Self-establishing communities enable cooperative metabolite exchange in a eukaryote

Essential revisions:The referees liked the new approach using plasmid loss to generate a synthetic yeast community exchanging nutrients, but questioned how the work may provide information on yeast physiology. This could be achieved, for example, using the Sigma1278b strain strategy mentioned by Reviewer #2.

We agree with referee 2 that BY4741, as a S288c derivative strain, has a long laboratory history that possesses (like other lab strains do as well) some metabolic defects. We therefore agree it can be interpreted as a subject of concern that all results shown were obtained in this background. However, only a single gene mutation (FLO1) prevents the assembly of biofilm-like structures in S228c (flocculation), so metabolically, the strain is completely capable of forming these types of communities (Smukalla et al., 2008). Although Sigma1278b is capable of biofilm formation in a growth-media dependent manner, it remains a very close relative to S288c, with a largely identical genome, similar metabolic capacities, and a long lab history on its own (http://wiki.yeastgenome.org/index.php/History_of_Sigma). We therefore chose to replicate the key experiments in a totally different yeast species, as these experiments not only exclude that the findings are specific to BY4741, but also indicate evolutionary conservation. In addition, we studied a flocculating derivative of BY4741, to exclude that ability of physical attachment influences the experiments outcome.

We chose S. pombe as the ideal candidate, being ~400 Myr evolutionary apart from S. cerevisiae, and additionally, as S. pombe offers a comparable set of laboratory techniques. We performed co-culture experiments using representative S. pombe auxotrophs and find that they, like S. cerevisiae cells, are also not able to form a co-growing community upon co-culturing (Figure 1B).

Next, we also replicated a segregation experiment in the same S. pombe strain as Figure 1B. Also in S. pombe, segregation allowed the cells to establish a cooperating community, as in S. cerevisiae (Figure 2—figure supplement 2).

In addition, to addressing whether this phenotype is the impacted by the biofilm forming capacities referred by Reviewer #2, we performed the co-culture experiments also in FLO⁺S. cerevisiae strain. The ability of FLO+ cells to physical attach to one another did not influence the outcome of the experiments. Complementary co-cultures of FLO+ auxotrophs did not co-grow in the absence of supplementation (Figure 1—figure supplement 1).

The most important implication for yeast physiology is that, our results being correct, cells within colonies continuously and extensively exchange anabolic metabolites among each other and do not, as the current assumption says, produce metabolites predominantly for themselves. We draw these conclusions from (i) finding metabolites to be highly concentrated between cells, (ii) demonstrating that cells take these up preferentially when they are available and (iii), showing that cells have a highly robust capacity for growing on the basis of this metabolite exchange.

We have extended the discussion about the following implications: that knowledge on ongoing metabolite exchange is essential for understanding the physiology of cells, as it shows that in a community, metabolic networks operate non-autonomously. For instance, this knowledge is essential to interpret the recent results by Shou et al and Murray et al, that mix yeast cells under the assumption that only feedback-modified cells exchange intermediate metabolites, whereas non-modified prototrophs do not (Momeni et al., 2013a, 2013b; Müller et al., 2014; Shou et al., 2007). These results are also essential for interpreting and improving results from metabolic modelling approaches such as flux balance analysis, that intuitively assume cell-autonomous functionality of intermediary metabolism and, at present, do not include transport reactions or assume regular metabolite exchange between cells. For biotechnology, the potential is huge as well, as SeMeCos offer a simple system for establishing cooperative communities; these could be very efficient for the production of cell-toxic substances, as the burden can be shared between cells.

Please also see the individual response to Reviewer #2 that is interested in the yeast ecological implications of the metabolite exchange.

There were also concerns about the lack of cell biological analyses and I would encourage you to address these satisfactorily with experiments and discussion.

We have replicated a series of experiments in this respect as requested by the reviewers. We have introduced plasmids containing fluorescent proteins coupled to auxotrophic markers and did reconstruct SeMeCo and visualise structural organisation of a micro-colony via confocal fluorescence microscopy. These experiments add appreciable visual evidence that metabolic heterogeneity in a SeMeCo colony is substantial (Figure 4A).

We next performed advanced image analyses from these microscopy images and show that the distance a metabolite needs to cross to reach a consuming auxotroph from a supplying prototroph is minimal (less than two yeast cell diameters) (Figure 4B). This shows that the community establishes a structure due to cooperation.

Please pay close attention to the point raised by Reviewer #3 on the exo versus endometabolome differences in the context of the colony and address it satisfactorily with experiments and discussion.

We have conducted several experimental controls to address the comments of Reviewer #3. First, we have re-isolated different auxotrophic cells from an established SeMeCo to test whether they, after metabolic reset by growing in supplemented media, behave like the original strains and fail to establish cooperation upon co-culturing. As illustrated, they behave like the original strains and fail to grow together upon co-culturing. This excludes that feedback regulatory systems could have acquired mutations while SeMeCos established, or being affected by other genetic alterations (Figure 2—figure supplement 3).

As another requested control by the reviewer, we co-cultured the auxotrophic strains in the composition as quantified in SeMeCo, and in a 1:1:1:1 ratio. As expected, a similar result as in the pairwise co-cultures is obtained. The result is also robust when co-culturing overnight in rich media before spotting to minimal media (Figure 2—figure supplement 4). We would like to note that these controls were, in another place, already part of the first submission, and apologize that they were not highlighted as such.

We also tested whether cell death in SeMeCo is different to that in wild-type cells. Both wild-type cells and SeMeCos did produce the same number of CFU's on complete media. Moreover, we counted the cell numbers with a CASY cell counter before spotting, and compared it with the number of colonies obtained. The numbers are close to a 1:1 ratio, confirming that cell death in exponentially growing wild-type and SeMeCo cells is marginal, and that virtually every cell can give rise to a new colony (Figure 3D). We would like to note that the Palková experiments that have been referred to (Váchová et al., 2012), are conducted in old, stationary colonies that establish over several days to a week, where cell death is expected to commence in the regions of the colony where there is no cell growth and the population ageing. In difference, all our experiments are conducted in continuously growing colonies and cultures, were old cells are constantly diluted and never reach a significant percentage of the community.

We also revised our point about the exo vs endometabolome and apologize that it created an unintentional confusion. It is not a problem that a low number of dying cells may contribute to the exometabolome – even in exponential cell growth a (very) low number of cell death might occur and this is a physiological situation in a yeast colony and cannot be prevented. There would also be no way to distinguish this by mass spectrometry or another analytical technique (physically it is the same metabolite, differential labelling is no option as it is not predictive which cell will die first), subsequently, we can only work indirectly, therefore we only use growth conditions where cell death is negligibly low.

Why is this point important at all? We have to exclude that nutrient leakage from dying cells is the main explanation for auxotrophic cell growth in our system. Nutrient recycling from dying cells can play a significant role in chronological ageing experiments, where cells are kept for days or weeks in stationary growth phase and cell death becomes substantial (i.e., studied by Palková and Longo labs). The phenotype, 'adaptive re-growth', can subsequently arise in these stationary ageing yeast cultures as a result of cell lysis. In stationary bacterial cultures, a similar phenotype also occurs, known as 'growth advantage in stationary phase' (GASP). In these stationary cultures, cell growth that is explained through recycling from dying cells, but the novo biomass is not formed. This is the clear difference between these scenarios and our experimental set-up: In our case we can say for certain that de novo synthesis explains cell growth, as we work under exponential growth conditions, where biomass and with it, the total amount of histidine, leucine, methionine, and uracil, doubles with every cell division. Recycling of nutrients from dying cells could – at best – maintain biomass, but never double it exponentially; so on minimal media, all new gained biomass is for certain explained by de novo synthesis. Together with detecting no relevant amount of cell death as expected for an exponential culture Figure 8, we conclude that nutrient release through cell death is overall not relevant for the growth of SeMeCos.

Reviewer #1:This is in interesting manuscript from Ralser and colleagues using a new approach to generating complex yeast community that exchange metabolites, and such exchange is essential for colony viability. The authors show that they can generate synthetic colonies that contain auxotrophs for 4 different amino acids/nucleobases. At one level this is exciting, but at another level, I am unclear what we have learned about yeast physiology from this work.Major concerns:1) The plasmid loss based approach to generating the SeMeCo colony is a clever trick, but how does it generate the SeMeCo. Delving into the mechanism of how gradual loss of plasmids in a mixed population (as opposed to mixing various auxotrophs) leads to SeMeCo will be a great strength.

We agree with the line of thinking of the reviewer, but believe the main question the reviewer asks is what prevents the formation of co- growth in the co-culture experiments. The growth of SeMeCo instead, as addressed in this paper, is clearly enabled though metabolite exchange of histidine, leucine, uracil, and methionine, as these are the only limiting metabolites for the auxotrophs to grow. We suspect that feedback control mechanisms might be responsible; but we have no clear idea of what activates them specifically upon co-culturing. We find this problem as interesting as the reviewer, however, its answer will become a complicated study, and will have to be addressed on its own in the future. We are currently applying for funding to extend this work.

2) The paper lacks a cell biological analysis of the phenomenon. For example, Muller, Murray use a fluorescence based assay to look at the position of the two genotypes in the colony. Admittedly, the situation here is more complex with a larger number of genotypes, but is still not insurmountable and some cell biological insight into colony organization will help

We have addressed this experimentally, please see Essential Revision Point #2 above. These new experiments form a new figure (Figure 4). We had originally restrained from the use of fluorescent markers as they might have introduced an additional cost though their synthesis in the establishment of a SeMeCo. However, what we can tell so far yeast seem to tolerate expression of the different GFP derivatives without changing the cell growth properties notably.

3) It is also not strictly correct to say that exchange of metabolites/cooperation does not happen in yeast, since the papers by Shou and Murray labs (cited in this paper as well) do generate a synthetic cross-feeding colony, except that these authors had to down-regulate feedback mechanisms involved in amino acid metabolism. What the current paper has done is to increase the complexity and this has been achieved using a different clever approach. This needs to be better articulated.

We apologize if our quoting of their work was misleading. We have carefully revised the paper.

Reviewer #2:

Major concerns:

1) I think that the main story of the paper is a bit underdeveloped still, and that the second part of the paper (where the authors measure stress resistanc, etc.) is perhaps less essential and a bit unconnected from the main story. To me, the most important question is how important and realistic the findings are for natural conditions. This particular experiment and the establishment of SeMeCos represents a scenario that is not very realistic in nature, where all the genes are maintained more or less stably in the genome instead of residing on unstable plasmids. The authors find that auxotrophic mutants that are mixed together do not seem able to develop some sort of mutualistic community. Moreover, in natural conditions, exchange of metabolites between yeast cells may not be very relevant, since if it were, one would expect that feral yeast colonies or communities would contain a very high rate of (different) auxotrophs? This also makes me wonder whether SeMeCo's develop mostly because of the cost of maintaining the plasmids, rather than a benefit of losing specific metabolic pathways to specialize in others?

The reviewer addresses the ecological implications of our study. In response to the general comment, we would like to bring up that there are two points here, and we think both are important for the natural live of S. cerevisiae. The questions a) is whether native yeast cells would exchange intermediary metabolites and possess the capacity to do that as described in our work, and b) whether this exchange would result in a situation, were losing biosynthetic genes provide an additional benefit to that, by forming communities were co-growth of different yeast genotypes becomes obligate.

To point a, our results give clear indication that the maintenance of an amino acid rich exometabolome establishes within yeast colonies as a normal property (Figure 1A) and the cellular uptake of these metabolites to support growth is a native yeast property as well (Figure 1D): The experiments were conducted with prototrophic yeast, grown on minimal media, were no pressure was given to maintain the exometabolome or to uptake the metabolites. Also we see throughout the paper that SeMeCos maintain growth and many physiological parameters that mimic that of colonies composed of genetically prototrophic cells.

The experiments in Figure 1 demonstrate that prototrophic yeast cells can flexibly switch between uptake and self-synthesis of histidine, leucine, uracil and methionine. The critical question for point B), is hence if cells would gain additional advantage by losing this flexibility, and enter a situation were co-growth of different genotypes becomes an obligate situation. One needs to see the answer to this question in the context of the natural life cycle of yeast. This involves a sexual cycle and the formation of endospora, important to endure long periods of starvation, and to spread between habitats. If a yeast cell would lose a prototrophic genotype, a single spora cannot any longer establish a colony on its own. Loss of genetic prototrophy would thus negatively affect the natural life cycle of yeast, with deleterious consequences to its survival, spreading between environments, and evolvability (interruption of genetic recombination, etc.). So even if exchanging metabolites is advantageous when cell grow in a community, there are good reasons for keeping the genes in the genome that allow metabolic flexibility when required and for completing the sexual cycle. The existence and activity of metabolite exchange strategies coupled with the ability to turn biosynthetic pathways on and off, is thus fully compatible with maintaining a prototrophic genome. Perhaps, higher organisms are a good example for that. As humans we obtain all amino acids through our food, yet we maintain biosynthetic pathways for eleven of them, so that we can flexibly activate these pathways when our diet or age requires that. For yeast, the majority (but not all) natural isolates keep genetic prototrophy, both in S. pombe and in S. cerevisiae (Jeffares et al., 2015; Liti et al., 2009). We have expanded the discussion about this point.

Last but not least, I also wonder whether the lab yeast strain BY4741, a derivative of S288C, which is known to be impaired in several mechanisms related to community formation (biofilms, pseudohyphal growth, mats…), is a wise choice for these experiments.

Some additional experiments may tell us much more about the real-life importance of metabolite exchange in yeasts. Firstly, I would argue to repeat the key experiments in a strain that is capable of forming communities, like for example Sigma1278b. At first sight, this might seem like a prohibitive amount of work, but it is not since it simply entails using some of the available auxotrophs (or generating a few new ones through transformation) to see if they can support each other in liquid suspension, colonies or biofilms/mats. If so, this would mute a lot of doubts (including whether the cost of maintaining a plasmid is a player). Second, I would also suggest transforming Sigma with (at least two of) the plasmids to investigate if and how this strain realizes SeMeCo's similar to S288c.

We have followed the suggestion of the reviewer, and have replicated the key experiments in (a) S. pombe, that is evolutionary distant to S. cerevisiae, and (b) in a FLO⁺S228c S. cerevisiae strain, to exclude that the findings are only relevant for S288c. Please see Essential Revision Point #1 above.

If the authors decide to follow this suggestion, I would also argue for the inclusion of constitutively expressed fluorescent markers, so that they can easily count the different lineages through cytometry and also follow their spatial distributions in colonies/biofilms (see second remark below).

We have now included experiments with fluorescent markers. Please see Essential Revision Point #2.

Last but not least, I think that the authors need to state more clearly what they think is the relevance of their findings in natural conditions – they seem to avoid this question a bit instead of being very explicit about it and discussing it in great detail. One very interesting bit of information would be if and how many auxotrophs are recovered from nature? Is there any data available? Or can the authors explore this themselves?

The answer to this has been included with point #1 above about the ecology.2) One of the most intriguing questions is whether the different auxotropic mutants arrange into some kind of spatial pattern in the colonies, so that complimentary strains (that can compensate each other's auxotrophies) would be in close proximity. This should be relatively easy to study by integrating a fluorescent marker in each of the plasmids (e.g. three dyes with a very different spectrum, like CFP, YFP and RFP).

See Essential Revision Point #2. These experiments have now been included. Complementary auxotrophs stay in close proximity. This does not, however, prevent macroscopic differences to establish.

3) It is unclear to me whether it is sufficient to keep cells in exponential phase to exclude the influence of dying cells (which could contribute nutrients, rather than through exchange between living cells). The fact that the "exometabolome" differs in composition from that inside of living cells may not be the ultimate proof, as it seems possible that dead cells do not only contribute metabolites that were free molecules when the cell was still alive, but also through (active or passive) breakdown of biomolecules after the cells disintegrate. Perhaps one way to tackle this is to let cultures grow well into stationary phase and/or add a certain number of dead cells and monitor if and how this changes the dynamics. Another way would be to investigate colonies, as it is known that cells in the center of colonies experience starvation and cell death (see for example work by Palkova and colleagues).

Please see Essential Revision Point #3, where these points have been addressed.4) The SeMeCo's seem to show convergence towards a stable state (with a defined and complex ratio of different auxotrophic mutants). Are we sure that this is indeed a stable state? And why do not all prototrophs disappear? Would it be worth letting some communities grow much longer to see what happens in the long term?

We have conducted a stability experiment for a SeMeCo community over a period of ~100 generations. As expected for a living ecological system, it is not 100% identical over this long period, but the variation is in the range not larger ~ 10% and restabilizes (Author response image 1).

Author response image 1

Download asset Open asset