Scatter plot of pRPL28 fluorescence versus cell size shows that the “low” subpopulation with the lower pRPL28 fluorescence had smaller cell size (lower FSC forward scatter signal), while the “high” subpopulation had a cell size and a pRPL28 fluorescence that were closer to pre-shift levels in YPD (Fig.  1d , Supplementary Fig.  3a ). To confirm that these differences in pRPL28 fluorescence between low and high subpopulations were in fact not due to cell size differences, we further normalised fluorescence by cell size (Supplementary Fig.  3b ). When the subpopulations were clustered based on cell size and cell size-normalised pRPL28 fluorescence, cell size-normalised pRPL28 fluorescence was still significantly different ( p  < 0.005; two-sided unpaired t test) between subpopulations for both proline and glutamine conditions, indicating variations in cell size alone could not account for the differences observed in pRPL28 fluorescence (Supplementary Fig.  3c , d ). Additionally, exposing Rpl28 tagged GFP strains to 4 h of proline treatment could also show the emergence of a “low” subpopulation with lower Rpl28-GFP levels (Supplementary Fig.  3e ), indicating that lower fluorescence in the “low” subpopulation was not due to post-transcriptional regulation of sfGFP alone.

To further investigate whether high and low subpopulations were truly isogenic, we sorted cells based on forward side scatter and GFP fluorescence, separated the two subpopulations, and performed the same shift from YPD to nitrogen limitation for high and low fractions separately. Given the significant overlap in cell size and pRPL28 intensity between both subpopulations at the onset of the shift (Fig.  2a ), sorting on both fluorescence and cell size allowed for a more precise separation at the early stage of differentiation (Supplementary Note  1 ). Regardless of their post-sorting classification into low or high subpopulations, both subpopulations re-exposed to rich media could regenerate both low and high fractions when shifted again, indicating that the heterogeneity observed was phenotypic and reversible when re-exposed to rich media (Supplementary Note  1 ). Extending the washing step and leaving the cells in PBS for 2 h did not affect the dynamics of heterogeneity (Supplementary Fig.  4 ).

a Contour plots of cell size (FSC-H) and pRPL28 fluorescence (BL1-H) over time for NLIM-PRO (dark purple) and NLIM-GLN (yellow). Pre-shift indicates growth in YPD prior to the shift. Contours indicate areas of higher density. b Calculation of subpopulation fractions for each timepoint and condition based on clustering applied to a . Example shown is NLIM-PRO after 4 h where two multivariate Gaussians were fitted to the data (Expectation Maximisation; Methods). Clustering was performed on cell size and pRPL28 fluorescence (contour plot) and is shown for both dimensions separately (histograms). Light orange and light red colours represent high and low fractions, respectively. c Based on this clustering, hard assignment (dashed line) was generalised to other conditions and timepoints to calculate subpopulation ratios (bar chart) of low versus high. Bar plot represents the mean of biologically independent samples ±SEM ( n  = 3) taken on different days. Significance scores denote the p value of unpaired one-sided t test with p  < 0.05 (*). P values are 0.0056, 0.0314 and 0.0329 for 4 h, 6 h and 8 h, respectively. d Volcano plot representation of transcriptomics data (NLIM-PRO) of subpopulations that were separated based on pRPL28 marker intensity and cell size, as represented in b . P values (two-sided) were adjusted for multiple hypothesis testing using the Benjamini–Hochberg procedure. Significant transcripts (indicated in green) are those with p adj < 0.05 and absolute log2 fold-change (FC) ≥ 1.

Low subpopulation is a daughter-specific reversible quiescent state

To investigate the dynamics of the differentiation between high and low subpopulations, we performed a time-course analysis for nitrogen-limited proline and glutamine conditions and monitored cell size and pRPL28 fluorescence every 2 h for 8 h (Fig.  2a ). The emergence of a second subpopulation could be seen for both conditions after 2 h and was more pronounced after 4 h. After that, we noticed a diverging outcome between conditions where cells in the nitrogen-limited proline (hereafter NLIM-PRO) condition maintained bimodality while cells in the glutamine conditions (NLIM-GLN) became more unimodal, but yet heterogeneous, in cell size and pRPL28 fluorescence over time (Fig.  2b , Supplementary Fig.  5a , b ). Bimodality also led to significant differences in the cell size coefficient of variation (CV), defined as the population cell size standard deviation divided by its mean, always higher for NLIM-PRO throughout the differentiation (Supplementary Fig.  6 ). To further evaluate the number of cells in each subpopulation, we clustered cells based on cell size and pRPL28 fluorescence (see Methods) and assumed two distinct subpopulations based on previous histograms (Fig.  1d ). Given the difficulty to reliably cluster subpopulation early and late into the shift, we assigned subpopulations based on the clustering performed on 4 h NLIM-PRO, which gave the clearest clustering (Supplementary Fig.  7 and Fig.  2b ). Significantly more cells were assigned to the low subpopulation for NLIM-PRO compared to NLIM-GLN from 4 h onwards (Fig.  2c ).

The strong and weak bimodality observed for proline and glutamine, respectively, echoed the growth score distributions computed from scRNAseq (Fig.  1c ). To confirm that the observed subpopulations were indeed the scRNAseq clusters and analyse the early response at the same time, we performed RNA sequencing on the sorted fractions separately (subpopRNAseq) 1 h post-shift (Supplementary Data  1 , Supplementary Data  2 , Fig.  2d ). Comparing genes that were significantly different ( p adj  < 0.05) between subpopulations and clusters in the NLIM-PRO condition revealed a good correlation (Spearman coefficient ρ  = 0.66) between the high subpopulation and cluster P1 as well as between the low subpopulation and cluster P2 despite differences in experimental set-up, laboratory strain and sampling time (1 h for subpopRNAseq vs 4 h scRNAseq) (Supplementary Fig.  8 ). SubpopRNAseq confirmed the upregulation of G1-phase daughter-specific markers DSE1 and DSE2 for the low subpopulation as seen in cluster P2 in the scRNAseq dataset (Fig.  2d , Supplementary Fig.  8 ). Genes involved in cytokinesis (HOF1, BUD4, ACE2, CHS2) or chromosome segregation (HCM1) were significantly ( p adj  < 0.05, Benjamini–Hochberg procedure) up-regulated in the high subpopulation as well as the mitotic exit regulators SPO12, DBF2 and CDC5 (Supplementary Data  2 ). KEGG enrichment of differentially expressed genes between high and low proline subpopulations revealed major differences in cell cycle progression, meiosis, MAPK signalling pathway and DNA repair (Supplementary Fig.  9 ). Further analysis of scRNAseq clusters 4 h into the shift shows that the low subpopulation had all the hallmarks of quiescence with a reversible growth arrest and upregulation of quiescence markers 4 as well as proteins involved in autophagy (Supplementary Note  2 , Supplementary Data  1 ).

Enrichment of the daughter-specific markers DSE1 and DSE2 in the low quiescent subpopulation prompted us to investigate subpopulation dynamics further. First, we sorted high and low subpopulations grown for 2 h in NLIM-PRO or NLIM-GLN and tracked their evolution after post-sorting resuspension in their original media composition (Supplementary Fig.  10 ). Sorted high fractions were able to regenerate the low subpopulation over time (Supplementary Fig.  10 ), further highlighting that the low subpopulation is a daughter-specific state arising from the high subpopulation. Once in the low state, the dynamics differed between NLIM-PRO and NLIM-GLN. In NLIM-PRO, most of the cells remained in this state and the mean subpopulation cell size and pRPL28 intensity slowly increased (Supplementary Fig.  10 and  11 ). Conversely, in the low NLIM-GLN subpopulation, the increase in cell size and pRPL28 intensity was more pronounced than for NLIM-PRO. Remarkably, the low NLIM-GLN subpopulation achieved similar levels in pRPL28 intensity and cell size to the high subpopulation 6 h post-sorting (Supplementary Figs.  10 and  11 ). To further validate our flow cytometry results, we also imaged and counted bud scars for each  sorted subpopulation in each condition using calcofluor white staining (Fig.  3a ). In both conditions, the low subpopulation primarily consisted of scarless daughters, whereas the high subpopulation was predominantly enriched in dividing mother cells (Fig.  3b ). After 2 h re-growth in their respective media, we noticed again a significant difference between conditions. In NLIM-PRO, ~80% of the cells were daughters, whereas in NLIM-GLN, this number was only 30% (Fig.  3b ). This observation suggests that low subpopulation cells in the NLIM-GLN condition were resuming growth and regenerating daughter cells significantly more than in NLIM-PRO. Taken together, our results show that low subpopulations arise in the NLIM state and growth either stalls or resumes according to the nitrogen source present.

a Microscopy imaging of sorted fractions exposed to a 2 h downshift. Bud scars were visualised using calcofluor white dye on an inverted microscope with ×40 magnification. p Fraction of unbudded cells calculated by microscopy. Over 100 cells were imaged for each subpopulation and condition. Error bars represent 95%-confidence intervals centred on the individual fraction of unbudded cells. Significance scores denotes p value with p  < 0.05 (*), p  < 0.01 (**), p  < 0.001 (***) based on the Fisher’s Exact test (one-tailed). Purple, light orange and light red bars represent unsorted (not gated), high and low fractions, respectively.

The cell size and pRPL28 fluorescence modality (i.e. unimodal or bimodal) observed above was conserved in nitrogen-replete conditions (10 mM), as well as intermediate concentrations, indicating that this was not a feature exclusive to nitrogen limitation but of the nitrogen source (Supplementary Fig.  12 ). We performed time-lapse microscopy of single-cells and could observe the emergence of three phenotypes in NLIM-PRO (Supplementary Fig.  13 ). Some cells (below 2% of the total population estimated from flow cytometry data; Supplementary Note  1 ) completely lost their fluorescence with cell size shrinkage, indicative of potential cell death, while most cells showed differences in pRPL28 intensity and cell size as observed during flow cytometry.

Finally, to assess if the observed heterogeneity was strain-specific, we analyzed the cell size modality in response to a downshift to NLIM-PRO or NLIM-GLN for 20 wild-types and laboratory S. cerevisiae and its close relative S. paradoxus isolated from diverse geographical locations worldwide 40 (Supplementary Data 3). We found that, similarly to BY4741 and FY4 strains, multiple subpopulations (sometimes more than two) could be formed in NLIM-PRO and in NLIM-GLN among wild-type isolates (Supplementary Fig.  14 ), showing that cell size modality is not the result of domestication but suggests broader ecological implications inherited from wild-type strains.

Quiescence heterogeneity is amino acid-specific

We next asked how a downshift to other nitrogen sources could affect quiescence heterogeneity. We performed a nitrogen downshift from rich media to 24 different nitrogen sources (20 amino acids and 4 non-proteinogenic sources) that can be transported across the plasma membrane 41 . For comparison, we also considered shifts to nitrogen-starved media (NSTARVE) and nitrogen-replete media (NREP, 10 mM) for a selection of nitrogen sources that were soluble under the conditions tested. Since cell size was an overall good discriminator of differentiation, we measured cell size heterogeneity 2 h, 4 h, 6 h and 8 h into the shift as a measure of quiescence heterogeneity. As seen previously for NLIM-PRO and NLIM-GLN, measuring cell size distribution 6 h into the shift was enough to assess the dynamics of the low subpopulation and capture differences between bimodal and unimodal conditions (Fig.  2a ). Nitrogen downshifts gave a wide range of heterogeneity profiles over time for NLIM (Supplementary Fig.  15 ) and NREP (Supplementary Fig.  16 ). Among those, downshift to either ammonium, arginine, asparagine, methionine, or serine gave unimodal profiles over time, while downshift to other nitrogen sources or nitrogen-starved media gave bimodal profiles. Notably, a shift to glutamate gave a bimodal followed by a more unimodal profile after 8 h. Direct proximity to glutamine through the GS-GOGAT nitrogen assimilation pathway could explain this hybrid response. Glutamate (bimodal after 2 h), slowly converting to glutamine (unimodal), would indeed lead to a bimodal followed by a unimodal response. We further quantified the bimodality in cell size distribution for each of the 25 different nitrogen conditions (24 nitrogen sources and NSTARVE) using Hartigan’s diptest bimodality scores (Methods) on three replicate flow cytometry experiments and found 14 NLIM conditions and 7 NREP conditions (out of 25 NLIM and 16 NREP tested, respectively) that gave significant ( p  < 0.05, Hartigan’s diptest) bimodality scores 6 h into the shift for all replicates (Fig.  4a , Supplementary Figs.  17 , 18 and 19a ). We noted that many of the nitrogen sources leading to unimodal growth were part of the “preferred” nitrogen source group that triggers the Nitrogen Catabolite Repression (NCR) programme, usually associated with better growth performances 41 , 42 . This was the case for arginine, glutamine, ammonium, serine, and asparagine but not for alanine and aspartate, which trigger NCR but were bimodal. To further investigate the connection between modality and growth, we also recorded growth performance across all conditions (Fig.  4b , Supplementary Fig.  19b ). Our results for NREP conditions showed a moderate-strong correlation (Spearman coefficient ρ  = 0.5–0.7) with previous growth rate performance measurements on different nitrogen sources in S. cerevisiae (Supplementary Fig.  20 ). As expected based on the pRPL28 growth marker, nitrogen sources sustaining higher maximal growth rate had a unimodal distribution, while nitrogen sources with a bimodal distribution had slower growth, suggesting a strong anti-correlation (Fig.  4c ; Spearman coefficient ρ  = −0.80 and −0.74 for NLIM and NREP, respectively) between cell size bimodality and maximal achievable growth rate, further validating the computational growth prediction and pRPL28 growth marker identified from scRNAseq. Comparison between NLIM and NREP conditions show that NREP could sustain higher growth (Supplementary Fig.  21a ) but did not show significant differences in modality across amino acids between NLIM and NREP (Supplementary Fig.  21b ). Together, our data suggest that amino acid quality and not quantity is important for modality and that population unimodality correlates with growth maximisation. Remarkably, the presence of alternative nitrogen sources yielded more extreme phenotypes than when nitrogen was absent. This includes a downshift to leucine which gave the most bimodal profile, or a downshift to either lysine or cysteine, which gave a significantly lower ( p  < 0.05 and p  < 0.0005, respectively; unpaired t test) maximal growth rate than when nitrogen was absent from the media (NSTARVE), further highlighting a strong connection between nitrogen sources and population dynamics.

a , b Bimodality scores (Hartigan’s diptest; Methods) and maximal growth rates across all NLIM conditions. Bar plot represents the mean ± SEM of a flow cytometry experiments for cells exposed to a 6 h downshift performed on different days ( n  = 3) and b growth curves of biological replicates ( n  = 3). Purple bars represent the NSTARVE condition where no nitrogen is present. Significance scores denotes conditions where all replicates had a p value (Hartigan’s diptest) of at most p  < 0.05 (*), p  < 0.005 (**) or p  < 0.0005 (***). Bimodality scores are shown as arbitrary units (abbreviated a.u.). c Tradeoff between bimodality and growth, shown in a , b , respectively. Regression lines are indicated in yellow and magenta for NLIM and NREP, respectively. Spearman correlation scores are indicated for each regression line. NLIM-CYS and NLIM-LYS were omitted from the plot due to poor growth performances. d , e Viability rates for cells exposed to a 4 h downshift and sorted based on size and pRPL28 intensities using FACS. Yellow and pink samples indicate high and low GFP fractions. Dark purple samples indicate cells passed through FACS but not gated. Error bars represent the mean ± SEM ( n  = 4). Significance scores denotes p-value of unpaired two-sided t test with p  < 0.05 (*), p  < 0.005 (**), p  < 0.0005 (***) and p  < 0.00005 (****). f , g Growth resumption in rich media (YPD) for sorted and unsorted fractions from e , f . f the shaded area represents the 95% confidence interval ( n  = 4) of the best fit using the ggplot2 function geom_smooth with “y ~ exp(x)” as fit function. g lag times were calculated as the time needed (discretized in 30 min interval) to achieve two doublings. Significance scores denote the p value of paired two-sided t test as in e . h Tradeoff between survival and growth resumption (measured as the inverse of the lag time). Black line represents the regression and the associated value the Spearman correlation. Circles and triangles represent NLIM and NREP conditions, respectively. d – h purple, light orange and light red colours represent unsorted (not gated), high and low fractions, respectively.

Heterogeneity follows a growth-viability tradeoff

Given the apparent growth defect associated with bimodal profiles, we asked what the evolutionary benefits of maintaining a low, quiescent subpopulation of smaller cell size could be. Previous research in S. pombe has shown that small-sized non-growing quiescent populations appear under nitrogen-starved conditions 43 with increased stress resistance 44 . With many subprocesses differentially regulated in quiescent cells linked to a general increase in survivability 4 , 9 , we investigated links between bimodal growth and cellular viability. To this end, we exposed the cells to a 20 h downshift in each of the 24 nitrogen sources, washed them and stored them in PBS for 20 days. As a measure of cellular viability, we used propidium iodide staining, routinely used in chronological lifespan assays. Propidium iodide permeates non-viable cells, with the proportion of stained cells used to estimate single-cell viability and chronological lifespan. We discovered strong differences in viability across conditions with nitrogen sources sustaining unimodal and faster growth such as glutamine, arginine, or ammonia generally displaying lower viability rates (Supplementary Fig.  22 ). NREP conditions lead to lower viability rates than NLIM showing that nitrogen source quantity could also influence viability (Supplementary Fig.  21c ). To further assess if increased viability during bimodal growth could be tied to one of the subpopulations, we recorded subpopulation-specific viability rates. We exposed cells to a 4 h downshift (5 NLIM and 3 NREP different conditions) and subsequently sorted subpopulations based on cell size and pRPL28 expression. As control, we also kept an unsorted fraction that passed flow cytometry but was not gated. We observed the greatest difference in viability in NREP-UREA, where the viability rate gradually decreased for all sorted and unsorted fractions over 30 days with the high subpopulation reaching a 33.7 ± 1.4% viability rate while, in comparison, the low subpopulation could maintain a remarkable 71.0 ± 0.2% viability rate (Fig.  4d ). In fact, the low fraction could achieve significantly better viability rates and chronological lifespan than the high and unsorted fraction in all conditions tested (Fig.  4e ).

We next sought to investigate how subpopulations would fare once optimal growth conditions would resume. To this end, we re-exposed each sorted fraction as well as the unsorted fraction to rich YPD media and monitored growth over 15 h. As a measure for growth resumption, we calculated a pseudo lag time, defined as the time it took to achieve two doublings (corresponding to a fourfold increase in OD 600 ). Remarkably, the high fraction could resume growth faster than the low and unsorted fractions. For NLIM-PRO, the high fraction achieved a lag time of 8.9 ± 0.3 h, while it took significantly longer ( p  < 0.05, paired t test) for the low fraction to resume growth with a lag time of 11.5 ± 0.7 h (Fig.  4f, g ). Again, the lag time extension observed for the low subpopulation was significantly longer than the high subpopulation for all conditions tested (Fig.  4g ), likely due to metabolic changes necessary to re-enter cell cycle progression. An extended lag time was also observed for the low subpopulation when grown in YPD on agar pads (Supplementary Fig.  23 ). Comparing viability and growth resumption rates, our data suggests a second tradeoff between long-term viability and growth maximization (Spearman correlation ρ  = −0.62, Fig.  4g ). In terms of population-wide resource allocation, investing in a subpopulation with higher viability and chronological lifespan is probably advantageous in fluctuating conditions and outweighs the associated population growth reduction. We indeed note that the low subpopulation investment seems to be mitigated during re-exposure to rich media, with the unsorted fraction showing a similar growth compared to the high subpopulation, especially under NREP conditions (Supplementary Fig.  24 ). Additionally, the increased lag phase in rich media for the low subpopulation could be a memory effect. Cells exposed to fluctuating environments often display history-dependent behaviours to prepare for recurring suboptimal conditions 45 . Therefore, maintaining a longer lag phase and not resuming growth straight after re-exposure to rich media could increase longevity as measured by viability extension and enhance survival chances, especially with reoccurring suboptimal conditions.

Viability and growth are distinct transcriptional states

We further characterised metabolic differences between growth and viability based on transcription factor (TF) nuclear intensity, where TF activity is proportional to nuclear intensity in addition to promoter affinity of the downstream targets 46 . To this end, we created a prototrophic transcription factor GFP fusion library (TF-GFP library) using Synthetic Genetic Array (SGA) (Methods). The library consists of 192 members containing a TF fused to GFP as well as a nuclear localisation tag (BFP) and a subpopulation marker (pRPL28-RFP). We performed a nitrogen downshift into proline or glutamine for each member of the TF-GFP library and analysed the resulting nuclear intensity over time for each condition and subpopulation using high-throughput microscopy (Fig.  5a ). TF nuclear intensities were computed from YeaZ segmented cells by overlapping nuclear localisation tag (BFP) and TF-GFP localisation (GFP), while subpopulation assignment was performed on cell size and pPRL28-RFP intensity (Methods, Fig.  5 a, ​ a,b). b ). For each GFP-tagged TF, we recorded cell size, mean cellular pRPL28-RFP and mean nuclear GFP for ~800,000 cells exposed to 30 min, 90 min, 150 min and 210 min of either NLIM-PRO or NLIM-GLN treatment (Source Data). The resulting cell size-pRPL28-RFP scatter plot obtained from microscopy was consistent with flow cytometry data (Fig.  5b , Supplementary Fig.  25 ). After 30 min, we found that already 157 and 116 transcription factors were significantly different ( p  < 0.05, two-tailed t test) between conditions and subpopulations (NLIM-PRO), respectively (Fig.  5 c, ​ c,d). d ). Because TF localisation can be noisy and highly variable over time 21 , we also ranked the nuclear intensities at each timepoint to highlight the most consistent TFs for each condition and subpopulation (Fig.  5 e, ​ e,f, f , Supplementary Data  4 ).

a Overview of the pipeline. A marker strain containing the subpopulation marker as well as a nuclear localisation tag was mated with one of the 192 transcription factor members of the GFP library (TF-GFP) using the Synthetic Genetic Array method (SGA). The improved library was then grown in rich media and shifted to nitrogen-limited conditions as described in the Methods. The library was imaged in PBS 30, 90, 150 and 210 min into the shift. The YeaZ algorithm 26 was used to segment cells based on neural networks. When possible, segmented cells were assigned to high and low clusters based on pRPL28 mScarlet fluorescence and cell size using the Expectation Maximisation (EM) algorithm. TF localisations were computed for each cell and each strain by overlapping the nuclear localisation signal (mTagBFP2) with the TF-GFP signal (Methods). b Scatter plot representing mean RFP fluorescence per cell and cell size 150 min into the shift, at the onset of differentiation for NLIM-PRO (dark purple) and NLIM-GLN (yellow). Cells were assigned to each cluster (low or high) based on their cell size and mean RFP intensity using the EM algorithm. c , d TF localisations for proline plotted against glutamine (population level) and high against low subpopulation (NLIM-PRO), 30 min into the shift. TF nuclear intensity was calculated as the mean GFP fluorescence over the nucleus, determined by the NLS-mtagBFP2. Significance score denotes p value of unpaired two-sided t test with p  < 0.05, with an adjusted mean (condition/subpopulation difference). TF with significant localisation scores are shown in green. e , f Relative TF nuclear localisation tracked over time for proline versus glutamine ( e ) and low versus high proline subpopulations ( f ). Relative localisations were normalised by the mean relative localisation of the timepoint. g Multi-channel ×50 imaging of Mig1-GFP for one representative of two experiments. NLS-mtagBFP2 represents the nuclear localisation and pRPL28-mScarletI was used to classify high and low subpopulations.

We first asked whether the TF library could capture established differences between proline and glutamine conditions. For example, Put3, activator of proline catabolic genes and under the control of the NCR programme 47 , was one of the most localised TF in proline compared to glutamine at bulk and at the low subpopulation levels (210 min; p  < 0.05, two-tailed t test) (Supplementary Data  4 , Supplementary Fig.  26a , b ). Ure2, one of the most consistent targets over time, had a higher nuclear intensity for the proline condition at the population level (Fig.  5 c, ​ c,e). e ). Ure2 is strongly associated with the NCR programme, sequestering the NCR activator Gln3 to the cytoplasm in presence of a preferred nitrogen source such as glutamine 48 . Dig2, a repressor of the Ste12 transcription factor, part of MAPK signalling, which was differentially activated between low and high subpopulation according to subpopulation RNA sequencing (Supplementary Fig.  9 ). To validate our high-throughput results, we further imaged Dig2-GFP strain during a shift to either NLIM-PRO or NLIM-GLN and could find it to be indeed more intense for proline (Supplementary Fig.  26f ). Conversely, Mig1, involved in respiration and gluconeogenesis gene repression 49 , had a higher nuclear intensity in glutamine, indicating that cells growing on glutamine could do fermentation despite nitrogen limitation (Fig.  5 e, ​ e,g). g ). This was also conserved at the low subpopulation level where Mig1 was consistently more intense for glutamine compared to proline (Supplementary Fig.  26e ). Consistent with this finding, we found that Snf1, a major Mig1 inhibitor, had higher nuclear intensity in the low proline subpopulation mimicking a low-glucose condition despite glucose present in abundance (Supplementary Fig.  26d ) 50 . This corroborates our previous findings that the low glutamine subpopulation is actively growing while the low proline subpopulation is more quiescent.

We next used the TF library to investigate differences between high and low subpopulations in NLIM-PRO. Among the most consistent targets, Tec1 had significantly higher nuclear intensity in the low subpopulation already after 30 min and onwards (Fig.  5 d, ​ d,f). f ). We found that Tec1 was specific to the low proline subpopulation since it was more intense when compared to the low glutamine, indicating an essential role in maintaining quiescence (Supplementary Fig.  26c ). Interestingly, Tec1 connects MAPK and TOR pathways to coordinate yeast development and has previously been reported as a positive regulator of chronological lifespan 51 , consistent with the improved chronological lifespan observed for the low subpopulation. Tec1, together with Ste12 of the MAPK pathway, both regulate genes required for filamentous and pseudohyphal growth 52 . Rcs1, another TF that specifically localises to the low proline subpopulation, is involved in iron homoeostasis and its nuclear localisation increases with DNA stress 53 (Fig.  5f , Supplementary Fig.  26g ). Other main targets included Rdr1 and Pdr8, both associated with ATP-binding cassette (ABC) family of transporters, important for resistance to drugs and other growth inhibitors and up-regulated in quiescent cells 9 , 54 .

Quiescent population is dormant and heterogeneous in ATP levels

Given important differences in transcription factor localisation for high and low subpopulations, we further studied how they could translate into physiological differences. Using metabolic sensors, we measured two key metabolic parameters, that are cellular energy status and metabolic activity. For energy status, we measured adenosine-triphosphate (ATP) levels using a FRET biosensor 55 (Fig.  6a ). For metabolic activity, we measured fructose-1,6-bisphosphate (FBP), shown to correlate with glycolytic flux 5 , 56 , using a riboswitch-based fluorescence sensor 57 (Fig.  6b ). To account for growth differences between subpopulations as well as cell-to-cell variability in sensor expression that could occlude the measurement, FBP signal was normalised to an RFP control placed under a constitutive promoter while the ATP sensor was ratiometric, providing the ratio between bound and unbound states (Fig.  6 a, ​ a,b b ).

a , b Overview of the ATP (yATP) FRET sensor 55 and FBP fluorescence sensor 57 . Both sensors are ratiometric, normalising out potential differences in sensor expression levels. For the ATP sensor, relative ATP levels are proportional to the inverse of the ratio of acceptor (590 nm) over donor (483 nm) emission from the same excitation (438 nm). For the FBP sensor, FBP levels can be estimated from the inverse of the ratio of GFP (riboswitch) over RFP (control) ratio. GFP and RFP are integrated in the same locus (URA3) and under the control of pTEF1 and pTEF2, respectively. c Scatter plot representing the emergence of the low subpopulation for NLIM-PRO (dark purple) and NLIM-GLN (yellow) based on the output of the ATP sensor (yATP) against cell size. d , e Quantification over time based on Expectation Maximisation clustering (Methods) of scatter plots as shown in c . Orange and red colours represent high and low fractions, respectively. Only timepoints where clustering could be performed are shown. Centre marks of the box plot represent the median, hinges mark the lower and upper quartiles and whiskers show all values that, at maximum, fall within 1.5 times the interquartile range. Indicated p-values were calculated from unpaired two-sided t test and represent single-cell heterogeneity performed on one representative of two experiments.

Analysis of ATP levels after 8 h of growth in NLIM media with glutamine or proline showed population-wide differences and cellular ATP levels generally lower in glutamine than proline media (Fig.  6c ). These data, together with the strongest growth profile observed in cells shifted to NLIM-GLN (Fig.  4b ), could further suggest fermentative rather than respiratory growth in NLIM-GLN. Remarkably, while ATP levels were unimodal at the population levels, scatter projection of ATP levels versus cell size revealed cellular bimodality for proline, with the emergence of two clusters corresponding to the low and high subpopulations (Fig.  6d ). When we clustered high and low subpopulations based on cell size (Supplementary Fig.  27a ), we found that from 4 h onwards, the low quiescent subpopulation showed a reduced but more heterogeneous ATP pool, while the high growing subpopulation maintained a homogeneous ATP content over the course of the experiment (Supplementary Fig.  27b ). This difference in ATP heterogeneity between low and high subpopulations was maintained for the other bimodal nitrogen sources (Supplementary Fig.  28a ). Similarly, conditions that sustained higher maximal growth rates (Fig.  4b ) had generally lower median ATP output at the population level (Supplementary Fig.  28b , c ). A high variability in the quiescent subpopulation could be consistent with bet-hedging where increased ATP variability could denote a strategy to further increase phenotypic diversity in preparation to environmental fluctuations. Similar observations were made for cells shifted from glucose to maltose, where cells resuming growth had a higher and more homogeneous ATP pool than cells that remained quiescent 29 . Conversely, the analysis of FBP levels did not show the same difference in variability between low and high subpopulations but showed a population-wide decrease over time instead (Fig.  6e ). This decrease was more marked for the low subpopulation suggesting a more dormant metabolism, consistent with previous studies on quiescence 6 , 58 .

Our results show that S. cerevisiae exposed to a nitrogen downshift display phenotypic bimodality with simultaneous presence of subpopulations following a tradeoff between chronological lifespan and growth capabilities. Concomitantly, these subpopulations operate at very distinct metabolic states where the low subpopulation is optimized for viability, and the high subpopulation is optimized for growth. With these fundamentally opposite metabolic states, individual cells need to commit to either state (specialisation) and phenotypic bimodality is therefore optimal given finite resource allocation. A similar phenotypic specialisation into a growing and a stress-resistant fraction has also been found when S. cerevisiae was exposed to growth inhibitors 35 . In proline, the low subpopulation is a terminal state enriched with daughter cells, while in glutamine, it can resume growth. Asymmetric generation of quiescent-specific daughter cells has also been observed during glucose exhaustion, at the onset of stationary phase 10 . Using subpopulation-based RNA sequencing and high-throughput TF imaging, we found that increased viability correlates with a more active MAPK signalling and DNA stress response associated with general chromosomal regulation, as well as organelle and morphological reorganisation. The seemingly more active MAPK signalling between subpopulations denotes an important crosstalk between nitrogen and carbon metabolism. More importantly, the low subpopulation had similarities with a low-glucose state despite glucose present in excess. Nuclear localisation of transcription factors involved in metal homoeostasis for the low proline subpopulation could be related to the non-growing phenotype, where upregulation of ion homoeostasis when growth stalls have been previously reported in a wide range of model organisms 9 , 59 , 60 .

We argue that in the context of evolution and given conservation in both wild and domesticated S. cerevisiae and S. paradoxus strains, cell size bimodality in response to a nitrogen downshift is a hard-wired regulatory mechanism and probably crucial to environmental interactions. We further hypothesise that amino acids could be used as an environmental cue for decision-making. In this aspect, glutamine could denote ideal growth conditions while leucine, which gave the strongest cell size bimodality (Fig.  4a ), could indicate conditions that necessitate a survival increase or dispersal. Supporting this theory, cells growing on leucine produce a distinctive banana-like aroma including isoamyl alcohol, isovaleric acid or isoamyl acetate 61 . While aroma production dependence on nitrogen sources is well documented in the yeast brewing industry 62 , its ecological purpose still remains obscure but past evidence suggests that aroma production could be used for attracting insects and facilitating dispersion 2 , 63 , 64 . Strong differentiation to quiescence for leucine coupled with increased aroma production could, therefore, facilitate dispersal through insect vectors directly linking the observed bimodality to a broader ecological context. Given that growth on leucine was more bimodal than growth without any nitrogen sources (NSTARVE), we further hypothesise that leucine could be used as an environmental cue for dispersion. Increasing viability and extending lifespan might be a strategic choice to allow better dispersion without needing to undergo sporulation and could offer several advantages. First, it could be seen as a lighter, more reversible commitment to dormancy. This could explain why diploid cells in the presence of fermentable carbon sources such as glucose undergo quiescence rather than sporulation when other essential nutrients such as nitrogen are limiting growth 65 . Second, the increased lag phase for the quiescent subpopulation could correspond to a delayed phenotypic outing and give a memory effect in fluctuating conditions to better withstand previously experienced suboptimal conditions 45 , 66 . We note that this is still unclear whether the increased lag phase could also be a by-product of increased viability where cells need time to exit quiescence. Yet, another advantage could be that quiescent cells remain metabolically active and could still be able to exchange metabolites with their environment. Metabolite exchange, amino acids in particular, is a key feature of yeast exo-metabolome and is important for phenotypic heterogeneity and metabolic specialisation 67 . Remarkably, while measuring the supernatant of yeast exposed to 8 h nitrogen downshifts, we could detect amino acids other than the nitrogen source used, even under low nitrogen contents (NLIM, 0.8 mM) (Supplementary Data  5 ), supporting a broader population-wide function based on amino acid exchange. Similarly, a recent study found that lysine was involved in cross-feeding interactions in ageing colonies on an agar plate resulting in phenotypic heterogeneity between young and old cells consuming and producing lysine 68 . Interestingly, we found that lysine was the amino acid that performed the worst in both growth and chronological lifespan. We could not detect any lysine throughout the different downshifts despite the observation that lys12 (lysine biosynthesis) knockout is able to establish syntrophic communities with lysine producers 69 , showing that lysine or its intermediate can be exchanged.

In light of a recent study showing that artificial consortia of S. cerevisiae auxotrophs have an extended chronological lifespan 70 , we show that such lifespan improvements could be explained by the heterogeneity we observed. Auxotrophs would typically be forced to rely heavily on amino acids produced by other members of the community and could, therefore, exhibit heterogeneity within an auxotrophy class, leading to increased lifespan. Finally, in terms of fitness burden, our results show that an investment in a small quiescent subpopulation with longer lifespan is beneficial since it results in a minor decrease in growth resumption capabilities with the lag phase for high and unsorted subpopulations being very similar if not the same in certain cases. This can explain other bet-hedging strategies such as antimicrobial resistance 71 . Our results could be relevant in future work to study amino acid sensing mechanisms or investigate potential division of labour across subpopulations and the role they play within a natural environmental context.

Single-cell RNA sequencing data analysis

Single-cell RNA sequencing data were obtained from ref. 23 . Raw counts data was downloaded from NCBI with accession number GSE125162. All data analysis was performed on R. Pre-processing was performed according to the author’s guidelines, except for the normalisation that was performed using the logNormCounts from the LTLA/Scuttle package ( https://github.com/LTLA/scuttle ). Dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP) was performed using the scater package 72 . To calculate single-cell growth scores, the log normalised scRNAseq reads were inputted into the calculateRates function from the growth regression model obtained from ref. 39 . To identify subpopulation-specific markers, DESeq2 73 was used to find genes that were differentially expressed between subpopulations. Detailed methodology is available as part of Supplementary Method  1 . All data analysis performed in this study are available on GitHub ( https://github.com/KiyanShabestary/2023-NLIM-heterogeneity ).

Strain creation

Two prototrophic strain versions based on the laboratory strain BY4741 ( MAT a his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ) 74 were made as subpopulation marker strains. One version had full prototrophy restored using the minichromosome pHLUM series 75 . A genome-integrated version was also created with HLUM fragments, PCR amplified from the respective minichromosomes with ~35 bp overhangs for homologous recombination targeting the HO locus. The fluorescent protein sfGFP was assembled under the control of the pRPL28 promoter, identified as a subpopulation marker through scRNAseq data analysis. Transcription termination was under the control of the tTDH1 terminator. The pRPL28 promoter part was amplified from genomic DNA taking 700 bp directly upstream of the coding sequence. Parts for sfGFP and terminators were obtained from the yeast MoClo Toolkit (YTK) 76 . Assembly was performed into a URA3 targeting vector using Golden Gate assembly as described in the YTK toolkit assembly. For the ATP sensor strain, a codon-optimised variant of yAT1.03 sensor 55 was ordered as GeneArt from ThermoFisher and cloned in a pYTK level 0 Type 3 plasmid. Using Golden Gate assembly, the fragment was placed under pTEF1 control and with the tENO1 terminator in a vector targeting ura3 . The plasmid was co-transformed with pHLM into BY4741. For the FBP sensor, the 2_riboswitch sensing unit was amplified from the pFBP-2_6.sensor 57 plasmid obtained from addgene (catalogue number 162800). The fragment was assembled into pYTK053 of the YTK collection, downstream of the promoter pTEF1 and mNeonGreen and upstream of the tSSA1 terminator. Both the FBP sensor and its control (mScarletI (RFP) under pTEF2 control and with tENO2 terminator) were genome-integrated at the ura3 locus, complemented with the pHLM minichromosome. For the TF library, a donor strain based on BY4742 (MATα hisΔ1 leu2Δ0 lys2Δ0 ura3Δ0 ) was used to create the prototrophic GFP library using SGA (see SGA section in Methods). As a subpopulation marker, mScarlet (RFP) was placed under pRPL28 control at the ura3 locus with tTDH1 terminator. For nuclear localisation visualisation, a nuclear localisation tag derived from the SV40 T-antigen 77 was added at the 5’ end of mTagBFP2 (BFP) placed under the control of pTEF2 and with the tENO2 terminator genome-integrated at the leu2 locus. Additionally, for haploid selection, a kanMX cassette was placed under the MATa-specific pSTE2 promoter (genome amplified) with tPGK1 terminator at the can1 locus. All assemblies were performed using Golden Gate assembly within the YTK framework 76 . Strains used in this study are indexed and described in Supplementary Data  3 .

Yeast transformation

Transformation into yeast was performed using the Lithium acetate protocol 78 . Overnight YPD yeast cultures were diluted in YPD (1:50, 5 ml per three transformations) in the morning and cultivated until they reached exponential growth (4–5 h). Cells were then washed once and resuspended in 0.1 M Lithium acetate (LiOAc, Sigma) to a final volume of 100 μl per transformation. The subpopulation marker plasmid was linearised using NotI (New England Biolabs). The linearised plasmid (500 ng) and the minichromosome (1 μl, >100 ng/μl), when appropriate, were mixed with boiled (5 min, 100 °C) salmon sperm DNA (10 μl, Invitrogen). Competent yeast cells were resuspended in the DNA/salmon sperm DNA mixture and then mixed with 260 μl 50% (w/v) PEG-3350 (Sigma) and 36 μl 1 M LiOAc. The transformation mixture was incubated at 42 °C for 25 min, resuspended in sterile water and plated on the appropriate selection medium. All strains were confirmed by cPCR followed by Sanger sequencing.

Growth conditions

Yeast cells were cultivated in sterile 14 ml cell culture tubes (Greiner Bio-One) grown at 30 °C in a Infors HT Multitron with 700 rpm shaking. Yeast extract-peptone dextrose (YPD) composed of 1% (w/v) Bacto Yeast Extract (Merck), 2% Bacto Peptone (Merck), and 2% Glucose (VWR) was used as rich pre-shift media. For the post-shift media, 1.7 g/l Yeast Nitrogen Base (YNB) without amino acid and ammonium sulfate (Sigma) and with 2% Glucose (YNB) was used with 0.8 mM (NLIM), 10 mM (NREP) or without (NSTARVE) nitrogen source. All amino acids were supplied from Sigma or Formedium. Both YPD and post-shift media were buffered with 50 mM phosphate buffer and adjusted to pH 6.0.

Medium shift

Pre-shift growth included an overnight step in YPD from a single colony stored on YNB agar plate followed by a 1:50 dilution in 5 ml YPD grown for 4 h until the end of exponential growth was reached (OD 600  = 0.8−1.0). Cells were then centrifuged for 6 min at 4 kG and washed in phosphate-buffered saline (PBS) solution twice. After the washing step, cells were resuspended in post-shift media (NLIM, NREP, or NSTARVE) to OD 600  = 0.4 in 14 ml cell culture tubes or 250 ml flasks (for subpopulation RNA sequencing) and grown at 30 °C with 700 rpm shaking (Infors).

Time-lapse microscopy

Time-lapse microscopy was performed using agarose pads. In short, cells were trapped in between a microscope slide and an agar pad as previously described 79 . Agarose pads were composed of the respective growth media and 1.5% (w/v) low melting agar (Sigma). To create the agar pad, 1 ml of agar/media mixture was pipetted on top of a microscope glass cover (22 mm × 22 mm, VWR). Another cover was placed on top to create a layer of even thickness. Approximately 50 mm 2 of the solidified agarose (one-ninth) was cut out to make an agarose pad and 2 μl of cells (OD 600  = 0.4 in PBS) was applied in the middle of each pad. The agarose pad was then placed upside down in an enclosed 35 mm cell imaging dish (Ibidi). Water was added to the enclosure to limit evaporation during the time-lapse. Imaging was performed on a Nikon Ti-2 Twin-Cam-TIRF with an environmental chamber to maintain temperature at 30 °C.

Subpopulation sorting

Subpopulations were sorted using fluorescence-activated cell sorting (FACS), performed on a BD FACSAria III Cell Sorter, based on GFP fluorescence (FITC-A, blue laser 488-530/30 nm) and morphology (SSC-A). Doublets and budding yeast exclusion were filtered out through FSC-W/FSC-H and SSC-W/SSC-H gatings. Prior to sorting, samples were filtered, and 20,000 events were used to adjust gating. Purity cheques were performed at the start of every sorting run to ensure accurate gating and no cross-contamination. Gating details are available as part of Supplementary Note  1 . For subpopulation RNA sequencing, pRPL28 marker strain culture grown overnight was diluted (1:50) in YPD (10 ml) and grown to exponential phase (4 h). Cells were washed twice with PBS as described above and exposed for 1 hour in post-shift media (post-shift resuspension OD 600  = 0.4). Prior to sorting, cells were centrifuged, resuspended in PBS and kept at 4 °C for the duration of sorting. After sorting in 15 ml tubes, cells from each sorted fraction were grouped and centrifuged (6 min at 4 kG) to remove the supernatant. Cell pellets (at least 4 million cells for each sample) were frozen in liquid nitrogen and stored at −80 °C until RNA extraction. Growth resumption and chronological lifespan experiments were performed in 96-well plates (Greiner). Cells were grown in YPD and exposed to a 4 h post-shift as described above. Cells were washed once in PBS and sorted in 15 ml tubes. Approximately 50 k cells were used per well for both subpopulation lifespan and growth resumption measurements.

Subpopulation RNA sequencing

Subpopulation RNA extraction, sequencing and data analysis was performed through Novogene sequencing services. Sorted cell pellets stored at −80 °C were thawed on ice. RNA was extracted with a RNAprep Pure Plant Plus kit (Tiangen). Messenger RNA was purified using poly-T oligos attached magnetic beads. After fragmentation, the first strand cDNA was synthesised using random hexamer primers, followed by a second strand cDNA synthesis. The library was quantified using real-time PCR and Qubit. Size distributions were calculated using Bioanalyzer analysis. Quantified libraries were pooled and sequenced on an Illumina platform (Novaseq 6000) with a paired-end 150 bp (PE150) method. Raw sequencing reads and count matrix are available in NCBI GEO under the accession number GSE235239.

RNA sequencing reads were filtered according to the following criteria: Reads with no adaptor contamination, no more than 10% of uncertain base (N) within the read, not >50% of the reads made of low-quality base reads (Base Quality Qscore less than 5). Low-quality reads represented less than 1% of total reads. HISAT2 80 was used to map the filtered reads to the genome. Reference genome (fasta file) and gene annotations (gtf file) used for alignments were obtained from the Ensembl database available at http://ftp.ensembl.org/pub/release-75 (December 2022).

Principal Component Analysis was performed on gene expression values (FPKM) to evaluate intergroup and intragroup variance and remove outliers. Differential gene expression was computed using DESeq2 73 and p values adjusted using the Benjamini–Hochberg procedure. KEGG enrichment analysis was performed using clusterProfiler 81 with adjusted p values obtained from DESeq2.

Growth measurements

Growth parameters were calculated for cultures grown in YPD and shifted to post-shift media (OD 600  = 0.1) as described above. Growth curves were obtained for 96-well plates (Greiner) recorded in a Tecan Spark microplate reader set at 30 °C with 200 rpm orbital shaking. Breathe-Easy sealing membrane (Sigma) was applied to reduce evaporation while maintaining gas transfer throughout the experiment. Bulk population maximal growth rates were calculated as follows: Data from the plate reader was blank normalised. The R package growthcurver ( https://github.com/sprouffske/growthcurver ) was used to obtain a smooth fit of the growth curve. Local growth parameter mu_log for each time interval was computed as the difference of the natural logarithmic of the ODs divided by the time interval. Finally, the maximal value mu_log within a time window excluding lag and stationary phases (between 5 h to 20 h in post-shift media) was taken as maximal growth rate. For growth resumption experiment in YPD following FACS-based sorting, cultures were resuspended to OD 600  = 0.05. Lag time was estimated as the time required (discrete interval) to reach two doublings (OD 600  = 0.2).

Bimodality quantification

Cell-to-cell heterogeneity of cultures grown in YPD and shifted to post-shift media (OD 600  = 0.4) was measured over time using an Attune Nxt flow cytometer (Invitrogen). Forward scatter (FSC-H) and GFP fluorescence (BL1-H, blue laser 480/10 nm) were used to measure cell size and GFP fluorescence heterogeneity, respectively. Doublets were excluded from the data based on FSC-H/FSC-A linear correlation (both log scale) using the FlowJo software. At least 10’000 events were recorded per sample/replicate. Cell size bimodality scores were computed using Hartigans’ diptest for unimodality using the R package diptest ( https://github.com/mmaechler/diptest ). The stat and pval output values were used as bimodality score and associated p-value. The R package flexmix 82 was used to perform Expectation Maximisation on cell size distributions to fit a Gaussian model for each subpopulation and estimate the respective mean and standard deviation for each subpopulation.

Chronological lifespan assay

Chronological lifespan was estimated as the percentage of cells remaining viable in PBS over time. For bulk population lifespan estimation, cultures (150 μl) exposed to a 18 h nitrogen downshift in 96-well plates were washed twice in PBS, resuspended in 200 μl PBS and 50 μl of cells/PBS mixture was added to 150 μl PBS in a 96-well plate. For subpopulation lifespan assessment following FACS-based sorting, cells exposed to a 4 h downshift were washed prior to sorting and 50 k cells in PBS were stored in 96-well plates (200 μl). One whole plate was used for each timepoint measurement to limit subsequent evaporation. Lifespan was measured up to 30 days after the end of the post-shift as indicated in text. Plates were sealed with Breathe-Easy sealing membrane (Sigma) and wrapped with aluminium foil and stored at 30 °C until cell viability measurement. Cell viability was measured on the basis of permeability to propidium iodide (PI, Merck) in apoptotic cells using flow cytometry (Attune Nxt, YL1-H, excitation 561 nm, emission filter 585/16 nm). Prior to the fluorescence assay, 1 μl of PI (1 mg/ml) was added to 200 μl cell/PBS mixture and gently mixed using a multi-channel pipette. For each sample/replicate, viability was thresholded based on viability measured in YPD.

Unbudded cells counting

Sorted fractions of cells exposed to a 2 h downshift and sorted via FACS were collected in PBS and either directly stained or transferred to fresh NLIM-PRO or NLIM-GLN media and incubated for 2 h at 30 °C and 700 rpm shaking before staining. Bud scars were stained using calcofluor white (Sigma-Aldrich) at a final concentration of 0.01 g/L. Cells were incubated for 15 min in the dark at room temperature and washed in PBS. Subsequently, 5 μl of cells were transferred to microscopy glass slides and imaged using a Nikon Ti microscope fitted with a Hamamatsu Flash 4 camera. Stained bud scars were visualised using a P4000 Cooled LED light source at 365 nm and filters for blue fluorescence. pRPL28-sfGFP fluorescence was captured using a 460 nm LED and green fluorescent filters. Counting was performed manually, to prevent bias due to differences in staining efficiency.

Metabolic sensors analysis

Metabolic sensors were used for in vivo measurement of ATP and FBP. Both sensors were genome-integrated at the URA3 locus. We used a FRET-based biosensor 55 for ATP and a riboswitch-based fluorescence sensor for FBP measurement 57 . For the ATP sensor, the ATP FRET signal was recorded using a 405 nm excitation laser, a 450/40 nm donor emission filter and a 525/50 nm acceptor emission filter (VL1 and VL2 channels). ATP levels were calculated by taking the ratio between the VL2 and VL1 channels (VL2-H/VL1-H). For the FBP sensor, the FBP signal was recorded using a 488 nm excitation laser and a 530/30 nm emission filter (BL1 channel) while the control RFP signal was measured using a 516 nm excitation filter and 620/15 emission filter (YL2 channel). With the FBP sensor fluorescence displaying a signal inversely proportional to FBP concentration. FBP levels were calculated by taking the ratio between the YL2 and BL1 channels (YL2-H/BL1-H). Signals were recorded on a Attune Nxt flow cytometer (Invitrogen). Expectation maximisation using the R package Rmixmod 83 with “Gaussian_pk_L_I” model selection was performed on cell size to compute subpopulation-specific ATP and FBP levels.

Library creation using SGA

A prototrophic version of the GFP collection 84 was created using the Synthetic Genetic Array method 85 . A transcription factor library (TF-GFP library) containing 192 members was created as follows. Selected strains from the GFP collection (MATa his3Δ1 met15Δ0 leu2Δ0 ura3Δ0 XXX -GFP-HisMX) were mated with a donor strain based on the laboratory strain BY4742. The donor strain contained a nuclear localisation marker subpopulation marker as well as kanamycin resistance placed under the MATa-specific pSTE2 promoter for haploid selection (MATα can1 ::pSte2-KanMX-tPGK1 his3Δ1 lys2Δ0 leu2 ::pTEF2-NLS-mtagBFP-tENO2-LEU2 ura3 ::pRPL28-mScarletI-tTDH1-URA3) (see Strain creation method section above). Mating was performed on solid agar plates in a 384 array using a Rotor pinning robot (Singer Instruments). Cells were then transferred to agar plates with minimal media (YNB) lacking amino acids to select prototrophic diploids. These diploids were then incubated in pre-sporulation media (YP with 1% potassium acetate) in liquid 96-well plates and grown for 24 h. Cells were then washed in PBS and resuspended in 1% potassium acetate sporulation media. After 5 days, haploid MATa spores were selected by transferring 50 µl of spores into 450 µl of synthetic media containing 50 µg/ml canavanine and 300 µg/ml G418. After 24 h, 50 µl were washed in PBS and resuspended in 450 µl of synthetic media lacking uracil, leucine, histidine, lysine and methionine containing 10 µg/ml canavanine and 300 µg/ml G418 to select prototrophic haploids overnight. This step was repeated in standard 96-well plates in total volumes of 150 µl and single colonies were selected by transferring cells onto rectangular agar plates using the Rotor pinning robot (Singer Instruments) and a 7 × 7 pinning protocol to select clonal populations from single colonies. Correct ploidy was confirmed using MATa- and MATα-specific primers as described in ref. 86 .

High-throughput microscopy

The transcription factor library consisting of 192 members was distributed along a 384 well plate μClear flat-bottom (Greiner) and imaged in a Nikon Ti-2 Twin-Cam-TIRF. Two images were taken at two different locations per library member using a bright field, blue, green and red fluorescent filter. To obtain maximal resolution without oil appliance, an ×40 optic with an additional ×1.5 lens was used. Looping through each unique well, an nd2 file was generated with a dimension of (fov(384) × channels (4) × Z stack (1) x X dimension x Y dimension)).

All data analysis was performed through a custom Python pipeline available on GitHub ( https://github.com/Benedict-Carling/YeaZ-Output-Analysis ). The generated nd2 file was used as an input for segmentation using the segmentation tool Yeast-Analyzer (YeaZ) 26 . The bright field segmentation parameters with a minimum seed distance of 1 and a threshold value of 0.5 were used. A scatter graph was generated for each cell identified by YeaZ with the x axis representing the cell size and the y-axis representing the mean Red fluorescence of the cell. To mitigate the influence of outliers, such as mis-segmentations in the scatter plot, we employed the KernelDensity utility from the Scikit-learn Python package 87 with the following arguments: algorithm was set to ball_tree, bandwidth set to 1, metric set to Euclidean and kernel set to linear. Expectation Maximisation (EM) was performed using the Gaussian Mixture utility from Scikit-learn on the filtered cells to assign cells to each subpopulation cluster. We performed EM with a confidence threshold of 0.85 to remove any manual intervention in the identification of the subpopulations.

To generate nuclear segmentation, the bounding box of each cell as identified by YeaZ was looped through using the blue channel (nuclear localisation marker). The image was smoothed using a Gaussian blur of sigma = 1. A mask of the top 15% brightest pixels was generated, and erosion and dilation were performed to remove isolated islands, returning a mask representing the nucleus of each cell. Single-cell transcription factors nuclear intensity were calculated by averaging the GFP signal over the nuclear mask. Subpopulation-specific scores were obtained by averaging the scores of each single-cell for a given subpopulation. For condition-specific scores, localisation scores were then averaged across subpopulations. To obtain the most consistent transcription factors over time, relative TF scores were computed at each point. For the relative TF score calculation, each TF score was normalised by the mean of all TF that were not part of the top 5 or bottom 5 transcription factor at a given timepoint.

Targets identified in the previous step were validated using inverted microscopy using a Nikon Ti (×60 magnification). Exposure time was kept constant for each channel (200 ms for RFP, 500 ms for BFP and 1 s for GFP) with 11 slices per z stack to capture the nucleus. Cells were exposed to a 30 min shift prior to imaging.

Statistical analysis and reproducibility

Heterogeneity in cell size and GFP measurements using flow cytometry were performed in triplicates and on separate days. Chronological lifespan measurements were performed at least on three biological replicates. Paired statistical analysis was performed using the t test function (unpaired, two-sided) in R. Statistical analysis of bimodal distributions was performed using Hartigan’s diptest. Significance scores for genes differentially expressed during subpopulation RNAseq and single-cell RNAseq were adjusted using DESeq2 (multiple hypothesis adjusted p value, Benjamini–Hochberg procedure). For significance testing of discrete data obtained from manual budscar quantification, we used Fisher’s exact testing, and error bars represent 95% confidence intervals.

Reporting summary

Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Source data, acknowledgements.

We thank Davina Patel (RLA lab) for helping with flow cytometry, Maria Portella (RLA Lab) for helping with the sensors, Larissa Zarate Garcia and the Imperial College Flow Cytometry Facility (Sir Alexander Fleming Building) for helping with cell sorting, David Bell (SynbiCITE) for helping with metabolomics, Mary Dunlop (Boston University) for fruitful discussions regarding cell tracking, Peter Thorpe (Queen Mary University) for providing members of the yeast GFP collection, Gianni Liti (Université de Nice) and Lars Steinmetz (EMBL Heidelberg) for providing laboratory as well as WT S. cerevisiae and S. paradoxus isolates. RLA received funding from BBSRC (BB/R01602X/1, BB/T013176/1, BB/T011408/1—19-ERACoBioTech- 33 SyCoLim), EPSRC (AI-4-EB), Yeast4Bio Cost Action 18229, European Research Council (ERC) (DEUSBIO—949080) and the Bio-based Industries Joint (PERFECOAT—101022370) under the European Union’s Horizon 2020 research and innovation programme. The Facility for Imaging by Light Microscopy (FILM) at Imperial College London is partially supported by funding from the Wellcome Trust (grant 104931/Z/14/Z). K.S. acknowledges a postdoctoral fellowship from the European Molecular Biology Organisation (EMBO) (ALTF 769-2021) and a UKRI-Marie Skłodowska-Curie Actions (MSCA) Postdoctoral Fellowship (UNICOH).

Author contributions

K.S. and R.L.A. conceptualised the study. K.S., C.K., B.C., and J.M. performed the experiments. C.K. and K.S. designed and created the TF-GFP library. B.C. and K.S. designed and created the high-throughput microscopy pipeline. J.S. cloned the metabolic sensors. K.S., C.K., and B.C. performed data analysis. K.S. wrote the draft. K.S., R.L.A., and M.S. supervised. All authors edited and reviewed the final manuscript.

Peer review

Peer review information.

Nature Communications thanks Murat Acar, Masamitsu Sato and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Code availability, competing interests.

The authors declare no competing interests.

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

Contributor Information

Kiyan Shabestary, Email: [email protected] .

Rodrigo Ledesma-Amaro, Email: [email protected] .

The online version contains supplementary material available at 10.1038/s41467-024-50602-8.