Structured cellular populations are complex, dynamic systems and their composition, expansion and internal structure are the result of interactions between the cells and their microenvironment. Cells absorb and metabolize nutrients and also produce and secrete metabolites, creating spatial gradients of nutrients and metabolites. Thus, cells at the outskirts of a multicellular assembly do not experience the same microenvironment as the cells deeply buried within. Reciprocally, cellular physiology is dependent on the cell’s position within a colony. Such variations in cellular physiology are consistently observed in a variety of multicellular systems – from bacterial and yeast colonies (Vulin et al., 2014; Cáp et al., 2012) to biofilms (Nadell et al., 2016) and tumors (Carmona-Fontaine et al., 2013; Delarue et al., 2014) – and are reflected by altered gene expression levels and cellular phenotypes as growth rates, nutrient uptake rates and metabolic activity. Such variations presumably emerge because of long-range metabolic interactions between cells, in that the cellular microenvironment at one position depends on the nutrient uptake rate at another position.

Notably, multicellular communities (Shapiro, 1998; Shou et al., 2007; Xavier and Foster, 2007) exhibit various adaptive benefits, including higher cell proliferation, improved access to resources and niches (Koschwanez et al., 2011), collective defence (e.g., against antagonists, drugs, antibiotics) (Nadell et al., 2016) resulting in optimization of population survival when confronted with averse physical, chemical, nutritional or biological challenges (Palková and Váchová, 2006). These examples indicate the importance of understanding the emergence and maintenance of complex spatial multicellular structures from ecological (Antwis et al., 2017; Gonzalez et al., 2012; Widder et al., 2016), medical (Bryers, 2008; Gilbert et al., 2018; Estrela and Brown, 2018) and evolutionary (Ratcliff et al., 2012; Nadell et al., 2010; Kim et al., 2008) perspectives. Yet, despite the obvious contrast between homogeneous environments and the pronounced environmental heterogeneity of microbial cellular assemblies, the majority of scientific research to date has either focused on single cells in homogeneous environments or populations of cells grown in batch or continuous liquid cultures. This is mostly due to the complexity of designing an experiment that would allow monitoring, over long time, the development of a spatially defined extended multicellular assembly. This is in particular the case for the widely used eukaryotic model organism yeast Saccharomyces cerevisiae, despite the numerous calls in recent reviews to study its nutrient sensing, signaling, and related growth and development control within the natural colony context. (Conrad et al., 2014; Broach, 2012; Horák, 2013).

As microorganisms in nature tend to live in multicellular communities, devising an experimental approach that captures this complexity while being easy to use and amenable to different experimental needs and conditions should further our understanding of complex gene regulatory networks in the context of microbial evolution and ecology.

Current direct observations of three-dimensional colonies and biofilms are cumbersome and often constrained by existing technologies (Nadell et al., 2016). For example, two-photon microscopy of sliced agarose-encapsulated yeast colonies was required to show that yeast cells may adopt different physiologies – and possibly different cell types – depending on their position within a colony (Cáp et al., 2012). In another example, nanospray desorption electrospray ionization mass spectrometry (nanoDESI MS) was used to study growing bacterial colonies on agar plates and showed a wide diversity and complexity of compounds that characterize microbial chemical ecology (Watrous et al., 2012; Traxler and Kolter, 2012). Such complex methodologies are not amenable to time-lapse imaging, nor to observation of the temporal variations in gene expression and growth rates of single cells over relevant time and length scales. An alternative is to grow microbial cells in microfluidic devices to spatially constrain the growth of the cells and to control the delivery of nutrients (Bennett and Hasty, 2009; Cookson et al., 2005; Robert et al., 2010; Ni et al., 2012; Llamosi et al., 2016). Microfluidic experimental research is typically designed to ensure that the cells being studied experience a homogeneous environment. This can be done at the single cell level, as it has been demonstrated in studies of aging of yeast (Jo et al., 2015) and bacteria (Yang et al., 2019) where single cells had to be trapped and kept under constant nutrient flow for long term observations to capture their death. Alternatively, a small cell assembly can be trapped in dead-end chambers under assumption that a quick diffusion of nutrients will keep the environment in chambers homogeneous. With that approach cell lineages were tracked for bacteria (Wang et al., 2010) (the widely used ‘mother machine’) and yeast (Xu et al., 2015), cells were subjected to fluctuating environments of different carbon sources to study non-genetic memory in bacteria (Lambert et al., 2014), and bacterial colonies were synchronized through quorum sensing and gas-phase redox signalling over centimeter-length scales to produce oscillating colony ‘biopixels’ (Prindle et al., 2011). Although effective for their specific applications, unfortunately, such devices do not fully capture emerging properties at a colony level, that is spatial variations in growth rates, microenvironments and phenotypes. Recently, there have been a few attempts to use microfluidics to study collective properties of bacterial colonies grown as a microcolony. Hornung et al. (2018) grew two-dimensional bacterial microcolonies in a 75 µm long device perfused with a very low concentration (up to 585 µM) of protocatechuic acid as the only carbon source from both sides of the cell chamber. They observed heterogeneous growth in agreement with a combination of a reaction-diffusion model and particle-based simulations. In a similar setup, a 60 µm long device perfused with a very low concentration of glucose (up to 800 µM) from one side was used to study the emergence of microscale gradients that resulted in metabolic cross-feeding between glucose-fermenting and acetate-respiring subpopulations of bacteria and antibiotic tolerance by slow growing subpopulation (Co et al., 2019). Wilmoth et al. used microwells up to 100 µm in diameter to look at spatial patterns of H1-Type VI secretion system (T6SS) mutants of Pseudomonas aeruginosa accompanied with an agent-based model depicting the two observed subpopulations (Wilmoth et al., 2018). They found that spatial constraints and local concentrations of growth substrates affect the spatial organization of cells. Finally, Liu et al. grew Bacillus subtilis biofilms perfused with glycerol and glutamate media. They discovered collective oscillations which emerge as a consequence of long-range metabolic co-dependence between cells in the interior and cells at the periphery of a biofilm, presumably to maximize the availability of nutrients and survival of interior cells (Liu et al., 2015). While the use of microfluidics gave rise to the discovery of collective properties of microbial assemblies and facilitated quantitative observations, such attempts are currently limited by inherent difficulties tied to fabrication of efficient microfluidic devices and choice of model organism. These limitations include small device dimensions (<100 µm) (Hornung et al., 2018; Co et al., 2019; Wilmoth et al., 2018), use of low nutrient concentrations (<1 mM) (Hornung et al., 2018; Co et al., 2019), limited scope of nutrient types (Hornung et al., 2018; Co et al., 2019; Liu et al., 2015) and in some cases inability to access single cell level measurements (Wilmoth et al., 2018). Therefore, it is challenging to apply such devices for the general case of the study of a large monolayer of cells in standard range and scope of nutrients, often used in biological research in liquid cultures. Additionally, it is tempting to reconstruct the emergence of gene expression landscapes on a global scale (e.g., within structured populations) from local (e.g., single cell) properties, given the extensive knowledge accumulated on single-cell gene regulatory networks. However, the variations in the microenvironment within a multicellular assembly and their interconnections with gene expression and cell metabolism are rather poorly known.

In attempt to overcome the above limitations in current methodologies and observe emerging properties at a colony level in larger dimensions and standard nutrient conditions, we developed a microfluidic device to grow thin, extended arrays of yeast cell monolayers that are perfused with nutrients from a single direction. We demonstrate the emergence of heterogeneous microenvironments and quantify spatial variation in cellular growth rate and the formation of gene expression landscapes for key metabolic genes involved in glucose transport and utilization, across the nascent 2D microcolony in 800 µm long cell chambers and up to 444 mM glucose concentration. Interestingly, the gene expression landscapes exhibited a high degree of spatial correlation over a wide range of glucose concentrations. Notably, we show that a growing extended assembly of cells presents a robust, steady state spatial structure, transitioning between fermentative (high glucose environment, fast growth, rapid glucose utilization) and respiratory (low glucose environment, slow growth, slow but efficient glucose utilization [Pfeiffer and Morley, 2014; Hagman and Piškur, 2015]) regimes, located close to and far from the nutrient source, respectively. This spatial structure emerges from the interplay between how cells individually adapt to the microenvironment and, at the same time, alter their surroundings as a result of their metabolic activity. Said differently, structured cells create and experience a spatially structured micro-environment through the interplay of nutrient diffusion and uptake without any obvious inherent biological program that would imply cell-cell communication and coordinated community action.

Here, we took an alternative point of view compared to traditional systems and single-cell biology. Rather than studying single-cell metabolic properties in a well-mixed, homogeneous environment, we designed a microfluidic chip to force yeast cells to grow and shape their microenvironment, solely by fixing the properties of the microenvironment at the boundary of the monolayer. This approach allowed us to measure simultaneously properties at both the single-cell scale and structured population scale and holds potential for establishing a quantitative link between these scales.

In standard batch culture, as cells exhaust the media, their adaptation time, limited by sensing, transcription and translation, may lag behind the decrease of glucose in the media. In contrast, in our 2D colony device, once steady-state gradient is settled, cells’ residence time within a given range of glucose that corresponds to stable HXT expression is significantly longer, assuring that even though cells are continuously pushed away, they spend sufficient time within a given concentration to equilibrate their response to the gradient and therefore faithfully report the glucose concentration in their environment. As example, at 1% glucose the mean velocity at Hxt7 expression peak, spanning over >200 µm (Figure 3a, varies between 5–15 µm/hour (Figure 2f), suggesting cells’ residence time of >10 hr, pointing to the advantage of working in this setup.

Specifically, we showed that cells self-generate nutrient landscapes that in turn influence cellular metabolism and gene expression profiles. This behavior, based on nutrient uptake adaptation, is generic and feeds back on the behavior of other cells through what we call non-specific long-range metabolic interactions. Indeed, the microenvironment sensed by cells a few hundred micrometres inside a colony is very different from the microenvironment experienced by the cells at periphery. The resulting patterns may emerge from the individual adaptive properties of the cells without the need to evoke specific higher-level community properties as cell-cell communication. Notably, gradients emerge over relatively short distances, and this process may possibly affect studies of cellular populations within microfluidics settings. More importantly, quantitative description of gene expression landscapes is critical if one wants to understand the establishment and behavior of cellular populations, whether these are as simple as yeast colonies or more complex, such as biofilms and complex microbial ecosystems in which several types of cells cohabit and may specifically communicate and interact. Indeed, in addition to the described long-range metabolic interactions, many other environmental and genetic determinants such as intercellular communication, cell surface properties, cell-cell adhesion strength and secretion of extracellular matrix components have been shown to participate in the emergence of the complex morphology (Nadell et al., 2016; Granek and Magwene, 2010; Flemming et al., 2016) and internal structure of microbial colonies in such complex situations. The nature of many of these interactions could also be studied using similar microfluidic devices to identify the relative contribution and relationship of environmental and genetic determinants to the metabolically generated microenvironment.

We have made notable advances in the study of emerging properties of yeast colony growth, microenvironment formation and gene expression compared to previously published studies (Cáp et al., 2012; Maršíková et al., 2017; Palková et al., 2014). These studies have shown fascinating differentiation and diversity within yeast colonies grown on agar but their relevance to study the dynamical emergence of complexity in microbial colonies is limited by their methodology (e.g., growth on a single specific medium with no dynamic control of environmental changes, two-photon microscopy, unsuitable for live time-lapse microscopy, obligation to section colonies etc.) which does not allow detailed spatiotemporal analysis of cellular growth, microenvironment and gene expression landscapes at a relevant single-cell scale. Our approach is designed to access the dynamics of large microbial colonies, and while we did not report it here, it is straightforward with microfluidics to dynamically change in frequency and composition the external environment, and as such to analyse how colonies adapt their internal organization to such stresses.

Our results are in most part in line with the knowledge of glucose metabolism obtained in batch culture. Yet, our methodology sheds quantitative description of the spatial expression of genes involved in the glucose metabolism and its correlation with the cell local growth rates. Our results show that even in the simple context studied here, reconstructing the microenvironment spatial structure from single-cell measurement is not trivial. A proper model should take into account how the growth rate and specific absorption rate vary with the glucose concentration and the microenvironment. Modeling the entire complexity of the microenvironment is hardly possible, even with today’s knowledge. Thus, we decided to take a different approach and use key genes involved in glucose metabolism to infer the glucose concentration gradient. We showed that different reporter genes consistently reported the same glucose gradient. We envision that the data extracted from relevant fluorescent reporters could be fed into an agent-based or mean-field models that take cell-cell interactions, mechanics and spatial diffusion of metabolites into account to fill the gap between data generated from single cells to data that is relevant to evolution and ecology, that is at the colony scale. We anticipate that linking local properties to macroscopic, global behavior will help to understand the architecture of microbial communities and how evolution shapes the development of these architectures through long-range metabolic interactions and possible inherent biological programs that coordinate it. Of note, in another rare attempt to study emergence of population level phenomena in yeast S. cerevisiae Campbell et al. (2015) looked at the synthetic ‘self-establishing communities’ that were able to cooperatively exchange metabolites. They inoculated on agar plate auxotrophic S. cerevisiae strain that had different auxotrophic markers on plasmids. As cells were dividing, some of the plasmids that complemented yeast auxotrophy and therefore rescued their growth were lost, resulting in a colony which is composed of yeast that are auxotrophic for a certain amino acid. However, they were able to grow because they used amino acids that were released in the environment by other yeast that were producing it, effectively generating a very heterogeneous colony that sustained growth through metabolite exchange. Interestingly, previous efforts to co-culture complementary auxotrophs had limited effectiveness in supporting co-growth in liquid cultures, indicating the importance of spatial structure in facilitating cooperation and makes our system very attractive for study of such phenomena (Wintermute and Silver, 2010).

Furthermore, while the spatial microenvironment is not fully characterized, we have shown that the emergence of gradients, and simultaneously gene expression landscapes, are robust and reproducible features of the colony. Moreover, the landscapes can be compared to extract correlation patterns and infer how gene regulatory networks act in synchronicity to establish the microenvironment within the colony. This approach may provide a relatively simple, yet effective method of screening for ‘organismic’ properties that have been shaped by evolution and are only relevant in a multicellular context.

Our future efforts to extend the application of this setup will be dedicated to the study of how the microenvironment dynamically changes when external conditions are altered, an uncharted territory at the scale of a multicellular assembly that is central to the understanding of microbial ecosystem resistance to stress, environmental fluctuations and adaptation. We anticipate that similar approaches could be used to study aging, cooperation and competition, cell memory or evolutionary dynamics, as well as quantitative characterization of (synthetic) ecological systems and mixtures of cells relevant to ecology and chemical biology.

All data presented in the figures, article, and supplementary information are available on Zenodo (http://dx.doi.org/10.5281/zenodo.2557396).

1. 1. Marinkovic Z
   2. Vulin C
   3. Acman M
   4. Song X
   5. Di Meglio JM
   6. Lindner AB
   7. Hersen P

   (2019) Zenodo

   Lab513/Landscape of Gene Expression Dataset.

   https://doi.org/10.5281/zenodo.2557396

1. 1. Antwis RE
   2. Griffiths SM
   3. Harrison XA
   4. Aranega-Bou P
   5. Arce A
   6. Bettridge AS
   7. Brailsford FL
   8. de Menezes A
   9. Devaynes A
   10. Forbes KM
   11. Fry EL
   12. Goodhead I
   13. Haskell E
   14. Heys C
   15. James C
   16. Johnston SR
   17. Lewis GR
   18. Lewis Z
   19. Macey MC
   20. McCarthy A
   21. McDonald JE
   22. Mejia-Florez NL
   23. O’Brien D
   24. Orland C
   25. Pautasso M
   26. Reid WDK
   27. Robinson HA
   28. Wilson K
   29. Sutherland WJ

   (2017) Fifty important research questions in microbial ecology

   FEMS Microbiology Ecology 93:fix044.

   https://doi.org/10.1093/femsec/fix044

   • Google Scholar
2. 1. Bennett MR
   2. Hasty J

   (2009) Microfluidic devices for measuring gene network dynamics in single cells

   Nature Reviews Genetics 10:628–638.

   https://doi.org/10.1038/nrg2625

   • PubMed
   • Google Scholar
3. 1. Bonander N
   2. Ferndahl C
   3. Mostad P
   4. Wilks MD
   5. Chang C
   6. Showe L
   7. Gustafsson L
   8. Larsson C
   9. Bill RM

   (2008) Transcriptome analysis of a respiratory Saccharomyces cerevisiae strain suggests the expression of its phenotype is glucose insensitive and predominantly controlled by Hap4, Cat8 and Mig1

   BMC Genomics 9:365.

   https://doi.org/10.1186/1471-2164-9-365

   • PubMed
   • Google Scholar
4. 1. Boyer D
   2. Mather W
   3. Mondragón-Palomino O
   4. Orozco-Fuentes S
   5. Danino T
   6. Hasty J
   7. Tsimring LS

   (2011) Buckling instability in ordered bacterial colonies

   Physical Biology 8:026008.

   https://doi.org/10.1088/1478-3975/8/2/026008

   • PubMed
   • Google Scholar
5. 1. Broach JR

   (2012) Nutritional control of growth and development in yeast

   Genetics 192:73–105.

   https://doi.org/10.1534/genetics.111.135731

   • PubMed
   • Google Scholar
6. 1. Bryers JD

   (2008) Medical biofilms

   Biotechnology and Bioengineering 100:1–18.

   https://doi.org/10.1002/bit.21838

   • PubMed
   • Google Scholar
7. 1. Campbell K
   2. Vowinckel J
   3. Mülleder M
   4. Malmsheimer S
   5. Lawrence N
   6. Calvani E
   7. Miller-Fleming L
   8. Alam MT
   9. Christen S
   10. Keller MA
   11. Ralser M

   (2015) Self-establishing communities enable cooperative metabolite exchange in a eukaryote

   eLife 4:e09943.

   https://doi.org/10.7554/eLife.09943

   • PubMed
   • Google Scholar
8. 1. Cáp M
   2. Stěpánek L
   3. Harant K
   4. Váchová L
   5. Palková Z

   (2012) Cell differentiation within a yeast colony: metabolic and regulatory parallels with a tumor-affected organism

   Molecular Cell 46:436–448.

   https://doi.org/10.1016/j.molcel.2012.04.001

   • PubMed
   • Google Scholar
9. 1. Carmona-Fontaine C
   2. Bucci V
   3. Akkari L
   4. Deforet M
   5. Joyce JA
   6. Xavier JB

   (2013) Emergence of spatial structure in the tumor microenvironment due to the warburg effect

   PNAS 110:19402–19407.

   https://doi.org/10.1073/pnas.1311939110

   • PubMed
   • Google Scholar
10. Preprint

    1. Co AD
    2. Vliet S
    3. Ackermann M

    (2019) Emergent microscale gradients give rise to metabolic cross-feeding and antibiotic tolerance in clonal bacterial populations

    bioRxiv.

    https://doi.org/10.1101/534149

    • Google Scholar
11. 1. Conrad M
    2. Schothorst J
    3. Kankipati HN
    4. Van Zeebroeck G
    5. Rubio-Texeira M
    6. Thevelein JM

    (2014) Nutrient sensing and signaling in the yeast Saccharomyces cerevisiae

    FEMS Microbiology Reviews 38:254–299.

    https://doi.org/10.1111/1574-6976.12065

    • PubMed
    • Google Scholar
12. 1. Cookson S
    2. Ostroff N
    3. Pang WL
    4. Volfson D
    5. Hasty J

    (2005) Monitoring dynamics of single-cell gene expression over multiple cell cycles

    Molecular Systems Biology 1:E1–E6.

    https://doi.org/10.1038/msb4100032

    • Google Scholar
13. 1. Delarue M
    2. Montel F
    3. Vignjevic D
    4. Prost J
    5. Joanny JF
    6. Cappello G

    (2014) Compressive stress inhibits proliferation in tumor spheroids through a volume limitation

    Biophysical Journal 107:1821–1828.

    https://doi.org/10.1016/j.bpj.2014.08.031

    • PubMed
    • Google Scholar
14. 1. Diderich JA
    2. Schepper M
    3. van Hoek P
    4. Luttik MA
    5. van Dijken JP
    6. Pronk JT
    7. Klaassen P
    8. Boelens HF
    9. de Mattos MJ
    10. van Dam K
    11. Kruckeberg AL

    (1999) Glucose uptake kinetics and transcription of HXT genes in chemostat cultures of Saccharomyces cerevisiae

    The Journal of Biological Chemistry 274:15350–15359.

    https://doi.org/10.1074/jbc.274.22.15350

    • PubMed
    • Google Scholar
15. 1. Edelstein AD
    2. Tsuchida MA
    3. Amodaj N
    4. Pinkard H
    5. Vale RD
    6. Stuurman N

    (2014) Advanced methods of microscope control using μmanager software

    Journal of Biological Methods 1:e10.

    https://doi.org/10.14440/jbm.2014.36

    • Google Scholar
16. 1. Estrela S
    2. Brown SP

    (2018) Community interactions and spatial structure shape selection on antibiotic resistant lineages

    PLOS Computational Biology 14:e1006179.

    https://doi.org/10.1371/journal.pcbi.1006179

    • PubMed
    • Google Scholar
17. 1. Flemming HC
    2. Wingender J
    3. Szewzyk U
    4. Steinberg P
    5. Rice SA
    6. Kjelleberg S

    (2016) Biofilms: an emergent form of bacterial life

    Nature Reviews Microbiology 14:563–575.

    https://doi.org/10.1038/nrmicro.2016.94

    • PubMed
    • Google Scholar
18. 1. Gilbert JA
    2. Blaser MJ
    3. Caporaso JG
    4. Jansson JK
    5. Lynch SV
    6. Knight R

    (2018) Current understanding of the human microbiome

    Nature Medicine 24:392–400.

    https://doi.org/10.1038/nm.4517

    • PubMed
    • Google Scholar
19. 1. Gonzalez A
    2. King A
    3. Robeson MS
    4. Song S
    5. Shade A
    6. Metcalf JL
    7. Knight R

    (2012) Characterizing microbial communities through space and time

    Current Opinion in Biotechnology 23:431–436.

    https://doi.org/10.1016/j.copbio.2011.11.017

    • PubMed
    • Google Scholar
20. 1. Granek JA
    2. Magwene PM

    (2010) Environmental and genetic determinants of colony morphology in yeast

    PLOS Genetics 6:e1000823.

    https://doi.org/10.1371/journal.pgen.1000823

    • PubMed
    • Google Scholar
21. 1. Hagman A
    2. Piškur J

    (2015) A study on the fundamental mechanism and the evolutionary driving forces behind aerobic fermentation in yeast

    PLOS ONE 10:e0116942.

    https://doi.org/10.1371/journal.pone.0116942

    • PubMed
    • Google Scholar
22. 1. Horák J

    (2013) Regulations of sugar transporters: insights from yeast

    Current Genetics 59:1–31.

    https://doi.org/10.1007/s00294-013-0388-8

    • PubMed
    • Google Scholar
23. 1. Hornung R
    2. Grünberger A
    3. Westerwalbesloh C
    4. Kohlheyer D
    5. Gompper G
    6. Elgeti J

    (2018) Quantitative modelling of nutrient-limited growth of bacterial colonies in microfluidic cultivation

    Journal of the Royal Society Interface 15:20170713.

    https://doi.org/10.1098/rsif.2017.0713

    • PubMed
    • Google Scholar
24. 1. Hovsepian J
    2. Defenouillère Q
    3. Albanèse V
    4. Váchová L
    5. Garcia C
    6. Palková Z
    7. Léon S

    (2017) Multilevel regulation of an α-arrestin by glucose depletion controls hexose transporter endocytosis

    The Journal of Cell Biology 216:1811–1831.

    https://doi.org/10.1083/jcb.201610094

    • PubMed
    • Google Scholar
25. 1. Jo MC
    2. Liu W
    3. Gu L
    4. Dang W
    5. Qin L

    (2015) High-throughput analysis of yeast replicative aging using a microfluidic system

    PNAS 112:9364–9369.

    https://doi.org/10.1073/pnas.1510328112

    • PubMed
    • Google Scholar
26. 1. Kim HJ
    2. Boedicker JQ
    3. Choi JW
    4. Ismagilov RF

    (2008) Defined spatial structure stabilizes a synthetic multispecies bacterial community

    PNAS 105:18188–18193.

    https://doi.org/10.1073/pnas.0807935105

    • PubMed
    • Google Scholar
27. 1. Koschwanez JH
    2. Foster KR
    3. Murray AW

    (2011) Sucrose utilization in budding yeast as a model for the origin of undifferentiated multicellularity

    PLOS Biology 9:e1001122.

    https://doi.org/10.1371/journal.pbio.1001122

    • PubMed
    • Google Scholar
28. 1. Lambert G
    2. Kussell E
    3. Kussel E

    (2014) Memory and fitness optimization of Bacteria under fluctuating environments

    PLOS Genetics 10:e1004556.

    https://doi.org/10.1371/journal.pgen.1004556

    • PubMed
    • Google Scholar
29. 1. Liu J
    2. Prindle A
    3. Humphries J
    4. Gabalda-Sagarra M
    5. Asally M
    6. Lee DY
    7. Ly S
    8. Garcia-Ojalvo J
    9. Süel GM

    (2015) Metabolic co-dependence gives rise to collective oscillations within biofilms

    Nature 523:550–554.

    https://doi.org/10.1038/nature14660

    • PubMed
    • Google Scholar
30. 1. Llamosi A
    2. Gonzalez-Vargas AM
    3. Versari C
    4. Cinquemani E
    5. Ferrari-Trecate G
    6. Hersen P
    7. Batt G

    (2016) What population reveals about individual cell identity: single-cell parameter estimation of models of gene expression in yeast

    PLOS Computational Biology 12:e1004706.

    https://doi.org/10.1371/journal.pcbi.1004706

    • PubMed
    • Google Scholar
31. 1. Maier A
    2. Völker B
    3. Boles E
    4. Fuhrmann GF

    (2002) Characterisation of glucose transport in Saccharomyces cerevisiae with plasma membrane vesicles (countertransport) and intact cells (initial uptake) with single Hxt1, Hxt2, Hxt3, Hxt4, Hxt6, Hxt7 or Gal2 transporters

    FEMS Yeast Research 2:539–550.

    https://doi.org/10.1111/j.1567-1364.2002.tb00121.x

    • PubMed
    • Google Scholar
32. 1. Maršíková J
    2. Wilkinson D
    3. Hlaváček O
    4. Gilfillan GD
    5. Mizeranschi A
    6. Hughes T
    7. Begany M
    8. Rešetárová S
    9. Váchová L
    10. Palková Z

    (2017) Metabolic differentiation of surface and invasive cells of yeast colony biofilms revealed by gene expression profiling

    BMC Genomics 18:814.

    https://doi.org/10.1186/s12864-017-4214-4

    • PubMed
    • Google Scholar
33. 1. Nadell CD
    2. Foster KR
    3. Xavier JB

    (2010) Emergence of spatial structure in cell groups and the evolution of cooperation

    PLOS Computational Biology 6:e1000716.

    https://doi.org/10.1371/journal.pcbi.1000716

    • PubMed
    • Google Scholar
34. 1. Nadell CD
    2. Drescher K
    3. Foster KR

    (2016) Spatial structure, cooperation and competition in biofilms

    Nature Reviews Microbiology 14:589–600.

    https://doi.org/10.1038/nrmicro.2016.84

    • PubMed
    • Google Scholar
35. 1. Ni M
    2. Decrulle AL
    3. Fontaine F
    4. Demarez A
    5. Taddei F
    6. Lindner AB

    (2012) Pre-disposition and epigenetics govern variation in bacterial survival upon stress

    PLOS Genetics 8:e1003148.

    https://doi.org/10.1371/journal.pgen.1003148

    • PubMed
    • Google Scholar
36. 1. Ohlmeier S
    2. Kastaniotis AJ
    3. Hiltunen JK
    4. Bergmann U

    (2004) The yeast mitochondrial proteome, a study of fermentative and respiratory growth

    Journal of Biological Chemistry 279:3956–3979.

    https://doi.org/10.1074/jbc.M310160200

    • PubMed
    • Google Scholar
37. 1. Otterstedt K
    2. Larsson C
    3. Bill RM
    4. Ståhlberg A
    5. Boles E
    6. Hohmann S
    7. Gustafsson L

    (2004) Switching the mode of metabolism in the yeast Saccharomyces cerevisiae

    EMBO Reports 5:532–537.

    https://doi.org/10.1038/sj.embor.7400132

    • PubMed
    • Google Scholar
38. 1. Palková Z
    2. Wilkinson D
    3. Váchová L

    (2014) Aging and differentiation in yeast populations: elders with different properties and functions

    FEMS Yeast Research 14:96–108.

    https://doi.org/10.1111/1567-1364.12103

    • PubMed
    • Google Scholar
39. 1. Palková Z
    2. Váchová L

    (2006) Life within a community: benefit to yeast long-term survival

    FEMS Microbiology Reviews 30:806–824.

    https://doi.org/10.1111/j.1574-6976.2006.00034.x

    • PubMed
    • Google Scholar
40. 1. Pfeiffer T
    2. Morley A

    (2014) An evolutionary perspective on the crabtree effect

    Frontiers in Molecular Biosciences 1:17.

    https://doi.org/10.3389/fmolb.2014.00017

    • PubMed
    • Google Scholar
41. 1. Prindle A
    2. Samayoa P
    3. Razinkov I
    4. Danino T
    5. Tsimring LS
    6. Hasty J

    (2011) A sensing array of radically coupled genetic 'biopixels'

    Nature 481:39–44.

    https://doi.org/10.1038/nature10722

    • PubMed
    • Google Scholar
42. 1. Ratcliff WC
    2. Denison RF
    3. Borrello M
    4. Travisano M

    (2012) Experimental evolution of multicellularity

    PNAS 109:1595–1600.

    https://doi.org/10.1073/pnas.1115323109

    • PubMed
    • Google Scholar
43. 1. Reifenberger E
    2. Boles E
    3. Ciriacy M

    (1997) Kinetic characterization of individual hexose transporters of Saccharomyces cerevisiae and their relation to the triggering mechanisms of glucose repression

    European Journal of Biochemistry 245:324–333.

    https://doi.org/10.1111/j.1432-1033.1997.00324.x

    • PubMed
    • Google Scholar
44. 1. Robert L
    2. Paul G
    3. Chen Y
    4. Taddei F
    5. Baigl D
    6. Lindner AB

    (2010) Pre-dispositions and epigenetic inheritance in the Escherichia coli lactose operon bistable switch

    Molecular Systems Biology 6:357.

    https://doi.org/10.1038/msb.2010.12

    • PubMed
    • Google Scholar
45. 1. Schneider CA
    2. Rasband WS
    3. Eliceiri KW

    (2012) NIH image to ImageJ: 25 years of image analysis

    Nature Methods 9:671–675.

    https://doi.org/10.1038/nmeth.2089

    • PubMed
    • Google Scholar
46. 1. Shapiro JA

    (1998) Thinking about bacterial populations as multicellular organisms

    Annual Review of Microbiology 52:81–104.

    https://doi.org/10.1146/annurev.micro.52.1.81

    • PubMed
    • Google Scholar
47. 1. Shou W
    2. Ram S
    3. Vilar JM

    (2007) Synthetic cooperation in engineered yeast populations

    PNAS 104:1877–1882.

    https://doi.org/10.1073/pnas.0610575104

    • PubMed
    • Google Scholar
48. 1. Tinevez JY
    2. Perry N
    3. Schindelin J
    4. Hoopes GM
    5. Reynolds GD
    6. Laplantine E
    7. Bednarek SY
    8. Shorte SL
    9. Eliceiri KW

    (2017) TrackMate: an open and extensible platform for single-particle tracking

    Methods 115:80–90.

    https://doi.org/10.1016/j.ymeth.2016.09.016

    • PubMed
    • Google Scholar
49. 1. Traxler MF
    2. Kolter R

    (2012) A massively spectacular view of the chemical lives of microbes

    PNAS 109:10128–10129.

    https://doi.org/10.1073/pnas.1207725109

    • PubMed
    • Google Scholar
50. 1. Verwaal R
    2. Paalman JW
    3. Hogenkamp A
    4. Verkleij AJ
    5. Verrips CT
    6. Boonstra J

    (2002) HXT5 expression is determined by growth rates in Saccharomyces cerevisiae

    Yeast 19:1029–1038.

    https://doi.org/10.1002/yea.895

    • PubMed
    • Google Scholar
51. 1. Vulin C
    2. Di Meglio JM
    3. Lindner AB
    4. Daerr A
    5. Murray A
    6. Hersen P

    (2014) Growing yeast into cylindrical colonies

    Biophysical Journal 106:2214–2221.

    https://doi.org/10.1016/j.bpj.2014.02.040

    • PubMed
    • Google Scholar
52. 1. Wang P
    2. Robert L
    3. Pelletier J
    4. Dang WL
    5. Taddei F
    6. Wright A
    7. Jun S

    (2010) Robust growth of Escherichia coli

    Current Biology 20:1099–1103.

    https://doi.org/10.1016/j.cub.2010.04.045

    • PubMed
    • Google Scholar
53. 1. Watrous J
    2. Roach P
    3. Alexandrov T
    4. Heath BS
    5. Yang JY
    6. Kersten RD
    7. van der Voort M
    8. Pogliano K
    9. Gross H
    10. Raaijmakers JM
    11. Moore BS
    12. Laskin J
    13. Bandeira N
    14. Dorrestein PC

    (2012) Mass spectral molecular networking of living microbial colonies

    PNAS 109:E1743–E1752.

    https://doi.org/10.1073/pnas.1203689109

    • PubMed
    • Google Scholar
54. 1. Widder S
    2. Allen RJ
    3. Pfeiffer T
    4. Curtis TP
    5. Wiuf C
    6. Sloan WT
    7. Cordero OX
    8. Brown SP
    9. Momeni B
    10. Shou W
    11. Kettle H
    12. Flint HJ
    13. Haas AF
    14. Laroche B
    15. Kreft JU
    16. Rainey PB
    17. Freilich S
    18. Schuster S
    19. Milferstedt K
    20. van der Meer JR
    21. Groβkopf T
    22. Huisman J
    23. Free A
    24. Picioreanu C
    25. Quince C
    26. Klapper I
    27. Labarthe S
    28. Smets BF
    29. Wang H
    30. Soyer OS
    31. Isaac Newton Institute Fellows

    (2016) Challenges in microbial ecology: building predictive understanding of community function and dynamics

    The ISME Journal 10:2557–2568.

    https://doi.org/10.1038/ismej.2016.45

    • PubMed
    • Google Scholar
55. 1. Wilmoth JL
    2. Doak PW
    3. Timm A
    4. Halsted M
    5. Anderson JD
    6. Ginovart M
    7. Prats C
    8. Portell X
    9. Retterer ST
    10. Fuentes-Cabrera M

    (2018) A microfluidics and Agent-Based modeling framework for investigating spatial organization in bacterial colonies: the case of Pseudomonas Aeruginosa and H1-Type VI Secretion Interactions

    Frontiers in Microbiology 9:33.

    https://doi.org/10.3389/fmicb.2018.00033

    • PubMed
    • Google Scholar
56. 1. Wintermute EH
    2. Silver PA

    (2010) Dynamics in the mixed microbial concourse

    Genes & Development 24:2603–2614.

    https://doi.org/10.1101/gad.1985210

    • PubMed
    • Google Scholar
57. 1. Xavier JB
    2. Foster KR

    (2007) Cooperation and conflict in microbial biofilms

    PNAS 104:876–881.

    https://doi.org/10.1073/pnas.0607651104

    • PubMed
    • Google Scholar
58. 1. Xu Z
    2. Fallet E
    3. Paoletti C
    4. Fehrmann S
    5. Charvin G
    6. Teixeira MT

    (2015) Two routes to senescence revealed by real-time analysis of telomerase-negative single lineages

    Nature Communications 6:7680.

    https://doi.org/10.1038/ncomms8680

    • PubMed
    • Google Scholar
59. 1. Yang Y
    2. Santos AL
    3. Xu L
    4. Lotton C
    5. Taddei F
    6. Lindner AB

    (2019) Temporal scaling of aging as an adaptive strategy of Escherichia coli

    Science Advances 5:eaaw2069.

    https://doi.org/10.1126/sciadv.aaw2069

    • PubMed
    • Google Scholar
60. 1. Youk H
    2. van Oudenaarden A

    (2009) Growth landscape formed by perception and import of glucose in yeast

    Nature 462:875–879.

    https://doi.org/10.1038/nature08653

    • PubMed
    • Google Scholar

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Emergence of metabolic landscapes in yeast monolayer colonies" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

We appreciate how your study uses a microfluidics setup to study the behavior of S. cerevisiae in a glucose gradient. The combination of this setup with GFP tagged version of hexose transporters, a glucose sensitive transcription factor (MIG1), and other reporters of metabolic state (PDC1, SHD2) allows studying spatial patterning as a function of glucose availability. The results show that patterns of expression of hexose transporters with different affinities are spatially arranged in a manner consist with a gradient of glucose availability with highest amounts of glucose available near the "source" (i.e. periphery of the artificial colony), and lowest levels near the "sink" (i.e. center of the artificial colony). Perhaps more interestingly, interpretation of the data suggests a relatively sharp transition in the colony between zones of growth and slow/no-growth, which they ascribe to zones of fermentative and aerobic metabolism.

Overall, the reviewers agree that this is a particularly elegant study. As you can see in the individual reports (below), reviewer 1 is significantly more positive than reviewers 2 and 3. However, after discussing the individual reviews, all reviewers question the amount of novel biological insight that is presented in your paper.

As such, I am afraid that we are not able to offer publication in eLife (unless of course you believe that we missed the major message and its biological relevance). On the other hand, everyone also agrees that the experiments and results are beautiful, and that the setup may be usable for other studies. As such, we wonder whether it would be conceivable to re-write your paper as a "Tools and resources" paper? If so, you would need to emphasize the design and operation of the system, detail its use for a broad range of studies and compare it to other, existing systems such as the ones describes by Hornung et al., 2018, and Wilmoth et al., 2018. The Hornung et al. paper in particular seems to use a rather similar setup…

Reviewer #1:

The authors applied a microfluidic device to study how long-range metabolic interactions influence the internal structure of simple yeast cell monolayers asymmetrically exposed to an external nutrient source. They showed spatial variation in nutrient levels, cellular growth, and expression of several metabolic genes to self-emerge. Notably, the gene expression landscapes exhibited a high degree of spatial correlation over different levels of external nutrient level and were indicative of a clear spatial transition between fermentative and respirative growth.

I really enjoyed reading this work. Although the experimental set-up is simple and the conclusions intuitive and straightforward, the depth of the analyses and the clarity of the interpretations are impressive. The authors combined growth measurements, single cell tracking, fluorescence based expression analysis with FACS and microscopy and advanced data representation to study spatial relations between nutrients, growth and metabolism with an unprecedented level of detail. This work will therefore undoubtedly provide inspiration to many other researchers interested in spatial heterogeneity of microbial communities. I therefore recommend publication in eLife.

Reviewer #2:

Marinkovic et al. address single-cell response to glucose gradients within structured communities. They present an intriguing method to look at metabolic landscapes in a microbial community. However, I do believe the novelty of the results need to be clarified significantly better, and the conclusions drawn must be discussed more critically. Alternatively, the work may be more suitable for a methods focused publication, as using the gene expression of the transporters to gauge the extracellular concentration of glucose within a colony is an interesting tool.

Major concerns:

A) The glucose gradients emerge in the end artificially, due to the design of the microfluidic chamber and due to the fact the yeast's consume glucose, and not due to community action. The method of the author's pictures of how the cells respond to different glucose availability as a function of glucose concentration in different areas of the closed channels. But the effects are the same if one cultivates the cells in batch cultures, and tunes glucose availability in this way. One does hence not learn anything about yeast communities, only, that yeast in batch culture and in a microfluidic chamber respond similarly to glucose depletion. Hence, is there anything new in the study, other than nicely illustrating the well-established fact that yeast activates different glucose transporter at different glucose concentration?

B) Batch cultures are used to tune gene expression sensors. But, in batch culture, glucose is constantly and rapidly depleted, while gene expression, needs some time to adjust. At any given time, in batch culture, the glucose concentration in the media, and the gene expression programme, are hence asynchronous (i.e. gene expression lags behind the glucose concentration). Chemostat experiments would be needed, to have a glucose concentration that corresponds to the gene expression programme at a particular point in time.

Method:

Figure 1: I have not found the mention of the exact number of dead-end chambers within one microfluidic chip (I'm guessing 16 from the figures?). However, looking at the n in different experiments (e.g. Figure 1, n=12, n>5 and n=9) I am wondering in how many separate experiments these were done, and how much of the collected data was discarded and for which reasons?

Was it excluded that nutrient concentrations from the previous chamber affected the later ones? Especially in the stages when the chambers were already filling up/cells were being carried away?

Results and Conclusion:

Subsection “Cellular metabolic activity creates gene expression landscapes”: "Both patterns demonstrate the formation and maintenance of a glucose gradient that emerges from cellular metabolic activity." The pattern of the gene expression demonstrates that the cells adapt to the glucose gradient that emerges through supply of fresh media from only one side. "Formation and maintenance by cellular metabolic activity" in the sense that, yes, the cells close to the opening do import the glucose they require seems a weak point to make. In other words, there is evidence for the contrary – there seems no different action of cells in the community as in the batch culture, when it comes to the expression of the transporters.

Discussion paragraph four: What is the dependence/independence of the different gene expression looked at here? Can we talk about synchronicity between networks if the studied genes are known to dependent on the same external factor?

Figure 6: Landscapes of gene expression is not shown at all glucose concentrations for all genes, why?

Reviewer #3:

This manuscript details the results of a study using a microfluidic system to study growth and gene regulation in monolayers of yeast cells. This system is presented as a tractable model for studying pattern formation in microbial colonies, via spatial and temporal measurement of cell division rates and measurement of gene expression via fluorescently tagged reporters.

The authors use this experimental setup, in combination with GFP tagged version of hexose transporters, a glucose sensitive transcription factor (MIG1), and other reporters of metabolic state (PDC1, SHD2) to study spatial patterning as a function of glucose availability.

Unsurprisingly, they find that patterns of expression of hexose transporters with different affinities are spatially arranged in a manner consist with a gradient of glucose availability with highest amounts of glucose available near the "source" (i.e. periphery of the artificial colony), and lowest levels near the "sink" (i.e. center of the artificial colony). Perhaps more interestingly, their interpretation of their data suggest a relatively sharp transition in the colony between zones of growth and slow/no-growth, which they ascribe to zones of fermentive and aerobic metabolism.

While I appreciate the elegance of the microfluidic colony model that is presented, the key findings are rather modest -- i.e. there are gradients of nutrient availability in microbial colonies. This doesn't really have the impact I expect from eLife papers.

Below I detail a number of concerns/comments about the study, the analyses, and whether the authors have adequate related their work to the larger literature in this area.

Detailed comments:

1) Several other recent papers (Hornung et al., 2018,; Wilmoth et al., 2018) have used microfluidic setups to study microbial colonies. The Hornung et al. paper is the most similar to the current experimental design.

2) If goal was to measure the rate of glucose uptake and availability it seems like there may be other more direct approaches than inferring this from hexose transporter expression, such as the use of 2-NBDG, a non-metabolizable, fluorescent glucose analog (Roy et al. 2015, doi: 10.1371/journal.pone.0121985).

3) The following paper, while not at the fine spatial scale provided by the microfluidics, presents a more complete and compelling view of metabolic differentiation in yeast colonies: Maršíková et al., 2017.

4) The spatial analyses focus primarily on peaks of maximum expression, but the data is potentially much richer and more interesting if the authors consider not only the global maxima but more complex spatial patterns. For example, in Figure 4B MIG1 nuclear localization appears to be multimodal. What is going on here?

5) There is very little effort to contextualize this work in the larger body of studies of yeast colony structure, physiology, or regulation of glucose responsive TFs. Illustrative of this, there are only three citations total in the entire discussion. There is only a single reference to the regulation of hexose transporters in the entire manuscript. Specific examples:

- Subsection “Cellular metabolic activity creates gene expression landscapes”: " HXT1 is a low-affinity glucose transporter mainly expressed under high-glucose conditions, while HXT7 is a high-affinity glucose transporter expressed under low-glucose conditions only" Citations?

- Subsection “Gene expression landscapes depend on the glucose source concentration”: "As HXT1 is mainly expressed under high-glucose conditions (> 1% w/vol glucose) in batch culture…" Citation?

- Subsection “Gene expression landscapes of other genes and transcription factor activity confirm the inferred glucose gradients”: "MIG1 is a key transcription factor involved in glucose repression that localizes to the nucleus in the presence of glucose, to repress genes that participate in parallel carbon metabolic pathways (e.g., galactose)." Citation?

- and in the same subsection: "…we examined the expression of PDC1 and SDH2, which are overexpressed in fermenting and respiring cells, respectively" Citation?

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "A microfluidic device for inferring metabolic landscapes in yeast monolayer colonies" for further consideration at eLife. Your revised article has been favorably evaluated by Naama Barkai (Senior Editor), a Reviewing Editor, and three reviewers.

We really appreciate the efforts of re-writing this manuscript as a "resource" paper. We feel that the paper has been much improved, but there are some remaining issues that need to be addressed through textual changes before acceptance, as outlined below:

1) We ask that the authors make it clear to the reader that what they call a "community action" does not imply that cells are responding differently in the community than they would as single cells in planktonic growth. The observed effects (gradients and response to these gradients) are merely the consequence of a high concentration of more or less static cells. So, perhaps calling it "population effect" would be more appropriate?

2) We would appreciate if you could insert a succinct discussion of the difference between a batch culture and this steady-state (where in principle, the cells could be better adapted as they are not always "running behind" in their response to a variable environment, as they would be in batch). (see reviewer 2's comments below for more details)

3) We again ask to include more references and to not overstate the advantage of the presented device (see reviewer 3's comment below).

4) Please provide more details about the reproducibility/variability and statistics of the results and analyses.

The above points summarize what we believe to be essential changes, and we only ask for textual changes instead of further experiments. We are including the full reviews for your reference below.

Reviewer #1:

In this revised manuscript, the authors – as requested- presented their work as a tool that can be generally applied to quantitatively and dynamically study growth landscapes, metabolic landscapes and gene expression landscapes within extended monolayers of cells. They extensively elaborated on the similarities and differences of their approach compared to other studies that applied agar colony models or microfluidics devices to study spatiotemporal patterns in colonies, and clearly pinpointed advantages of their approach. A figure was added to clarify the experimental set-up. Also my other (minor) comments were addressed to a sufficient extent. I therefore recommend publication.

Reviewer #2:

Although the authors have put efforts in their revision, they have not addressed my two main points. I'm hence wondering if I have not expressed myself clearly enough, so it's perhaps my own fault – or it also could be a different use of language between disciplines, so that the authors did simply not understand my questions. Long story short, I don't want to sound negative as it may be my own use of language, but the revision has not addressed the two simple concerns I had. Perhaps the other reviewers could comment if they agree or disagree with me, I'm happy if my comments are ignored, in case my concerns are not shared.

I think both points still are relevant, though. The first one could be fixed simply in writing. 'Community action' implies to me, and I guess to many others in the field, that there are interactions between the cells that emerge in the community – implying that cells would behave differently in the community as if they are not in the community. When it comes to glucose consumption, this seems not to be the case – there seems nothing which indicates that the individual cell's glucose import would be depending on interaction terms between the cells. The manuscript is written as if glucose gradients would emerge because of 'community action', but in fact, they emerge from cells that consume the glucose, in a way that does not require a different action as if cells would act independently to one another. In other words, the cells deplete the glucose in the MF channels, and within the channel, this creates a gradient, to which the cells respond. This alone does not indicate that cells act differently in the community. as they would act in Isolation.

The second point may remain relevant, however. The sensors are tuned in batch culture. But in batch culture, one expects a time differential, between the glucose concentration present at a point in time, and the average gene expression program activated at that same moment (as gene expression needs time to adjust, but glucose is constantly depleted. I.e. if gene expression needs 30mins to adjust to a changed glucose concentration, the gene expression program reflects the glucose concentration that was in the batch 30mins earlier before the timepoint was taken). I had suggested the authors to control this in chemostats, where gene expression and concentration are a time-wise aligned. I agree this would have been time-consuming to do and could hence be difficult for the authors to do. But it would have been great if they could have come up at least with some idea, and quantify how big this effect is; One would assume the differential between glucose concentration and gene expression are strongest in the early and late exponential phase. If chemostats are too difficult, one could suggest also much simpler surrogate experiment, for instance. One could, for instance, re-supplement exhausted media with glucose, perform a time-course, and estimate in this way the time needed until the gene expression has re-adjusted to the supplied glucose level. The sensor data would need to be adjusted by the time differential.

Reviewer #3:

This revised manuscript has been appropriately rewritten as a "Tools and Resource" article.

1) The authors overstate the uniqueness of their approach. For example, in the Introduction in reference to previous studies using similar microfluidics designs:

" While the use of microfluidics gave rise to the discovery of interesting collective properties of microbial assemblies, such attempts were too specific and had to deal with some of the limitations like small device dimensions (<100 100 μm), use of low nutrient concentrations (<1 mM), limited scope of nutrient types, inability to access single cell level – and therefore cannot be transposed to the general case of the study of a large monolayer of cells in standard range and scope of nutrients employed in biological research."

Few of these critiques hold up. For example, low nutrient concentrations and limited scope of nutrient types are simply specifics factors applied to the different organisms study. This critique applies equally to the system presented here. Furthermore, the Hornung et al. article certainly allows access to the single cell level. The text should be revised to reflect a more accurate summary of this earlier related work.

2) The rebuttal states "We agree that more contextualization was necessary so we expanded our Introduction and Discussion part, as well as provided additional citations, including all the ones proposed by the reviewer" I can find no references our discussion of prior literature at the line numbers provided.

3) One of the points the authors make several times regards the reproducibility of their assays. For example: "We demonstrate a novel capacity to reproduce and

quantify…" and "…we found the growth pattern was highly reproducible across parallel chambers.." However in the eLife "transparent reporting form" the authors state "We did not use statistical inference or comparison between data sets." How then did the authors estimate reproducibility/repeatability?

[Editors’ note: the author responses to the first round of peer review follow.]

We thank the editors and reviewers for their careful review of our work. We concur that our work does not provide novel genetic network mechanistic insights but reveals and quantifies time- and length-scales at play in the emergence and the maintenance of gene expression and cellular growth patterns in a growing multicellular assembly. It also provides an elegant and straightforward method for quantitative spatiotemporal exploration of microbial colonies, as you outlined in your editorial decision. As such, and following your recommendations, we rewrote our article as a “Tools and Resources” submission, outlining the methodology and discussing how this could be used to gain novel insights in a broad range of multicellular systems. We changed the title, the abstract, rewrote part of the introduction and discussion sections and added one figure with methodological details of the microfluidic system and a few supplementary figures.

Reviewer #1:

The authors applied a microfluidic device to study how long-range metabolic interactions influence the internal structure of simple yeast cell monolayers asymmetrically exposed to an external nutrient source. They showed spatial variation in nutrient levels, cellular growth, and expression of several metabolic genes to self-emerge. Notably, the gene expression landscapes exhibited a high degree of spatial correlation over different levels of external nutrient level and were indicative of a clear spatial transition between fermentative and respirative growth.

I really enjoyed reading this work. Although the experimental set-up is simple and the conclusions intuitive and straightforward, the depth of the analyses and the clarity of the interpretations are impressive. The authors combined growth measurements, single cell tracking, fluorescence based expression analysis with FACS and microscopy and advanced data representation to study spatial relations between nutrients, growth and metabolism with an unprecedented level of detail. This work will therefore undoubtedly provide inspiration to many other researchers interested in spatial heterogeneity of microbial communities. I therefore recommend publication in eLife.

We do thank the reviewer for his positive assessment of our work. In light of the other reviewers comments, we recall here that our goal was indeed to get a quantitative analysis of cellular growth and gene expression landscapes in an extended monolayer of cells. We did not expect to gain novel mechanistic insights on, say, the glucose genetic regulatory network, which has been studied in details in the past. On the contrary, we used this knowledge (e.g., on HXTs) to extract the lengthscales on which gradients emerged as a function of the external glucose concentration. Providing with a quantitative, detailed, description of these landscapes is the main scientific contribution of our work. Methodologically, our contribution is probably broader, since we describe a tool for researchers interested in the emergence of metabolic gradients in multicellular systems (microbial colonies, biofilms, multicellular tissue, tumor, etc…). As suggested by the editors, we now addressed this in the discussion part of our paper.

Reviewer #2:

Marinkovic et al. address single-cell response to glucose gradients within structured communities. They present an intriguing method to look at metabolic landscapes in a microbial community. However, I do believe the novelty of the results need to be clarified significantly better, and the conclusions drawn must be discussed more critically. Alternatively, the work may be more suitable for a methods focused publication, as using the gene expression of the transporters to gauge the extracellular concentration of glucose within a colony is an interesting tool.

We thank the reviewer for his careful review. Our goal was both to provide with a simple method to study the emergence of gradients within multicellular systems (in contract with classic studies that tend to study isolated single cells) and to obtain a quantitative description of an extended monolayer of yeast cells. As such, our analysis does not bring novel mechanistic understanding of the biology of yeast, but rather focus on bringing quantitative data that could be used to model yeast colonies.

Major concerns:

A) The glucose gradients emerge in the end artificially, due to the design of the microfluidic chamber and due to the fact the yeast's consume glucose, and not due to community action. The method of the author's pictures of how the cells respond to different glucose availability as a function of glucose concentration in different areas of the closed channels. But the effects are the same if one cultivates the cells in batch cultures, and tunes glucose availability in this way. One does hence not learn anything about yeast communities, only, that yeast in batch culture and in a microfluidic chamber respond similarly to glucose depletion. Hence, is there anything new in the study, other than nicely illustrating the well-established fact that yeast activates different glucose transporter at different glucose concentration?

We disagree. The gradients do not appear artificially but because of the yeast cells. The design of the microfluidic chamber does not influence the gradients at all.If there were no cells, no gradients would be observed. We would argue that the gradients emerge precisely because there is community action. Cells at different positions consume nutrients at different rates and this is the main driving force of nutrient availability within a colony. Community structure (density, spatial structure, metabolic and genetic heterogeneity, etc.) will determine how, where and when different emergent properties arise. To put it differently, the spatial description of a yeast community cannot be learnt from single cell /batch analysis. Therefore, we do learn (quantitatively) something; the spatial pattern that emerges from single cell programs when cells are together, shaping a metabolic environment at a much larger scale than their own size.

Note that the gradient of glucose is not known inside the cell monolayer (and this is not mentioning the other nutrients) and cannot be inferred from existing models. Methods set aside, the scientific novelty of our study is that we now know quantitatively where and when different glucose transporters are expressed in space and time in a community context. It is a good fundament for further studies and models which could have hardly been extrapolated from batch culture glucose transporter expression data.

B) Batch cultures are used to tune gene expression sensors. But, in batch culture, glucose is constantly and rapidly depleted, while gene expression, needs some time to adjust. At any given time, in batch culture, the glucose concentration in the media, and the gene expression programme, are hence asynchronous (i.e. gene expression lags behind the glucose concentration). Chemostat experiments would be needed, to have a glucose concentration that corresponds to the gene expression programme at a particular point in time.

We are not sure to understand this comment. Our goal is NOT to study a particular gene in a chemostat condition. On the contrary, our method embraces the complexity of a dynamically growing cell monolayer and propose to observe and analyze its emergence and maintenance in terms of landscapes of growth and gene expression.

If this comment is directed to the way we quantified HXTs level of expression in batch culture, we do agree that the most elegant way to do this would be by using a chemostat, or even better, a microfluidic device that can trap single cells subjected to perfusion of media with a constant nutrient concentration. However, we made sure to measure glucose depletion in batch culture and adjusted our experiments accordingly (number of inoculated cells and transfer to a fresh medium) so that cells do not experience a significant drop in glucose concentration. We show in Figure 6—figure supplement 1 the concentration of glucose in supernatant right before flow cytometry measurements, and it was more than 90% of initial glucose concentration.

Method:

Figure 1: I have not found the mention of the exact number of dead-end chambers within one microfluidic chip (I'm guessing 16 from the figures?). However, looking at the n in different experiments (e.g. Figure 1, n=12, n>5 and n=9) I am wondering in how many separate experiments these were done, and how much of the collected data was discarded and for which reasons?

We used different microfluidic chips but most of the experiments presented here contain 5 – 10 chamber in parallel. The number of experiments is the sum of all “growing monolayers” for a given condition, observed in several, separated experiments (2-3). Sometimes it can happen that either no or not enough cells are loaded in one or two chambers and therefore they remain empty or take a long time to fill up compared to other well-loaded chambers. Or, it can happen that during the growth some chambers get clogged by cells and then we do not take these chambers into account for analysis. In the end, we took all growing monolayers into account and did not discard any data in the analysis we provide.

Was it excluded that nutrient concentrations from the previous chamber affected the later ones? Especially in the stages when the chambers were already filling up/cells were being carried away?

Yes. The dimensions of the feeding channels and the flow rate (5μL/min) ensures that the concentration seen by the cells in different chambers are independent. In line with this consideration, we did not observe any decays of the growth rate of cells or of the speed of the monolayer front, nor any drift in the gene expression landscapes from one chambers to the next.

Results and Conclusion:

Subsection “Cellular metabolic activity creates gene expression landscapes”: "Both patterns demonstrate the formation and maintenance of a glucose gradient that emerges from cellular metabolic activity." The pattern of the gene expression demonstrates that the cells adapt to the glucose gradient that emerges through supply of fresh media from only one side. "Formation and maintenance by cellular metabolic activity" in the sense that, yes, the cells close to the opening do import the glucose they require seems a weak point to make. In other words, there is evidence for the contrary – there seems no different action of cells in the community as in the batch culture, when it comes to the expression of the transporters.

We are not sure we understand the point made by the reviewer. We do not intend to show that there are “surprising” differences between what is known of yeast biology and what occurs in the monolayers. Our point, is to say that our (quantitative) knowledge of single cell (obtained e.g. in batch culture) is not enough to infer quantitatively the landscape of gene expression observed. On the contrary, we USE the landscape of gene expression to infer the glucose gradients.

Discussion paragraph four: What is the dependence/independence of the different gene expression looked at here? Can we talk about synchronicity between networks if the studied genes are known to dependent on the same external factor?

Again, this is a difference of perspective. Given that the global microenvironment is not known, it is not obvious that the different genes that we studied display a high level of coordination in their spatial patterns. For example, you could imagine that the ethanol, pH, metabolite by products, or more simply the varying dilution rate of cells due to their variable growth rate, may affect the gene expression levels. We just show that this is the case in an experimental context that is relevant for multicellular systems.

Figure 6: Landscapes of gene expression is not shown at all glucose concentrations for all genes, why?

We did not show the landscapes that were flat or non-informative. The point was to show that the landscapes were spatially correlated and adopt a structure with two main regions displaying different metabolic activity.

Reviewer #3:

This manuscript details the results of a study using a microfluidic system to study growth and gene regulation in monolayers of yeast cells. This system is presented as a tractable model for studying pattern formation in microbial colonies, via spatial and temporal measurement of cell division rates and measurement of gene expression via fluorescently tagged reporters.

The authors use this experimental setup, in combination with GFP tagged version of hexose transporters, a glucose sensitive transcription factor (MIG1), and other reporters of metabolic state (PDC1, SHD2) to study spatial patterning as a function of glucose availability.

Unsurprisingly, they find that patterns of expression of hexose transporters with different affinities are spatially arranged in a manner consist with a gradient of glucose availability with highest amounts of glucose available near the "source" (i.e. periphery of the artificial colony), and lowest levels near the "sink" (i.e. center of the artificial colony). Perhaps more interestingly, their interpretation of their data suggest a relatively sharp transition in the colony between zones of growth and slow/no-growth, which they ascribe to zones of fermentive and aerobic metabolism.

We understand that the reviewer found our analysis “unsurprising”, but it was not our intention to show a surprising effect (which could have been seen in a different context, for example by changing the frequency and concentration of nutrients supply or type, by perturbing gene networks or by controlling one or more additional nutrients, to name a few. Some of which are currently a subject of investigation in the lab using this microfluidic system). Here our goal was to propose a quantitative description of these landscapes. This is not only about observing two populations. It is about observing the continuum of populations and the emergence of a spatially stable structure which results from growth landscape, metabolic landscape and gene expression landscape. Methodologically, we propose a novel method that can be used to study these landscapes dynamically, which no other systems proposed so far.

While I appreciate the elegance of the microfluidic colony model that is presented, the key findings are rather modest -- i.e. there are gradients of nutrient availability in microbial colonies. This doesn't really have the impact I expect from eLife papers.

This is not the key finding of the article. It was expected that we observed gradients, but we did not know their spatial extent and their stability and maintenance. Consider, that it is not possible (to date) to propose a model that would estimate this gradient for any metabolite. This is due to our limited quantitative data that describe multicellular systems, and our limited knowledge about multicellular effects (potential synergistic or antagonistic effects, interactions with other nutrients, effects of secreted metabolites etc…). Here, we do propose quantitative measurements of these landscapes, providing in return real data to fine tune computational models. We also are able to use this to infer the shape of the glucose gradients and estimate the number of cells that can grow in such conditions. Additionally, we show that the gradients are very different depending on the amino acid concentration (Figure 2—figure supplement 1), indicating that the global microenvironment, which is also composed of complex gradients needs to be taken into account to properly model the dynamics of cellular assemblies. It is one thing to assume that there are gradients. It is not the same to propose a robust experiment that allows to quantitatively study them. This is even more important if we want to get closer to the context that occurs in natural, non-laboratory, environments where cells often find themselves in everchanging environments composed of many interacting cells of same or different type.

Below I detail a number of concerns/comments about the study, the analyses, and whether the authors have adequate related their work to the larger literature in this area.

Detailed comments:

1) Several other recent papers (Hornung et al., 2018,; Wilmoth et al., 2018) have used microfluidic setups to study microbial colonies. The Hornung et al. paper is the most similar to the current experimental design.

This article specifically deals with “small” chambers. There are other articles that try to model colonies, the originality of our system is that we are able to let the cells create the gradients themselves through metabolic activity over almost 1000 fold concentration range of nutrients spanning from very high concentrations that are commonly used in laboratory and that appear in nature to complete lack of nutrients. And additionally, we can track on global and single cell level the activity of cells in space and time. It is thus different and applicable to a wider range of contexts and questions. In the revised manuscript, we changed the Introduction / Discussion to mention other articles that started to address the problem of spatial organization of multicellular assemblies in microfluidics even there are only a few to our knowledge.

2) If goal was to measure the rate of glucose uptake and availability it seems like there may be other more direct approaches than inferring this from hexose transporter expression, such as the use of 2-NBDG, a non-metabolizable, fluorescent glucose analog (Roy et al. 2015, doi: 10.1371/journal.pone.0121985).

We did consider the use of 2-NBDG, however it is not as straightforward as proposed by the reviewer. As a non-metabolizable glucose analog, it is likely to be progressively absorbed by cells until they cannot absorb it anymore, leaving it to freely diffuse further in the colony. Here, glucose (and other metabolizable nutrients) show a decrease of concentration precisely because they are used and converted in biomass. Using an analog to infer the glucose concentration would likely not work.

3) The following paper, while not at the fine spatial scale provided by the microfluidics, presents a more complete and compelling view of metabolic differentiation in yeast colonies: Maršíková et al., 2017.

The interesting differentiation patterns shown in this paper and in other papers published by the same group have one shortcoming, which is that they are observed in colonies grown on glycerol or glycerol/ethanol supplemented agar plates as the major carbon source. As they are not fermentable, it is likely that the insight in this major component of yeast metabolism is limited. This method also has limited spatiotemporal scale as noted by the reviewer and inability to dynamically change environments or provide constant environments. The advantage is that it is possible to perform analysis like RNAseq, which is nicely demonstrated in this paper. In our experiments we use media that has precisely quantified components (yeast nitrogen base, complete supplement mixture of amino acids and additional carbon source at precise concentration) as opposed to media based on yeast extract (basically yeast growing on yeast, which nota bene, would be an interesting possibility to study the effects of cannibalization of dead yeast in our microfluidic colony system). Also, they do not systematically quantify the landscape of gene expression, but they showed evidence that there are different metabolic areas in a colony. We did mention this study in our revised article.

4) The spatial analyses focus primarily on peaks of maximum expression, but the data is potentially much richer and more interesting if the authors consider not only the global maxima but more complex spatial patterns. For example, in Figure 4B MIG1 nuclear localization appears to be multimodal. What is going on here?

We do agree and this is the main point that we make by quantifying the full landscape of expression of many genes in similar conditions that emerges from the metabolic activity of cells. With that regards, the conclusions may not be surprising but they take place in a physiological setting. Regarding MIG1, we do not observe a multimodal localization, but a sharp transition from a localized to a cytoplasmic state. This is used as an indicator of a shift in glucose concentration.

5) There is very little effort to contextualize this work in the larger body of studies of yeast colony structure, physiology, or regulation of glucose responsive TFs. Illustrative of this, there are only three citations total in the entire discussion. There is only a single reference to the regulation of hexose transporters in the entire manuscript. Specific examples:

We agree that more contextualization was necessary so we expanded our Introduction and Discussion part, as well as provided additional citations, including all the ones proposed by the reviewer.

- Subsection “Cellular metabolic activity creates gene expression landscapes”: " HXT1 is a low-affinity glucose transporter mainly expressed under high-glucose conditions, while HXT7 is a high-affinity glucose transporter expressed under low-glucose conditions only" Citations?

See above

- Subsection “Gene expression landscapes depend on the glucose source concentration”: "As HXT1 is mainly expressed under high-glucose conditions (> 1% w/vol glucose) in batch culture…" Citation?

See above

- Subsection “Gene expression landscapes of other genes and transcription factor activity confirm the inferred glucose gradients”: "MIG1 is a key transcription factor involved in glucose repression that localizes to the nucleus in the presence of glucose, to repress genes that participate in parallel carbon metabolic pathways (e.g., galactose)." Citation?

See above

- and in the same subsection: "…we examined the expression of PDC1 and SDH2, which are overexpressed in fermenting and respiring cells, respectively" Citation?

See above

[Editors' note: the author responses to the re-review follow.]

We really appreciate the efforts of re-writing this manuscript as a "resource" paper. We feel that the paper has been much improved, but there are some remaining issues that need to be addressed through textual changes before acceptance, as outlined below:

1) We ask that the authors make it clear to the reader that what they call a "community action" does not imply that cells are responding differently in the community than they would as single cells in planktonic growth. The observed effects (gradients and response to these gradients) are merely the consequence of a high concentration of more or less static cells. So, perhaps calling it "population effect" would be more appropriate?

We did not intend to imply a “community” effect (with the meaning proposed by reviewer 2) and have exchanged all relevant instances of “community” with the word “population” as you suggested. Notably, the cells themselves are not static, i.e. they continuously grow and therefore passively pushed to be washed away from the channel. However, a steady state gradient is achieved so that cells, transiently occupying a geographic location along the gradient axes, exhibit a steady state response, mirroring the established stable gradient.

2) We would appreciate if you could insert a succinct discussion of the difference between a batch culture and this steady-state (where in principle, the cells could be better adapted as they are not always "running behind" in their response to a variable environment, as they would be in batch). (see reviewer 2's comments below for more details)

i) In standard batch culture, as cells exhaust the media, their adaptation time, limited by sensing, transcription and translation (on the order of 30 minutes; response time referred to by reviewer #2), may lag behind the decrease of glucose in the media. In contrast, in our 2D colony device, once steady-state gradient is settled, cells’ residence time within a given range of glucose that corresponds to stable Hxt expression is significantly longer, assuring that even though cells are continuously pushed away, they spend sufficient time within a given concentration to equilibrate their response to the gradient and therefore faithfully report the glucose concentration in their environment. As example, at 1% glucose the mean velocity at Hxt7 expression peak, spanning over >200μm (Figure 3A, varies between 5-15μm/hour (Figure 2F), suggesting cells’ residence time of >10 hours, pointing to the advantage of working in this setup.

ii) In our batch culture experiments (modified from a previously published protocol reported by Youk and van Oudenaarden, 2009), we intently used very diluted cultures to assure that within the experiment span cells both adapt to the nutrient condition and do not consume sufficiently to change significantly the glucose concentration. Thus, we deem the chemostat experiment is not useful.

We addressed point (i) by inserting the paragraph above to the Discussion section, and point (ii) in a succinct manner in the Materials and methods section of our revised manuscript.

3) We again ask to include more references and to not overstate the advantage of the presented device (see reviewer 3's comment below).

We apologize to the reviewer for the confusion about the references that we actually added in the revised manuscript but not pointing to the correct line numbers. The lines were referring to the lines in the original manuscript that the reviewer pointed out. In the current revised manuscript they can be found in subsections “Cellular metabolic activity creates gene expression landscapes” and “Gene expression landscapes of other genes and transcription factor activity confirm the inferred glucose gradients”.

Indeed, our method compliments other approaches cited in the manuscript and expands researchers’ arsenal for quantitative observation of 2D colony growth. We addressed this in the revised manuscript, toned down the advantages of our device and discussed its limitations in view of other published devices more clearly (see Introduction section).

4) Please provide more details about the reproducibility/variability and statistics of the results and analyses.

Each result is represented with standard deviation as error bars as descriptors of how disperse measurements are. When we mentioned the reproducibility of our experiments, it is based on the fact that error bars are small compared to the range of variation of the mean fluorescence peak location when the control parameter (here the external glucose concentration) is varied. Given the significant in-between concentration differences as compared to the error bars we see no added value in computing further statistics such as p-values. All figure captions contain the number of samples and the description of the error bars displayed. In addition, the complete dataset used in this study is available on an online Zenodo archive.

Author details

1. Zoran S Marinkovic

   1. Laboratoire Matière et Systèmes Complexes, UMR 7057 CNRS and Université de Paris, Paris, France
   2. U1001 INSERM, Paris, France
   3. CRI, Université de Paris, Paris, France

   Contribution

   Data curation, Investigation, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7720-2802

2. Clément Vulin

   1. Laboratoire Matière et Systèmes Complexes, UMR 7057 CNRS and Université de Paris, Paris, France
   2. Institute of Biogeochemistry and Pollutant Dynamics, ETH Zürich, Zürich, Switzerland
   3. Department of Environmental Microbiology, Eawag, Dübendorf, Switzerland

   Contribution

   Investigation, Writing—original draft

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3914-0708

3. Mislav Acman

   1. Laboratoire Matière et Systèmes Complexes, UMR 7057 CNRS and Université de Paris, Paris, France
   2. CRI, Université de Paris, Paris, France

   Contribution

   Visualization, Methodology

   Competing interests

   No competing interests declared

4. Xiaohu Song

   U1001 INSERM, Paris, France

   Contribution

   Software, Image analysis

   Competing interests

   No competing interests declared

5. Jean-Marc Di Meglio

   Laboratoire Matière et Systèmes Complexes, UMR 7057 CNRS and Université de Paris, Paris, France

   Contribution

   Supervision, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0446-4550

6. Ariel B Lindner

   1. U1001 INSERM, Paris, France
   2. CRI, Université de Paris, Paris, France

   Contribution

   Supervision, Writing—original draft, Writing—review and editing

   For correspondence

   ariel.lindner@inserm.fr

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-5015-6511

7. Pascal Hersen

   Laboratoire Matière et Systèmes Complexes, UMR 7057 CNRS and Université de Paris, Paris, France

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing

   For correspondence

   pascal.hersen@univ-paris-diderot.fr

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-2379-4280

• 2,942

  Page views

• 406

  Downloads

• 17

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.