The Role of Ancestral Duplicated Genes in Adaptation to Growth on Lactate, a Non-Fermentable Carbon Source for the Yeast Saccharomyces cerevisiae

The cell central metabolism has been shaped throughout evolutionary times when facing challenges from the availability of resources. In the budding yeast, Saccharomyces cerevisiae, a set of duplicated genes originating from an ancestral whole-genome and several coetaneous small-scale duplication events drive energy transfer through glucose metabolism as the main carbon source either by fermentation or respiration. These duplicates (~a third of the genome) have been dated back to approximately 100 MY, allowing for enough evolutionary time to diverge in both sequence and function. Gene duplication has been proposed as a molecular mechanism of biological innovation, maintaining balance between mutational robustness and evolvability of the system. However, some questions concerning the molecular mechanisms behind duplicated genes transcriptional plasticity and functional divergence remain unresolved. In this work we challenged S. cerevisiae to the use of lactic acid/lactate as the sole carbon source and performed a small adaptive laboratory evolution to this non-fermentative carbon source, determining phenotypic and transcriptomic changes. We observed growth adaptation to acidic stress, by reduction of growth rate and increase in biomass production, while the transcriptomic response was mainly driven by repression of the whole-genome duplicates, those implied in glycolysis and overexpression of ROS response. The contribution of several duplicated pairs to this carbon source switch and acidic stress is also discussed.


Introduction
Saccharomyces cerevisiae, the budding yeast, uses glucose as its main carbon source to obtain energy, as with many other organisms. Similar to numerous other yeast species, it can alternate between fermentative and oxidative metabolism, depending on the availability of glucose and oxygen [1]. In addition, this species is able to use alternative carbon sources by adapting growth profiles [2]. This growth adaptation/versatility has been of high interest for biotechnology, from the fuel industry to biomedical/pharmaceutical industries, driving the interest in metabolism adaptation through genome modifications (i.e., adding new transporters or enzymes from closely related species or bacterial species of interest) (reviewed in [3]). This interest also relies on the tolerance to industrial bioprocess duplicates, especially by WGDs. This transcriptomic response induced a cellular response that explains the observed phenotypic changes, and highlights the importance of other stress responses, such as ROS response similar to those observed with osmotic or ethanol stresses, or the chaperone implication at the adaptive process. These aspects are discussed in this paper.

Lactic Acid as a Non-Fermentable Carbon Source Affects S. cerevisiae Growth Parameters
To test the ability of the yeast to overcome a challenge with a non-fermentable carbon source as lactic acid/lactate, the yeast Saccharomyces cerevisiae strain Y06240 was evolved first through a daily 1% bottlenecking in YPD for 100 passages (~6.6 generations per passage;~660 generations). One of the evolved populations was then selected and subjected to ALE (adaptive laboratory evolution) to YPL (stressing media with 1% yeast extract, 2% peptone, and 3% lactic acid as the sole carbon source, at pH 5.5) by daily passages of 10% of the population for another 10 passages (~3.3 generations per passage;~33 generations on adaptive evolution), also maintaining a control line in YPD under the same conditions; both lines were in triplicate. After evolution, cells from the fossil records at different time points were recovered and subjected to phenotypic characterization through determination of growing parameters with Bioscreen c, and transcriptomic profiling by RNAseq ( Figure 1A; see Material and Methods Section for further details).
We determined the maximum growth rate (r, Figure 1B, and Supplementary Materials Figure S1A) and carrying capacity (k, Figure 1C and Supplementary Materials Figure S1B) for each population in the evolved and challenged media (YPD and YPL), from OD 600 time course log 2 -adjusted model as implemented in Growthcurver [32]. Both parameters were used as a measurement of the populations' fitness in each medium. The starting yeast population (t 0 ) showed a sigmoidal OD 600 curve with a maximum growth rate (r ± s.d.) of 0.261 ± 0.055 h −1 in YPD and of 0.175 ± 0.043 h −1 in YPL, being significantly lower in the second as expected due to the acidic stress (Wilcoxon rank test, p-value = 0.0175; Figure 1B and Supplementary Materials Figure S1A). The evolved population t 100 also showed a sigmoidal curve in YPD, but with a lower growth rate of 0.195 ± 0.049 h −1 (Wilcoxon rank test, p-value = 5.83 × 10 −4 ), and when challenged to YPL growth rate, was even lower (r = 0.141 ± 0.030 h −1 ; Wilcoxon rank test, p-value = 5.349 × 10 −2 ), not observing statistical differences among media at this time point (Wilcoxon rank test, p-value = 0.165). At t 110 , the control population (Dat 110 , being evolved in YPD) showed also sigmoidal curve in YPD recovering growth rate (r = 0.257 ± 0.062 h −1 ) compared to t 0 (Wilcoxon rank test, p-value = 5.827 × 10 −4 ), but showed a reduced growth rate when challenged to YPL (r = 0.120 ± 0.016 h −1 ; Wilcoxon rank test, p-value = 4.15 × 10 −4 ). Whereas YPL-adapted t 110 population (Lat 110 ) had reduced growth rate in YPL (r = 0.104 ± 0.019 h −1 ; Wilcoxon rank test, p-value = 3.04 × 10 −6 ) compared to t 100 challenged to YPL, and showed a higher response when challenged to the ancestral medium, YPD (r = 0.262 ± 0.032 h −1 ; Wilcoxon rank test, p-value = 3.04 × 10 −6 ; Figure 1B).

Figure 1.
Saccharomyces cerevisiae laboratory evolution under diversification and adaptation to YPL and phenotypic effects on growing parameters. (A) Scheme of experimental evolution setup, with 1% bottlenecks during the first 100 passages (population diversification stage), and adaptive laboratory evolution to YPL or YPD under 10% population bottlenecks for additional 10 passages; (B) Growth rate (r, in h −1 , as inferred from Growthcurver) for each time point and lines series, with the indication of the evolved and challenged media, note that the t0 populations were not evolved, but reared in YPD medium; (C) Carrying capacity (k), as maximum OD600 (as inferred from Growthcurver) for each time point and lines series, following the same criteria as in (B). Statistical differences among means were determined with *, **, *** or **** when the probabilities are p < 0.05, p < 0.005, p < 0.001, and p < 0.0001, respectively, using a Wilcoxon rank test.

Transcriptional Response of S. cerevisiae to Lactic Acid/Lactate as Non-Fermentable Carbon Source
The genetic transcriptional response of the yeast cells to a non-fermentable carbon source, as the lactic acid/lactate, was determined by RNA sequencing with comparison between time points and conditions ( Figure 1A). The transcripts were mapped to a total Figure 1. Saccharomyces cerevisiae laboratory evolution under diversification and adaptation to YPL and phenotypic effects on growing parameters. (A) Scheme of experimental evolution setup, with 1% bottlenecks during the first 100 passages (population diversification stage), and adaptive laboratory evolution to YPL or YPD under 10% population bottlenecks for additional 10 passages; (B) Growth rate (r, in h −1 , as inferred from Growthcurver) for each time point and lines series, with the indication of the evolved and challenged media, note that the t 0 populations were not evolved, but reared in YPD medium; (C) Carrying capacity (k), as maximum OD 600 (as inferred from Growthcurver) for each time point and lines series, following the same criteria as in (B). Statistical differences among means were determined with *, **, *** or **** when the probabilities are p < 0.05, p < 0.005, p < 0.001, and p < 0.0001, respectively, using a Wilcoxon rank test.

Transcriptional Response of S. cerevisiae to Lactic Acid/Lactate as Non-Fermentable Carbon Source
The genetic transcriptional response of the yeast cells to a non-fermentable carbon source, as the lactic acid/lactate, was determined by RNA sequencing with comparison between time points and conditions ( Figure 1A). The transcripts were mapped to a total of 6692 genes. At t 0 , when challenged to YPL (compared to YPD), a deregulation of 1283 genes (FDR < 0.005 and |log 2 FC| > 1) was observed, being almost equally distributed among up-regulated (N = 628; log 2 FC > 1), and down-regulated genes (N = 655; log 2 FC < −1; Exact binomial test: p-value = 0.47; Figure 2). The evolved population t 100 challenged with YPL, showed 1015 deregulated genes, not being evenly distributed between up-(N = 615) and down-regulated genes (N = 400; Exact binomial test: p-value = 1.59 × 10 −11 ; Figure 2). When comparing transcriptomic response to YPL between t 100 and t 0 , only 431 of the de-regulated genes were also altered at t 0 , with 243 up-regulated genes at t 0 ( Figure 2). This is significantly higher than the number of down-regulated genes at both populations (N = 138; Fisher's exact test: odds ratio = 1.83, p-value = 3.49 × 10 −7 ). −1; Exact binomial test: p-value = 0.47; Figure 2). The evolved population t100 challenged with YPL, showed 1015 deregulated genes, not being evenly distributed between up-(N = 615) and down-regulated genes (N = 400; Exact binomial test: p-value = 1.59 × 10 −11 ; Figure 2). When comparing transcriptomic response to YPL between t100 and t0, only 431 of the de-regulated genes were also altered at t0, with 243 up-regulated genes at t0 (Figure 2). This is significantly higher than the number of down-regulated genes at both populations (N = 138; Fisher's exact test: odds ratio = 1.83, p-value = 3.49 × 10 −7 ). Indeed, to point out the high reliability between the transcriptomic data obtained here, we investigated how many genes were up-regulated at t100 but down-regulated at t0, observing only four genes, whereas nine genes were observed in the opposite direction, being down-regulated at t100 after being up-regulated at t0. This indicates that in general, the regulation sense is kept under our evolutionary experiment, meaning that up-regulated genes at t0 are also up-regulated at t100 and that down-regulated genes at t0 are also down-regulated at t100 (Figure 2).
To understand which transcriptomic changes drive the adaptation process, we analysed the transcriptome of the population t110, which had been grown for 10 passages in YPL, having lactic acid/lactate as the sole carbon source. In this early adaptation, 2075 genes altered their expression profile when compared to the growth in the YPD control media, showing a huge cellular reprogramming. Contrary to what was observed in the t0 and t100 populations, this adapted population showed 1157 down-regulated genes, being significantly higher than the number of up-regulated genes (N = 918; Exact binomial test: p-value = 1.69 × 10 −7 ).
Comparing the altered genes at t0, t100, and t110, we identified a core set of 317 transcriptionally altered genes, responding to lactic acid/lactate stress (  Indeed, to point out the high reliability between the transcriptomic data obtained here, we investigated how many genes were up-regulated at t 100 but down-regulated at t 0 , observing only four genes, whereas nine genes were observed in the opposite direction, being down-regulated at t 100 after being up-regulated at t 0 . This indicates that in general, the regulation sense is kept under our evolutionary experiment, meaning that up-regulated genes at t 0 are also up-regulated at t 100 and that down-regulated genes at t 0 are also down-regulated at t 100 ( Figure 2).
To understand which transcriptomic changes drive the adaptation process, we analysed the transcriptome of the population t 110 , which had been grown for 10 passages in YPL, having lactic acid/lactate as the sole carbon source. In this early adaptation, 2075 genes altered their expression profile when compared to the growth in the YPD control media, showing a huge cellular reprogramming. Contrary to what was observed in the t 0 and t 100 populations, this adapted population showed 1157 down-regulated genes, being significantly higher than the number of up-regulated genes (N = 918; Exact binomial test: Comparing the altered genes at t 0 , t 100 , and t 110 , we identified a core set of 317 transcriptionally altered genes, responding to lactic acid/lactate stress ( Figure 2), with 267 showing the same expression profile. More specifically, 94 genes were down-regulated and 173 genes were up-regulated at all time points. Interestingly, only 181 of the 2803 responding genes showed a different expression profile (Figure 2), 1072 genes retained the same profile and 1550 genes were profile and point specific.

Many Cellular Processes Are Altered When S. cerevisiae Is Challenged with Lactic Acid/Lactate
To shed light on how the yeast changed its cellular response when the environment suddenly changes the carbon source, we analysed what cellular processes were altered in the different populations according to the Gene Ontology terms (GO) using the R package clusterProfiler [34]. At t 0 up-regulated genes were enriched in functional categories, including "aerobic respiration", "drug metabolic process", "oxidation-reduction processes" and "ATP metabolic process", whereas fundamental cell processes were down-regulated, including "cytoplasmic translation", "ribosome biogenesis", "ribosome assembly", "rRNA metabolic process", and "RNA transport" (Figure 3). suddenly changes the carbon source, we analysed what cellular processes were altered in the different populations according to the Gene Ontology terms (GO) using the R package clusterProfiler [34]. At t0 up-regulated genes were enriched in functional categories, including "aerobic respiration", "drug metabolic process", "oxidation-reduction processes" and "ATP metabolic process", whereas fundamental cell processes were down-regulated, including "cytoplasmic translation", "ribosome biogenesis", "ribosome assembly", "rRNA metabolic process", and "RNA transport" (Figure 3). For the population t100, we observed a large overlap with t0 for up-regulated genes, including "aerobic respiration", "drug metabolic process" or "oxidation-reduction processes", and found some functional groups related to mitochondria as "mitochondrial translation", "mitochondrial gene expression" and "oxidative phosphorylation" that were down-regulated in t0 (Figure 3), whereas t100 down-regulated gene GO enrichment included "organophosphate catabolic process", "fructose transmembrane transport", "glucose transmembrane transport", and "carbohydrate catabolic process" that were almost exclusive for this time point (Figure 3).
In the population t110, up-regulated genes were enriched for many cellular processes shared with populations, t0 and t100, like "oxidation-reduction process", "purine-containing compound metabolic process" or "mitochondrion organization"; having three specific functional categories: "energy derivation by oxidation of organic compounds", "generation of precursor metabolites and energy" and "purine ribonucleotide metabolic process" (Figure 3), whereas down-regulated gene enrichment was partially shared with t0, including "cytoplasmic translation", ribosome biogenesis" or "ribosome assembly"; also showing a specific signature that included "rRNA processing", "RNA export from nucleus", and "ribosome localization" (Figure 3). For the population t 100 , we observed a large overlap with t 0 for up-regulated genes, including "aerobic respiration", "drug metabolic process" or "oxidation-reduction processes", and found some functional groups related to mitochondria as "mitochondrial translation", "mitochondrial gene expression" and "oxidative phosphorylation" that were down-regulated in t 0 (Figure 3), whereas t 100 down-regulated gene GO enrichment included "organophosphate catabolic process", "fructose transmembrane transport", "glucose transmembrane transport", and "carbohydrate catabolic process" that were almost exclusive for this time point (Figure 3).
In the population t 110 , up-regulated genes were enriched for many cellular processes shared with populations, t 0 and t 100 , like "oxidation-reduction process", "purine-containing compound metabolic process" or "mitochondrion organization"; having three specific functional categories: "energy derivation by oxidation of organic compounds", "generation of precursor metabolites and energy" and "purine ribonucleotide metabolic process" (Figure 3), whereas down-regulated gene enrichment was partially shared with t 0 , including "cytoplasmic translation", ribosome biogenesis" or "ribosome assembly"; also showing a specific signature that included "rRNA processing", "RNA export from nucleus", and "ribosome localization" (Figure 3).

The Implication of Duplicated Genes in the Transcriptional Response of S. cerevisiae to Lactic Acid/Lactate
As indicated previously, the budding yeast keep around a third of its genome in duplicate, as a result of the ancestral whole-genome duplication, genome rearrangement and reduction, and coetaneous small-scale duplication events that took place~100 MYA. Thus, these ancestral duplicates can be studied depending on the molecular mechanism they originate from, being divided into WGDs (whole-genome duplicates) and SSDs (small scale duplicates) [5,6,21,22,35]. For this study, we used a starting set of 1101 duplicated pairs split into 548 WGD and 553 SSD pairs, to analyse their contribution to the transcriptional and cellular response observed to lactic acid/lactate as the sole carbon source. Of this data set, we were able to identify expression data from 1094 WGDs and 1097 SSDs with a differential response at each time point ( Figure 4A). (A) Distribution of fractions of deregulated genes, depending on their classification as singletons or duplicates, and within duplicates, depending on their origin, WGDs (whole-genome duplicates) or SSDs (small-scale duplicates); (B) Distribution of expression level as the logarithm of fold-change (log2FC) for singletons and duplicates, with duplicates also being divided into WGDs and SSDs. Statistical differences among means were determined with *, ** or ***, corresponding to the probabilities p < 0.05, p < 0.005, and p < 0.001, respectively, using a Wilcoxon rank test.
This trend was also observed in up-regulated (N WGD = 178, N SSD = 140; Fisher's exact test: odds ratio F = 1.27, p-value = 0.0476; Figure 4A). Previously we determined that expression level (fold-change, FC) also contribute to duplicates preservation and to their functional innovation through response to different stresses [24]. In this case, we observed at t 0 that duplicated genes exhibited a higher absolute log fold-change (median = 1.536) than singletons (median = 1.388; Wilcoxon test: p-value = 4.58 × 10 −7 ) of the de-regulated genes. Furthermore, this was also the case when separating out into up-and down-regulated genes ( Figure 4B). Considering the origin of those duplicates, down-regulated WGDs (median = −1.561) were more down-regulated than SSDs (median = −1.431; Wilcox test: p-value = 0.045; Figure 4B) whereas no differences among duplicate types were observed in up-regulated genes (median WGD = 1.549, median SSD = 1.631; Wilcox test: p-value = 0.818). At population t 100 we did not observe any difference between expression levels of altered genes, regardless of the comparison carried out ( Figure 4B). Whereas at population t 110 , one significant difference with up-regulated duplicates showed a significantly higher expression fold-change (median = 1.658) than singletons (median = 1.538; Wilcoxon test: p-value = 2.448 × 10 −4 ), with no difference detected for this time-point ( Figure 4B).

The Cellular Re-Programming in Response to Lactic Acid/Lactate Is Driven through Duplicated Genes
We identified the duplicated genes that were altered at the core gene set, hence those that were altered in all populations (t 0 , t 100 , and t 110 ). We found that 148 duplicated genes belong to the core category, with 124 having the same expression profile at all three populations (N up = 67 and N down = 57). This core set gene showed enrichment of down-regulated duplicates (Fisher's exact test: odds rate F = 3.164, p-value = 5.0 × 10 −8 ; Figure 2) but the up-regulated genes were distributed among duplicates and singletons as would be expected (Fisher's exact test: odds rate F = 1.298, p-value = 0.1015). When investigating what functional categories were affected, we observed that altered core genes were enriched mainly for functions involved in precursor metabolites and energy, such as "oxidation-reduction process", "generation of precursor metabolites and energy", "aerobic respiration" and "mitochondrial respiratory chain complex assembly".

Chaperones, Heat Shock Proteins and Stress-Related Proteins Responding to Lactic Acid/Lactate as Carbon Source
Among the lactate quick response genes (those up-regulated at t 0 and t 100 ) we found: Gre1 (YPL223C) a WGDs responsive to stress; Fyv5 (YCL058C) a gene de novo originated from non-coding sequence involved in ion homeostasis and required for survival after exposure to killer toxins; and Fyv4 (YHR059W) a protein of unknown function also required for survival to killer toxins.
Among the highly expressed genes responding to acute lactate exposure (Lat 110 population) we found several heat shock proteins:

Metabolic Evolution of YPL Adapted S. cerevisiae Populations
To determine similarities among the different populations after the experimental evolution, we measured the metabolic distance among the three populations, this being the grade of sharing of cellular GO processes to determine proximity of the populations in terms of metabolism (see Materials and Methods section). The metabolic distance (MD) was determined for all the transcriptionally altered genes (including duplicates and singletons), as well as for the up-and down-regulated genes between pairs of the t 0 , t 100, and t 110 populations. As expected, we observed that the most related metabolic distance of the populations for the whole altered gene set were t 0 and t 100 since the population t 100 was neutrally evolved from the population t 0 to increase its genetic variability. The same was true for the up-regulated gene set but surprisingly not for the down-regulated gene ( Table 1). When we analysed only the duplicated genes, we observed that the closest metabolic distance for all comparisons was between t 0 and t 110 populations. Table 1. Metabolic distances between evolved populations in response to lactic acid/lactate as sole carbon source.

Discussion
Yeast, like many other organisms, uses glucose as the main carbon source to obtain energy, and is able to modify its metabolism depending on the available carbon sources. The budding yeast undergoes fermentative metabolism of glucose to obtain ATP and other basic metabolic precursors from the complex YPD medium (containing 2% dextrose) having ethanol, glycerol, and lactate/lactic acid as the main by-products. These by-products can become potent control molecules that limit the development of other microorganisms by exerting different kinds of stresses or are used as alternative carbon sources to obtain energy through respiration. In this work, we determine the contribution of duplicated genes in response to the carbon source switch from glucose to lactic acid/lactate and how this transcriptional switch affected yeast growth.

Lactic Acid as Non-Fermentable Carbon Source Affects S. cerevisiae Growth Parameters
Lactic acid, an organic acid produced by some yeasts and lactic bacteria through sugar media fermentation, is usually produced as its conjugated anion lactate due to intracellular pH (pKa = 3.68). This weak organic acid compound has a wide range of industrial applications ranging from food preservation to pharmaceutical production to plastic production. Moreover, it is a biological stressor, by either forcing a growth arrest and a metabolic rewire of the target organism due to the lack of glucose in the medium or by acidic stress induced by the dissipation of pH gradient through the plasma membrane [28,36]. In addition, lactate has been recently classified as a signaling molecule in several human diseases, including cancer, deserving further research on its metabolism and transcriptional effects [37][38][39][40].
The presence of this weak acid in the culture media (either in small or at factory scales) induces in a first step a cell-cycle arrest, resuming cell growth after several hours. This transient growth arrest is lost when cells are pre-adapted to weak acids, such as lactic acid or acetic acid, implying a durable transcriptomic rewire [41]. In our work, we saw this growth arrest with a significant decrease of growth rate when switching carbon source from glucose to lactic acid, this decrease was also observed in the lactic acid short-adapted lines (population YPL t 110 , Figure 1B). However, as indicated, cells are able to resume cell growth after several hours, time allocated to transcriptional activity, and cell growth and reproduction, which at the end indicates an increase of the carrying capacity of the culture. Similarly, our YPL-adapted lines showed a significant increase of carrying capacity compared to the non-adapted ones (control lines), indicating successful growth resumption (see Figure 1C and Supplementary Materials Figure S1B). In other works, we demonstrated that this short adaptive experimental evolution (only 33 generations) also affected growth rate, with a significant reduction when challenged to glucose-deprived media using other non-fermentative carbon sources [26,27]. However, when using glycerol or ethanol as nonfermentative carbon sources, growth was resumed without a significant increase of biomass as observed with lactic acid/lactate ( Figure 1C and Supplementary Materials Figure S1B). These results would indicate that in this case, we are not yet fixing adaptive mutations, as expected if a genotypic switch took place after the adaptive process [42][43][44][45][46][47][48][49]. But the increase of carrying capacity in the YPL adapted lines is also in agreement with an increase of TCA fluxes, as explained in [50]. Indeed, we observed phenotypic heterogeneity in the evolved cells, which decreased after adaptive evolution (see Figure 1, and Supplementary Materials Figure S1). It seems that our populations show heterogeneous phenotypes linked to transcription heterogeneity in lactic acid/lactate medium, as observed previously to adaptation to other functional trade-offs [51,52].

Lactic Acid as a Non-Fermentable Carbon Source Affects S. cerevisiae Transcriptomic Response
The improvement of lactic acid production from sugar fermentation in S. cerevisiae has been achieved by alcohol fermentation inhibition, by increasing methyltransferases, and by modulation of Jen1 and Ady2 monocarboxylic permeases [53][54][55]. These two transporters (repressed by glucose) mediate the export of acetate, formate, propionate, and lactate from the cytoplasm reducing the internal acidic stress and increasing the release of lactic acid. However, these two transporters also work in the reverse direction, being involved in the use of lactic acid/lactate as a carbon source by the budding yeast [54]. In this work, we forced the cell to use lactic acid/lactate as the sole carbon source, finding that both transporters Jen1 and Ady2 were overexpressed since the first challenge, and that ethanol reduction to acetaldehyde was also increased, redirecting the central metabolism to obtain energy (see Figure 5). Among the responding genes to lactate in the core set (altered genes shared by all three populations), we found Idp2 (YLR174W; WGD), Cyb2 (YML054C; singleton), Mls1 (YNL117W; SSD), Acs1 (YAL054C; SSD), and Rgi2 (YIL057C; WGD) genes implied in the utilization of non-fermentable carbon sources, with affected growth parameters even under vegetative growth and decreased lifespan [58,59]. It has been demonstrated that under glucose limiting conditions, the budding yeast undergoes a significant transcriptional switch, with a fast response under an acute situation or with a slow response under chronic glucose limitation [56,57]. As indicated previously, the budding yeast can use several non-fermentable carbon sources that are also products of glucose metabolism, such as ethanol, glycerol, acetate, and lactate, inducing a transcriptional regulation involving overexpression of membrane transporters and activation of transcription factors for the use of these non-fermentable carbon sources [58,59]. In this work, we show that the complete replacement of glucose (dextrose) by a non-fermentable carbon source such as lactic acid/lactate in the growing medium (from YPD to YPL) drives to a complete transcriptomic re-wiring affecting central metabolism genes that would affect growing parameters (see Figures 3 and 5).
Similarly, when the hybrid yeast Zygosaccharomyces parabailii was subjected to lactic acid supplementation with a medium rich in glucose (4%), a transcriptomic rewire was observed [60]. This species shows a weak acid stress tolerance by modulating cell wallrelated genes, including the transcription factors Haa1, Aft1/Aft2 (repression of these genes to avoid lactic acid uptake), by the induction of genes involved in oxidative stress response and iron homeostasis, and by amplification by gene duplication at a small scale of formate dehydrogenase (FDH) genes, identifying such SSDs directly linked to lactic acid tolerance with differential gene expression between copies [60]. The budding yeast, S. cerevisiae, undergoes a different strategy (when tested under acidic stress in the presence of glucose), inducing ROS response, affecting iron homeostasis, and expending large amounts of ATP through Pdr12 ABC transporter (YPL058C) that catalyses lactic acid [41]. In our case, we observed that under media with no glucose, the budding yeast forces the entry of lactate with over-expression of transporters Jen1 and Ady2, and re-wires the central metabolism to the production of energy via pyruvate and acetate synthesis from lactate through up-regulation of a set of duplicated genes, including some related to ROS response and heat-shock family proteins (see some of the duplicates in Figure 5).

Ancient Duplicates Direct the Transcriptomic Response
As indicated previously, gene duplication is a major force in evolution, sourcing new genetic material and novel functions, being related to the radiation of angiosperms and developmental complexity in animals [61][62][63][64][65]. In this study, we were interested in deciphering the role of the anciently duplicated genes during the adaptation to a challenging environment, the YPL medium containing lactic acid/lactate as the sole carbon source, by analysing results of an adaptive laboratory experiment.
Experimental evolution has been a common tool in the past years to achieve insight into molecular mechanisms and cellular responses underlying adaptation [66][67][68]. As shown, we found that (Figures 2 and 4) duplicates direct the transcriptomic response to an acidic environment with lactic acid/lactate as the sole carbon source after experimental evolution. Duplicates significantly alter their expression profile compared to singletons, showing also significantly higher expression fold-change than singletons (Figure 4). When studying the transcriptional response, a core set of genes was identified, with an enrichment of the WGDs. Whole-genome duplicates implied in glycolysis, glucose metabolism, and hexose catabolic process were repressed. We also observed down regulation of Sam2 (YDR502C) involved in direct lactic acid tolerance and production [54]. Up-regulated WGDs were involved in "oxidation-reduction process", "generation of precursor metabolites and energy", "aerobic respiration" and "mitochondrial respiratory chain complex assembly" (Figure 3). Of these pairs, we also observed a discordant trend (discordant category as in [23,24]) with opposite transcriptional response to duplicated pairs, including SSD pairs such as Acs1/Acs2, Ald4/Ald5, and Adh1/Adh2, and the WGD pair Pyc1/Pyc2 (Figure 5), and only one SSD gene pair, Yat1/Yat2, was up-regulated.

Lactic Acid/Lactate as the Sole Carbon Source Induces More Than a Single Stress Response
Wild-type yeasts can grow at pH values ranging from 2.5 to 8.5, with growth and fermentation kinetics not affected between pH 3.5 and 6.0. As indicated previously, lactic acid/lactate is purely a respiratory substrate for Saccharomyces cerevisiae that affects growth parameters by reducing growth rate and increasing biomass production. This lactate consumption increases the pH of the medium by decreasing the basal respiratory rate, mainly due to a decrease in ATP consumption linked to the maintenance of intracellular pH and reduction of the number of mitochondria per cell [69]. The intracellular pH is also a tightly controlled factor, with great differences among cell compartments (vacuole, mitochondria, nucleus, Golgi network, peroxisomes, and secretory vesicles), that alters the cellular set-up and cell physiology affecting multiple regulatory levels simultaneously [31,70]. A decrease of the internal pH will affect enzymes activity, and also by changing interaction between residues of amino acid side chains conformational stability of proteins and interactions between proteins will be affected. In addition, oxidation-reduction potential (transference of electrons and expenditure of NADPH) is also dependent on internal pH, affecting ROS response-related proteins (reviewed in [70]). Moreover, acidification of cytosol will affect other structural molecules as lipids or coupled phosphates affecting intracellular compartment membranes and DNA (reviewed in [70] and references within). Each of these steps seems to be involved in the induction of acidic, oxidative, and DNA damage stresses, which were observed in this work (see Figure 3). In addition, the conformational instability of proteins might be responsible for the observed up-regulation of a long list of heat shock proteins, chaperones, and co-chaperonins (see Section 2.6). This deserves further study to determine their implication on system robustness, as this response was only observed after chronic exposure to lactic acid/lactate. The implication of altered duplicates on such responses (to different stresses) still deserves further research, but some light has been shed on how duplicates were selected, preserved, and still able to innovate following Ohno's dilemma [71,72].

Growth Characterization under Lactic Acid Use as Sole Carbon Source
Evolved yeasts at t 0 , t 100 , and t 110 were characterized for growing ability response to carbon source, either in ancestral medium (YPD) or challenged to YPL ( Figure 1A). Cells from fossil records were recovered in the corresponding growing media, by seeding 5 to 50 µL of glycerol stock into 5.0 mL of fresh medium in 50 mL conic vials. After 24 h at 28 • C and 200 rpm, cells were diluted till OD 600~0 .1 in 200 µL volume and growth was recorded in a Bioscreen c MBR plate reader system (Oy Growth Curves Ab Ldt., Helsinki, Finland) taking OD 600 measurement every 15 min, with brown filter and continuous shaking at 28 • C. Each line and time points were challenged in both YPD and YPL with at least 5 technical replicates.
Growth rate (r) and carrying capacity (k) were determined from the logistic adjustment of corrected OD 600 time data series in each well with Growthcurver package in R [32]. Means were compared among treatments and lines with Wilcoxon rank test in R [73].

RNAseq and Transcriptomic Profiling
Evolved yeasts at t 0 , t 100 , and t 110 were characterized for their transcriptional response to carbon source switch or their adaptation to this acidic environment ( Figure 1). As described previously, cells from fossil records were recovered in the corresponding growing media, and after reaching OD 600~0 .6 (~16 h at 28 • C and 200 rpm), challenged to YPD and YPL, in triplicate. Cultures were stopped at OD 600~0 .6 (by placing them in an ice bath), cells harvested by centrifugation and total RNA extracted using RNEasy kit (Qiagen) following manufacturer instructions. RNA quality was checked with QuBit 4, and those samples with RIN >8 were used for library construction.
Raw reads were analysed using FastQC report, cleaned with CutAdapt, and trimmed for quality and length (Pred score inferior to 20 and size less than 40 nt were discarded). Reads were aligned with Bowtie2 (up to two mismatches accepted) to the reference S288c strain genome (only CDS; assembly R64) [26]. Statistical assessment of differential gene expression was carried out with edgeR, setting false discovery rate (FDR) at <0.005, and applying BY correction for p-value (0.005) [75]. Comparisons were conducted between time points and media, taking into account the original transcriptomic background.

Identification of Duplicated Genes Involved in the Usage of Lactic Acid as Sole Carbon Source and Their Response to Acidic Stress
Differentially transcribed genes were further analysed as singletons and duplicates (homologs), duplicates were split into two, according to their mechanism of origin (WGDs or SSDs). Whole-genome duplicates (WGDs; 555 pairs) were extracted from the reconciled YGOB list v.7 (Yeast Gene Order Browser version 7 (August 2012); http://wolfe.gen.tcd.ie/ /ygob; [76]), while small-scale duplicates (SSDs; 560 pairs) were identified by previously carried out best reciprocal blast searches [21,22,26].
After classification of differentially expressed genes as singletons or duplicates (WGDs or SSDs), an enrichment analysis of gene ontology (GO) terms was performed using the R package cluster Profiler and p-value cut-off of <0.01 [34], to determine the contribution of each set in the transcriptomic response to acidic stress.

Metabolic Distance between Populations
The metabolic distance was calculated as previously described [27]. Briefly, lists of GO process terms enriched for transcriptionally altered genes between two populations (i and j) were compared by calculating the number of shared process terms (SPij), and the number of enriched terms for transcriptionally altered genes only in one of the populations but not in the other (Pi and Pj), with metabolic distance (MDij) calculated as: where Min[Pi, Pj] is the number of cellular processes enriched for transcriptionally altered genes for the population with the minimum number of such processes. Metabolic distance varies between 0 and 1, with smaller values indicating the closeness of populations, and a value of 1 corresponding to two complete divergent populations, as there is no overlap in the enriched process terms shared.

Conclusions
Lactic acid/lactate is an alternative non-fermentative carbon source that also induces acidic stress to yeast cells. These cells undergo cell growth arrest, with a significant reduction of growth rate, to adjust their transcription profiling. After this transcriptional rewiring, cell growth arrest is resumed thereby producing an increase of biomass (measured as carrying capacity), even after a short-term adaptive process (33 generations). The transcriptional rewire of central metabolism includes down-regulation of WGDs implied in glycolysis, and up-regulation of several WGDs and SSDs involved in transcription adjustment, in the oxidative stress response and showing the implication of several heat shock proteins and chaperones. With this work, we add a further step in determining duplicated genes' central role in acute and chronic response to stress, ultimately with their continued involvement in the adaptation process.