Convergent targeting of conserved regulatory networks during thermal evolution across Saccharomyces

12 Whether evolution follows predictable genetic paths across species remains a central 13 question in evolutionary biology, particularly as rising temperatures reshape species 14 distributions worldwide. Despite its importance, the genetic basis of thermal adaptation 15 remains poorly understood across divergent species. Here, we use the yeast genus 16 Saccharomyces as a comparative model to investigate how species with contrasting thermal 17 niches adapt to rising temperatures. We combined experimental evolution under 18 progressively increasing temperatures for up to ~600 generations with whole -genome 19 sequencing of 256 evolved genotypes, followed by transcriptomic, functional, and 20 physiological analyses across eight species. Despite large differences in ancestral thermal 21 tolerance and evolutionary outcomes, selection repeatedly targeted the same conserved 22 regulatory pathways across species. Independent lineages accumulated de novo mutations in 23 central growth and stress response networks, particularly in TORC1, PKA, and MAPK 24 signaling pathways, revealing striking molecular convergence across species occupying 25 distinct thermal environments . However, these shared genetic targets produced divergent 26 transcriptional and physiological responses depending on species background, indicating that 27 thermal adaptation primarily proceeds through rewiring of conserved regulatory hubs rather 28 than changes in temperature -specific enzymes . Cold-tolerant species frequently lost 29 mitochondrial DNA during thermal evolution , yet loss alone was insufficient to reproduce 30 the adaptive thermal phenotypes of evolved populations. Together, our results show that 31 adaptation to increasing temperature is driven by predictable changes in conserved regulatory 32 networks, while species-specific constraints shape divergent phenotypic outcomes . These 33 findings reveal both the predictability and contingency of evolutionary responses to rising 34 temperature across species. 35

Thermal adaptation , experimental evolution, molecular convergence , 36 mitochondrial genome loss, Saccharomyces 37 38 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 3

39 Thermal adaptation is a fundamental process that shapes the ecology, distribution, and 40 evolution of organisms 1,2. With global climate change driving more frequent and intense 41 thermal extremes, understanding how organisms adapt at the molecular level is cr ucial for 42 predicting their evolutionary trajectories and resilience3,4. 43 Species of the genus Saccharomyces provide a powerful model for studying thermal 44 adaptation, given their broad diversity in thermal tolerance and ecological niches. The genus 45 spans from cold-adapted species such as S. eubayanus, S. arboricola, and S. kudriavzevii to 46 warm-adapted species such as S. cerevisiae and S. paradoxus 5–8. This natural thermal 47 diversity, combined with experimental tractability in the laboratory , enables controlled 48 experimental evolution studies that allow direct observation of evolutionary changes over 49 hundreds of generation s9–11. Thermal performance curves (TPCs), which describe the 50 relationship between temperature and organismal performance, provide a quantitative 51 framework for linking evolutionary change to fitness -related traits and a common currency 52 for comparing adaptive responses across species12–15. Previous work has shown that different 53 Saccharomyces species evolve distinct TPCs when exposed to constant versus rising 54 temperatures5. Here, we describe the molecular basis of these changes. 55 We previously evolved populations of eight Saccharomyces species under constant (25 ºC) 56 and progressively increasing temperature regimes , ranging from 25 to 40 ºC, for up to 600 57 generations, to assess their evolutionary potential in adapting to future warming5. We found 58 that TPCs varied significantly between species, revealing two main trajectories: i) Warm -59 tolerant species showed an increase in both optimum growth temperature and thermal 60 tolerance, consistent with a “hotter is wider” evolutionary trajectory; ii) Cold-tolerant species 61 on the other hand evolved larger thermal breadth and higher thermal limits, but suffered from 62 reduced maximum performance overall, consistent with a generalist or “a jack of all 63 temperatures is a master of none” trajectory5. 64 Despite these well -characterized phenotypic trajectories, the molecular mechanisms 65 underlying thermal adaptation across species remain poorly understood. Thermal adaptation 66 may involve changes in regulatory pathways, protein stability, membrane composition, and 67 organellar function 7,16–21. Although growing evidence indicates that temperature is a 68 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 4 dominant selective force under ongoing climate warming 22, it remains unclear whether 69 adaptation to rising temperatures proceeds through predictable molecular solutions 70 (convergent evolution) or is instead constrained by species -specific physiological and 71 evolutionary histories (divergent evolution) 23,24. While phenotypic adaptation is often 72 repeatable24, growing evidence from experimental evolution and comparative genomic 73 studies suggests that the underlying genetic basis can be highly contingent, with convergence 74 frequently emerging at the level of molecular pathways rather than individual genes 23,25–27. 75 Consequently, we still lack an integrative understanding of how conserved cellular signaling 76 pathways, organelle function, and organism performance interact to shape evolutionary 77 trajectories under thermal stress. Especially mitochondrial function represents a putative yet 78 understudied mechanism of thermal adaptation, given its central role in energy production, 79 metabolic regulation, and stress responses, as well as its known sensitivity to temperature28,29. 80 Here, we use the budding yeast genus Saccharomyces as a comparative evolutionary model 81 to address these gaps in our understanding of the molecular and regulatory bas is of thermal 82 adaptation. By combining whole-genome sequencing across eight species , functional 83 analyses of conserved growth-stress signaling pathways, mitochondrial manipulations , and 84 thermal performance assays, we ask whether climate warming drives convergent molecular 85 evolution and how it manifests at the phenotypic and physiological levels . Our integrative 86 approach reveals that while increasing temperature repeatedly targets conserved signaling 87 hubs, adaptive responses are ultimately shaped by species -specific regulatory and energetic 88 constraints, with important implications for predicting evolutionary responses to climate 89 change. 90

91 Genetic changes associated with thermal evolution 92 To explore the genetic basis of TPCs evolution under increasing temperature conditions in 93 eight Saccharomyces species, we analyzed de novo single-nucleotide polymorphisms (SNPs) 94 across 256 evolved genotypes and their ancestral strains (Table S1 and S2). The 256 evolved 95 genotypes correspond to individual clones isolated at the end of experimental evolution, from 96 populations across eight Saccharomyces species, including between one to three strains per 97 species. Evolution was performed in four independent replicate populations per strain and 98 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 5 temperature regime (constant and increasing), and two individual clones (genotypes) were 99 sequenced per population. The genomes of e ach evolved genotype were compared to their 100 corresponding ancestral strain, allowing for the identification of de novo mutations that arose 101 during experimental evolution. Across all genomes, a total of 378 SNPs were identified, with 102 variation in the number and genomic location across species, strains, genotypes, and 103 temperature conditions (Figure 1 and S1, Table S2C). 104 105 Figure 1. Genetic changes occurring in experimental evolution in constant vs. increasing 106 temperatures in Saccharomyces. (A) Total number of de novo SNPs per species and 107 experimental evolution condition (constant vs. increasing temperature). (B) Distribution of 108 SNPs across coding and non-coding regions per species. (C) Parallel evolution across species 109 revealed by independent mutations in shared genes, displayed as an UpSet plot. Species are 110 grouped by thermal tolerance (blue = cold -tolerant, orange = intermediate, red = warm -111 tolerant). Set size bars show the number of mutations per species. Intersection size bars show 112 the number of mutations shared by two or more species. The intersection matrix shows which 113 species (lines) share mutations. 114 115 Genotypes evolved under increasing temperatures accumulated nearly twice as many de novo 116 SNPs as those evolved under constant temperature (248 vs. 130 SNPs; Figure 1A and S1A), 117 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 6 suggesting stronger selection pressures under increasing temperature regimes. SNPs were 118 detected in both coding and non-coding regions, but mutations in coding sequences were on 119 average three times more likely under both evolutionary conditions (Figure 1B and S1B). A 120 subset of these mutations was synonymous or located in regulatory regions, suggesting 121 possible functional consequences (Figure 1B and S1B). 122 To test whether mutational accumulation differed among species and temperature regimes, 123 we fitted a generalized linear mixed -effects model using the number of de novo SNPs per 124 evolved clone as the response variable, with species, evolutionary condition (constant vs. 125 increasing temperature), and their interaction as fixed effects, and strain as a random effect. 126 We detected significant effects of species (Wald test, p -value = 0.027) and evolutionary 127 condition (Wald test, p-value < 0.001) , as well as a strong species x condition interaction 128 (Wald test, p-value < 0.001, Table S2F), indicating that the increase in the number of SNPs 129 under thermal stress varies across species. These results suggest that evolution in increasing 130 temperatures leads to faster accumulation of de novo SNPs than in constant temperature, but 131 that the magnitude of genetic change is strongly modulated by the species’ genetic 132 background. 133 To determine whether these mutations affect biological processes or signaling pathways, we 134 performed GO term and KEGG pathway enrichment analyses . For genotypes evolved in 135 increasing temperature, we found enrichment for genes related to cellular processes, positive 136 regulation of cell growth, regulation of invasive growth in response to glucose limitation, and 137 regulation of filamentous growth (Table S2G). Genotypes evolved in increasing temperature 138 also showed an enriched K EGG pathway related to longevity regulation, involving the 139 TORC1, PKA, and MAPK signaling pathways (Table S2G). Meanwhile, for genotypes 140 evolved at constant temperature , we found enrichment for genes involved in the regulation 141 of biological processes and primary metabolic processes, such as adenylate cyclase-142 modulating G protein -coupled receptor signaling pathways , and Ras protein signal 143 transduction (Table S2G). 144 Notably, 13 genes repeatedly accumulated independent mutations across multiple strains and 145 species under increasing temperature conditions , including species with different natural 146 temperature tolerances (e.g., the warm-tolerant species S. cerevisiae and the cold -tolerant 147 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 7 species S. kudriavzevii). These included BMH1, SNF4, CYR1, and PRR2, which are involved 148 in cellular signaling pathways (e.g., TORC1 and MAPK). This indicates repeated targeting 149 of the same functional components across genetically and ecologically vastly different 150 species, suggesting convergent molecular evolution under thermal selection (Figure 1C). In 151 comparison, genotypes evolved under constant temperature showed only four genes with 152 independent mutations (shared by three species: S. cerevisiae, S. uvarum, and S. kudriavzevii; 153 Figure S1C). 154 Differential activation of signaling pathways 155 To evaluate the functional implications of the pathways detected by SNP enrichment analysis 156 (TORC1, PKA, MAPK), we measured the expression of reporte d canonical target genes 157 (TORC1: RPS6A, CRF1; PKA: SSA4, TPK1; MAPK: SLT2, RLM1)21,30–34 by qPCR in S. 158 cerevisiae (the most warm-tolerant species) and S. eubayanus (the most cold-tolerant 159 species)5, upon exposure to benign and high temperatures (Figure 2 and S2, Table S3A). 160 161 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 8 Figure 2. Functional validation of TORC1/PKA/MAPK outputs across species and 162 temperature conditions. (A) Log2 normalized gene expression for ancestral S. cerevisiae 163 (SC-A, red) and S. eubayanus (SE-A, blue) shown as the relative difference when exposed 164 to 25 ºC vs. the species-specific high temperature (40 ºC for SC; 34 ºC for SE). (B) Log2 165 normalized expression of evolved genotypes relative to their respective ancestral strains at 166 high temperature (SC 40 ºC, top; SE 34 ºC, bottom). EC: genotype evolved at a constant 25 167 ºC; EI: genotype evolved in increasing temperature. In (A) and (B), the dashed line marks no 168 change. Plotted values are the average of three biological replicates . Asterisks denote 169 significant differences in gene expression within species in (A) and evolved -ancestral 170 differences in (B) (Welch’s t-test, * p < 0.05, ** p < 0.01, *** p < 0.001, ns = non-significant). 171 (C) Heatmap of temperature effect s on gene expression for evolved genotype s relative to 172 their ancestors, computed as (expression at high temperature minus expression at 25 ºC in 173 EC or EI) minus ( expression at high temperature minus expression in the ancestor). Blue 174 indicates a smaller response to temperature (dampened plasticity); red indicates a greater 175 response (enhanced plasticity). Asterisks mark significant temperature x evol ved genotype 176 interaction terms from OLS models with HC3 robust standard errors. Genes are grouped by 177 pathway (TORC1: RPS6A, CRF1; PKA: SSA4, TPK1; MAPK: SLT2, RLM1). 178 179 First, we examined gene expression in the ancestral strains (before experimental evolution). 180 At 25 ºC , S. cerevisiae and S. eubayanus ancestors differed in several pathway readouts 181 (Figure S2A). S. cerevisiae showed higher expression than S. eubayanus in RPS6A, SLT2, 182 SSA4, and RLM1. S. eubayanus exceeded S. cerevisiae in CRF1 and TPK1 (Welch test, p-183 value < 0.05, Table S3B). This indicates species-specific baselines for TORC1/PKA/MAPK 184 output under benign temperature conditions. 185 When ancestral strains were exposed to their species-specific high temperature ( maximum 186 temperature reached in experimental evolution : 40 ºC for S. cerevisiae , 34 ºC for S. 187 eubayanus), S. cerevisiae showed robust induction of expression in CRF1, RLM1, SLT2, 188 SSA4, and TPK1 (log₂ ≈ +3.6, +3.7, +2.0, +5.5, +2.3; Welch tests, p < 0.05), but no significant 189 change in RPS6A. In contrast, S. eubayanus showed significant repression of CRF1 (log₂ ≈ 190 −2.9; Welch tests, p < 0.05) but no statistically significant expression differences in the other 191 genes (Figure 2A; Table S3C). A species x temperature interaction test (Ordinary Least 192 Squares (OLS) model with HC3 robust standard errors) confirmed that the magnitude of the 193 heat response (high temperature vs. 25 ºC) differed between species for multiple genes 194 (CRF1, SSA4, TPK1, RPS6A, p < 0.05, Table S3C). These patterns indicate a stronger 195 transcriptional heat response in the warm -tolerant species (S. cerevisiae) and a limited or 196 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 9 suppressed response in the cold -tolerant species (S. eubayanus ), consistent with their 197 contrasting thermal ecologies. 198 We next asked whether evolved genotypes (isolated from the endpoint of experimental 199 evolution) diverged transcriptionally from their ancestors when assayed at high temperature 200 (Figure 2B, Table S3D). For this, w e compared genotypes evolved at constant (EC) and 201 increasing temperature (EI) to the ir respective ancestors. In S. cerevisiae , EI genotypes 202 showed consistent down -regulation of gene expression relative to their ancestor across all 203 genes at 40 °C (SSA4: log₂ = -5.22; RPS6A: -3.32; CRF1: -3.89; SLT2: -2.03; TPK1: -2.30; 204 Welch tests, p < 0.05) except for RLM1, where gene expression increased (log₂ = +1.37, p = 205 0.026, Figure 2B, Table S3D). In contrast, expression in S. eubayanus EI genotypes was 206 strongly up-regulated relative to the ancestor for all genes at 34 °C , with log₂ normalized 207 expression from +6.69 to +11.26 across RLM1, SLT2, TPK1, SSA4, CRF1, and RPS6A (all p 208 ≤ 0.007, Figure 2C, Table S3D). Notably, we did not observe any differences between the 209 constant-evolved genotypes ( EC) and their ancestors when assayed at high temperature 210 (Table S3D, Welch test, p > 0.05). This demonstrates that the major functional shift at high 211 temperature is specific to EI genotypes. Together with the enrichment of mutations in 212 TORC1/PKA/MAPK ( Figure 1 ), these results suggest that evolution under increasing 213 temperature elicits a stronger and directionally opposite remodeling of central signaling 214 pathway activity in warm- vs. cold-tolerant yeast species, with repression in S. cerevisiae and 215 hyperactivation in S. eubayanus. 216 We also measured the transcriptional response of the evolved genotypes (EI and EC) at 25 217 ºC ( Figure S2B , Table S3E). S. cerevisiae EI genotype s did not show any expression 218 differences to their ancestor, whereas S. eubayanus EI genotype s showed widespread 219 constitutive up-regulation (Welch test, p < 0.05). Thus, in S. eubayanus, the evolved shift 220 was observable already at 25 ºC, whereas in S. cerevisiae this manifested only at higher 221 temperatures. 222 For each type of genotype (ancestral, EC, EI), we calculated the temperature effect as the 223 change in expression between the high temperature and 25 ºC (Figure S2C). We then tested 224 whether evolution altered thermal regulatory plasticity by comparing the temperature effect 225 found in evolved genotypes with that found in their ancestral strains . Specifically, we 226 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 10 subtracted the expression change in the ancestor from the change observed in EC and EI 227 genotypes (Figure 2C ). This metric is positive when evolved genotypes show enhanced 228 plasticity relative to their ancestors, and negative when they show reduced or dampened 229 plasticity (Figure 2C and S2C, Table S3F). In S. cerevisiae, plasticity was dampened after 230 evolution: EC genotypes showed a smaller change in expression between high temperature 231 and 25 ºC for five genes ( RPS6A, CRF1, SSA4, TPK1, RLM1, p < 0.05), and EI genotypes 232 showed a smaller change for SSA4 and SLT2 (p < 0.05). In contrast, in S. eubayanus, the 233 change in expression was enhanced, but not significantly different from the ancestral strain 234 in both genotypes (EI and EC) , indicating that major gene expression changes in S. 235 eubayanus are constitutive rather than temperature-dependent (elevated baseline expression 236 at both temperatures) . Together, these analyses show that increasing temperature evolution 237 drives opposite regulatory responses in the two species: reduced output and plasticity in S. 238 cerevisiae versus elevated, constitutive output in S. eubayanus. These functional patterns are 239 congruent with the genomic enrichments in TORC1/PKA/MAPK and support a model in 240 which increasing temperature evolution reprograms central signaling differently in warm- vs. 241 cold-tolerant Saccharomyces species. 242 Copy number variation associated with thermal evolution 243 We next evaluated copy number variation (CNV) across all 256 evolved genotypes from all 244 eight species to identify structural genomic changes associated with thermal adaptation 245 (Figure S3 and Table S4). After filtering out telomeric regions and retaining only significant 246 events (|log2| ³ 0.5, FDR < 0.05), we detected only a small number of condition -exclusive 247 CNVs across all species (21 and 22 for increasing and constant temperature conditions, 248 respectively, Table S4C), indicating that large genomic rearrangements were rare during 249 thermal evolution. 250 The amplitude of CNV events was generally modest, with median |log2| values around 0.7-251 1.0 in both evolutionary regimes ( Figure S3A), suggesting moderate copy number shifts 252 rather than whole-chromosome aneuploidies35,36. CNV amplitudes did not differ significantly 253 between constant and increasing temperature conditions (Wilcoxon rank-sum test, p = 0.33), 254 and gains were more frequent than losses across most species in both conditions ( Figure 255 S3B). 256 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 11 The total number of CNV events varied among species and temperature regimes (Figure 257 S3C). Under increasing temperature conditions, CNVs were scattered across multiple 258 chromosomes, with a modest recurrence of events in chromosome XV in S. kudriavzevii, S. 259 mikatae, and S. cerevisiae (Table S4C ). Under constant temperature, CNVs were 260 predominantly located on chromosomes X and XII, with both genetically different S. 261 eubayanus strains sharing a CNV on chromosome X, suggesting a species -specific 262 recurrence. Interestingly, chromosome IX contained CNVs in both temperature conditions 263 but in different species (Table S4C), pointing at independent structural changes rather than 264 convergent evolutionary events. 265 Overall, CNVs were generally rare and species -specific, with no evidence of recurrent, 266 convergent events across species, suggesting that thermal evolution in Saccharomyces is 267 primarily driven by point mutations and regulatory changes rather than by large-scale copy-268 number variation, within the timeframe of this evolution experiment. 269 Loss of mitochondrial genome as an adaptive strategy under thermal stress 270 We detected a consistent loss of mitochondrial DNA (mtDNA) in genotypes evolved under 271 increasing temperature conditions ( Figure 3A and S4A, Table S4D ). Clear differences 272 emerged between species with different thermal tolerances: all three cold-tolerant species (S. 273 eubayanus, S. kudriavzevii , and S. arboricola ) consistently showed a complete loss of 274 mtDNA under increasing temperature conditions in all 31 genotypes we sequenced . In 275 contrast, species with intermediate thermal tolerance (S. uvarum and S. jurei) showed mixed 276 results, with some genotypes retaining and others losing their mtDNA (all genotypes of the 277 strains Su-1 and Sj -1 lost mtDNA, only 72% of Sj -2 genotypes lost mtDNA, and all Su -2 278 genotypes retained mtDNA ( Figure 3A and S4A, Table S4E). Warm-tolerant species ( S. 279 mikatae, S. paradoxus, and S. cerevisiae) also showed a mixed pattern, with all Sm-2 and Sp-280 2 genotypes completely los ing their mtDNA, and 88% Sp-1, 15% Sc-1, and 57% Sc -3 281 genotypes losing their mtDNA. On the other hand, all Sm-1 and Sc-2 genotypes retained their 282 mtDNA. Notably, u nder constant -temperature conditions, mitochondrial loss was rare, 283 observed only in the Sk-1 strain and in a single Sp-2 genotype. Together, these results reveal 284 a strong association between thermal regime, species-specific thermal tolerance, and mtDNA 285 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 12 stability, with cold-tolerant species showing a markedly higher propensity for mtDNA loss 286 when adapting to increasing temperature. 287 288 Figure 3. The effect of mtDNA loss on t hermal performance across Saccharomyces 289 species. (A) Frequency of mitochondrial genome loss (petite formation) across 290 Saccharomyces species grouped by thermal tolerance , according to Molinet & Stelkens 291 (2024). Cold -tolerant species: S. eubayanus , S. arboricola , S. kudriavzevii ; Intermediate 292 species: S. uvarum and S. jurei ; warm -tolerant species: S. mikatae , S. paradoxus , S. 293 cerevisiae. (B) Differences in maximum performance (Pmax) and upper thermal limit (CTmax) 294 between genotypes (Evolved vs Ancestral, Petit e vs Ancestral). Violin plots show the 295 distribution of the difference in P max and CT max across strains, with internal boxplots 296 indicating medians and interquartile ranges. The dashed line indicates the performance of 297 ancestral strains , set at 1. Asterisks indicate significant differences between evolved and 298 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 13 petite strains relative to their ancestors (Wilcoxon test; * p < 0.05, ** p < 0.001). (C) Heatmap 299 showing the relative effects of mtDNA loss on thermal parameters across strains. Values 300 represent the ratio between petite and ancestral strains for each parameter (petite/ancestral) . 301 Values 1 indicate an increase relative to the ancestors. 302 Hierarchical clustering was applied to strains based on Euclidean distance of scaled values 303 to highlight similarities in response patterns. Asterisks denote significant differences between 304 petite and ancestral strains based on non -overlapping confidence intervals. Color scale 305 indicates the magnitude and direction of change. 306 307 To evaluate the role of mitochondria in thermal tolerance and to test whether mtDNA loss or 308 gain alters the shape of thermal performance curves (TPCs) , we generated mitochondrial -309 deficient (petit e) strains from each ancest or and compared their growth performance 310 (maximum growth rate, µmax) with ancestral and evolved genotypes. All strain types 311 (ancestral, evolved, and petite) were grown at 11 temperatures ranging from 16 to 40 ºC 312 (Table S5A). We then fitted TPCs for each strain type (ancestral, evolved, and petite) within 313 each strain, plotting the µmax against temperature (Figure S5). We first tested whether growth 314 performance differed among strain types (ancestral, evolved, and petit e) across the whole 315 temperature range using a linear -mixed-effects model that included temperature as a spline 316 term to capture the shape of the TPC. The model revealed a strong effect of temperature on 317 growth rate (F3,2055 =339.0, p < 0.001) and a significant main effect of strain type (F2,2055 = 318 6.30, p = 0.0019), indicating overall differences in thermal performance among ancestral, 319 evolved, and petite strains. The strain type x temperature interaction was marginal (F6,2055 = 320 1.06, p = 0.068), suggesting similar curve shapes ( Table S5B ). Pairwise post hoc 321 comparisons showed that these differences were most pronounced at higher temperatures 322 (>25 ºC; Table S5C). 323 Then, we applied a linear model to each strain separately to determine whether the observed 324 differences were general across the genus or strain-specific. Significant effects of strain type 325 were detected in several cold - and intermediate-tolerant strains (e.g. S. eubayanus Se-1, S. 326 uvarum Su-1, S. kudriavzevii Sk-2, and S. jurei Sj-1), indicating altered growth performance 327 in mtDNA-depleted strains (Table S5D and S5E). However, other cold- and intermediate-328 tolerant strains (S. eubayanus Se-2, S. kudriavzevii Sk-1, and S. uvarum Su-2), together with 329 all strains of the warm-tolerant species ( S. cerevisiae and S. paradoxus ), showed no 330 significant performance differences between ancestral and petit strains, suggesting a limited 331 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 14 impact of mitochondrial function on thermal performance ( Table S5D and S5E). These 332

suggest that mitochondrial function affects the shape of TPCs, but that this is strongly 333 strain-dependent. 334 Analyzing individual thermal performance parameters (T opt: optimum temperature, CT max: 335 critical thermal maximum, CT min: critical thermal minimum, P max: maximum performance, 336 Tbr: thermal breadth) across all strains and species showed a significant decrease in petite 337 strains of 10.4% in Pmax and 1.8% in CTmax, relative to their ancestors (Wilcox test, p-value 338 = 5.00x10-3 and 0.018, respectively , Figures 3B and S4B, Tables S5F and S5G), but no 339 significant differences in T opt, CT min, or Tbr. This suggests that, in general, mtDNA loss 340 constrains both maximal performance and upper thermal limits. This dissociation indicates 341 that the adaptive thermal phenotypes of evolved genotypes require additional regulatory and 342 metabolic reprogramming beyond the loss of respiratory function. 343 However, global comparisons can mask strain-specific responses. We therefore evaluated the 344 effects of mtDNA loss on thermal parameters also at the strain level (Figure 3C, Table S5H). 345 Here, the effects of mtDNA loss were highly strain-dependent, revealing four distinct 346 response groups. First, three warm-tolerant petite strains (Sc-2, Sc -3, Sp -1) showed no 347 changes in any thermal parameters relative to their ancest ors, indicating that mitochondrial 348 function does not measurably influence TPC shape or thermal limits in these genetic 349 backgrounds. Second, one warm-tolerant and two intermediate-tolerant petite strains (Sm-1, 350 Su-1, and Sj -1, respectively) displayed a consistent pattern of reduced performance and 351 reduced thermal tolerance, characterized by simultaneous decreases in Pmax, CTmax, and Topt. 352 This pattern indicates that, in these backgrounds, mtDNA loss not only lowers maximal 353 growth performance but also shifts the TPC toward lower temperatures and reduces heat 354 tolerance. Third, Se-1 represented a distinct case in which mtDNA loss in the ancestral strain 355 was associated with a partial shift toward higher thermal tolerance despite a cost in 356 performance: Pmax decreased, while both CTmax and Tbr increased. In this strain, mtDNA loss 357 resulted in an expansion of the thermal range and an elevated upper thermal limit, 358 accompanied by reduced maximal performance (Figure S5 ). Thus, Se -1 petite ancestral 359 strains showed a phenotype similar to that of evolved genotypes, i.e., consistent with the 360 generalist or “a jack of all temperatures is a master of none” pattern5. Fourth, mtDNA loss in 361 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 15 Sc-1 and Sp -2 was characterized by an increase in T opt coupled with a decrease in T br, 362 consistent with a more specialized TPC (a higher optimum but narrower thermal breadth) 363 rather than a broader tolerance to high temperatures. 364 Together, these results demonstrate that mitochondrial function plays an important, but 365 highly strain -dependent role in shaping thermal performance. While the loss of mtDNA 366 consistently reduced overall performance (Pmax), its impact on thermal optima, upper thermal 367 limits, and thermal breadth varied substantially across strains. Thus, mtDNA loss alone does 368 not explain the evolutionary shifts in thermal tolerance across all species and strains. 369

370 Temperature is a key factor influencing cellular physiology, ecological niches, and species 371 distributions12. Using a comparative framework across eight Saccharomyces species 372 spanning the genus’s full thermal niche , we show that thermal adaptation to sustained 373 temperature increase is characterized by a dual pattern: convergent selection on conserved 374 molecular targets coupled with divergent regulatory and physiological outcomes. By 375 integrating experimental evolution, genome -wide analyses, functional validation, 376 mitochondrial manipulations, and thermal performance assays, our study reveals how climate 377 warming can drive predictable molecular responses while simultaneously amplifying natural 378 species- and strain-specific constraints. 379 Convergent molecular targets under thermal selection 380 Our experimental evolution regime under increasing temperature repeatedly targeted 381 conserved signaling pathways regulating growth, metabolism, and stress responses, 382 particularly TORC1, PKA, and MAPK (Figure 4A). Although evidence for parallel evolution 383 at the molecular level in nature is relatively rare (but see, e.g., 37–39), microbial experimental 384 evolution studies often report the independent emergence of mutations in the same gene, 385 protein complex, or regulatory pathway 11,25,40–44. However, this pattern has largely been 386 limited to experiments initiated from clonal populations. Here, we show that this convergence 387 extends across species with contrasting thermal niches: across the Saccharomyces genus, 388 thermal selection consistently targets central regulatory hubs rather than primarily affecting 389 temperature-specific enzymes. This reveals cross -species predictability at the level of 390 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 16 conserved regulatory pathways without strict gene -level identity, highlighting shared 391 molecular constraints on thermal adaptation across the Saccharomyces genus. The repeated 392 targeting of the same signaling hubs across species, despite distinct genomic backgrounds, 393 suggests non-random selection on conserved regulatory architectures rather than mutation -394 rate-driven coincidence16,45. 395 396 Figure 4. Conserved signaling pathways and mitochondrial feedback shape divergent 397 thermal adaptation trajectories in Saccharomyces. (A) Schematic representation of 398 conserved growth -stress sign aling pathways recurrently targeted during experimental 399 evolution under increasing temperatures. Independent lineages of Saccharomyces species 400 accumulated mutations in key components of the TORC1, PKA, and MAPK signaling 401 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 17 networks, suggesting these pathways are central regulatory hubs under thermal selection. 402 Only major nodes relevant to repeatedly mutated genes and pathway integration are shown. 403 These conserved pathways coordinate cellular processes, including growth, stress responses 404 (STRE), autophagy, ribosome biogenesis, and cell wall integrity, illustrating how thermal 405 evolution repeatedly targets shared molecular architectures rather than temperature-specific 406 enzymes. (B) Conceptual model illustrating the interaction between mitochondrial function 407 and conserved signaling pathways during thermal adaptation. Mitochondrial dysfunction, 408 including mitochondrial genome loss, alters cellular energy status, membrane potential, and 409 redox balance, generating signals (e.g., changes in ATP levels and reactive oxygen species) 410 that feed back into central regulatory circuits such as TORC1 and RAS -PKA signaling. 411 Retrograde signaling components (e.g., Rtg1/3 and Rtg2) further link mitochondrial status to 412 nuclear gene expression. These interactions modulate, but do not solely determine, adaptive 413 outcomes, contributing to strain - and species-specific differences in thermal performance. 414 Genes with mutations identified in evolved genotypes are shown in red. The mitochondrial 415 image in B was generated with AI. Adapted from 16,33,46. 416 417 Recurrent mutations in key regulatory genes such as CDC25, CYR1, BMH1, and SNF4 418 highlight the importance of nodes that integrate nutrient sensing, growth, and stress signaling. 419 These genes occupy central positions within their respective pathways and are well-suited to 420 mediate trade -offs between proliferation and stress resistance 33,46–48. Importantly, many 421 strain-specific mutations not shared across species mapped to the same signaling networks, 422 suggesting that thermal adaptation can proceed through multiple genetic routes that converge 423 at the level of pathway output. Together, these patterns support a model in which thermal 424 evolution is constrained by the architecture of conserved signaling systems, favoring repeated 425 modification of the same regulatory modules across independent lineages. 426 Divergent regulatory outcomes despite molecular convergence 427 While genomic analyses revealed strong convergent evolutionary patterns in TORC1, PKA, 428 and MAPK pathways, our functional validation demonstrated that this convergence does not 429 translate into uniform regulatory outcomes. Instead, when assayed near the thermal limits of 430 warm- and cold -tolerant species , evolution under increasing temperature reprogrammed 431 pathway activity in opposite directions , depending on the species background. Ancestral 432 genotypes (i.e., before experimental evolution) of warm- and cold-tolerant species already 433 differed markedly in baseline signaling output and in their transcriptional responses to heat 434 stress, reflecting distinct regulatory architectures shaped by their thermal adaptation. 435 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 18 These differences in pathway activity were further accentuated by experimental evolution. In 436 the warm -tolerant species S. cerevisiae, genotypes evolved under increasing temperature 437 showed reduced pathway output and dampened transcriptional plasticity at high temperature, 438 consistent with regulatory buffering of an already heat-adapted organism. In contrast, in the 439 cold-tolerant species S. eubayanus , evolved genotypes showed strong , often constitutive 440 upregulation of the same pathway targets, indicating a shift toward sustained stress signaling 441 rather than plastic regulation. Thus, although thermal selection repeatedly targets the same 442 signaling hubs, adaptive solutions exploit different regions of the regulatory landscape 443 depending on species -specific constraints. This functional divergence despite shared 444 molecular targets underscores the importance of genetic background in shaping adaptive 445 trajectories. Rather than converging on a single optimal signaling state, thermal adaptation 446 appears to involve context -dependent rewiring of conserved pathways, reflecting different 447 balances between growth and stress resistance. 448 Mitochondrial function as a context-dependent constraint for thermal adaptation 449 A second major axis of divergence in our study involves mitochondrial function. Loss of 450 mtDNA occurred frequently in populations evolved under increasing temperature and was 451 universal in cold-tolerant species. Similar patterns of mtDNA instability or loss under thermal 452 stress have been reported in previous experimental evolution studies in yeast, suggesting that 453 thermal stress can destabilize mitochondrial maintenance or select against respiratory 454 metabolism16,19. Mitochondria play a central role in energy production, redox balance, and 455 stress signaling49, and high temperatures are known to challenge mitochondrial integrity by 456 increasing membrane fluidity, disrupting electron transport, and elevating reactive oxygen 457 species50–52. Consistent with this, multiple studies have shown that mitochondrial genotype 458 and function strongly modulate thermal tolerance 19,29,53–56. In cold -tolerant species, 459 mitochondrial dysfunction may represent a recurrent, possibly adaptive, response to 460 increasing temperature stress . Loss of respiration may reduce oxidative stress or mitigate 461 metabolic imbalance at elevated temperatures , as has been proposed in other contexts of 462 thermal or oxidative stress53. 463 Despite prevailing mtDNA loss, our functional assays demonstrate that mitochondrial 464 deficiency alone is insufficient to cause significant fitness gains or adaptive shifts in thermal 465 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 19 tolerance. Across species, loss of mtDNA in ancestral strains consistently reduced maximal 466 growth performance, indicating a substantial energetic cost. Moreover, our analyses revealed 467 that petite strains exhibited, on average, a reduction in Ctmax, rather than an increase, relative 468 to their ancestral counterparts, arguing against a simple or universally adaptive role for 469 mtDNA loss during thermal evolution. These results are in line with recent work showing 470 that mitochondrial perturbations can constrain thermal tolerance by limiting bioenergetic 471 capacity or disrupting mitochondrial-nuclear coordination57. 472 Importantly, effects of mtDNA loss on thermal parameters w ere highly strain -dependent: 473 Only a single cold-tolerant petite strain (Se-1) showed partial phenotypic convergence toward 474 the heat tolerance of the genotypes evolved in increasing temperature, resembling the 475 characteristic generalist profile or “a jack of all temperatures is a master of none” phenotype. 476 A subset of warm -tolerant strains (Sc -2, Sc-3, and Sp -1) showed no detectable changes in 477 any thermal parameter following mtDNA loss, while other warm-tolerant strains (Sc-1 and 478 Sp-2) showed a thermal specialization rather than a “hotter is wider” phenotype. 479 These results indicate that mitochondria act as modulators rather than primary drivers of 480 thermal adaptation. Although mtDNA loss is a recurrent outcome of evolution under 481 increasing temperature in yeast 16,19, its phenotypic consequences are highly context -482 dependent, ranging from neutral to deleterious, and only rarely mimicking the adaptive shifts 483 observed after experimental evolution. Differences in mitochondrial –nuclear coordination, 484 redox homeostasis, or retrograde signaling likely determine whether mitochondrial 485 dysfunction alleviates or exacerbates thermal stress 58. Notably, several of the conserved 486 signaling pathways repeatedly targeted by thermal selection are known to sense 487 mitochondrial status and metabolic flux46, suggesting that mitochondrial function feeds back 488 into the same regulatory circuits that shape adaptive thermal responses. 489 Implications for predicting evolutionary responses to climate warming 490 Together, our findings support a multi-layered model of thermal adaptation in which climate 491 warming imposes shared molecular constraints while generating divergent physiological 492 outcomes. Increasing temperature consistently selects for changes in conserved signaling 493 hubs, but the adaptive consequences of these changes depend on species-specific regulatory 494 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 20 architectures and energetic constraints. As a result, similar molecular signatures can give rise 495 to distinct thermal performance curves and adaptive strategies, even within a single genus. 496 From an ecological perspective, our results highlight the challenges of predicting 497 evolutionary responses to climate warming. Adaptive potential cannot be inferred from 498 molecular changes alone, nor from species-level classifications, but instead emerges from the 499 interaction between conserved stress -response pathways, organelle function, and genetic 500 background. The frequent loss of mtDNA under thermal stress further underscores the role 501 of energetic trade-offs in shaping evolutionary trajectories, particularly in environments with 502 recurrent high temperatures. 503 More broadly, our study suggests that climate change is likely to promote both convergent 504 molecular evolution and divergent ecological outcomes, reshaping microbial diversity 505 through a mosaic of adaptive solutions. Integrating genomic, functional, and physiological 506 approaches will therefore be essential for understanding and predicting how organisms 507 respond to ongoing global warming. 508

509 Strains 510 We used strains from a n experimental evolution study that exposed populations of eight 511 Saccharomyces species (S. cerevisiae, S. paradoxus, S. mikatae, S. jurei, S. kudriavzevii, S. 512 arboricola, S. uvarum, and S. eubayanus) to gradually increasing temperature (25-40 ºC) and 513 constant temperature (25 ºC) for up to 600 generations 5. We selected two genotypes from 514 each replicate evolution line per strain and temperature condition. We used four replicate 515 lines per strain from a total of 16 strains (2 per species), and the two temperature conditions, 516 yielding a total of 256 genotypes. These genotypes, together with their respective ancestral 517 strains, were sequenced for whole -genome analysis ( Table S1). Ancestral strains represent 518 the starting point before experimental evolution (16 strains), evolved genotypes under 519 constant temperature ( EC) were used as thermal control (128 genotypes), and evolved 520 genotypes under increasing temperature ( EI) represent heat -adapted lineages (128 521 genotypes). All strains were maintained on YPD agar (1% yeast extract, 2% peptone, 2% 522 glucose, and 2% agar) and stored at -70 ºC in 20% glycerol stocks. 523 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 21 Whole genome analysis 524 Genomic DNA was obtained for whole -genome sequencing using the Thermo Scientific 525 KingFisherTM Duo Prime Purification system, and sequence d at the Earlham Genomics 526 facilities ( https://www.earlham.ac.uk/) using the LITE pipeline 59. The quality of the raw 527 sequencing reads was assessed using FastQC 0.12.1 528 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and MultiQC 1.22.260, both 529 before and after trimming and quality filtering using fastp 0.23.461. Reads were aligned 530 against reference genomes of each species (Table S2A) using BWA 0.7.1862. BAM files were 531 sorted using SAMtools 1.2063. Duplicate reads were removed using Picard 2.27.5 532 (https://broadinstitute.github.io/picard/), and mapping quality was assessed using QualiMap 533 2.2.164 (Table S2B). 534 SNPs and CNV analysis 535 Variant calling and filtering w ere performed using GATK version 4.3.0.065. Variants were 536 called per sample using HaplotypeCaller (default settings), generating g.vcf files. Variant 537 databases were constructed using GenomicsDBImport, and genotypes were called using 538 GenotypeGVCFs with the -G Standard Annotation option. SNPs and InDels were extracted 539 and filtered out separately using SelectVariants. We then applied recommended filters with 540 the following options: QD 60.0, MQ 4.0, MQRankSum < -12.5, 541 ReadPosRankSum < -8.0. This vcf file was further filtered by removing missing data using 542 the option -max-missing 1, filtering out sites with a coverage below the 5th or above the 95th 543 coverage sample percentile using the options -min-meanDP and -max-meanDP, and a 544 minimum site quality of 30 (-minQ 30) in VCFtools 0.1.1666. Sites with a mappability score 545 less than 1 , as calculated by GenMap 1.2.0 67, were filtered using bedtools 2.18 68. As an 546 additional filtering step, the ancestral and evolved files were intersected using BCFtools 547 1.2069, and variants with shared positions were extracted from the vcf files of the evolved 548 genotypes. Annotation and effect prediction of the variants were performed with SnpEff 70. 549 To enable cross-species comparison, gene identities were assigned based on species-specific 550 genome annotations when available, or inferred by BLAST -based homology against the S. 551 cerevisiae reference genome. Gene ontology analysis was performed using the tools provided 552 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 22 by the DA VID Bioinformatic s Resource71, selecting categories with a significant 553 overrepresentation using an FDR < 0.05. 554 CNVs were called using CNVkit (whole-genome mode)36,72 from the Illumina WGS data of 555 all ancestral and evolved genotypes within each species/strain background. To ensure 556 identical binning across samples, including the mitochondrial chromosome, we first created 557 an accessibility BED file from the species reference (FASTA files). Next, we generated a 558 single whole-genome bin set per species using 1 -kb tiles and used the same BED file and 559

for every sample of that species. For each BAM file, we computed coverage in the 560 1-kb bins, normalized to the species reference and ancestral strain, and segmented into 561 contiguous genomic regions with similar normalized log 2 copy-ratio profiles . For each 562 sample, we computed per-segment confidence intervals and t-tests, and merged these metrics 563 back with the segments. Segments were kept as significant if all of the following held: the 564 95% CI on log2 fold-change did not cross zero, Benjamini-Hochberg FDR on the t-test < 0.05 565 (within sample), did not correspond to telomeric regions, and the effect size was biologically 566 meaningful for a diploid genome (amplitude thresholds corresponding to ± 1 copy: gain if 567 log2 ³ +0.50, loss if log 2 £ -1.00; +2 copies, when log 2 ³ +1.00). For gains and losses 568 separately, we reduced all significant segments into consensus intervals and computed a 569 sample-by-region presence matrix, calling a region “present” in a sample if ³ 50% of the 570 consensus interval overlapped that sample’s significant segment. Regions were then 571 classified as exclusive to the evolved under increasing temperature genotypes (present in the 572 ³1 evolved genotype, absent in the evolved genotypes under constant temperature) or 573 exclusive to the evolved under constant genotypes (present in the ³1 evolved genotype, 574 absent in the evolved genotypes under increasing temperature). 575 To quantify mtDNA copy change per sample, we used the corrected bin -level files (.cnr). 576 After excluding bins with zero depth and extreme outliers, we compute d the median log 2 577 ratio across mtDNA bins and subtracted the median across nuclear DNA bins from the same 578 sample, yielding a relative mtDNA log2 (mt – nuclear). Samples were heuristically classified 579 as: mtDNA loss when the relative mtDNA log2 was £ -1.0 (³ 2x lower than nuclear), mtDNA 580 depletion when the relative mtDNA log 2 was between -1.0 and -0.6, mtDNA normal -like 581 when the relative mtDNA log2 was between -0.6 and +0.4, and mtDNA enrichment candidate 582 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 23 when the relative mtDNA log 2 was ³ +0.4. mtDNA loss was visual ly verified in the 583 Integrative Genomics Viewer (IGV) software , and by growing the genotypes on YPG agar 584 (1% yeast extract, 2% peptone, 2% glycerol, 2% agar) , where growth requires a functional 585 mitochondrial respiratory chain and thus fails in mitochondrial genome -deficient (petite) 586 genotypes. 587 Generation of rho0 strains 588 Strains lacking mtDNA (rho 0) were generated by treating ancestral strains (rho +) with 589 ethidium bromide73. Briefly, cells were grown overnight in YPD at 25 ºC. Then, 100 µL of 590 the culture was diluted in 1 mL of YPD and incubated at 25 ºC until it reached the exponential 591 growth phase (OD600nm 0.4-0.6). Cells were washed and resuspended in 0.1 M potassium 592 phosphate buffer at a concentration of 106 cells/mL. Ethidium bromide was added to a final 593 concentration of 10 µg/mL and incubated for 8 h. After incubation, 200 μL of the cell culture 594 was plated on YPD agar and incubated at 25 °C for 3-4 days. The mtDNA loss was evaluated 595 by growing the colonies in YPG agar and by standard colony PCR using primers previously 596 described for mtDNA sequences29 (Table S3). Because of the mutagenic nature of ethidium 597 bromide, we controlled for the potential effect of spurious nuclear mutations by using three 598 rho0 colonies (replicates) from each ancestral strain (Table S1D). We did not obtain rho 0 599 colonies for S. mikatae yHAB336 (Sm -2), S. jurei NCYC3947 (Sj -2), and S. arboricola 600 ZP960 (Sa-1), despite repeated attempts using the ethidium bromide protocol, suggesting 601 strain-specific resistance to mitochondrial genome loss under these conditions. 602 Thermal performance curves analysis 603 Strains were phenotypically characterized under microculture conditions as previously 604 described5. Mitotic growth was measured in 96 -well plates at 11 different temperatures: 16, 605 18, 20, 23, 25, 28, 31, 32, 34, 37, and 40 °C. Inocula were prepared by growing one colony 606 of each strain for 24 h at 25 °C in 96 -well plates with 200 μL YPD in each well. Cultures 607 were then diluted to an initial OD600nm of 0.1 in fresh YPD for growth at the 11 temperatures 608 for 24-48 h until all strains had entered the stationary phase. The next day, these cultures were 609 used to inoculate a new 96-well plate with 200 μL of YPD per well, at an initial OD600 nm of 610 0.1, and kinetic growth parameters were measured. Growth curves were obtained by 611 measuring OD600nm every 30 min in a TECAN Sunrise instrument at all 11 temperatures for 612 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 24 24-48 h , until all strains had entered the stationary phase. Three independent OD 613 measurements were taken per strain (i.e. technical triplicates). Maximum growth rate ( μmax) 614 was determined as previously described 74. For this, μmax was calculated following a 615 smoothing procedure on ln -transformed OD -values and using the discrete derivative, as 616 previously described75 in R version 4.3.1. 617 We fitted thermal performance curves to μmax of each strain using the cardinal temperature 618 model with inflection (CTMI) as previously described6, using equation 1. 619 620 𝑃 = 0 𝑖𝑓 𝑇 ≤ 𝐶𝑇!"# 𝑜𝑟 𝑇 ≥ 𝐶𝑇!$% 621 𝑃 = 𝑃&'( × .𝐷 𝐸1 2 𝑖𝑓 𝐶𝑇!"# < 𝑇 < 𝐶𝑇!$% 622 𝐷 = (𝑇 − 𝐶𝑇!$% ) × (𝑇 − 𝐶𝑇!"# )) (1) 623 𝐸 = .𝑇&'( − 𝐶𝑇!"# 2624 × 7.𝑇&'( − 𝐶𝑇!"#2 × .𝑇 − 𝑇&'( 2625 − .𝑇&'( − 𝐶𝑇!$% 2 × .𝑇&'( + 𝐶𝑇!"# − 2𝑇2: 626 627 Where: 628 CTmax is the temperature above which no growth occurs. 629 CTmin is the temperature below which no growth occurs. 630 Topt is the temperature at which Pmax equals its optimal value (Popt). 631 632 CTMI parameters were estimated using nonlinear regression in R version 4.3.1. Thermal 633 tolerance and T br were obtained using the calc_params function from the rTPC package 76, 634 where thermal tolerance corresponds to CTmax minus CTmin, and thermal breadth (Tbr) is the 635 temperature range across which performance is above 80% of optimal. The adequacy of fit 636 of TPCs was checked by the proportion of variance explained by the model (R 2) and by the 637 residual sum of squares (R SS). Relative thermal param eters were obtained by normalizing 638 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 25 each parameter in the petite and evolved strains to their corresponding ancestral value s and 639 used to construct a heatmap with hierarchical clustering. Hierarchical clustering was 640 performed on strains using Euclidean distance and complete linkage. Heatmap visualization 641 was generated using the pheatmap package. We considered thermal parameters to differ 642 significantly among strains if their 95% confidence intervals did not overlap. 643 RNA extraction and qPCR assay 644 Gene expression of target genes in the TORC1 (RPS6A, CRF1), PKA (SSA4, TPK1), and 645 MAPK (SLT2, RLM1) signaling pathways was evaluated for ancestral strains, and genotypes 646 evolved under increasing and constant temperatures in the species S. cerevisiae and S. 647 eubayanus (strains Y12 (Sc -1) and CBS12357 (Se -1), respectively ). Gene expression 648 analysis was performed by qPCR from exponential cultures grown at 25 ºC for both species, 649 34 ºC for S. eubayanus strains, and 40 ºC for S. cerevisiae strains. These temperatures 650 correspond to the maximum reached during the experimental evolution for each strain. This 651 design intentionally probed each species near its upper thermal limit, allowing comparison 652 of regulatory responses under ecologically relevant heat stress. Cells were first grown in 5 653 mL of YPD for 24 h at 25 ºC and then diluted to an initial OD 600 nm of 0.1 in fresh YPD for 654 growth at the three different temperatures of interest, until they reached the exponential phase 655 (OD600 nm of ~0.6). Cells were collected, immediately frozen in liquid nitrogen, and stored at 656 -70 ºC until RNA extraction. RNA was extracted using the YeaStar TM RNA Kit (Zymo 657 Research) according to the manufacturer’s instructions. Then, genomic DNA traces were 658 removed by treating samples with DNase I (Zymo Research), and total RNA was recovered 659 using the RNA Clean and Concentrator TM kit (Zymo Research) . Concentrations of the 660 purified RNA were determined using a UV -Vis spectrophotometer and verified on 1.5% 661 agarose gels. The RNA extractions were performed in three biological replicates. 662 cDNA was synthesized using 200 units of RevertAid M -MulV RT (Thermo Scientific), 0.5 663 µg of Oligo (dT)18 primer, and 0.5 µg of RNA in a final volume of 20 µL according to the 664 manufacturer’s instructions. The qPCR reactions were carried out using Maxima SYBR 665 Green/ROX qPCR Master Mix (Thermo Scientific) in a final volume of 12 µL, containing 666 0.3 µM of each primer and 1 µL of the cDNA previously synthesized. The qPCR reactions 667 were carried out in two technical replicates per biological replicate using a Step One Plus 668 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 26 Real-Time PCR System (Applied Biosystems) under the following conditions: 95 ºC for 10 669 min, followed by 40 cycles at 95 ºC for 15 s and 60 ºC for 1 min. The genes and primers used 670 are listed in Table S6. The relative expression of the target genes was quantified using the 671 mathematical method described by77 and normalized with three housekeeping genes (ACT1, 672 UBC6, and RPN2), following established procedures 78,79. For each sample, noremalization 673 was based on the median Ct value of the two technical replicates . Statistical analysis was 674 performed on the log-transformed, normalized expression per sample80. 675 Statistical analysis 676 Statistical analysis and data visualization were performed in R version 4.3.1. To test whether 677 the accumulation of de novo SNPs differed among species and evolutionary regimes, we 678 modeled the number of SNPs detected per evolved genotype using a generalized linear 679 mixed-effects model (GLMM). Because SNP counts are discrete and overdispersed, we fitted 680 a negative binomial model with a log link function. Species, evolutionary condition (constant 681 vs. increasing temperature), and their interaction were included as fixed effects, while strain 682 was included as a random effect to account for non-independence among genotypes derived 683 from the same ancestral strain. Models were fitted using the glmmTMB package. Statistical 684 significance of mixed effects was assessed using Type II WALD c2 tests. When significant 685 interactions were detected, post hoc contrasts comparing evolutionary conditions within each 686 species were performed using estimated marginal means (emmeans), with p-values adjusted 687 for multiple testing using the Benjamini -Hochberg false discovery rate (FDR). Model 688 assumptions were evaluated by inspecting residual distributions and dispersion parameters. 689 For qPCR analyses and to keep comparisons statistically coherent, we used two 690 complementary normalizations: for comparison between ancestral S. cerevisiae and S. 691 eubayanus at 25 ºC, all samples were calibrated to the S. cerevisiae ancestral strain at 25 ºC; 692 for all within-species contrasts (25 ºC vs high temperature; ancestral vs evolved at a given 693 temperature), each species was calibrated to its own ancestral strain at 25 ºC or high 694 temperature condition (40 ºC for S. cerevisiae and 34 ºC for S. eubayanus). All gene-level 695 tests were run per gene and (when applicable) per species using Welch’s t -test on log 2 696 normalized expression values and corrected using Benjamini -Hochberg false discovery rate 697 (FDR). To test whether the change from 25 ºC to high temperature differs between S. 698 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 27 cerevisiae and S. eubayanus , ordinary least squares (OLS) models with robust (HC3) 699 standard errors were fit as log2 normalized expression ~Species * Temperature. To test the 700 change in temperature sensitivity in evolved genotypes, we first computed the temperature 701 effect as expression at high temperature minus expression at 25 ºC for each strain. Then, we 702 tested whether evolution altered this temperature sensitivity by fitting per -species OLS 703 models with HC3 as log2 normalized expression ~Species * Evolved genotype. An adjusted 704 p-value < 0.05 was considered significant. 705 To evaluate whether mtDNA loss affected thermal performance, we analysed μmax for 706 ancestral, evolved, and petit strains across 11 assay temperatures (16–40 ºC). First, we tested 707 overall differences in thermal performance across all strains using a linear mixed -effects 708 model fitted with the nlme package (version 3.1-168) as μmax = Genotype x ns(Temperature, 709 df = 3) + (1|Strain), where Genotype had three levels (Ancestral, Evolved, Petit e), and 710 ns(Temperature, df = 3) denotes a natural spline function used to capture the non-linear effect 711 of temperature. Models were fitted by restricted maximum likelihood (REML), and 712 heteroscedasticity in the fitted values was modelled using a power variance structure 713 (varPower(form = ~fitted(.))). The significance of fixed effects (Genotype, Temperature, and 714 their interaction) was assessed by Type III ANOV A. We then determined whether genotype 715 effects were strain -dependent by fitting separate linear models for each strain: μmax = 716 Genotype × ns(Temperature, df = 3). Pairwise contrasts between genotypes (Ancestral -717 Evolved, Ancestral -Petite, Evolved -Petite) were computed using the emmeans package 718 (version 1.11.2 -8) with Benjamini -Hochberg false discovery rate correction. Model 719 assumptions (normality, independence, and homoscedasticity) were verified with the 720 check_model() function from the performance package (version 0.15.1) and were met in all 721 cases. 722 To compare thermal parameters among genotypes, we used a non-parametric Kruskal-Wallis 723 test followed by Benjamini -Hochberg-corrected Wilcoxon pairwise comparisons, because 724 residuals did not meet normality assumptions. All parameters were previously normalized to 725 the mean of the ancestral genotype within each strain to allow direct comparison across 726 species. 727 Data availability statement 728 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 28 All fastq sequences were deposited in the National Center for Biotechnology Information 729 (NCBI) as a Sequence Read Archive under the BioProject accession number 730 PRJNA1270780. The code used for analyses and plotting is available in our GitHub 731 repository, https://github.com/j-molinet/TPC_genomic_paper. All other data are included in 732 the manuscript and/or in the supporting information. 733 ACKNOWLEDGMENTS 734 We thank Chloé Haberkorn for her help in the laboratory. Computation and data handling 735 were enabled by resources provided by the National Academic Infrastructure for 736 Supercomputing in Sweden (NAISS) , partially funded by the Swedish Research Council 737 through grant agreement no. 2022 -06725, under Project NAISS 2024/22-917 and 2024/23-738 411. RK, JM, and CG were supported by the Swedish Research Council (2022 -03427) and 739 the Knut and Alice Wallenberg Foundation ( 2017.0163 and 2024.0216). JM and PV were 740 funded by the Agencia Nacional de Investigación y Desarrollo (ANID) Fondecyt Iniciación 741 grants 11260235 and 11240649, respectively, and by the ANID Iniciativa Científica Milenio 742 program ICN17_022. We acknowledge Fundación Ciencia & Vida for providing 743 infrastructure and laboratory space. 744 AUTHOR CONTRIBUTIONS 745 Conceptualization: J.M., R.S.; Investigation: J.M., R.S.; Methodology: J.M. , C.G. , P.V .; 746 Formal Analysis: J.M., R.S. , P.V .; Resources: R.S.; Visualization: J.M.; Original Draft 747 Preparation: J.M., R.S. All authors have read and agreed to the published version of the 748 manuscript. 749 COMPETING INTEREST 750 The authors declare no conflict of interest. 751

752 1. Kontopoulos, D., Smith, T. P., Barraclough, T. G. & Pawar, S. Adaptive evolution shapes 753 the present-day distribution of the thermal sensitivity of population growth rate. PLoS 754 Biol. 18, e3000894 (2020). 755 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 29 2. Bello, M. D. & Abreu, C. I. Temperature structuring of microbial communities on a 756 global scale. Curr. Opin. Microbiol. 82, 102558 (2024). 757 3. Martin, R. A., da Silva, C. R. B., Moore, M. P. & Diamond, S. E. When will a changing 758 climate outpace adaptive evolution? WIREs Clim. Change 14, e852 (2023). 759 4. Edelsparre, A. H., Fitzpatrick, M. J., Saastamoinen, M. & Teplitsky, C. Evolutionary 760 adaptation to climate change. Evol. Lett. 8, 1–7 (2024). 761 5. Molinet, J. & Stelkens, R. The evolution of thermal performance curves in response to 762 rising temperatures across the model genus yeast. Proc. Natl. Acad. Sci. 122, 763 e2423262122 (2025). 764 6. Salvadó, Z. et al. Temperature Adaptation Markedly Determines Evolution within the 765 Genus Saccharomyces. Appl. Environ. Microbiol. 77, 2292–2302 (2011). 766 7. Pinto, J., Balarezo -Cisneros, L. N. & Delneri, D. Exploring adaptation routes to cold 767 temperatures in the Saccharomyces genus. PLOS Genet. 21, e1011199 (2025). 768 8. Gonçalves, P., Valério, E., Correia, C., Almeida, J. M. G. C. F. de & Sampaio, J. P. 769 Evidence for Divergent Evolution of Growth Temperature Preference in Sympatric 770 Saccharomyces Species. PLOS ONE 6, e20739 (2011). 771 9. Johnson, M. S. et al. Phenotypic and molecular evolution across 10,000 generations in 772 laboratory budding yeast populations. eLife 10, e63910 (2021). 773 10. Natalino, M. & Fumasoni, M. Experimental approaches to study evolutionary cell 774 biology using yeasts. Yeast 40, 123–133 (2023). 775 11. Ament-Velásquez, S. L. et al. The Dynamics of Adaptation to Stress from Standing 776 Genetic Variation and de novo Mutations. Mol. Biol. Evol. 39, msac242 (2022). 777 12. Angilletta Jr., M. J. Thermal Adaptation: A Theoretical and Empirical Synthesis. (Oxford 778 University Press, 2009). doi:10.1093/acprof:oso/9780198570875.001.1. 779 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 30 13. Knies, J. L., Kingsolver, J. G. & Burch, C. L. Hotter Is Better and Broader: Thermal 780 Sensitivity of Fitness in a Population of Bacteriophages. Am. Nat. 173, 419–430 (2009). 781 14. Malusare, S. P., Zilio, G. & Fronhofer, E. A. Evolution of thermal performance curves: A 782 meta‐analysis of selection experiments. J. Evol. Biol. 36, 15–28 (2023). 783 15. Mesas, A., Jaramillo, A. & Castañ eda, L. E. Experimental evolution on heat tolerance 784 and thermal performance curves under contrasting thermal selection in Drosophila 785 subobscura. J. Evol. Biol. 34, 767–778 (2021). 786 16. Huang, C.-J., Lu, M.-Y ., Chang, Y .-W. & Li, W.-H. Experimental Evolution of Yeast for 787 High-Temperature Tolerance. Mol. Biol. Evol. 35, 1823–1839 (2018). 788 17. Mondal, S. et al. Crucial plant processes under heat stress and tolerance through heat 789 shock proteins. Plant Stress 10, 100227 (2023). 790 18. De Smet, I. et al. Evolutionary tuning of molecular charge state of UBP24 shapes 791 eukaryotic responses to high temperature. Preprint at https://doi.org/10.21203/rs.3.rs -792 7157930/v1 (2025). 793 19. Longan, E. R. & Fay, J. C. Experimental evolution of Saccharomyces uvarum at high 794 temperature yields elevation of maximal growth temperature and loss of the 795 mitochondrial genome. MicroPublication Biol. eCollection 2023, (2023). 796 20. Caspeta, L. et al. Biofuels. Altered sterol composition renders yeast thermotolerant. 797 Science 346, 75–78 (2014). 798 21. Fay, J. C., Alonso -del-Real, J., Miller, J. H. & Querol, A. Divergence in the 799 Saccharomyces Species’ Heat Shock Response Is Indicative of Their Thermal Tolerance. 800 Genome Biol. Evol. 15, evad207 (2023). 801 22. Hansen, J. E. et al. Global Warming Has Accelerated: Are the United Nations and the 802 Public Well-Informed? Environ. Sci. Policy Sustain. Dev. 67, 6–44 (2025). 803 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 31 23. Rêgo, A. et al. Repeatability of evolution and genomic predictions of temperature 804 adaptation in seed beetles. Nat. Ecol. Evol. 9, 1061–1074 (2025). 805 24. Lässig, M., Mustonen, V . & Walczak, A. M. Predicting evolution. Nat. Ecol. Evol. 1, 806 0077 (2017). 807 25. Tenaillon, O. et al. The molecular diversity of adaptive convergence. Science 335, 457–808 461 (2012). 809 26. Kryazhimskiy, S., Rice, D. P., Jerison, E. R. & Desai, M. M. Microbial evolution. Global 810 epistasis makes adaptation predictable despite sequence-level stochasticity. Science 344, 811 1519–1522 (2014). 812 27. Blount, Z. D., Lenski, R. E. & Losos, J. B. Contingency and determinism in evolution: 813 Replaying life’s tape. Science 362, eaam5979 (2018). 814 28. Keefe, M. S., Levitt, D. E., Vellers, H. L., Benjamin, C. L. & Sekiguchi, Y . Mitochondrial 815 adaptations from heat acclimation – A narrative review. J. Therm. Biol. 133, 104283 816 (2025). 817 29. Li, X. C., Peris, D., Hittinger, C. T., Sia, E. A. & Fay, J. C. Mitochondria-encoded genes 818 contribute to evolution of heat and cold tolerance in yeast. Sci. Adv. 5, eaav1848 (2019). 819 30. Cañonero, L. et al. Heat stress regulates the expression of TPK1 gene at transcriptional 820 and post-transcriptional levels in Saccharomyces cerevisiae. Biochim. Biophys. Acta BBA 821 - Mol. Cell Res. 1869, 119209 (2022). 822 31. Kumar, S., Mashkoor, M., Balamurugan, P. & Grove, A. Yeast Crf1p is an activator with 823 different roles in regulation of target genes. Yeast 41, 379–400 (2024). 824 32. Kessi-Pérez, E. I. et al. Indirect monitoring of TORC1 signalling pathway reveals 825 molecular diversity among different yeast strains. Yeast 36, (2019). 826 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 32 33. Broach, J. R. Nutritional control of growth and development in yeast. Genetics 192, 73–827 105 (2012). 828 34. Jung, U. S. & Levin, D. E. Genome -wide analysis of gene expression regulated by the 829 yeast cell wall integrity signalling pathway. Mol. Microbiol. 34, 1049–1057 (1999). 830 35. Bendixsen, D. P. et al. Reproductive Isolation due to Divergent Ecological Selection Is 831 Accompanied by Vast Genomic Instability in Experimentally Evolved Yeast Populations. 832 Mol. Ecol. 34, e70110 (2025). 833 36. Talevich, E., Shain, A. H., Botton, T. & Bastian, B. C. CNVkit: Genome -Wide Copy 834 Number Detection and Visualization from Targeted DNA Sequencing. PLOS Comput. 835 Biol. 12, e1004873 (2016). 836 37. Colosimo, P. F. et al. Widespread Parallel Evolution in Sticklebacks by Repeated Fixation 837 of Ectodysplasin Alleles. Science 307, 1928–1933 (2005). 838 38. Schrider, D. R., Hahn, M. W. & Begun, D. J. Parallel Evolution of Copy -Number 839 Variation across Continents in Drosophila melanogaster. Mol. Biol. Evol. 33, 1308–1316 840 (2016). 841 39. Chen, L., DeVries, A. L. & Cheng, C. -H. C. Convergent evolution of antifreeze 842 glycoproteins in Antarctic notothenioid fish and Arctic cod. Proc. Natl. Acad. Sci. 94, 843 3817–3822 (1997). 844 40. McDonald, M. J. Microbial Experimental Evolution – a proving ground for evolutionary 845 theory and a tool for discovery. EMBO Rep. 20, e46992 (2019). 846 41. Saxer, G. et al. Mutations in Global Regulators Lead to Metabolic Selection during 847 Adaptation to Complex Environments. PLoS Genet. 10, e1004872 (2014). 848 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 33 42. Rêgo, A. et al. Adaptation to complex environments reveals pervasive trade -offs and 849 genetic targets with pleiotropic effects. 2024.01.24.577006 Preprint at 850 https://doi.org/10.1101/2024.01.24.577006 (2025). 851 43. Lenski, R. E. Experimental evolution and the dynamics of adaptation and genome 852 evolution in microbial populations. ISME J. 11, 2181–2194 (2017). 853 44. Lind, P. A., Farr, A. D. & Rainey, P. B. Evolutionary convergence in experimental 854 Pseudomonas populations. ISME J. 11, 589–600 (2017). 855 45. Hoitinga, P. K. & Birkeland, S. Pathway-level convergence: an underexplored aspect of 856 convergent evolution. Trends Genet. 41, 853–867 (2025). 857 46. Liu, Z. & Butow, R. A. Mitochondrial Retrograde Signaling. Annu. Rev. Genet. 40, 159–858 185 (2006). 859 47. Kim, D., Hwang, C. Y . & Cho, K. -H. The fitness trade -off between growth and stress 860 resistance determines the phenotypic landscape. BMC Biol. 22, 62 (2024). 861 48. Kvitek, D. J. & Sherlock, G. Whole Genome, Whole Population Sequencing Reveals 862 That Loss of Signaling Networks Is the Major Adaptive Strategy in a Constant 863 Environment. PLoS Genet. 9, e1003972 (2013). 864 49. Malina, C., Larsson, C. & Nielsen, J. Yeast mitochondria: an overview of mitochondrial 865 biology and the potential of mitochondrial systems biology. FEMS Yeast Res. 18, (2018). 866 50. Zhang, M., Shi, J. & Jiang, L. Modulation of mitochondrial membrane integrity and ROS 867 formation by high temperature in Saccharomyces cerevisiae. Electron. J. Biotechnol. 18, 868 202–209 (2015). 869 51. Guyot, S. et al. Surviving the heat: heterogeneity of response in Saccharomyces 870 cerevisiae provides insight into thermal damage to the membrane. Environ. Microbiol. 871 17, 2982–2992 (2015). 872 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 34 52. Guaragnella, N. et al. The role of mitochondria in yeast programmed cell death. Front. 873 Oncol. 2, 70 (2012). 874 53. Baker, E. P. et al. Mitochondrial DNA and temperature tolerance in lager yeasts. Sci. Adv. 875 5, eaav1869 (2019). 876 54. Rikhvanov, E. G. et al. Do mitochondria regulate the heat -shock response in 877 Saccharomyces cerevisiae? Curr. Genet. 48, 44–59 (2005). 878 55. Paliwal, S., Fiumera, A. C. & Fiumera, H. L. Mitochondrial -Nuclear Epistasis 879 Contributes to Phenotypic Variation and Coadaptation in Natural Isolates of 880 Saccharomyces cerevisiae. Genetics 198, 1251–1265 (2014). 881 56. Zubko, E. I. & Zubko, M. K. Deficiencies in mitochondrial DNA compromise the 882 survival of yeast cells at critically high temperatures. Microbiol. Res. 169, 185 –195 883 (2014). 884 57. Wang, J.-T. J., Ng, P. L. P., Powers, M. E., Rha, C. H. & Brem, R. B. The role of mitotype 885 variation and positive epistasis in trait differences between Saccharomyces species. 886 Genetics 232, iyaf233 (2026). 887 58. de Pinto, M. C., Locato, V ., Paradiso, A. & De Gara, L. Role of redox homeostasis in 888 thermo-tolerance under a climate change scenario. Ann. Bot. 116, 487–496 (2015). 889 59. Perez-Sepulveda, B. M. et al. An accessible, efficient and global approach for the large-890 scale sequencing of bacterial genomes. Genome Biol. 22, 349 (2021). 891 60. Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results 892 for multiple tools and samples in a single report. Bioinformatics 32, 3047–3048 (2016). 893 61. Chen, S., Zhou, Y ., Chen, Y . & Gu, J. fastp: an ultra-fast all-in-one FASTQ preprocessor. 894 Bioinformatics 34, i884–i890 (2018). 895 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 35 62. Li, H. & Durbin, R. Fast and accurate long -read alignment with Burrows –Wheeler 896 transform. Bioinformatics 26, 589–595 (2010). 897 63. Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 898 2078–2079 (2009). 899 64. Okonechnikov, K., Conesa, A. & García-Alcalde, F. Qualimap 2: advanced multi-sample 900 quality control for high-throughput sequencing data. Bioinformatics 32, 292–294 (2016). 901 65. DePristo, M. A. et al. A framework for variation discovery and genotyping using next -902 generation DNA sequencing data. Nat. Genet. 43, 491–498 (2011). 903 66. Danecek, P. et al. The variant call format and VCFtools. Bioinformatics 27, 2156–2158 904 (2011). 905 67. Pockrandt, C., Alzamel, M., Iliopoulos, C. S. & Reinert, K. GenMap: ultra -fast 906 computation of genome mappability. Bioinformatics 36, 3687–3692 (2020). 907 68. Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic 908 features. Bioinformatics 26, 841–842 (2010). 909 69. Danecek, P. et al. Twelve years of SAMtools and BCFtools. GigaScience 10, giab008 910 (2021). 911 70. Cingolani, P. et al. A program for annotating and predicting the effects of single 912 nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster 913 strain w1118; iso-2; iso-3. Fly (Austin) 6, 80–92 (2012). 914 71. Huang, D. W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative analysis of 915 large gene lists using DA VID bioinformatics resources. Nat. Protoc. 4, 44–57 (2009). 916 72. Molinet, J. et al. Wild Patagonian yeast improve the evolutionary potential of novel 917 interspecific hybrid strains for lager brewing. PLoS Genet. 20, e1011154 (2024). 918 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 36 73. Fox, T. D. et al. [10] Analysis and manipulation of yeast mitochondrial genes. in Methods 919 in Enzymology vol. 194 149–165 (Academic Press, 1991). 920 74. Ibstedt, S. et al. Concerted Evolution of Life Stage Performances Signals Recent 921 Selection on Yeast Nitrogen Use. Mol. Biol. Evol. 32, 153–161 (2015). 922 75. Molinet, J. et al. Natural Variation in Diauxic Shift between Patagonian Saccharomyces 923 eubayanus Strains. mSystems 7, e0064022 (2022). 924 76. Padfield, D., O’Sullivan, H. & Pawar, S. rTPC and nls.multstart: A new pipeline to fit 925 thermal performance curves in r. Methods Ecol. Evol. 12, 1138–1143 (2021). 926 77. Pfaffl, M. W. A new mathematical model for relative quantification in real-time RT–PCR. 927 Nucleic Acids Res. 29, e45 (2001). 928 78. Vandesompele, J. et al. Accurate normalization of real-time quantitative RT-PCR data by 929 geometric averaging of multiple internal control genes. Genome Biol. 3, research0034.1 930 (2002). 931 79. Teste, M.-A., Duquenne, M., François, J. M. & Parrou, J.-L. Validation of reference genes 932 for quantitative expression analysis by real -time RT-PCR in Saccharomyces cerevisiae. 933 BMC Mol. Biol. 10, 99 (2009). 934 80. Taylor, S. C. et al. The Ultimate qPCR Experiment: Producing Publication Quality, 935 Reproducible Data the First Time. Trends Biotechnol. 37, 761–774 (2019). 936 937 Supplementary Information 938 Figure S1. Detailed distribution of de novo SNPs across strains and evolutionary 939 conditions in Saccharomyces. (A) Total number of SNPs per strain, grouped by evolutionary 940 condition (constant vs. increasing temperature). (B) Distribution of SNPs across coding and 941 non-coding regions per strain. The upper panel corresponds to strains evolved under 942 increasing temperature, while the lower panel shows strains evolved under constant 943 temperature. (C) Parallel evolution across species under constant temperature, displayed as 944 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 37 an UpSet plot. Species are grouped by thermal tolerance (blue = cold -tolerant, orange = 945 intermediate, red = warm-tolerant). 946 Figure S2. Additional validation of TORC1/PKA/MAPK outputs across species and 947 evolutionary conditions. (A) Log2 normalized expression for ancestral S. cerevisiae (SC-948 A, red) and S. eubayanus (SE-A, blue) across the six genes at 25 ºC. Values are relative to 949 SC-A. (B) Log2 normalized expression of evolved genotypes at 25 ºC relative to their 950 species’ ancestral strains at the same temperature (SC top; SE bottom). EC: genotype evolved 951 at a constant 25 ºC; EI: genotype evolved under increasing temperatures. For (A) and (B), 952 the dashed line marks no change. Plotted values correspond to the average of the biological 953 replicates (n=3). Asterisks denote between-species differences in (A) and evolved -ancestral 954 differences in (B) (Welch’s t-test, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). 955 Genes are grouped by pathway (TORC1: RPS6A, CRF1; PKA: SSA4, TPK1; MAPK: SLT2, 956 RLM1). (C) Change in expression between each species’ high temperature and 25 °C for A, 957 EC, and EI (computed as high temperature minus 25 °C; high temperature is 40 °C for SC 958 and 34 °C for SE). The horizontal dashed line marks no change. Asterisks denote differences 959 within genotype, per gene and species (Welch’s t-test, * p < 0.05, ** p < 0.01, *** p < 0.001, 960 high vs 25 °C). 961 Figure S3. Copy number variation associated with thermal evolution in Saccharomyces. 962 (A) Distribution of CNV amplitudes (absolute log₂ copy -number change) for all significant 963 CNV events detected across species under constant and increasing temperature evolution 964 regimes. Violin plots show the density of CNV amplitudes, with embedded boxplots 965 indicating median and interquartile range. CNV amplitudes did not differ significantly 966 between evolutionary conditions (Wilcoxon rank-sum test, p = 0.33), indicating comparable 967 magnitudes of copy-number changes across regimes. (B) Total number of CNV events per 968 species under constant and increasing temperature conditions. Bars represent the number of 969 condition-exclusive CNVs detected in each species, highlighting substantial species-specific 970 variation but no consistent increase in CNV burden under increasing temperature evolution. 971 (C) Proportion of CNV gains and losses per species under constant and increasing 972 temperature conditions. Stacked bars show the relative contribution of copy -number gains 973 and losses within each species and condition, illustrating that gains were generally more 974 frequent than losses across taxa. 975 Figure S4. Thermal parameters across ancestral, evolved, and petit strains. (A) 976 Frequency of mitochondrial genom e loss (petite formation) across Saccharomyces species. 977 (B) Differences in thermal parameters between genotypes (Evolved vs Ancestral, Petit e vs 978 Ancestral). Violin plots show the distribution of differences in thermal parameters across 979 strains, with internal boxplots indicating medians and interquartile ranges. The dashed line 980 indicates the ancestral strains set at 1. Asterisks indicate significan t differences between 981 evolved and petit strains relative to their ancestors (Wilcoxon test; * p < 0.05, ** p < 0.01, 982 ns not significant). 983 Figure S5. TPCs of ancestral, evolved, and petit genotypes per strain. Changes in the 984 maximum growth rates as a function of temperature for each strain, representing the TPCs. 985 Ancestral TPC is in grey, evolved TPC under increasing temperature is in red, and petit e 986 TPCs are in green, orange, and purple. TPCs were constructed using the maximum growth 987 rate of three technical replicates and were fitted using the CTMI. 988 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint 38 Table S1. (A) Ancestral strains used in this study. (B) Evolved genotypes under constant 989 temperature (25 ºC) used in this study. (C) Evolved genotypes under increasing temperatures 990 used in this study. (D) Rho0 strains generated in this study. 991 Table S2. (A) List of reference genomes used in this study. (B) Bioinformatics summary 992 statistics. (C) Total number of SNPs. (D) List of SNPs identified in evolved genotypes under 993 increasing temperatures. (E) List of SNPs identified in evolved genotypes under constant 994 temperatures. (F) Effects of species and evolutionary regime on the accumulation of de novo 995 SNPs. (G) GO and KEGG pathway enrichment analysis for genes with de novo mutations. 996 Table S3. (A) Relative gene expression for target genes of TORC1, PKA, and MAPK 997 pathways in S. cerevisiae and S. eubayanus strains at 25 ºC and higher temperature s. (B) 998 Baseline cross-species comparison at 25 ºC. Log2 normalized expression per gen (ancestral 999 strains), Welch’s tests (two sided), 95% CIs, Hedges’g, and BH -FDR. (C) Ancestral 1000 temperature effect within species (High – 25 ºC) and Species*Temperature interaction. Per 1001 species and gene: High – 25 ºC mean difference (Welch) with 95% CIs and FDR; OLS (HC3) 1002 interaction estimates testing whether SC and SE differ in their temperature response. (D) 1003 Evolved vs. Ancestral at high temperature. Per species and gene: A vs. EC and A vs. EI at 1004 high temperature (SC 40 ºC; SE 34 ºC): Log 2 mean differences, 95% CIs, Hedges’g, Welch 1005 p-values, BH-FDR. (E) Evolved vs. Ancestral at 25 ºC. Per species and gene: A vs. EC and 1006 A vs. EI at 25 ºC. Log2 mean differences, 95% CIs, Hedges’g, Welch p-values, BH-FDR. (F) 1007 Change in temperature responsiveness after evolution (difference -in-differences within 1008 species). Per species and gene: estimates of “(High-25 in EC/EI)-(High-25 in A)” alongside 1009 OLS (HC3) Temperature x Evolution interaction coefficients, 95% CIs, and BH-FDR (these 1010 underpin in Fig. 2C). 1011 Table S4. (A) Per sample CNV calls (gain/loss) with CI/FDR, log2, and size thresholds. (B) 1012 Condition-exclusive consensus CNVs (Increasing vs. Control evolved conditions) with 1013 recurrence and median log 2. (C) Condition-exclusive consensus CNVs (Increasing vs. 1014 Control evolved conditions) with recurrence and median log2, excluding telomeric regions 1015 (10 kb). (D) Per-sample mitochondrial status. (E) CNV summary by evolutionary condition. 1016 Table S5. (A) Estimated maximum growth rate for each strain and temperature. (B) Linear 1017 mixed-effects model results for growth rate as a function of temperature and strain type. (C) 1018 Pairwise comparison of growth rate between strain types at each temperature. (D) Linear 1019 mixed-effects model results for growth rate for each strain. (E) Pairwise comparison of 1020 growth rate between strain types for each strain. (F) Estimated thermal parameters of the 1021 TPCs for each strain and strain type. (G) Kruskal-Wallis rank sum test and Wilcoxon test 1022 with Benjamini-Hochberg correction for multiple pairwise comparisons for relative thermal 1023 parameters for each strain type. (H) Overlapping of confidence intervals of thermal 1024 parameters. 1025 Table S6. List of primers used in this study. 1026 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2026. ; https://doi.org/10.64898/2026.03.23.708575doi: bioRxiv preprint