Kinetic analysis of yeast-yeast interactions in 1 oenological conditions

Fermentation by microorganisms is a key step in the production of traditional food products 19 such as bread, cheese, beer and wine. In these fermentative ecosystems, microorganisms 20 interact in various ways, namely competition, predation, commensalism and mutualism. 21 Traditional wine fermentation is a complex microbial process performed by Saccharomyces 22 and non- Saccharomyces yeast species. To better understand the different interactions occurring 23 2 within wine fermentation, isolated yeast cultures were compared with mixed co-cultures of one 24 reference strain of S. cerevisiae with one strain of four non- Saccharomyces yeast species 25 ( Metschnikowia pulcherrima, M. fructicola, Hanseniaspora opuntiae and H. uvarum ). In each 26 case, we studied population dynamics, resource consumption and the production of metabolites 27 from central carbon metabolism. This deep phenotyping of competition kinetics allowed us to 28 identify the main mechanisms of interaction between species. S. cerevisiae competed with H. 29 uvarum and H. opuntiae for resources although both Hanseniaspora species were characterized 30 by a strong mortality either in isolated or mixed fermentations. M. pulcherrima and M. 31 fructicola displayed a negative interaction with the S. cerevisiae strain tested, with a decrease 32 in viability in co-culture, probably due to iron depletion via the production of pulcherriminic 33 acid. Overall, this work highlights the importance of measuring cell populations and their 34 metabolite kinetics to understand yeast-yeast interactions. These results are a first step towards 35 ecological engineering and the rational design of optimal multi-species starter consortia using 36 modeling tools. 37 38

In natural or anthropized environments, microbial species are part of an ecosystem and interact 42 positively or negatively, forming a complex network. Until recently, process optimization in 43 agriculture or food processing was mostly based on the selection of single strains. However, 44 this paradigm is now being challenged and the scientific community is increasingly seeking to 45 exploit and optimize consortia of several strains and/or species. Indeed, many studies have 46 shown that more diverse anthropized environments have many advantages in terms of 47 resilience, disease resistance or yield (Barot et al., 2017). Efforts are now being made to design 48 = 0.17 ± 0.04 g.L -1 .h -1 ). The four mixed cultures had intermediate Vmax values between those 123 of Sc and the highest Vmax of all non-sacc cultures (Fig. 1a). Mixed cultures containing 124 Metschnikowia species had significantly higher Vmax values than those containing 125 Hanseniaspora species (Fig 1a). Although we did not monitor all cultures until the exhaustion 126 of glucose and fructose, it was however possible to estimate the capacity of a given species to 127 complete fermentation by estimating the amount of CO2 produced during the first 300 hours. 128 Sc fermentations finished after around 220 hours with a CO2maxSc = 88.2 ± 2.2 g.L -1 . We 129 therefore can make the hypothesis that all cultures that produced more than 80g CO2.L -1 (90% 130 of Sc maximum) within 300 hours will be able to complete fermentation. Under this 131 assumption, all mixed cultures, but not isolated non-Saccharomyces cultures, would eventually 132 complete fermentation. Among the latter cultures , both Hanseniaspora species had the highest 133 CO2max (Hu 30 ± 0.4 g.L -1 , and Ho 46 ± 0.6 g.L -1 ) followed by Metschnikowia species (Mp 22 134 ± 0.4 g.L -1 and Mf 20 ± 1 g.L -1 ). 135 136

Population Kinetics 137
We also looked at population dynamics in each culture (Fig. 2) and determined the maximum 138 growth rate of the population (µ), the maximum population size, also termed carrying capacity 139 (K) and the relative abundance of each species after 300 hours of mixed culture, corresponding 140 in our case to the end of the monitoring period (Table. 1). Fermentations with S. cerevisiae 141 alone went through an exponential growth rate (µSc = 0.15 ± 0.02 h -1 ) and reached a maximum 142 population of around 1.5*10 8 cells.mL -1 (KSc = 1.55 ± 0.15 10 8 cells.mL -1 ) that remained constant 143 until the end of the fermentation. Fermentations with either Hanseniaspora species alone had a 144 growth dynamic similar to Sc at the beginning of the fermentation but a higher growth rate (µHo 145 = 0.19 ± 0.03 h -1 , µHu = 0.62 ± 0.18 h -1 ). On the opposite, their stationary phase was quite 146 different from that of Sc and characterized by a higher cell mortality with a population drop of 147 . CC-BY-ND 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint about 70% by the end of the process. Fermentations performed by Metschnikowia species in 148 isolation had growth dynamics mostly similar to Sc fermentations: a similar growth rate (µmp = 149 0.18 ± 0.03 h -1 , µMf = 0.17 ± 0.2 h -1 ), no mortality during the stationary phase but a much 150 reduced maximum population (KMp = 0.57 ± 0.01 10 6 cells.mL -1 , KMf = 0.8 ± 0.05 10 6 cells.mL -151 1 ). In most cases, mixed cultures displayed an intermediate pattern between the two 152 corresponding isolated cultures (Fig. 2). However, mixed or isolated cultures with 153 Metschnikowia displayed different cell mortality rates during the stationary phase: in the case 154 of ScvsMp fermentations, only the S. cerevisiae population decreased significantly during the 155 stationary phase, while in ScvsMf fermentations, both subpopulations significantly decreased. 156 As a measure of fitness, we also followed the variations of S. cerevisiae frequency along the 157 fermentation. In all mixed cultures, S. cerevisiae was found dominant (frequency > 50%) in the 158 end, increasing significantly during fermentation from 10% initially to frequencies varying 159 between 50% (ScvsMp) and 96% (ScvsMf) ( Table. 1) 160

Sugar and NAS consumption 161
We then looked at the final concentration of resources: sugars (fructose and glucose) and NAS 162 (sum of all assimilable nitrogen sources, i.e. ammonium and amino-acids) (Fig. 3, Table 1). In 163 Sc fermentations, less than 0.1% of the initial concentration of both sugars remained (Fig. 3a). 164 As seen in the paragraph concerning CO2 production, non-Saccharomyces species in 165 monocultures did not complete fermentation in the 300h period and left respectively 45% of 166 sugars for Ho, 67% for Hu, 68% for Mf and 71% for Mp. Furthermore, all species except H. 167 opuntiae preferentially consumed glucose (see supplementary figure 1). Sugar consumption 168 was higher in mixed cultures than in single non-Saccharomyces species cultures (Table 1). 169 However, it was still lower than in Sc species cultures, also with a preference for glucose. This The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint The consumption of nitrogen assimilable sources (NAS, amino-acids and ammonium) 173 displayed the same pattern (Fig. 3b). NAS were almost entirely consumed both in Sc isolated 174 cultures and in all co-cultures, whereas in non-Saccharomyces isolated cultures the fraction of 175 NAS consumed varied between 84% and 94%. However, the preference for different nitrogen 176 sources varied with each species (Fig. 3c). Both Hanseniaspora species had similar behaviors, 177 consuming only half of the available ammonium, 90% of histidine and 89% or 79% of arginine 178 (Fig. 3c). Metschnikowia species presented a similar pattern. It was possible to classify these 179 non-Saccharomyces species preferences for the various NAS. The resulting ranking by order 180 of preference was glutamine, methionine, glutamate, valine, threonine, serine, tryptophan, 181 alanine, histidine, arginine, aspartate, glycine and, surprisingly last, ammonium. 182 183

Metabolite production 184
In parallel with must resources consumption monitoring, we also investigated the production 185 of metabolites from CCM (Central Carbon Metabolism): ethanol, glycerol, succinate, pyruvate, 186 acetate and alpha-ketoglutarate (Table 1). These measurements of metabolite production were 187 taken after 300 hours when sugars consumptions were quite different from one culture to 188 another depending on their dynamics. To allow figures comparison, we computed the 189 production yield (total production / sugar consumption) for each culture and, from these data, 190 we then estimated this yield relatively to that of Sc in isolated culture (Fig. 4). The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint Finally, all mixed cultures seemed to have a lower succinate yield than both corresponding 197 isolated cultures (but not significantly after correction for multiple tests). 198 199 For each fermentation, the total production of metabolites resulted from the combination of 200 species yields, total sugar consumption and respective population dynamics during 201 fermentation. Therefore, differences observed in the total productions of mixed cultures were 202 the consequences of additive or subtractive effects observed for these 3 components. 203 Considering ethanol, its total production was directly linked to the consumption of resources 204 and all mixed cultures were equivalent to Sc fermentations (Table 1, supplementary figure S2). 205 The case of glycerol was more interesting. Indeed, even if the average sugar consumption was 206 lower in ScvsHu and ScvsMp mixed cultures than in isolated Sc culture, the total production of 207 glycerol was significantly higher than that of the corresponding isolated cultures 208 (GlycerolSCvsMp = 6.1±0.1 g.L -1 , GlycerolSc = 5.3±0.4 g.L -1 , GlycerolMp = 3.7±0.2 g.L -1 ). This 209 resulted from the positive combination of the greater glycerol yield by Hanseniaspora and 210 Metschnikowia and their population dynamics. For all other metabolites, the total production 211 of mixed cultures was not significantly different from the corresponding isolated cultures 212 (Table 1)  production of glycerol that seemed to be superior in mixed cultures whereas their sugar 230 consumption was inferior; this is characteristic of a transgressive interaction (often referred to 231 as over-yielding), i.e. a situation in which the ecosystem performance is higher than that of the 232 best-yielding species present (when cultivated alone). This glycerol overproduction in mixed 233 cultures has already been observed in previous works (Ciani and Ferraro, 1996 The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint unclear. A possible explanation could be linked to unmonitored resources. For instance, the 246 third most important resource in the synthetic medium was lipids, available as phytosterols. 247 Concurrently, it has been shown that some Metschnikowia species were unable to import lipids 248 from the medium, depending instead on their own lipid synthesis (see Fig S1). For all yeast 249 species, lipid synthesis requires oxygen and is therefore impossible in anaerobic conditions. 250 From these data, one can hypothesize that Metschnikowia cells were able to produce their own cerevisiae, we could not evidence any major antagonistic phenomena. For almost all assays, 291 mixed cultures performance stood always between that of the corresponding isolated cultures 292 (Fig 1, 2 & 3), with the exception of the total production of glycerol. Moreover, despite the 293 differences in yield and interactions between species, the rapid dominance of S. cerevisiae 294 (increasing from 10% to at least 50% during the fermentation) resulted in mixed cultures that 295 S. cerevisiae has a much better fitness than the non-Saccharomyces strains studied in this paper. 300 Therefore, if we want mixed culture behavior to deviate from that of S. cerevisiae monoculture, 301 it must be ensured that non-Saccharomyces cells dominate the culture as soon as possible. To 302 achieve this, two conceivable options are currently tested: either to reduce the proportion of S. with various initial conditions and identify optimal strategies depending on one or several given 314 criteria. Using these approaches could limit the number of necessary tests, potentially saving a 315 lot of time and money and opening the way to a more methodical ecological engineering. The 316 development of such mathematical models will only be possible thanks to a deep tracking of 317 population dynamics to understand underlying mechanisms of growth and mortality. 318 Obviously, it is also critical to validate this approach by i) first extending the number of species 319 co-cultured with S. cerevisiae, ii) investigating intra-specific variability and strain-strain 320 The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint Initial cultures (12 h, in 50 ml YPD medium, 28 °C) were used to inoculate fermentation media 344 at a total density of 10 6 cells/mL; therefore, for mixed culture the S. cerevisiae cells density was 345 0.1x10 6 /mL and the non-Saccharomyces cells density was 0.9x10 6 /mL. Fermentations were 346 carried out in a synthetic medium (SM) mimicking standard grape juice (Bely et al., 1990). The 347 SM used in this study contained 200 g/L of sugar (100 g glucose and 100 g fructose per liter) 348 and 200 mg/L of assimilable nitrogen (as a mix of ammonium chloride and amino acids). The 349 concentrations of weak acids, salts and vitamins were identical to those described by Seguinot   The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint

fermentation. A) Final concentration of sugar (average ± standard deviation). B) Final 547
concentration of NAS (average ± standard deviation). C) Percentage of consumption of each 548 NAS in each type of fermentation represented as a color gradient from green (<75 %) to red (> 549 75%) . Sc, Mp, Mf, Ho, Hu The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/363531 doi: bioRxiv preprint