Cell-cell metabolite exchange creates a pro-survival metabolic environment that extends lifespan

Metabolism is fundamentally intertwined with the ageing process. We here report that a key determinant of cellular lifespan is not only nutrient supply and intracellular metabolism, but also metabolite exchange interactions that occur between cells. Studying chronological ageing in yeast, we observed that metabolites exported by young, exponentially growing, cells are re- imported during the stationary phase when cells age chronologically, indicating the existence of cross-generational metabolic interactions. We then used self-establishing metabolically cooperating communities (SeMeCos) to boost cell-cell metabolic interactions and observed a significant lifespan extension. A search for the underlying mechanisms, coupling SeMeCos, metabolic profiling, proteomics and genome-scale metabolic modelling, attributed a specific role to methionine consumer cells. These cells were enriched over time, adopted glycolytic metabolism and increased export of protective metabolites. Glycerol, in particular, accumulated in the communal metabolic environment and extended the lifespan of all cells in the community in a paracrine fashion. Our results hence establish metabolite exchange interactions as a determinant of the ageing process and show that metabolically cooperating cells shape their metabolic environment to achieve lifespan extension.


Introduction
indicating the existence of cross-generational metabolic interactions. As this result implied that metabolite exchange interactions could be impact CLS, we boosted metabolite export and metabolic interactions through the use of self-establishing metabolically cooperating (SeMeCo) communities, a yeast community model that allows the tracing of metabolite consumer or producer cells of four different metabolic pathways in their sixteen possible combinations (metabotypes) 41 . We observed a significant extension of the yeast chronological lifespan when cell-cell metabolic interactions are boosted. In the search for the underlying mechanisms, we coupled lifespan assays with proteomics, metabolomics and genome-scale metabolic flux analysis, and discovered a role for the extracellular metabolic environment that is created by the cooperating communities. We find that cells cooperating for the biosynthesis of methionine generate a protective metabolic environment, in which methionine consumers obtain a more glycolytic metabolism and overflow glycolytic products, glycerol in particular. The exometabolome created this way, in turn, extends the lifespan of all cells in the community via a paracrine effect. Our results show that widespread metabolome changes, occurring when cells cooperate metabolically, create a pro-survival metabolic environment leading to extension of their own lifespan. Ultimately, these findings establish cell-cell metabolic interactions and generated exometabolomes as a longevity modulating mechanism.

Yeast cells establish cross-generational metabolite exchange interactions during chronologic ageing
As cells sense extracellular metabolites and feedback inhibit their own metabolite synthesis pathways when grown in rich media 25 common lab strain in which four artificially introduced metabolic biosynthetic deficiencies (auxotrophies) in three amino acids (his3Δ1, leu2Δ, met15Δ) and one nucleobase (ura3Δ) 42 were repaired through genomic integration of the wild-type alleles 38 .
The prototrophic cells, metabolically competent for the biosynthesis of the four metabolites (wild-type cells) were grown in batch culture through exponential, early stationary and stationary phases (1, 2 and 8 days of culture respectively) (Fig 1a). Initially, cells grow exponentially (E), consuming glucose supplemented to the culture media, followed by a decline in growth rate as cells transition from diauxic shift to stationary phase (early stationary phase, ES), and then enter stationary phase (S) once preferred carbon sources are exhausted. While exponential cells are mostly glycolytic, they then start consuming released products of glucose catabolism (like ethanol or glycerol) during the diauxic shift, before entering stationary phase, when cells arrest growth and mostly use oxidative phosphorylation to generate ATP (Fig. 1a). In order to evaluate the intracellular metabolome of chronologically ageing prototrophic cells, we used a targeted LC-MS/MS method 43 . The concentration profile of intracellular amino acids, nucleotides as well as glycolysis and tricarboxylic acid (TCA) intermediates was specific to the growth phase; the profiles clustered in a principal component analysis (PCA) according to growth phase (Fig. 1b   i)). The metabolite concentration changes measured reflected the known metabolic transitions from exponential to the stationary phase [44][45][46] . Consistent with a shift from fermentation to oxidative metabolism, the overall concentration of glycolytic metabolites decreased, while we detected an increase in the concentration of TCA derived metabolites (Fig. 1b ii)). Moreover, reflecting the ceasing of growth, the concentration of nucleotides decreased during early stationary and stationary phases 46 . Interestingly, a differentiated picture was obtained for intracellular amino acids. While the concentration of overall amino acids did increase in the stationary phase, we observed a spread in the concentration range (Fig. 1b ii)), unpaired two- decreased, most likely reflecting their role in interconversion reactions in the biosynthesis of other metabolites, including other amino acids and pyruvate (Fig. 1c), unpaired two-sided Wilcoxon Rank Sum test and multiple testing correction using the BH method, adjusted p-values in Supplementary File 2).
As amino acids can be exported by yeast cells into the surrounding environment 25,41,[47][48][49] , we therefore continued our analysis with a quantification of the extracellular amino acid pools using a targeted LC-MS/MS method 50 . Despite having inoculated our cells in a minimal medium lacking amino acid supplements, we found that by mid-log phase (exponential phase) yeast cells had produced and exported amino acids to reach significant concentrations in the medium.
Further, 14 of the 19 analysed amino acids are increased by more than >10% in the extracellular medium in the stationary compared to the exponential phase medium (Fig. 1c inlet). Indeed, only glutamate and aspartate were reduced in the extracellular metabolome of stationary cells when compared to the medium formed by exponential cells (Fig. 1c, unpaired two-sided Wilcoxon Rank sum test and multiple testing correction with BH method, adjusted pvalues in Supplementary File 2). The source of these metabolites can be metabolite export as well as cell death in the stationary phase. As most of the metabolites were already increasingly detected in the exponential and early stationary phases where cell death is negligible (~95% live cells) (Supplementary Fig. 2a) and also because in stationary cultures, the concentration of metabolites did not correlate with the number of live cells (Supplementary Fig. 2b), we concluded that the main source of metabolites is export during the exponential and early stationary phases.
Amino acids are sensed and efficiently uptaken by actively growing yeast cells 25,31,47 . We therefore asked if cells during the stationary phase, no longer actively proliferating, would uptake the amino acids previously produced in the exponential phase (Fig. 1d). We exploited 13 C-glucose isotope labelling to test for the consumption, by stationary cells, of metabolites that had been produced during the exponential phase. We cultured wild-type yeast cells on SM media supplemented with 12 C-glucose or 13 C-glucose for 48h, a duration which ensured that the glucose in the media had been exhausted -catabolized into unlabeled ( 12 C) and labelled ( 13 C) metabolites, respectively. Then we swapped the media between labelled and unlabelled cells. In parallel, we set control cultures growing on SM media supplemented with 13 C-glucose, which were then swapped into SM media only supplemented with amino acids (without glucose), to allow distinguishing if intracellular amino acid levels were a direct result of import, or indirectly derived from catabolism of imported carbohydrates. Levels of fully labelled ( 13 C) or unlabeled ( 12 C) intracellular amino acids (from glucose catabolism) were quantified using a targeted LC-MS/MS method 50 (Fig. 1e i)). Growth on SM media supplemented with 12 C-glucose or 13 Cglucose did not change cell growth parameters prior or post swap (Supplementary Fig. 3a- Notably, cells in exponential phase synthesise sufficient amounts of amino acids so that they can be exported and then uptaken by neighbouring cells, as shown by the increased intracellular levels of 13 C-or 12 C-containing amino acids in cultures initially grown on 12 C-or 13 Cglucose, respectively, or when unlabeled amino acids were added to cells previously cultured in 13 C-glucose in the control cultures ( Fig. 1e ii), Supplementary Fig. 3c). Hence amino acids that are produced and exported during the exponential growth phase are taken up by yeast cells during the stationary phase, indicating that yeast cells establish cross-generational metabolite exchange interactions during chronological ageing.

Metabolite exchange interactions extend lifespan in yeast communities
We next questioned what impact the exchange of metabolites might have on chronological lifespan. The export and import of metabolites cannot be prevented without imposing major metabolic constraints on cells. We overcame this issue by choosing to boost metabolite exchange interactions instead and made use of self-establishing metabolically cooperating communities (SeMeCos) 41 . SeMeCos exploit the segregation of plasmids that encode for 8 essential metabolic enzymes, to stochastically introduce auxotrophies (metabolic deficiencies), upon which cells can only continue proliferation by exchanging metabolites. Because plasmid segregation continues until a maximum amount of auxotrophic cells is reached, SeMeCos boost metabolite exchange interactions within the communities. Indeed, compared to wild-type cell communities, SeMeCos are characterised by increased metabolite export, an increase in extracellular metabolite concentrations, and increased metabolic interactions 38 . Despite boosting metabolite exchange interactions, SeMeCos still exploit the native metabolite export and import capacities of yeast cells and do not have artificially altered metabolic pathways or metabolite sensing properties 38,41,51 (Fig. 2a).
Analysing chronological lifespan of SeMeCos (Fig. 2b), we observed that in comparison to the isogenic wild-type strain, SeMeCos were long-lived, as assessed by monitoring colony forming units (CFUs) over time. SeMeCos lost more CFUs immediately after reaching the stationary phase, but in later time-points contained more CFUs and were alive after the wild-type cultures lost viability (Fig. 2c, unpaired two-sided t-test, p-value = 0.00661 at day 18 of culture; CLS pvalues listed in Supplementary File 4). To have an independent assessment of survival, we also monitored cell viability using Live/Dead TM cell staining assays. At late timepoints, SeMeCos also contained significantly more viable cells (Supplementary Fig. 4a, unpaired two-sided ttest; CLS p-values listed in Supplementary File 4). Finally, we exploited the situation where due to the higher cell density and proximity, metabolite exchange is amplified in colonies compared to liquid cultures. Yeast cells survived much longer in colonies than in liquid culture (~65 vs 20 days). Moreover, also in the colony, SeMeCos had a significantly longer CLS than the isogenic wild-type cells (Fig. 2c, unpaired two-sided t-test, p-value = 0.0338 at day 65 of growth, CLS p-values listed in Supplementary File 4). We ruled out that the difference in lifespan between SeMeCos and wild-type was explained by differences in pH, a common confounder of lifespan experiments 52 (Supplementary Fig. 4b, unpaired two-sided t-test, pvalues listed in Supplementary File 5). Moreover, our data suggests that SeMeCo cells were not long-lived due to amino acid starvation, a known lifespan extending intervention 53

The lifespan extension in SeMeCos is mediated by a paracrine mechanism
Next, we tested if the lifespan extension is associated with specific metabotypes, i. e. with To test the contribution of the individual auxotrophies to the lifespan extension, we next generated additional versions of the SeMeCo communities, in which each one of the auxotrophic markers (HIS3, LEU2, URA3 or MET15) was genomically repaired, and hence, only three plasmids segregate ('3p-SeMeCos') (Fig. 3b). The genomic repair of HIS3, LEU2 or URA3 did not significantly change the lifespan of SeMeCos, whereas the 3p-SeMeco in which the MET15 locus was no longer segregating had significantly shorter lifespan (Fig. 3c Methionine and other sulphur containing amino acids have repeatedly been linked to ageing, and typically it was a methionine restriction that caused a lifespan extension in model organisms [55][56][57][58][59][60] . Interestingly however, the high prevalence of met15Δ cells in SeMeCos and the high concentration of methionine produced by cells in the growth media did not suggest that an underlying methionine restriction would apply. To confirm that the fundamental mechanism of our observation is not methionine restriction, we conducted a control experiment, where we supplemented met15Δ cells with 2g/L of methionine. Despite the high methionine levels, we observed a robust lifespan extension in met15Δ cells (Supplementary Fig. 6, unpaired twosided Wilcoxon Rank Sum test, p-value at 28 days of culture = 3.27e-05; p-values across CLS are listed in Supplementary File 9). In search of an alternative explanation, we found evidence for a paracrine effect. We performed an independent CLS experiment, comparing SeMeCos and '3p-SeMeCos' unable to segregate the MET15 locus, confirmed the dependency of the lifespan extension on the organic sulphur cycle pathway (Supplementary Fig. 7

Lifespan extension in cooperating communities is mediated by an exometabolome rich in protective metabolites
To explore the cell-extrinsic factors that mediate the lifespan extension phenotype, we started by dissecting the metabolic changes emerging when cells metabolically interact. First, we simulated the likely flux changes using a community-adapted version of the flux balance analysis (FBA) that allows monitoring the exchange of metabolites between cells 38 . We . This result opened the possibility that it is not only methionine exchange itself, but the metabolic changes introduced by the metabolic cooperation that cause the lifespan extension. In order to get a deeper understanding of the pathways involved, we continued with proteomics analysis. We extracted proteins from the communities, generated tryptic peptides, and analysed them using microLC-SWATH-MS 61 and processed the data with DIA-NN 62 . We then performed differential protein expression analysis comparing otherwise identical SeMeCos that differ in the segregation of the MET15 marker (SeMeCos vs MET15-SeMeCos). We measured proteomes during exponential phase (day 1), early stationary (day 2) and stationary (day 8) growth phases (Fig. 4a). We consistently quantified 1951 proteins, around half the typically expressed yeast proteome 63  and met15Δ cells interact (Fig. 4d, Supplementary Fig. 13). Continuing with a pathway-centric analysis of the proteome did point our attention to glycolysis. Enzymes associated with the glycolysis pathway were generally upregulated in the communities that contained MET15 segregants (Fig. 4e). Moreover, an increase in the expression of glycolytic metabolites in stationary cells that typically rely on oxidative phosphorylation for energy production 44 was somewhat a surprise ( Fig. 5a-b). Indeed, both glycolytic activity and glycolytic overflow metabolites are associated with chronological ageing. While glucose restriction itself extends lifespan 65 , the glycolytic overflow metabolites ethanol and acetate both shorten lifespan 52,66 , but another glycolytic overflow metabolite, glycerol, increases CLS 67 . We speculated that a change in release of such metabolites might very well change the lifespan of cells that share a common metabolic environment. We therefore measured ethanol, acetate and glycerol in the exometabolome of the different SeMeCos and wild-type communities during CLS. In the stationary phase, levels of all three metabolites were higher in the communities where MET15 and met15Δ segregants interacted. Most striking changes were observed for glycerol, whose levels were ~8 fold increased, whilst ethanol and acetate levels were ~2 fold higher ( Fig. 5b i), unpaired two-sided t-test, p-values in Supplementary File 15). In order to explain the sources of the increase in glycerol, we studied the intracellular metabolome. SeMeCos revealed concentration changes in upper and lower glycolytic metabolites across all growth phases: the most significant changes were however detected in the glycerol-associated three carbon phosphates (G3P, DHAP, and PEP) in the stationary phase. These were increased specifically in the communities where MET15 and met15Δ cells interacted (Fig. 5b ii), unpaired two-sided ttest, p-values in Supplementary File 15). In parallel, we conducted oxygen consumption (OC) analysis. We found that the OC was reduced in the communities containing the MET15 segregants (Fig. 5c, unpaired two-sided Wilcoxon Rank Sum test, p-values in Supplementary File 16). The three results were all consistent with the accumulation of glycerol in the extracellular medium: glycolytic enzymes and glycerol precursors were up, while respiratory metabolism, required for the use of a non-fermentable carbon source as glycerol, was reduced.
To test whether an accumulation of glycerol could be associated with extending lifespan of cooperating communities, we performed a CLS assay where cells were grown in SM media supplemented with glycerol. Glycerol supplementation extended lifespan to 62 days of culture as compared to previously observed (Fig. 3c) <20 days in wild-type and MET15-SeMeCos. The SeMeCos also profited from the glycerol treatment, albeit the relative gain was lower than in wild type cells (mean fold change survival to wild-type in early stationary phase of 5.060%, 0.006% and 0.286% in SeMeCos, wild-type and MET15-SeMeCos, respectively, at 62 days of culture) (Fig. 5d, unpaired two-sided Wilcoxon Rank Sum test, p-values in Supplementary File

17).
While these results demonstrated that glycerol accumulation is beneficial, the glycerol increase alone might not sufficiently reflect the complexity of the yeast exometabolome. To validate if the community-created exometabolome is indeed mediating the lifespan extension, we hence complemented these results with a media swap experiment. We cultured wild-type communities in SM media until the early stationary phase (48h of culture) and then transferred them to a SeMeCo exometabolome (48h culture media generated in parallel). Control cultures were

Discussion
The classical view of the metabolic network is the one of a biochemical network operating inside the cell. However, with increased understanding of single cell properties, microbial landscapes and phenotypic heterogeneity this view is rapidly evolving 68 . The individual cell is increasingly seen to be part of a metabolic environment spanning across cells, and thus, metabolite exchange interactions between cells are an essential part of metabolism 32,69,70 . In microbes, metabolic networks span not only over single, but also over multiple species that interact within microbial communities 36,49,71-73 . The degree of metabolite exchange within those communities appears to be extensive. For instance, a majority of microbes are uncultivable outside their community environments, with metabolic co-dependencies being one of the key reasons 74 .
Another interesting observation is that from all 12,538 microbial communities sequenced as part of the Earth Microbiome project 75 only 6 contained no amino acid auxotrophs 38 . The interactions between amino auxotrophs and prototrophs is hence a common situation in microbial communities, which also shows that amino acids are effectively exchanged all the time. In ecology, metabolite exchange interactions can lead to competition or cooperativity 21,23 , but in any case, they have fundamental physiological implications. For instance, we have previously shown that cells that uptake lysine from the environment mount better protection against oxidants 37 , or that the presence of auxotrophs enriches metabolic environments and increases drug tolerance 38 . Despite being fundamentally important in modulating cellular processes that also impact on ageing -in particular to growth rate, metabolic signalling, and stress toleranceto our knowledge, the impact of metabolic intercellular interactions has barely been studied in the context of cellular ageing and lifespan. Indeed, studying the physiological impact of metabolite exchange interactions is technically challenging. Metabolite exchange interactions between cells are not captured by many typical single-cell techniques, such as microscopic imaging or single cells RNA sequencing, nor does the concentration of a metabolite explain whether it was produced or consumed by the analysed cell. Moreover, the export and import of metabolites can not be prevented without imposing major metabolic constraints. We herein In studying metabolite exchange interactions in the context of chronological lifespan in yeast, we made two observations that triggered our curiosity. The first was that metabolites exported during the exponential phase, label the cells during the stationary phase, reflecting their import by post-mitotic cells. In the batch-culture, there exists hence a 'cross-generational' exchange of metabolites. Ecologically speaking, the batch culture might have been seen as an artificial situation. However, in natural yeast colonies, old and young yeast cells co-exist in close physical distance 76 . That means that metabolite exchange interactions are even more likely in a natural colony than in batch culture. It is consistent with this notion, that a switch from liquid culture to colony growth tripled the chronological lifespan of our cells. To us, this result hence implies there could be extensive metabolic interactions that apply to both growth and ageing phases of the yeast cell communities.
The second observation was that upon boosting metabolic interactions by using the SeMeCo model, a significant extension of the CLS was achieved. Studying the metabotype composition within SeMeCos attributed a special role to the methionine biosynthetic pathway that is part of the organic sulphur cycle. We observed that cells that had segregated the MET15 plasmid comprised the highest proportion of long-lived cells in ageing communities. Sulphur amino acids include methionine, cysteine, homocysteine and taurine 77 and have previously been associated with lifespan extension. Methionine restriction in particular can extend lifespan in a number of organisms [55][56][57][58][59][60] , prevent the development of a variety of diseases 78 and influence response to anti-cancer therapies 79,80 . Our results differed, however, from many of these studies in a fundamental aspect, as they did not indicate the lifespan extension is caused by methionine restriction. We hence speculated that another mechanism could be at play in the communal cells. The key observation which eventually led to a better understanding of the mechanism, was that the presence of the MET15 segregants did not only increase their own lifespan, but also the lifespan of the other cells found within SeMeCos. This result strongly suggested that only part of the answer is to be found in the intracellular metabolic reconfiguration in the MET15 segregants, and that we need to search for a paracrine effect, like a change in the extracellular metabolite pool, to understand the lifespan extension of the entire community.
In order to identify the metabolic changes, we combined metabolite profiling, proteomics and genome-scale metabolic modelling. We detected widespread metabolic changes in the communities containing MET15 segregants, but were most intrigued by an upregulation of glycolytic enzymes, in a growth phase where typically oxidative metabolism dominates. In following this, we confirmed a decrease in oxygen consumption and an 8-fold increase in the glycolytic overflow metabolite glycerol. Glycerol is known as a protective and pro-survival metabolite 67 , and also in our hands, significantly extends the lifespan of both wild-type and SeMeCo communities. Glycerol stimulates several survival-associated processes, including osmoregulation, lipid biogenesis, cell wall integrity 81  In summary, we uncover a protective metabolic paracrine effect occurring in metabolically interacting eukaryotic microbial communities. Glycolytic methionine consumer cells enrich the intercellular space for the pro-survival metabolite glycerol, increasing the survival of their producer counterparts and overall community longevity. Impairment or inability to metabolically interact drives cellular dysfunction, which accompanies ageing and disease, therefore dissecting the metabolic dynamics and emerging metabolic environment when cells metabolically interact will aid the development of therapies targeting these processes. Often lifespan extension is associated with restriction conditions, but our data shows that a differentiated view is also necessary, as simple nutritional interventions like the exchange of amino acids can have broad changes in the metabolic network dynamics, reflected in the exometabolome, and alter lifespan this way. Future investigations are necessary to determine how broadly this situation impacts on other nutritional and/or metabolic contexts influencing lifespan.

SeMeCo generation and culture
The generation and culture of SeMeCos was performed as previously described 38 . The pH, pL, pU and pM plasmid used to generate a SeMeCo strain in the BY4741 background are described in Table S3. All SeMeCo strains were cultured in minimal synthetic (SM) media, composed of yeast nitrogen broth without amino acids (YNB, 6.  Table S1.
For CLS assays where cells were grown on glycerol, SM was supplemented with 3% Glycerol grown on solid SM supplemented with glucose prior to being cultured in SM supplemented with glycerol from pre-culture stage onwards.

Knockout strains culture
Knockout strains cultures followed the exact same procedure as described for SeMeCo generation and culture. In the case of metabolic knockout mutants (met15∆), cells were grown on SM media supplemented with the metabolite for which the strain was biosynthetically impaired (2g/L L-methionine), with respective wild-type controls also being cultured in SM supplemented with the metabolite. Strain details are in Table S1.

Isotope tracing
Wild-type yeast cells were cultured in SM media supplemented either with 12 C-glucose ( 12 C-glu; Sigma #G8270) or 13 C-glucose ( 13 C-glu; Sigma #389374), during 48 hours, then media was swapped for tracing amino acid export/import, using targeted metabolomics 50  were determined from growth curves using the R 'grothcurver' package 87 .

Conventional and High-throughput Colony Forming Unit (CFU) assays
Conventional CFU analysis was performed as described previously 88

Oxygen consumption measurements
Ten mL of CLS cultures were collected during exponential (day 1) and early stationary phase

Sample preparation
Ageing cultures, at several time points reflecting different growth phases, were sampled and 400 uL of each culture were quenched in 1600 uL dry-ice-cold methanol, into a 48-deep-well plate. This suspension was spun down (600 g, 3 min, 4°C), and the supernatant was discarded by inversion, followed by a short spin (600 g, 1 min, 4°C) to ensure complete removal of the SN.
Cell pellets were immediately placed on dry ice and then transferred to −80 °C until analysis.
Intracellular metabolites were then extracted as described 91 . Briefly, 140 μl of 10:4 MeOH/water were added and vortexed. Then, 50 μl chloroform was added, followed by 50 μl water and 50 μl chloroform with thorough mixing in between. Phases were separated by centrifugation at 3,000 g for 10 min. The aqueous phase was recovered and used without further conditioning. One microlitre was injected for HPLC-MS/MS analysis. Before analysis by HPLC-MS/MS the order of samples was randomised and during analysis a quality control sample (QC) was assessed every 24 samples.

Sample acquisition
Metabolites were resolved on an Agilent 1290 liquid chromatography system by HILIC coupled to an Agilent 6470 triple quadrupole instrument operating in dynamic MRM mode, as previously described 43 . In short, the gradient program started at 30% B (100 mM ammonium carbonate) and was kept constant for 3 min before a steady increase to 60% B over 4 min. Solvent B was maintained at 60% for 1 min before returning to initial conditions. The column was washed and Extracellular amino acids and uracil data from wild-type in exponential phase are a re-analysis of data in 38 ; experiments, including cell culture, metabolite extraction and sample acquisition, were performed in parallel.

Sample preparation
Frozen SN in 96 deep-well plates (collected as described above for amino acid and uracil analysis) were defrosted and kept shaking using plate shaker for 20 minutes 900 rpm room temperature, just before the filtration, using a multiscreen filtered plate with 0.45 um durapore (https://phenomenex.blob.core.windows.net/documents/863d86a0-3aba-4591-979b-bf54b1188038.pdf) and a Welch vacuum pump.

Sample acquisition
The target compounds were quantified using a Shimadzou Prominance HPLC

Sample preparation
Ageing cultures, at several time points reflecting different growth phases, were sampled and 500 uL of each culture were collected into a 96-deep-well plate. Samples were centrifuged at 4,000 g for 3 min and supernatants (SN) were discarded. Samples were centrifuged again at 4,000 g for 1 min to fully remove any residual SN. Cell pellets were immediately placed on dry ice before being stored at -80ºC, until all samples were collected. Sample preparation for proteomics was performed as previously described 61

Sample acquisition
The digested peptides were analysed on a nanoAcquity (Waters) (running as 5 µl min −1 microflow liquid chromatography) coupled to a TripleTOF 6600 (SCIEX). Protein digest (2 µg Post-processing data analysis was conducted in R 92 .

Constructing auxotroph-prototroph community metabolic models
The community metabolic models were reconstructed using the approach from our previous study 38 . Briefly, the genome-scale metabolic model of S. cerevisiae 93,94 was used to create auxotrophic strain models by switching off respective metabolic reactions. Then the reactions from auxotroph (H, L, U and/or M) and prototroph (WT) models were combined to make the community, using the compartment per guild approach, where both strains were treated as separate compartments and metabolic exchange between strains were allowed. The community biomass was the combined biomass of all strains. The Cobra toolbox 95 was used to perform the model simulations.

Data analysis and statistics
All statistical analyses were done in R (R Core Team, 2015) 92 using specific packages as indicated throughout the methods section. For the basic data manipulation and visualisation we used the R tidyverse package compilation and for statistical analysis we used the R ggpubr package. Hypothesis testing to assess means of population differences were mainly done using   100 .

Data availability
The data supporting the findings of this study are available within the paper, its Supplementary Information and will be deposited within publicly accessible repositories (before formal acceptance). The proteomic datasets generated during the current study that are relevant to data shown in Fig. 4 and Supplementary Fig 10-

Acknowledgements
We