Mitochondrial-nuclear coadaptation revealed through mtDNA replacements in Saccharomyces cerevisiae

Background Mitochondrial function requires numerous genetic interactions between mitochondrial- and nuclear- encoded genes. While selection for optimal mitonuclear interactions should result in coevolution between both genomes, evidence for mitonuclear coadaptation is challenging to document. Genetic models where mitonuclear interactions can be explored are needed. Results We systematically exchanged mtDNAs between 15 Saccharomyces cerevisiae isolates from a variety of ecological niches to create 225 unique mitochondrial-nuclear genotypes. Analysis of phenotypic profiles confirmed that environmentally-sensitive interactions between mitochondrial and nuclear genotype contributed to growth differences. Exchanges of mtDNAs between strains of the same or different clades were just as likely to demonstrate mitonuclear epistasis although epistatic effect sizes increased with genetic distances. Strains with their original mtDNAs were more fit than strains with synthetic mitonuclear combinations when grown in media that resembled isolation habitats. Conclusions This study shows that natural variation in mitonuclear interactions contributes to fitness landscapes. Multiple examples of coadapted mitochondrial-nuclear genotypes suggest that selection for mitonuclear interactions may play a role in helping yeasts adapt to novel environments and promote coevolution.

Analysis of phenotypic profiles confirmed that environmentally-sensitive interactions between 23 mitochondrial and nuclear genotype contributed to growth differences. Exchanges of mtDNAs 24 between strains of the same or different clades were just as likely to demonstrate mitonuclear 25 epistasis although epistatic effect sizes increased with genetic distances. Strains with their 26 original mtDNAs were more fit than strains with synthetic mitonuclear combinations when 27 grown in media emulating their original ecological niches. 28 Conclusions: This study shows that natural variation in mitonuclear interactions contribute to 29 fitness landscapes and thus provide a platform for natural selection to promote coevolution. in energy production, nutrient and redox sensing, and signaling pathways are governed by both 39 the mitochondrial and nuclear genomes. Known genetic interactions between these distinct 40 organellar DNAs are required for basic mitochondrial functions including mitochondrial DNA 41 (mtDNA) replication and repair, transcription and translation of mitochondrial genes, assembly 42 and function of the respiratory chain complexes and creation of a mitochondrial membrane 43 potential that drives energy production and maintains mitochondrial homeostasis [1,2]. These 44 genetic interactions have shaped the evolution of both genomes [3][4][5]. 45 distances and phenotypes. Using mitochondrial coding SNPs from available mtDNA sequences 115 (Table S2), we observed a statistically significant positive mtDNA genotype-phenotype 116 correlations in only 3 of 15 tested conditions (Fig. S2). In addition to the relatively low numbers 117 of mitochondrial SNPs that could have contributed to these results, it is likely that the reticulate 118 mtDNA ancestries in Saccharomyces [49,51,52,62]} and selection for specific mitochondrial 119 alleles disrupt positive genotype-phenotype correlations. 120 In populations, detection of epistatic interactions between different loci is limited to those 121 with large effect as most statistical approaches are hampered by multiple testing and power 122 issues [63]. Here, we created a mitonuclear strain collection by precisely exchanging mtDNAs 123 between 15 parental strains to create 225 unique mitochondrial-nuclear genome combinations 124 (15 mtDNA × 15 nuclear genomes) (Fig. 1B). The mitonuclear strains (including biological 125 replicates of each genotype and reconstructed parental genotypes) were grown in 15 different 126 conditions. These environments were 5 different media (oak exudate (SOE), synthetic grape 127 must (SGM), maple sap media (MSy) and lab media containing glucose (CSM) and 128 ethanol/glycerol (CSMEG), each grown at three temperatures (20, 30, and 37°C), resulting in 129 67,727 fitness estimates. Distribution of colony sizes varied among conditions, with the widest 130 range of phenotypes seen in media requiring mitochondrial respiration for growth under high 131 temperature stress (CSMEG 37) (Fig. 1C). 132 The balanced design of the mitonuclear strain collection allows for direct testing of 133 mitonuclear epistasis. Highly significant mitonuclear interaction terms were observed in each 134 condition (ANOVA on the random effect model y ij = µ ~ mt i + n j + (mt×n) ij + ε ij , P < 0.0001, 135 Table S3) and could explain 10.8% to 31.5% of the total phenotypic variances (Fig. 1C, Table  136 S4). Mitonuclear interactions explained the highest proportion of phenotypic variances in media requiring mitochondrial respiration (CSMEG), consistent with the idea that physically interacting 138 structural subunits of the oxidative phosphorylation machinery would be particularly sensitive to 139 genetic variation that altered the efficacy of ATP production. Growth in high and low 140 temperatures generally increased the effects of mitonuclear interactions (Fig. 1D) Fig S3), providing both fitness 149 advantages and disadvantages. All strains harboring synthetic mitonuclear combinations could 150 support growth via mitochondrial respiration indicating that strong mitonuclear incompatibilities, 151 such have been observed in interspecific hybrids [24][25][26]64], did not exist in these strains. Still, 152 our strain collection reveals that mitonuclear interactions provide a substantial phenotypic 153 landscape upon which selection can act. 154

Mitonuclear coadaptation revealed in ecologically relevant environments 155
Selection for mitonuclear interactions could contribute to phenotypic differences between 156 different clades of yeast. We reasoned that if selection for specific mitonuclear allele pairings 157 occurred during the initial expansions of S. cerevisiae, exchanging mtDNAs between strains 158 within the same clade should have less of an effect than between clades. Our collection contained 159 five strains from a Wine/European clade, and single representatives from West African, Sake and 160 North American clades, providing 10 within-and 18 between-clade mtDNA exchanges ( Fig. 2A). 161 For each exchange, we tested whether mitonuclear interactions explained growth differences 162 between strains harboring original (nDNA i /mtDNA i and nDNA j /mtDNA j ) and synthetic 163 (nDNA i /mtDNA j and nDNA i /mtDNA j ) mitonuclear combinations using fixed effects two-way 164 ANOVAs (Table S7). Within clade mtDNAs exchanges were just as likely to show significant 165 mitonuclear epistasis as between clade exchanges (P >0.05 for Χ 2 tests, Table S8, Fig. 2B,). 166 However, mitonuclear interactions produced larger effect sizes (measured as the absolute 167 differences between the changes in growth in two strains following the exchange of their 168 mtDNAs) when mtDNAs were exchanged between clades in 4 (of 15) conditions (Fig. 2C). This 169 is consistent with a model where populations maintain separate mitonuclear alleles that diverge 170 over time. However, mitonuclear effects were greater in the within-clade mtDNA exchanges in 171 one environment and were not different in the remaining ten environments. Taken together, these 172 results demonstrate that natural genetic variation in mitonuclear interactions will alter 173 phenotypes but genetic distance alone cannot necessarily predict the effects of these interactions 174 in all environments. 175 Previously, we showed that a subset of the strains harboring original mitonuclear 176 combinations conferred higher growth rates than strains with synthetic combinations at elevated 177 temperature and under respiratory conditions [58]. In this larger survey of mitonuclear 178 combinations, strains with their original mtDNAs did not show significant fitness advantages 179 over introduced mtDNAs, except in SGM media at 30°C (Fig. S3). Because fitness effects of 180 mitonuclear interactions are sensitive to environmental conditions (see Table S5) and because 181 any selection for optimal mitonuclear interactions would have occurred in ecologically relevant 182 conditions, it is perhaps not meaningful to compare average fitness effects for all strains in the 183 same environment. We reasoned that the fitness advantage of original mitonuclear pairs observed 184 in SGM media might be explained by the overrepresentation of strains isolated from 185 fermentation sources (5 of the 15 parental strains) whose mitonuclear genomes were coadapted 186 for an alcoholic fermentation niche. 187 To determine if the original mitonuclear genome pairings were coadapted to isolation 188 habitats, we paired the parental strains with media emulating their respective isolation habitats 189 and asked whether each nuclear background preferred its own (coadapted) or foreign mtDNA 190 ( Fig. 3). Coadapted mitonuclear combinations were generally preferred in strains derived from 191 fermentation isolates when tested in a synthetic grape must media. Strains with nuclear 192 backgrounds from Sake (Y12), West African (DBVPG6044) and Wine/European (L-1528) 193 clades had higher growth rates when paired with their original mtDNAs than with any foreign 194 mtDNAs (Fig. 3A). Synthetic mitonuclear combinations in strains derived from the nuclear 195 background of a wine isolate BC187 showed statistically significant advantages over the 196 coadapted mitonuclear combination, though the growth differences were small. Interestingly, no 197 preference was seen for the coadapted mitonuclear combination when the nuclear background 198 originated from a strain with mosaic ancestry (YIIc17_E5). In this case, synthetic mitonuclear 199 genotypes were just as likely to show fitness advantages as disadvantages. It is possible that a 200 recent hybridization event interrupted ancestral coadapted mitonuclear complexes in this mosaic 201 parental strain. 202 Beneficial coadapted mitonuclear genome combinations could also be observed in other 203 strains. The original mitonuclear genotype derived from an oak (Quercus) isolate (YPS606) grew 204 better when paired with its own mtDNA than with 8 others (Fig 3B). The experimental model 205 strain W303, maintained under laboratory conditions since the 1970s [65], had growth 206 advantages with its own mtDNA than any other mtDNA when tested in common laboratory 207 media CSM and CSMEG (Fig. 3B). The original mitonuclear genotypes from 2 clinical isolates 208 of Wine/European ancestries were more likely to show growth benefits when grown at high 209 temperatures (Fig. 3C). The original mitonuclear genotype from a soil isolate of Wine/European 210 ancestry (DBVPG1373) was less likely to show fitness advantages in grape must media, 211 suggesting that this isolate may have undergone different ecological pressures from the 212 fermentation isolates of the same clade. Taken together, these data suggest that selection for 213 mitonuclear interactions has occurred repeatedly during the population expansion of S. cerevisiae, 214 shaped by adaptation to new ecological niches. 215

216
Our study provides insight into the coevolution of mitochondrial-nuclear genotypes in S. 217 cerevisiae yeasts. We found that many isolates contain coadapted mitonuclear genotypes that 218 provide growth advantages in ecologically-relevant media. Our results suggest ancestral and 219 recent mitonuclear coadaptations, suggesting mitonuclear fitness is undergoing constant selection. Coadapted mtDNA-nuclear genome pairings provided fitness advantages in media emulating 236 isolation habitats, but did not necessarily provide fitness advantages in other environments. Our 237 study provides a substantial resource for further research into mitonuclear coadaptations. 238

Strains 240
Yeast strains used or created for this study are provided ( Table S1). The parental yeast strains 241 [69] were purchased from the National Collection of Yeast Cultures (SGRP strains). 242

Creation of a mitonuclear strain collection
To generate strains with 225 unique mitonuclear genotypes, mtDNAs were transferred 276 between 15 divergent S. cerevisiae isolates using karyogamy-deficient matings ( Figure 1B). 277 Each parental strain (MATa ura3 ρ + ) was mated to NAB32 (MATα KAR1-1 ade2 arg8 ρ 0 ) on 278 solid YPD media. When zygotes were visible under a compound microscope (2-6 hours), mating 279 mixtures were diluted into YPD liquid media and incubated for 2 hours to promote cell division. 280 Cell mixtures were plated for single colonies on media that selected against the mtDNA donor 281 strain (CSM-URA). Colonies were printed to selective media (YPEG, CSM-ADE and CSM-282 ARG) and mitochondrial cargo strains were identified as respiring colonies with auxotrophies of 283 the KAR1-1 recipient strain. Cargo strains were mated with rho 0 derivatives of each parental 284 strain, followed by selecting and screening for respiring haploids with the genotype of the 285 parental strains. At least 2 biological replicates for each mitonuclear genotype were isolated from 286 independent matings. As controls, each of the original mtDNAs were reintroduced to the rho 0 287 parental strains to recreate parental mitonuclear combinations. Rho 0 strains were generated using 288 ethidium bromide [74]. Strain names and genotypes for the complete 15×15 mitonuclear strain 289 collection are found in Table S1. 290

Phenotyping 291
Strains were phenotyped by spotting cells in high density arrays using a BM3-BC colony 292 processing robot (S&P Robotics) using a 4 x 8 block design. Arrays were printed from YPD to 293 test media, acclimated for 2 days at 30°C, reprinted to test media (CSM,CSMEG,MSy,SGM,294 and SOE) and incubated at 20°, 30° and 37°C. Strains with nuclear backgrounds strains SK1 and 295 322136S were omitted as they were flocculant. Arrays were photographed over 96 hours at 18 296 time points. Colony sizes from each image were determined using gitter [75] and were fed 297 through a custom R script pipeline (modified from [44]). Colony spots that failed circularity 298 measures were omitted from further analyses. Colonies from the two outermost rings of each 299 plate were removed to avoid edge effects. Colony spots were used to fit growth curves using 300 logistic regression and outliers were omitted. Fitness parameters (minimum and maximum 301 colony size) were estimated from logistic growth curves and normalized to a reference strain 302 (DAU2) included on each array. To correct for unequal numbers of technical replicates for 303 strains with a shared genotype (i.e. biological replicates), a random subset of technical replicates 304 was selected with equal numbers of each biological replicate. The difference between maximum 305 and minimum colony size was used as a proxy for fitness. 306

Statistical Analysis 307
Statistical analyses were performed with the lme4 package in R (version 3.6.0). Eight 308 mitochondrially-encoded genes were obtained from available mitochondrial sequences (Table  309 S2). Each gene was aligned to that of the reference genome S288c, and was subsequently 310 concatenated to generate a genetic distance matrix based on the numbers of SNP differences 311 between each strain pairing using Geneious version 2020.0.5 [76] . Pairwise fitness differences 312 among 15 parental strains were computed and plotted against genetic distance to estimate the 313 mitochondrial genotype-phenotype correlation for all conditions. 314 The significance of nuclear, mitochondrial genetic components, environments and their 315 interactions were tested using random effect models (lmer) by comparing the full model n + mt + 316 e + (n × e) + (n × e) + (mt × n)+ (mt × n × e) with a model lacking the evaluated term. Within 317 individual environments, random effects models were used to determine the significance of 318 nuclear, mitochondrial genotypes and mitonuclear interactions. Variance component analysis 319 was performed in each condition (VarCorr), in which the contributions of each genetic term to 320 phenotypic variance were estimated from the full model: n + mt + (mt × n).
To estimate the frequency of mitonuclear epistasis when exchanging mtDNA strains from 322 the same or two different subpopulations, two-way ANOVAs were used to test the significance 323 of mitonuclear interactions among each mtDNA exchange such that each test included 4 324 genotypes: two parental mitonuclear genotypes and two synthetic genotypes derived from the 325 exchange of mtDNAs between the parental strains. The mitonuclear effect size in these 326 exchanges was determined as the absolute differences between the change in growth for each 327 mitonuclear genotypes were compared to synthetic mitonuclear genotypes within the same 600 nuclear background. Growth rates were normalized to the parental mitonuclear genotype. Lines 601 connecting the original (parental) and synthetic mitonuclear are colored to indicate a statistically 602 significant reduction (red), increase (black), or no change (gray). See Table S7 Table S3. Mitochondrial × nuclear interactions (mt × n) within single environments. To 610 determine the significance of each term, ANOVAs comparing the full model with a model 611 lacking the indicated term were evaluated. Each factor was treated as a random effect. Full 612 model: n + mt + (mt × n). 613 Table S4. Phenotypic Variance Analysis to show the amount of variance contributed by 614 mitochondrial, nuclear background, and mitonuclear interactions in all growth conditions. 615 Table S5. Mitochondrial × nuclear × environment interactions (mt × n × e) To determine the 616 significance of each term, ANOVAs comparing the full model with a model lacking the 617 indicated term were evaluated. Each factor was treated as a random effect. Full model: n + mt + 618 e + (mt × n) + (n × e) + (mt × e) + (mt × n × e). 619 Table S6. Phenotypic variance of strains with original, synthetic, and all possible mitonuclear 620 combinations across all growth conditions. Student's t-test were performed to determine whether 621 the difference among strains with original and synthetic combinations is statistically significant. 622  cerevisiae parental strains were plotted against mitochondrial genetic distances. Pairwise SNP 634 differences from concatenated mitochondrial coding sequences (COX1, COX2, COX3, ATP6, 635 ATP8, ATP9, COB, VAR1) were used to generate distance matrix (Table S2). Regression lines 636 for significant correlations are shown in red. 637 Figure S3. Coadapted mitonuclear combinations rarely provide growth advantages overall. 638 Average growth rates of strains with original (yellow) or synthetic (gray) mitonuclear genome