Yeast strains used or created for this study are provided (Table S1). The parental yeast strains  were purchased from the National Collection of Yeast Cultures (SGRP strains).
Media included the standard laboratory media YPD (10g/L yeast extract, 20g/L peptone, 20g/L glucose). YPEG (10g/L yeast extract, 20g/L peptone, 30mL/L ethanol, 30mL/L glycerol), Complete synthetic media (CSM) (800mg/L CSM premix (Sunrise Science), 6.7g/L yeast nitrogen base without amino acids (YNB w/o AA), 20g/L glucose), and CSMEG (CSM containing 30mL/L ethanol and 30mL/L glycerol instead of glucose). CSM media lacking specific amino acids were prepared by replacing CSM premix with those formulated to omit specific amino acid drop out mixes (CSM-Adenine, CSM-Uracil, CSM-Arginine), as recommended by the manufacturer (Sunrise Science). Additional media included SOE, a synthetic oak exudate  containing 1g/L yeast extract, 1.5g/L peptone, 10g/L sucrose, 5g/L fructose, 5g/L glucose). Media emulating maple sap (MSy)  contained 1.75g/L YNB w/o AA, 800mg/L CSM premix, 0.5g/L allantoin, and a locally-produced maple syrup added to a final concentration of 2˚ Brix.
Synthetic Grape Must (SGM) [78, 79] contained 125g/L glucose, 125g/L fructose, and 460 mg/L ammonium chloride, supplemented with vitamin, mineral salts, and amino acid mixes. To mimic the normal concentrations of these acids commonly found in must at grape maturity, 3 g/L malic acid, 3 g/L tartaric acid, and 0.3 g/L citric acid, as well as anaerobic factors (15 mg/L ergosterol and 5 mg/L sodium oleate dissolved in (1:1,v/v) ethanol: Tween 80) were added to media, then the pH was adjusted to 3.0 using NaOH. The vitamin mix used in SGM media contained 20 mg/L myo-inositol, 2 mg/L nicotinic acid, 1.5 mg/L calcium pantothenate, 0.25 mg/L thiamine HCl, 0.25 mg/L pyridoxine HCl, 0.003 mg/L biotin, and mineral salts (750 mg/L KH2PO4, 500 mg/L K2SO4, 250 mg/L MgSO4 · 7H2O, 55 mg/L CaCl2 · 2H2O, 200 mg/L NaCl, 4 mg/L ZnSO4, 1 mg/L CuSO4 ·5H2O, 1 mg/L KI, 0.4 mg/L CoCl2·6H2O, 1 mg/L H3BO3, 1 mg/L NaMoO4·2H2O). The amino acid mix used in SGM media contained 19 amino acids (612.6 mg/L L-proline, 505.3 mg/L L-glutamine, 374.4 mg/L L-arginine, 179.3 mg/L L-tryptophan, 145.3 mg/L L-alanine, 120.4 mg/L L-glutamic acid, 78.5 mg/L L-serine, 759.2 mg/L L-threonine, 48.4 mg/L L-leucine, 44.5 mg/L L-aspartic acid, 44.5 mg/L L-valine, 37.9 mg/L L-phenylalanine, 32.7 mg/L L-isoleucine, 32.7 mg/L L-histidine, 31.4 mg/L L-methionine, 18.3 mg/L L-tyrosine, 18.3 mg/L L-glycine, 17.0 mg/L lysine, 13.1 mg/L L-cysteine,35 mg/L uracil, 35 mg/L adenine hemisulfate).
For solid media, agar was added to 2% prior to autoclaving. Due to agar disintegration in low pH, 4% agar solution was prepared separately and added to SGM mixture (1:1, v/v) after autoclaving.
Creation of a mitonuclear strain collection
To generate strains with 225 unique mitonuclear genotypes, mtDNAs were transferred between 15 divergent S. cerevisiae isolates using karyogamy-deficient matings (Figure 1B). Each parental strain (MATa ura3 r+) was mated to NAB32 (MATa kar1-1 ade2 arg8 r0) on solid YPD media. When zygotes were visible under a compound microscope (2-6 hours), mating mixtures were diluted into YPD liquid media and incubated for 2 hours to promote cell division. Cell mixtures were plated for single colonies on media that selected against the mtDNA donor strain (CSM–URA). Colonies were printed to selective media (YPEG, CSM–ADE and CSM–ARG) and mitochondrial cargo strains were identified as respiring colonies with auxotrophies of the kar1-1 recipient strain. Cargo strains were mated with rho0 derivatives of each parental strain, followed by selecting and screening for respiring haploids with the genotype of the parental strains. At least 2 biological replicates for each mitonuclear genotype were isolated from independent matings. As controls, each of the original mtDNAs were reintroduced to the rho0 parental strains to recreate parental mitonuclear combinations. Rho0 strains were generated using ethidium bromide . Strain names and genotypes for the complete 15×15 mitonuclear strain collection are found in Table S1.
Strains were phenotyped by spotting cells in high density arrays using a BM3-BC colony processing robot (S&P Robotics) using a 4 x 8 block design. Arrays were printed from YPD to test media, acclimated for 2 days at 30°C, reprinted to test media (CSM, CSMEG, MSy, SGM, and SOE) and incubated at 20°, 30° and 37°C. Strains with nuclear backgrounds strains SK1 and 322136S were omitted as they were flocculant. Arrays were photographed over 96 hours at 18 time points. Colony sizes from each image were determined using gitter  and were fed through a custom R script pipeline (modified from ). Colony spots that failed circularity measures were omitted from further analyses. Colonies from the two outermost rings of each plate were removed to avoid edge effects. Colony spots were used to fit growth curves using logistic regression and outliers were omitted. Fitness parameters (minimum and maximum colony size) were estimated from logistic growth curves and normalized to a reference strain (DAU2) included on each array. To correct for unequal numbers of technical replicates for strains with a shared genotype (i.e. biological replicates), a random subset of technical replicates was selected with equal numbers of each biological replicate. The difference between maximum and minimum colony size was used as a proxy for fitness.
Statistical analyses were performed with the lme4 package in R (version 3.6.0). Eight mitochondrially-encoded genes were obtained from available mitochondrial sequences [50, 82-84] (Table S2). Alignments and identification of SNPs and nucleotide diversities were performed using Geneious version 2020.5 .
The significance of nuclear, mitochondrial genetic components, environments and their interactions were tested using random effect models (lmer) by comparing the full model n + mt + e + (n × e) + (n × e) + (mt × n)+ (mt × n × e) with a model lacking the evaluated term. Within individual environments, random effects models were used to determine the significance of nuclear, mitochondrial genotypes and mitonuclear interactions. Variance component analysis was performed in each condition (VarCorr), in which the contributions of each genetic term to phenotypic variance were estimated from the full model: n + mt + (mt × n).
To estimate the frequency of mitonuclear epistasis when exchanging mtDNA strains from the same or two different subpopulations, two-way ANOVAs were used to test the significance of mitonuclear interactions among each mtDNA exchange such that each test included 4 genotypes: two parental mitonuclear genotypes and two synthetic genotypes derived from the exchange of mtDNAs between the parental strains. The mitonuclear effect size in these exchanges was determined as the absolute differences between the change in growth for each nuclear genotype (ΔΔ=½(ΔnDNAi/mtDNAi®nDNAi/mtDNAj) – (ΔnDNAj/mtDNAi® nDNAj/mtDNAj) ½).
To test the effect of disrupting naturally occurring mitonuclear combinations, one-way ANOVA was used to test for significant difference in fitness between original strains and 195 synthetic strains in each condition. Within each nuclear background, the significance in fitness difference between the strain with coadapted mitonuclear combinations and each of its 14 synthetic derivatives was investigated using a fixed effect model.
We performed association tests between 198 mitochondrial SNPs (Table S2) and the phenotypes listed in Figure 3 using the 15 strains sharing a common nuclear background in each media type. ANOVAs were conducted in R using a simple linear model (phenotype ~ SNPi_) and Bonferroni corrections were applied to determine significance thresholds (P < 0.00025).