Design of a new model yeast consortium for ecological studies of enological fermentation

Wine fermentation involves complex microbial communities of non-Saccharomyces yeast species besides the well-known Saccharomyces cerevisiae. While extensive research has enhanced our understanding of S. cerevisiae, the development of multi-species fermentation starters has led to increased interest in yeast interactions and the role of microbial diversity in winemaking. Consequently, molecular methods have emerged to identify the different species at different stages of the winemaking process. Model microbial communities or consortia, which provide simplified systems resembling natural microbial diversity, offer opportunities to investigate population dynamics and understand the role of community diversity in ecosystem performance. Here, this work aims to design a yeast consortium reflecting the diversity of wine yeasts and to develop a method for accurately tracking their population dynamics during fermentation. We developed and characterized a six-species consortium, with S. cerevisiae, Hanseniaspora uvarum, Starmerella bacillaris, Metschnikowia pulcherrima, Lachancea thermotolerans and Torulaspora delbrueckii. By tagging each yeast species with distinct fluorescent markers, the study enables real-time monitoring of individual species within the consortium using flow cytometry. We have carried out a complete analysis of this consortium, studying the evolution of populations over time and examining factors such as metabolite production and fermentation kinetics. In addition, the yeast consortium was used to test the diversity-function relationship as a proof of concept. We sought to determine the impact of the initial evenness on communities’ performances subjected to osmotic stress. To this end, ten randomly designed consortia with varying initial species proportions were followed in enological fermentation with 200 and 280 g/L of initial sugars. Although no demonstrable effect of the initial evenness was shown, the usefulness of this consortium has been demonstrated. Moreover, the initial proportion of certain species on population dynamics and metabolite production were observed. The consortium presented in this work and available to the scientific community can contribute to future work trying to decipher multispecies dynamics and exploring the role of yeast diversity in wine fermentation.


Introduction
Fermented foods and beverages are consumed worldwide, and their production relies on diversified microbial communities that include numerous bacteria and yeasts (Furukawa et al., 2013;Wolfe & Dutton, 2015;Tamang et al., 2016).Wine alcoholic fermentation involves different yeast species from the Saccharomyces, Hanseniaspora, Pichia, Metschnikowia, Starmerella, Torulaspora, Lachancea, or Rhodothorula genera (Barata et al., 2012;Drumonde-Neves et al., 2021).Research has mainly been focused on optimizing the process with Saccharomyces cerevisiae, as it is the main species carrying out the alcoholic fermentation.Thus extensive knowledge has been provided on S. cerevisiae growth, and its influence on wine aroma as well as the genetic basis of its suitability for fermentation (Lambrechts & Pretorius, 2000;Sablayrolles, 2009;Peter et al., 2018).However, the development of multispecies starters, with one or two non-Saccharomyces strains alongside S. cerevisiae (Roudil et al., 2020), has spurred research on interactions between yeast species and on the role of wine microbial diversity in the process (Ivey et al., 2013;Ciani & Comitini, 2015;Bordet et al., 2020).
Consequently, an increasing number of studies on wine microbial ecology, ranging from investigation on the factors influencing yeast and bacterial diversity in grape must, such as climate and vine management (Bokulich et al., 2014;Grangeteau et al., 2017;Bagheri et al., 2018), to the mechanisms of interactions involved between the different species (Bordet et al., 2020).Studies on population dynamics have highlighted mechanisms of competition for nutrient, secretion of toxic compounds or involvement of cellular contact (Holm Hansen et al., 2001;Medina et al., 2012;Kemsawasd et al., 2015).Molecular mechanisms and phenotypic changes caused by interactions have also been studied through transcriptomic analysis or description of physical interactions of cell wall proteins (Brückner et al., 2020;Mencher et al., 2021;Conacher et al., 2022).In addition, interactions and population dynamics of yeasts can be modulated by environmental conditions, such as the different stress factors they encounter during winemaking.Indeed, across the alcoholic fermentation yeasts are exposed to harsh conditions, encompassing low pH, low oxygen availability, or high sugar content (Heard & Fleet, 1988;Bauer & Pretorius, 2000;Varela et al., 2021).The varied environmental conditions and complex microbiota involved in fermentation makes it difficult to draw conclusion.
Model microbial communities allow working on simplified systems that mimic the diversity found in natural communities (Roy et al., 2014;Blasche et al., 2017).These consortia include species found in natural communities, and are used in controlled environments to foster knowledge on ecological drivers of population dynamics as well as decipher the role of community diversity in its functioning (Wolfe et al., 2014;Venturelli et al., 2018;Piccardi et al., 2019;Niccum et al., 2020).Several methods can be used to evaluate the evolution of population in such consortia, such as differential culture media, quantitative PCR or metabarcoding.Cytometry has recently proven useful to monitor population dynamics in multispecies communities (Rigottier-Gois et al., 2003;Kylilis et al., 2018) and model wine yeast pairwise cocultures and consortia (Longin et al., 2017;Conacher et al., 2020).
The present study aimed to develop a novel six-species consortium representative of wine yeast diversity as well as the method to discriminate its different species population.This consortium consisted of one strain of six different species, modified with a specific fluorescent tag, at different initial abundance.We also verified our ability to accurately measure species proportions in complex mixtures with mock communities and finally used this novel model consortium to test the diversity-function relationship as a proof of concept.

Yeast genetic modification
Genes coding for fluorescent proteins were integrated into the genome in fusion of TDH3 gene (or its orthologue in non-Saccharomyces).Cassettes used for transformation were amplified from plasmid matrices listed in Supplementary Table 1 with a high fidelity enzyme, either the KAPA HiFi kit (Cape Town, South Africa) or Phusion High-Fidelity DNA polymerase (Thermo Fisher Scientific, Vilnius, Lithuania), using primers specific to each species, as listed in Supplementary Table 3. PCR amplification products were purified with the PCR clean-up Kit (Machery Nagel, Düren, Germany, REF 740609).In addition, S. cerevisiae MTF4798 was tagged with a second fluorescent protein, in fusion of ENO2, another gene involved in glycolysis.TDH3 has a strong promoter and is expressed continuously during fermentation in S. cerevisiae.Similarly, ENO2 is also expressed throughout S. cerevisiae fermentation, even though to a lesser extent (Puig & Pérez-Ortín, 2000;Peng et al., 2015).Yeast transformation was carried out by electroporation, as described by Pourcelot et al. (2023).Transformants were selected on YPD agar supplemented with the appropriate antibiotic: 800 µg/mL of hygromycin B for S. bacillaris, 100 µg/mL of nourseothricin for S. cerevisiae, T. delbrueckii, and L. thermotolerans, and 200 µg/mL of geneticin for H. uvarum and S. cerevisiae.Integration at the locus was verified with two independent PCR (Supplementary Table 4).To check that transformation did not modify cells behavior, growth of WT strains and transformants were compared by microplate assay (Supplementary Figure 1).

Synthetic must
Synthetic grape must was prepared according to Bely et al., (1990) with 425 mg/L of yeast assimilable nitrogen (as a mixture of amino-acids and ammonium), supplemented with 5 mg/L of phytosterols (Betasitosterol, Sigma, St-Louis, US, MO) and a final pH of 3.3.Total sugar concentration was either 200 g/L (S200, with 100 g/L of glucose and 100 g/L of fructose), or 280 g/L (S280, with 140 g/L of glucose and 140 g/L of fructose).Synthetic must was pasteurized at 100°C for 15 min before use.

Cell inoculation
Cell cultures were prepared with a first propagation in 5 mL YPD for 18 hours at 28°C, and a second propagation of 1 mL in 50 mL of synthetic must S200 for 24 hours at 28°C with agitation.Cell concentration of each pre-culture was measured by flow cytometry before inoculation.Appropriate cell culture volume was used to start fermentation with a final cell concentration of 10 6 cells/mL.In the case of consortia, the volume of different species cultures were pooled to obtain the desired initial abundance (Table 2).Cell suspensions were centrifuged at 4415 g for 5 min and washed in physiological water (9 g/L of sodium chloride, Sigma), and suspended in 5 mL of synthetic must for inoculation (either S200 or S280).

Scale of fermentation
Fermentations were carried out either in 1 L or 250 mL scale fermentation.One liter fermentation were run in 1.2 L fermenters (200mL headspace) with automated weighing every 20 min to evaluate CO 2 weight loss.Fermenters filled with 1 L of synthetic grape must were aerated for 40 min before inoculation.An internally developed software was used to measure in real time the rate of CO 2 production (g/L/h) and fermentations were stopped when it was lower than < 0.02 g/L/h, or after 312 hours if fermentation had not finished yet.At 0,4,6,12,24,48,72,168 hours of fermentation and when fermentation stopped or after 312 hours, 6 mL were sampled for population and metabolite analysis.
Two-hundred and fifty mL fermentations were run in 300 mL fermenters with manual weighing twice to thrice a day.Fermenters were aerated for 20 min before inoculation.At the start and end of fermentation, 5 mL were sampled for both population and metabolite analysis.At 21, 45, 120, 168, 210, 290 h, 1.5 mL were sampled for population analysis only.

Inoculation and consortia composition
Precultures of each strain were done by inoculating 5 mL of YPD with colony grown on a YPD agar plate and incubated for 18 hours at 28°C.One milliliter of the preculture was then propagated in 50 mL in synthetic must S200 in 125 mL Erlenmeyer flasks and incubated for 24 hours at 24 °C with shaking (250 rpm).Volume of cell culture of each species was calculated after measuring cell concentration of each strain culture by flow cytometry.Consortia were made by mixing the adequate volume of each species in individual 50 mL Falcon tubes for each fermenter.The cell mixes were centrifuged for 5 min at 4415 g and the cell pellet was rinsed in physiological water once (NaCL at 9 g/L) then thoroughly ressuspended in 5 mL of synthetic must (S200 or S280 accordingly).Fermenters were inoculated at a total 10 6 cell/mL rate with this final cell suspension.
Composition of the different consortia used in this study can be found in Table 2.The 6-species consortia (Co) had the following initial proportions of the six species were: 35% H. uvarum (3.5 × 10 5 cells/mL), 25% M. pulcherrima (2.5 × 10 5 cells/mL), 20% S. bacillaris (2 × 10 5 cells/mL), 10% S. cerevisiae (1 × 10 5 cells/mL), 5% T. delbrueckii (0.5 × 10 5 cells/mL), 5% L. thermotolerans (0.5 × 10 5 cells/mL).The consortia used for testing the diversity-functionnality relationship included eleven consortia with the five fluorescent species (noted C02 to C12).These consortia had varying initial relative abundance of the five different species (except for S. cerevisiae which had an initial abundance of 5% in C02 to C11, and 15 % in C12) encompassing different levels of evenness (Table 2).Diversity was evaluated with the Shannon index with the following formula: െ ∑ ‫‬ ln ሺ‫‬ ሻ where ‫‬ is the relative abundance of species ݅, and ݊ the number of species.C12 composition is adapted from the 6-species consortium but without M. pulcherrima.Two consortia consisting of equal initial abundance for all species, with S. cerevisiae (C07) or without (C13), as well as a control monoculture with S. cerevisiae alone (noted C01) were also included.Cell number of each species was monitored by flow cytometry with the Attune NxT™ Thermofisher® Flow Cytometer (Life Technologies, Singapore).Each population tagged with one or two fluorescent protein was detected with a specific set of channels described in Table 3.To validate the method, accuracy of the numbering of the 6 species was made with mock communities.The mock communities were constructed by incubating single species cultures overnight in synthetic must at 28°C with agitation.Subsequently, one mL of culture was centrifuged at 4415 g for 5 min and the resulting cell pellet was resuspended in 1 mL of PBS (130BmM NaCl, 2.6BmM KCl, 7BmM Na 2 HPO 4 , 1.2BmM KH 2 PO 4 , pH 7.4; Sigma).Cell suspensions at a concentration of 10 6 cells/mL were prepared in PBS and combined to create mock communities in 96-well microplates (200 µL of final volume).In total, 30 communities were tested, each included one species with theoretical abundance of either 1, 5, 10, 50 or 90% while the five other species were in equal proportions.Cell numbering was performed both immediately after preparation of the mock communities in the microplate and after 2 hours at room temperature.
Table 3:  At the different sampling timepoints, 200 µL of sample were ressuspended in 200 µL of PBS after centrifugation in 96-wells microplate (V-bottom) at 4415g for 5 min.Samples were diluted in PBS before reading so that the cell concentration was approximately 1 to 5.10 5 cells/mL.Total cell count was measured with the FSC and SSC channels (488 nm laser, 488/10 filter).To restore the fluorescence, cells suspended in PBS were kept at room temperature for 2 hours.This incubation step likely enables the maturation of fluorescent proteins which is oxygen-dependent (Hansen et al., 2001;Zimmer, 2002), while oxygen is rapidly depleted during fermentation (Bardi et al., 1999;Moenne et al., 2014).Moreover, we observed a decrease in fluorescence of eGFP under the control of the ENO2 promoter in S. cerevisiae near the end of fermentation, reducing the accuracy of measurement after 144 hours of fermentation (up to 5% of inaccurate events at 144 hours).This was corrected with an incubation step in YPD for 1 hour (Breslow et al., 2008) followed with an incubation step of 1h in PBS.These steps did not impair total cell count or viability.Viability was assessed after staining cells with 1 µg/mL propidium iodide (PI, stored at 4°C protected from light; Calbiochem).Cell number and percentage were obtained from gatings done on the Attune NxT software.

Metabolites analysis
For the metabolite analysis, 5 mL of samples were centrifuged 5 min at 4415 g at 4°C to get rid of cells.Supernatant aliquots were kept at -18°C before being diluted with 2.5 mM sulfuric acid (H 2 SO 4 , Merck, Darmstadt, Germany) at 1:6, and centrifuged again at 16 000 g for 5 min.Diluted samples were kept at -18°C until analysis.Extracellular metabolite and sugar concentrations (glucose, fructose, ethanol, glycerol, acetic acid, lactic acid, pyruvic acid, and α-ketoglutaric acid) were determined by highperformance liquid chromatography (HPLC), following the method described in Deroite et al., (2018).Analyses were run on the HPLC (HPLC 1260 Infinity, Agilent Technologies, Santa Clara, California, USA) using a Rezex ROA ion exclusion column (Phenomenex, Torrance, CA, USA) at 60°C, with a flow rate of 0.6 mL/min of 2.5 mM H 2 SO 4 .Concentrations of acetic acid, lactic acid and α-ketoglutaric acid have been measured with a UVmeter at 210 nm and other compounds with a refractive index detector and chromatograms were analyzed on OPEN LAB 2X software.Fermentation samples were analyzed with technical duplicates and data analysis were carried out on the mean of both replicates.

Data analysis
Mock communities were analyzed by comparing theoretical cell concentration to the observed cell concentration, as measured by cytometry.
All fermentations were performed in biological triplicates, on three different runs.Fermentations of the six monocultures and the 6-species model consortium were compared based on 10 fermentation parameters.They included four CO 2 kinetics parameters -latency (time needed to reach 5 g/L of total produced CO 2 ), time of fermentation (tF, time when CO 2 production rate < 0.02 g/L/h), maximum CO 2 production rate (Vmax) -and six metabolic parameters, namely the yields of ethanol, glycerol, and acetic, α-ketoglutaric, lactic and pyruvic acids.Metabolite yields were calculated with the formula: Fermentations of the 5-species consortia were compared with the six metabolites yieds and CO 2 yield.Differences in yields between the S200 and S280 conditions were visualized by calculating the difference between both conditions with the following formula: (2) ߂ ௌଶ଼,ௌଶ ൌ ೄమఴబ ି ೄమబబ ೄమబబ Data were analyzed with R studio software (version 4.3.1, Core Team, 2023).Mock communities' data were analyzed with the 'ggpmisc' package (Aphalo, 2023).ANOVA were performed to compare the different monocultures and 6-species consortium parameters, followed by a post hoc Tukey test when significant using the `rstatix` package (Kassambara, 2023).For the 5-species consortia, comparison between both sugar conditions were done with t-test for consortia C01 to C13, on the following parameters: residual sugars, yield in CO 2 , and end yields in ethanol, glycerol, and acetic, lactic, αketoglutaric and pyruvic acids

Construction of consortium and cytometry numbering optimization
The model community that we used in this study was constituted of 6 different species at different initial abundance: Saccharomyces cerevisiae (10% of total population), Hanseniaspora uvarum (35%), Metschnikowia pulcherrima (25%), Starmerella bacillaris (20%), Lachancea thermotolerans (5%), and Torulaspora delbrueckii (5%).The species and their proportions were determined based on a literature analysis of 18 articles, and the species initial abundance was determined from the average of the initial abundance found in a subselection of 11 articles (Combina et al., 2005;Hierro et al., 2006;Di Maro et al., 2007;Zott et al., 2008;Ocón et al., 2010;Maturano et al., 2015;Ghosh et al., 2015;Bagheri et al., 2017b;Simonin et al., 2018;Mateus et al., 2020;González-Alonso et al., 2021).These articles needed to include at least one sampling without inoculation and without enrichment steps, and the identification had to be at the species level.They encompassed different world regions and winemaking practices (grape variety, conventional and organic, etc).Despite the clear influence of these factors on yeast diversity, we found little relationship between the regions and practices with the presence of specific species, which is in line with the conclusions of a recent meta-analysis (Drumonde-Neves et al., 2021).Among the most frequent species, we decided to select only one species per genera with the assumption that it might increase phenotypic diversity and the overall usefulness of our model consortium.Thus, the six following species were selected: S. cerevisiae, H. uvarum, M. pulcherrima, S. bacillaris, T. delbrueckii and L. thermotolerans (Supplementary Figure 2A).Then, the initial abundance for each species was determined from their natural initial abundance in must and early stages of fermentation (Supplementary Figure 2B).
The consortium was designed to track individually six species (H.uvarum, L. thermotolerans, M. pulcherrima, S. bacillaris, S. cerevisiae, Torulaspora delbrueckii) during fermentation.To this end, our strategy was to label the different species with different fluorescent proteins.The fluorescent proteins were selected based on their excitation and emission spectra to minimize overlaps.Genes of the fluorescent proteins were successfully integrated by homologous recombination in fusion of the TDH3 gene into the genomes of H. uvarum (mTagBFP2), S. bacillaris (eGFP), L. thermotolerans (mCherry) and T. delbrueckii (mCitrine).S. cerevisiae was tagged with mCherry fused with TDH3 and eGFP with ENO2 (Table 1).For M. pulcherrima, the WT strain was used since we did not manage to integrate genes at the locus even when using long homology sequences, which is in accordance with previous studies (Gordon et al., 2019;Moreno-Beltrán et al., 2021).As shown on Figure 1, which displays derived dot plots and gates used in this study, the different transformants could be well discriminated.L. thermotolerans and S. cerevisiae are distinguished by the mCherry fluorescence and S. cerevisiae by the additional GFP fluorescence.T. delbrueckii (eGFP-tagged) cells could be separated from S. bacillaris (mCitrine-tagged) using the violet laser, not needing any specific optical configuration, as reported for YFP (Yellow Fluorescent Protein) (Marcus & Raulet, 2013).Initially, the T-Sapphire fluorescent protein was also selected, but its signal overlapped with the mTagBFP2 signal and was not used further.In addition, accurate numbering of the different yeast populations was verified by comparing measured to theoretical cell concentration in mock communities (Figure 2).Mock communities are constructed with known composition and used to validate microbial quantification methods, with metagenomics or cytometry, since they (Tourlousse et al., 2021;van de Velde et al., 2022;Rué et al., 2023).In this study, 6-species yeast mock communities were constructed from overnight cultures in S200 and included one species with theoretical abundance of either 1, 5, 10, 50 or 90% while the five other species were in equal proportions.In total, 30 communities were tested in microplates, with a final cell concentration of 10 6 cells/mL, and the concentration of each species was measured by flow cytometry.When species concentrations were numbered immediately after the mock assembly, the observed and theoretical cell concentrations were well correlated (R 2 ranging between 0.8 and 0.99 , Figure 2A).However, during fermentation, a loss of cell fluorescence was observed after 24 hours, especially for the mCherry-tagged cells.To restore fluorescence, samples needed to stay at least 2 hours in PBS at room temperature.This step was also tested on the mock community, and except for M. pulcherrima that was under-estimated after the 2 hours (R 2 Mp = 0.73, Figure 2B), the other species were still accurately numbered (R 2 ≥ 0.90).Monocultures and the 6-species model consortium (Co) were characterized with fermentation in synthetic grape must (S200) in 1 L-scale fermentation.All fermenters were inoculated with a total cell concentration of 10 6 cells/mL and initial abundance of the six species in the 6-species consortium (Co) were 35% H. uvarum, 25% M. pulcherrima, 20% S. bacillaris, 10% S. cerevisiae, 5% T. delbrueckii, 5% L. thermotolerans.The dynamic of species abundance in the consortium are shown in Figure 3 and population dynamics of monocultures in Supplementary Figure 3.In the consortium, the early fermentation phase was characterized by a rapid growth of H. uvarum that reached 65% of the total population at 12 h, with a maximum population of 6.7 log of cells/mL (Figure 3, Supplementary Table 5).On the contrary, S. bacillaris and M. pulcherrima abundance dropped under 5% after 12 hours, even though they were inoculated at 20 and 25%, respectively.L. thermotolerans, showed a little growth at 24h, reaching 6.3 log of cells/mL, (Supplementary Table 5), but quickly decreased after 48 hours.After 24 h, a drop of viability was observed (live cells < 75%) associated with the drop in abundance of NS species, while S. cerevisiae became dominant in the consortium and represented 41% of the total population and 58% of the live population (Figure 3).Maximum population (8 log) in the consortium was reached after 48 h, with almost only S. cerevisiae remaining alive(Figure 3).

Fermentation of the 6-species consortium and monocultures
Fermentation capacity of the 6-species consortium and its member species alone was evaluated through CO 2 production (Figure 4).The consortium performance was similar to that of S. cerevisiae alone, with a slightly longer latency and fermentation time (latency = 25.3±1.9h and tF = 158.5±6.1 h compared to 18.7±0.7 and 120.6±3.9 h respectively.The maximum CO 2 rate (Vmax) of consortium Co was significantly different from any monocultures.It reached 1.74 ± 0.0591 g/L/h (Figure 4B), which was between the maximum CO 2 rate of S. cerevisiae (2.16 ± 0.02 g/L/h), T. delbrueckii (1.11 ± 0.11 g/L/h) and L. thermotolerans (1.01 ± 0.06 g/L/h) alone.The consortium reached the same total CO 2 production as S. cerevisiae (93.08 ± 1.06 g), while M. pulcherrima, H. uvarum, and S. bacillaris alone showed the lowest CO 2 production (Figure 4B).MaxCO2 = maximum CO 2 produced (in g/L); tF = fermentation time (time to reach CO 2 production rate < 0.02 g/L/h), Vmax = maximum rate of CO 2 production, and Latency = time necessary for total CO 2 production to 5 g/L.ANOVA results are indicated by the p-value.Statistical groups determined with post-hoc Tukey tests are indicated with lowercase letters.Co: Consortium, Hu: H. uvarum, Lt: L. thermotolerans, Mp: M. pulcherrima, Sb: S bacillaris, Sc: S. cerevisiae, Td: T. delbrueckii.

Impact of consortium's diversity on osmotic stress response
The aforementionned consortium was developped as a tool to further study ecological drivers of the wine microbial community.As a first application, we focused on determining how the initial abundance of species would affect the fermentation process, especially when submitted to an osmotic stress.Twelve yeast consortia (numbered C02 to C13) with varying evenness of five species (excluding M. pulcherrima due to its underestimation), alongside a S. cerevisiae monoculture (C01) were tested.Fermentations were conducted in 250 mL-scale of synthetic grape must with two sugar concentration: 200 g/L (S200) and 280 g/L (S280).Furthermore the consortium C13 did not contain Saccharomyces cerevisiae.
In all consortia initially containing S. cerevisiae, this species largely dominated after 120 hours in both sugar conditions and constituted the only live cells, except in consortium C03 (Figure 6, Supplementary Table 6).Unlike the others, in the C03 consortium there was 17% remaining of T. delbrueckii at 120h in condition S280, which led to a delayed dominance of S. cerevisiae that reached over 95% of total cells only after 168 hours instead of 120 hours in the S200 condition.S. cerevisiae monoculture was not much affected by the different sugar conditions and reached its maximum population of 8.12 log at 21 hours (Supplementary Table 6).On the other hand, observations revealed variations in the dynamics of non-Saccharomyces populations depending on the initial sugar concentration (Figure 6) with significant differences in population abundance between S200 and S280 found at 45 hours.In most consortia, H. uvarum reached a similar maximum proportion at 21 h for both sugar conditions but decreased more sharply at 45h in condition S280 (Figure 6).Similarly, L. thermotolerans tended to decrease more in condition S280, except in consortium C13 (without S. cerevisiae), while T. delbrueckii tended to maintain a higher proportion in condition S280 as compared to S200 (Figure 6).At 45 hours in consortium C03, the decrease in abundance between S200 and S280 of L. thermotolerans partially matched the increase in abundance of T. delbrueckii (6% less L. thermotolerans and 15% more T. delbrueckii in condition S280).The C03 consortium also showed a longer persistence of T. delbrueckii with S280 compared to the other consortia.Interestingly, in the S. cerevisiae-free consortium (C13), while both L. thermotolerans and T. delbrueckii were present at the end of fermentation, L. thermotolerans dominated the consortium after 21 hours of fermentation and tended to maintain higher abundance in S280.Regarding S. bacillaris, its population decreased very quickly and its proportion was less than 5% after 21h in all consortia, for both sugar concentrations.Fermentation performance was also assessed, through metabolite and CO 2 yields.The differences in yield of the main metabolite from carbon metabolism between S280 and S200 are shown monocultures in Figure 7. Positive values indicate higher yield in condition S280, while negative values indicate lower yield in condition S280.
Higher initial sugar concentrations led S. cerevisiae to produce more acetic acid (increase by 75 % of yield in acetic acid) coupled to a lower production in α-ketoglutaric acid (decrease by 60 % of yield in αketoglutaric acid, orange diamond on Figure 7).
In addition, we also observed a lower yield in ethanol and lactic acid for S. cerevisiae monoculture with high initial sugars, although very small for ethanol (decrease of 1.6 % for the ethanol yield and 25% for the lactic acid yield).We did not observe differences for yield of pyruvic acid for both initial sugar conditions in the S. cerevisiae monoculture.We next set out to compare the response to osmotic stress in the consortia.The differences in metabolite yield between the two sugar conditions are shown in Figure 7.The different consortia exhibited a wide range of responses, especially regarding the production of acetic and pyruvic acids.Contrary to S. cerevisiae monoculture, consortia tended to have lower pyruvic acid yield in condition S280, but differences were significant only in consortia C07, C11 and C13.Regarding acetic acid, its overproduction in condition S280 was greatly reduced in all consortia compared to S. cerevisiae monoculture since the highest increase was of 31 % in consortium C02.In the consortium inoculated with a higher proportion of S. cerevisiae (C12), 55 % more acetic acid was produced in condition S280 than at S200.In addition, differences in yield of acetic acid in consortia were only significant for consortia C02, C11 and C12.On the contrary, the yield in α-ketoglutaric acid remained lower in condition S280 compared to S200.Differences ranged from 53% lower to 68% lower in consortia C11 and C03 respectively.Only in consortium C02 was it not significantly different between both sugar conditions.This could suggest a prevailing importance of S. cerevisiae concerning this metabolite since the difference between both S280 and S200 was often close to that observed in S. cerevisiae monocultures, as indicated by the orange diamonds (Figure 7).Regarding ethanol, consortia showed an overall lower yield with S200 compared to S. cerevisiae monocultures (between 0.482 ± 0.001 in C05 and 0.494 ± 0.003 in C10.Surprisingly, even though glycerol is an essential osmolyte involved in osmotic stress resistance, our results did not show significant differences in glycerol yields between both sugar conditions (Figure 7).One possible explanation for this observation may be that the sugar difference between both conditions was not enough to cause noticeable change in glycerol yield.
Effect of initial composition on consortia behavior.Consortia C02 to C11 were designed to represent variable α diversity, through increasing evenness of the initial relative abundance of non-Saccharomyces species (Table 2).Hence, the Shannon index of the different consortia ranged from 0.895 in C02 to 1.449 in C11, except C07 that presents an equal proportion of each NS species, and a 1,515 Shannon index.These 10 consortia also included the same initial abundance of S. cerevisiae (5%), to avoid changes being related only to S. cerevisiae.We were interested in knowing whether an increasing diversity in the yeast consortia would help maintain the fermentation performance during the osmotic stress.However, we found no demonstrable effect of the initial diversity on the different parameters related to fermentation performance, such as yield in CO 2 , yield in metabolites or residual sugars (Supplementary Figure 5).However, the initial composition still influenced the behavior of the consortia, especially for the difference in acetic and α-ketoglutaric yields between both sugar conditions (Supplementary Figure 6).

Discussion
In the present study, we constructed a new model community of yeast species that can be numbered by flow cytometry thanks to the expression of different fluorescent protein.To validate the accuracy of this counting method, we employed mock communities.Then, we characterized the performances of the 6-species community by assessing various fermentation parameters.Finally, to further prove its usefulness, twelve consortia were tested in enological conditions with two differents sugar concentrations to evaluate the influence of varying initial eveness.
In the first part of this work, we tested whether the strategy of species quantification during fermentation was applicable and accurate.Five out of six species were successfully tagged with a set of four fluorescent proteins that can readily be discriminated with a flow cytometry equiped with the appropriate optical configuration.We then assembled 30 mock communities including S. cerevisiae, H. uvarum, M. pulcherrima, S. bacillaris, L. thermotolerans, and T. delbrueckii with abundance ranging from 1 to 90 %.In the range of concentrations tested here, our results indicate that the concentration of cells measured by cytometry highly correlates with the theroretical value).Hence, the quantification strategy could be validated on a consortium including 5 to 6 species.It should still be noted that the accuracy in numbering of M. pulcherrima decreased with longer handling time, which could lead to its understimation and should be taken into account when including this species.
During fermentation, species succession of the 6-species community described in this study reflected trends found in both natural communities and laboratory cocultures.For example, as observed in our data, several studies have reported a fast multiplication of H. uvarum in early fermentation, including studies in cocultures and consortia (four species) grown in synthetic grape must (Lleixà et al., 2016;Harlé et al., 2020), but also at industrial scale non-inoculated fermentations (Combina et al., 2005).However, in industrial conditions, this increase of H. uvarum can be limited when treating must with SO 2 , which is a chemical commonly used as an antimicrobial and antioxidant agent, promoting S. cerevisiae (Grangeteau et al., 2017).After 24 hours, M. pulcherrima and S. bacillaris were detected under 5 % in our consortium.This early decline of M. pulcherrima in fermentation has been observed previously (Combina et al., 2005;Chasseriaud et al., 2018), whereas several studies have found S. bacillaris able to multiply and maintain a relatively high population in cocultures.For instance, in semi-industrial scale fermentations, S. bacillaris could be found in mid fermentation.However, this was related to a later onset of S. cerevisiae (Ocón et al., 2010).For other model consortia containing 8 species, tested in both synthetic and natural grape must, S. bacillaris remained present up to the mid or end of fermentation (Bagheri et al., 2017).Discrepancies on the presence of S. bacillaris during fermentation might be due to different media composition, as was demonstrated with different initial sugar and nitrogen content (Lleixà et al., 2016).
Regarding L. thermotolerans and T. delbrueckii, both were detectable up to 72 hours in our consortium despite their low inoculation ratio, showing a rather good persistence.In a four species consortium, these two species were able to reach a sizable population despite the presence of S. cerevisiae.However, their inoculation ratio was much higher (Conacher et al., 2020).Similarly, L. thermotolerans was also found at the end of fermentation of Chenin Blanc grape must, in a consortium of eight species (Bagheri et al., 2017).In natural communities, their maintenance is harder to evaluate since they are present in only low abundance.However, they have been found mid-fermentations (Sternes et al., 2017) and at the end of fermentation in some occasions (Díaz et al., 2013;Simonin et al., 2018).
As for S. cerevisiae, it was the main viable species after 24 hours, and likely caused the early drop in viability observed between 12 and 24 hours.After 72 hours, it was the only viable species present in the consortium.These results would be somewhat consistent with previous works on cocultures that found that non-Saccharomyces, such as L. thermotolerans, T. delbrueckii or M. pulcherrima, populations tended to decrease once S. cerevisiae reached a high cell number or a high cell number of live S. cerevisiae was added (Nissen et al., 2003;Kapsopoulou et al., 2005;Comitini et al., 2011;Taillandier et al., 2014).This phenomenon is also evidenced in sequential inoculation when non-Saccharomyces stop growing or even die after the addition of S. cerevisiae (Benito et al., 2015;Binati et al., 2020;Seguinot et al., 2020).
In addition, our data presented species-related features, such as high glycerol and α-ketoglutaric acid production for M. pulcherrima or S. bacillaris monocultures.These characteristics have previously been described in monocultures in synthetic must (Englezos et al., 2018;Mbuyane et al., 2022) and in mixed fermentaions (Comitini et al., 2011;Englezos et al., 2017;Binati et al., 2020;Seguinot et al., 2020).Similarly, the great production of lactic acid observed here in L. thermotolerans monocultures is already known, even though it varies greatly between strains and conditions (Benito et al., 2015;Vaquero et al., 2020).Concerning the 6-species community, its metabolite production and fermentation kinetics seemed mostly influenced by S. cerevisiae but was intermediate with non-Saccharomyces, which is commonly observed in cocultures (Renault et al., 2013;Bagheri et al., 2017;Harlé et al., 2020).Our second experiment also highlighted the influence of the different species on the fermentationa performance in consortia leading to variation in the response to osmotic stress more or less influenced by the presence of S. cerevisiae.
Taken together, our results confirm the applicability of the combined use of fluorescently tagged cells and cytometry to follow several species individually in oenological fermentations.Indeed, this method has been successfully applied in mixed fermentations of one non-Saccharomyces and S. cerevisiae (Petitgonnet et al., 2019;Harlé et al., 2020), but rarely included more species (Conacher et al., 2020).Other studies including more species in their consortium employ cultivation or molecular biology methods to follow the microbial dynamics (Lleixà et al., 2016;Bagheri et al., 2017;Chasseriaud et al., 2018).These methods allow enumeration in natural community without specific tagging of the different species but involve heavier experiments and a longer time between sampling and results.
As a proof-of-concept, we employed the yeast community with only the 5 fluorescently tagged species to investigate the effect of different initial eveness of consortia on their response to two sugar concentrations (S200 and S280).Altogether, our results indicated a tendency of T. delbrueckii and S. cerevisiae to be promoted in a complex community during a fermentation with high sugar content, while H. uvarum and L. thermotolerans were negatively impacted.T. delbrueckii is described as relatively osmotolerant, to both salts and reducing sugars (Lages et al., 1999;Hernandez-Lopez et al., 2003), and has been found able to maintain a high population in natural grape musts with 250 to 360 g/L of sugars (Bely et al., 2008;Belda et al., 2015).Thus, the fact that it tended to be promoted at S280 in our study was to be expected.On the other hand, we found that S. bacillaris abundance decreased very quickly even in the S280 condition.This is somewhat surprising given this species is osmotolerant and could be expected to be promoted in high sugar must (Tofalo et al., 2012;Csoma et al., 2020).Its osmotolerant nature means it is often isolated in fermentations of high sugar must, such as icewine or botrytised wine that can usually contain at least 350 g/L of reducing sugar (Magyar & Soós, 2016).It can even dominate early and middle stages of such fermentations (Nisiotou et al., 2007;Magyar & Soós, 2016).Our inconsistent results might be due to the use of synthetic must with high nitrogen content that could limit the establishment of S. bacillaris in a community, in favor of faster growing species.Fewer data are available for H. uvarum or L. thermotolerans in conditions of high-sugar fermentation.In botrytised must, H. uvarum was shown to decrease more quickly than in healthy must, but other factors than the initial osmotic stress, such as presence of other species, may also explain these results since this observation stayed true even when there was no difference in initial sugar (Nisiotou et al., 2007;Contreras et al., 2015).Another interesting highlight from these data are the differences of population dynamics in the different consortia, which shows that initial abundance of the different species can influence their growth under different conditions.This could be explained by differences in response of the different species to the osmotic stress in the first hours, since different coping mechanisms seem to occur in T. delbrueckii and S. cerevisiae for example (Renault et al., 2009;Tondini et al., 2020).
In the osmotic stress condition, the most striking difference between consortia and S. cerevisiae monocultures was the overproduction acetic acid.S. cerevisiae overproduced acetic acid in the S280 condition, while the acetic acid overproduction was smaller in consortia and varied significantly between consortia.This may be partly due to the presence of non-Saccharomyces since several studies have reported a positive effect of mixed fermentation of S. cerevisiae with non-Saccharomyces, namely S. bacillaris and M. pulcherrima, in reducing the acetate yield (Rantsiou et al., 2012;Contreras et al., 2015;González-Royo et al., 2015;Mbuyane et al., 2018).Another explaination might also lie in the population changes between both sugar conditions.With S280, H. uvarum decreased more quickly while T. delbrueckii stayed longer.T. delbrueckii is known to produce low amount of acetic acid (Bely et al., 2008;Renault et al., 2009), while H. uvarum is associated with high production in acetic acid (Ciani & Maccarelli, 1998;Capece et al., 2005;Mendoza et al., 2019).Their difference in dynamics may therefore influence the overall response.However, the consortia and range of initial abundance tested here are too limited to allow us to properly disentangle the combined effect of each species on the production of metabolites in consortia in both sugar conditions.
Despite the effet of composition of the communities on microbial dynamics and metabolite production, we found no correlation to the initial eveness.The relatively low complexity of our environment may not be sufficient to generate a set of ecological sub-niches that are all the more covered as microbial diversity increases (Hooper et al., 2005;Shade et al., 2012).For example, Schmidt et al., (2020) observed differences in S. cerevisiae strains dynamics in natural must with 200 or 280 g/L of sugars, but no difference in synthetic must.In addition, the range of diversity tested in this study might be limited since it varied only based on initial abundance while the number of species (and therefore species richness) was constant.For example, Ruiz et al., (2023) showed a negative effect of an increasing richness in wine yeast communities.They observed that S. cerevisiae could not finish fermentation in some communities due to the higher probability of the presence of antagonistic species.Diversity effect has also been observed in microcosms, where fermentation performance was reduced in the highest and lowest diversity context, due to a lower establishment of S. cerevisiae (Boynton & Greig, 2016).Thus, further investigation on microbial diversity effect in winemaking might need an increased number of species.Model communities could also include bacterial species to follow both the alcooholic and malolactic fermentations.

Conclusion
In this study, we successfully developed a 6-species consortium of fluorescently tagged wine yeasts that allow accurate real time tracking of species subpopulations during fermentation.The consortium was also fully characterized in terms of population dynamics and phenotype, such as metabolite production and fermentation kinetics.The strains constructed in this work are available to the scientific community at the CIRM-levures.Such microbial consortium could be a valuable resource for applications in yeast ecology and support future research aimed at investigating the influence of stress factors relevant to the fermentation process, particularly when using multispecies starters.Factors such as oxygen availability, nitrogen levels, temperature and the use of SO 2 are critical parameters influence the presence of non-Saccharomyces, often with species-specific effects.Consequently, the trends observed within the 6-and 5-species communities and the our methodology could be applied and tested under a wider spectrum of environmental conditions.Ultimately, the insights gained from studying population dynamics might provide strategies to mitigate fermentations by managing the yeast population.
Description of the set of channels (with dichroic filters details) used for the detection of the different populations considered in this study.Voltages were set using fluorescent and wild type cells to ensure a proper signal of both.The fluorescence phenotype of each strain is indicated with + (fluorescent) and -(non-fluorescent) on the corresponding channel.Viability was measured

Figure 1 :
Figure 1: Cytometry set up with an example of sample at inoculation.A: YL2 (600DLP-620/15)/BL1 (495DLP-530/30) dot plot is used to separate GFP positive and mCherry positive cells.B: VL2 (495DLP-512/25)/SSC dot plot derived from the GFPPos gate on A, is used to separate mCitrine (-) from eGFP signal (+).C: VL1 (417LP-440/50)/SSC dot plot derived from the GFPNeg gate on A, is used to separate non-fluorescent cells from BFP positive cells.D: Hierarchy of gating and percentage of each population.% Total indicates the % of the total population of events in the corresponding gate.

Figure 2 :
Figure 2: Comparison of observed population and theoretical population in 30 mock communities for immediate measure (A) and after 2 hours in PBS (B).Correlation factor and linear equation of observed population in function of theoretical populations are indicated for each species.Dashed line: log10(Observed cell number) = log10(Theoretical cell number).Hu: H. uvarum, Lt: L. thermotolerans, Mp: M. pulcherrima, Sb: S. bacillaris, Sc: S. cerevisiae, Td: T. delbrueckii.Characterization of the 6-species model consortium Population dynamics in the consortium and monocultures.

Figure 4 :
Figure 4: Fermentation kinetics of monocultures and consortium Co. A: CO 2 production rate kinetics of each monoculture (solid line) and consortium Co (dashed line).Smoothing with the Loess method was applied on the results of triplicates.B: Kinetics parameters obtained from CO 2 loss.

Figure 7 :
Figure 7: Metabolite yield changes in the different consortia (calculated by dividing the difference between measures for S280 and S200 by the measure for S200).Both sugar conditions were compared by T-test.Gray bars: p-value > 0.05; *, **, ***: p-value inferior to 0.05, 0.01, 0.0001 respectively (in blue when delta is positive, yellow when delta is negative).Orange diamonds indicate the values observed in the S. cerevisiae monoculture.C02 to C11 initially contained 5% of S. cerevisiae, C12 initially contained 15% of S. cerevisiae and C13 did not include S. cerevisiae.

Table 2 :
Composition (in %) and diversity index of the different consortia tested in this study.