Allododecaploid yeasts: synthetic hybrids of six species

Polyploidy generates diversity by increasing the number of copies of each chromosome. Many plants, animals, fungi, and other eukaryotes are ancient or recent polyploids, including some of the best-known evolutionary radiations, crops, and industrial organisms. Polyploidy facilitates differentiation and adaptation to new environments, but the tools to test its limits are lacking. Here we develop an iterative Hybrid Production (iHyPr) method to produce allododecaploid yeast strains with a base ploidy of 12n. Chromosomal instability increased dramatically as additional copies of the genome were added. These six-species hybrids rapidly improved their fitness during adaptive laboratory evolution. This new method for making synthetic hybrids will enable basic research on polyploidy, cancer, and chromosome biology, as well as more applied research on biofuels, bioproducts, and synthetic biology. One sentence summary We constructed six-species synthetic hybrids and showed that they were chromosomally unstable but able to adapt rapidly.


Main text
Polyploidy is a widespread phenomenon in biology. Organismal life cycles manipulate ploidy to promote dispersal, environmental robustness, and species diversification (1). One of the major drivers of evolutionary innovation is through whole genome duplication (2,3). Polyploidization promotes genome instability and increases mutation rates, both of which can facilitate short-term adaptation to new environments (4). The karyotypic diversity generated by polyploidization can lead to tumorigenesis in some cancer types (5), but it can also be exploited to develop yeast strains to generate new bioproducts (6).
The baker's yeast Saccharomyces cerevisiae has been one of the most important model organisms to study polyploidy in the context of evolution (7), its effects on mutation rate (4), and as a model of how cancer progresses as clonal populations adapt through driver mutations (8). Experimental evolution assays in S. cerevisiae and comparisons of the genomes of industrial Saccharomyces interspecific hybrids have also uncovered many of the same genetic mechanisms that occur during the clonal expansion of tumorigenic cells, such as aneuploidy and loss-of-heterozygosity (9,10).
To experimentally test the limits of chromosome biology and polyploidy in yeasts, we developed an iterative Hybrid Production (iHyPr) method to construct a series of higherorder interspecific hybrids. First, we used two differentially marked HyPr plasmids, which each encode a drug-inducible HO gene (homothallic switching endonuclease) that promotes mating-type switching, to efficiently generate and select for two-species hybrids (11). Next, using two additional differentially marked HyPr plasmids, we crossed these two-species hybrids to construct three-species and four-species hybrids. Finally, we constructed six-species hybrids using three different crossing schemes ( Figure 1A, Figure   S1).
In all three schemes, diploid genomes were successfully introduced from each of the six parental species, but some species contributed many fewer chromosomes than others to the final strains. Thus, even though the base ploidy of the final strains was allododecaploidy (12n), a ploidy level acquired by a handful of plants and animals in the wild (12,13), none were euploid ( Figure S2). During construction, chromosome losses were widespread and outnumbered gains (t-test p-value < 1.66*10 -2 ) ( Figure S2). These aneuploidies rose dramatically as the number of species donating genomes increased (linear regression r 2 = 0.78, p-value = 3.33*10 -6 ) ( Figure 1B, Figure S2). Complete chromosomal aneuploidies were much more common than aneuploidies caused by unbalanced translocations (97.31% versus 2.69% of total detected chromosomal aberrations). Mitochondrial genome inheritance had a significant impact on the retention of the nuclear genome (ANOVA multifactor p-value = 1.21*10 -3 ), with the mitochondrial genome donor tending to also contribute more nuclear chromosomes ( Figure S3-S5).
To determine if the inherent chromosomal instability of these six-species hybrids could be harnessed, we tested how these new six-species hybrids altered their kinetic parameters during adaptive laboratory evolution (ALE) for an estimated 80 generations in a medium containing glucose or in another medium containing xylose, a sugar poorly metabolized by most Saccharomyces species (14). To provide some growth capability upon which to improve, we chose a S. cerevisiae parental strain that had been engineered by inserting xylose utilization genes into Chromosome IV (15,16). Ancestral six-species hybrids grew slowly, and despite differing from each other in chromosomal composition ( Figure S2), single-colony isolates of all 12 ALE replicates outperformed their ancestors in culturing conditions identical to the ALE (Wilcoxon rank sum test, p-value = 3.51*10 -4 ).
This improvement in maximum growth rate on xylose occurred regardless of whether they were evolved on xylose or glucose, and many evolved strains outperformed the S. cerevisiae reference strain ( Figure 1C). In microplate culturing conditions, evolved hybrid populations grew as much as 71% faster on xylose than the reference S. cerevisiae strain, and populations evolved on xylose outperformed those evolved on glucose (Wilcoxon rank sum test, p-value = 1.29*10 -2 ) (Table S2, Figure S6), giving further support to the utility of iHyPr and ALE.
Higher-order synthetic hybrids allow investigators to rapidly combine traits from many different parents, such as the xylose consumption of the cellulosic biofuel reference strain of S. cerevisiae with traits from other Saccharomyces species. Both ALE selection regimes improved performance on xylose, probably partly due to the removal of interspecific incompatibilities through chromosomal loss and partly due to conditionspecific effects ( Figure 1C, Figure S2, Figure S6). The karyotypic diversity generated allowed for selection from a great deal of phenotypic diversity ( Figure S2), a process of great interest for many industrial applications. The fitness improvements observed here also mimic the genetic processes seen in tumor cells escaping antitumorigenic treatments, where polyploidy drives genomic instability and evolution (8). In summary, higher-order allopolyploids constructed with iHyPr rapidly acquire genome aberrations that involve multiple species and can rapidly adapt to new environmental conditions. This new technology pushes the yeast cell toward its limits in pursuit of basic research questions in chromosome biology and evolutionary genetics, as well as potential industrial applications.

Supplementary Text
Figures S1-S6 Tables S1-S6 References   Chromosomes were colored according to their species designation, with height representing copy number, using the sppIDer pipeline (17). For an extended explanation and other schemes, see Figure S1. Arrows mark hybridization steps. For additional intermediary and six-species hybrid nuclear and mitochondrial genomes with higher resolution, see Figure S3, S4. B) Number of chromosomal aberrations inferred versus the number of species contributing genomes ( Figure S2, S5). Synthetic hybrids generated from each independent scheme are represented with different shapes and colors. A linear regression line and the 95% confidence interval of the fit are represented with a discontinuous blue color and grey shadow, respectively. C) Box plots for the individual evolved colonies isolated from YPX or YPD plates after ALE and their synthetic hybrid ancestors. Kinetic parameters were tested in 3 mL YPX on a rotating culture wheel, identically to how they were evolved for 80 Generations. The average values (n=6) of maximum specific growth rates (µ, defined as (ln(OD2)-ln(OD1))/(T2-T1)) for the S. cerevisiae reference strain (black line, yHRW135 was derived from yHRW134 by plasmid loss), ancestor (light blue dots), and evolved six-species strains (red dots) are shown (Table S1). Different shapes indicate the media in which the synthetic six-species hybrids were evolved. Color points differentiate the ancestor from the evolved hybrids. Additional kinetic parameters from microplate experiments performed on evolved populations are shown in Figure S6 and Table S2.   Tables  Table S1  Table S2  Table S3  Table S4  Table S5  Table S6  References  Yeast strains and maintenance The reference strain chosen for improvement was GLBRCY101, a haploid derivative of a Saccharomyces cerevisiae strain engineered and aerobically evolved for the consumption of xylose (GLBRCY73) (16,15). Representative strains were selected from five additional Saccharomyces species based on published nuclear and mitochondrial genomes (Table S3). These six parent strains were used to generate the six-species hybrids. Yeast strains were stored in cryotubes with YPD (1 % yeast extract, 2 % peptone, and 2 % glucose) and 15 % glycerol at -80 °C. Routine cultures were maintained in YPD plus 2 % agar plates at 24 °C.

Two new Hybrid Production (HyPr) plasmids
We previously published two HyPr plasmids with natMX (pHCT2) and hphMX (pHMK34) resistance cassettes (11). Following our previously described methodology, we amplified the ble (ZEOcyn resistance) and nptII (G418 resistance) coding regions for marker swaps to generate pHRW32 and pHRW40 plasmids, respectively. The new HyPr plasmids enabled complex, iterative crossing schemes without adding extra steps to remove one of the two HyPr plasmids between the hybridization steps.

Saccharomyces transformation with HyPr plasmids
Before transforming GLBRCY101 with a HyPr plasmid, we removed its nuclear KanMX cassette by swapping the KanMX marker to TkMX (18). Next, we transformed this strain using a short DNA fragment designed to allow the TkMX gene to be removed via homologous recombination and selecting for successful marker loss on synthetic complete (SC) + FUdR medium (0.17 % yeast nitrogen base, 0.5 % ammonium sulfate, 0.2 % complete drop out mix, 2 % glucose, and 50 µg/ml 5-fluorodeoxyuridine). S. cerevisiae yHWA85 and representative strains of Saccharomyces paradoxus, Saccharomyces mikatae, Saccharomyces kudriavzevii, Saccharomyces arboricola, and Saccharomyces uvarum were transformed with one of the four HyPr plasmid versions (Table S3). Transformation of yeast strains was done using the lithium acetate/PEG-4000/carrier DNA method, as previously described (19), with previously described modifications for particular species (11). Cells were suspended in YPD, followed by incubation at each species' optimal temperature (i.e. 30 °C for S. cerevisiae and room temperature for all others, including hybrids) for 3 h before being plated on selective medium. Selective media consisted in YPD plus 2 % agar plates with the appropriate selectable drug(s). S. cerevisiae yHWA85 was first diploidized using the HyPr plasmid pHRW40, creating yHRW134 for subsequent crosses. The generation of this diploid strain occurred in one step, which was confirmed by polymerase chain reaction (PCR) amplification of the MAT loci (see below). The experimental reference strain yHRW135 was derived from yHRW134 by screening for spontaneous plasmid loss.
iHyPr (iterative HyPr) method for sequentially generating higher-order hybrids Following the HyPr method to facilitate mating-type switch (11), we pre-cultured strains with differentially marked HyPr plasmids in the presence of doxycycline to express the endonuclease encoded by HO, which is under a Tet-ON promoter; each plasmid also contains the full machinery for inducible expression of the promoter. To generate the sixspecies hybrids yHRWh36 and yHRWh39, we first forced three separate pairs of species to mate, generating two-species hybrids. In each case, once the three two-species hybrids were generated, we crossed two of those two-species hybrids to create a fourspecies hybrid, which finally was crossed with the last two-species hybrid to generate the six-species hybrid. To generate the six-species hybrid yHRWh56, two separate twospecies hybrids were separately crossed with diploid Saccharomyces strains from other species to create two separate three-species hybrids, which were then mated to generate the six-species hybrid ( Figure S1). Before each cross, parental strains were transformed with differentially marked HyPr plasmids (Table S3, Figure S1) and treated with doxycycline in YPD at room temperature, except for S. cerevisiae where 30 °C was applied. As previously described (11), the doxycycline triggers the expression of the Ho endonuclease, which cuts one or more MATa/MATa idiomorphs and generates matingcompatible strains that behave as either MATa or MATa. A sample of each culture was combined in a 1-mL Eppendorf tube and patched on a YPD plate. After 2-3 days, a sample was taken with a toothpick and streaked on a YPD plate supplemented with the corresponding drugs to select for successful matings. In contrast to the original HyPr method, we pre-cultured the new hybrid in YPD with one of the two drugs used during the selective medium step, and that hybrid was then crossed with another strain containing one or two of the other HyPr plasmids not used previously. During these subsequent steps, we expected (and phenotypically verified) the loss of the HyPr plasmid containing the drug-resistance cassette not under selection. This approach and the additional HyPr plasmids made for this study facilitated the iterative crosses required to make six-species hybrids by avoiding the steps of plasmid removal and minimizing the number of generations between crosses ( Figure S2).
Mating-type and PCR-RFLP confirmation of strains Diploidization of the S. cerevisiae strain was confirmed by PCR at the mating-type locus. Hybrid statuses were confirmed by restriction fragment length polymorphism (RFLP) analysis. We used the Standard Taq Polymerase (New England Biolabs, Ipswich, MA), and the primers listed in Table S4. Genomic DNA was extracted using the phenol:chloroform method on a strain grown from pre-culture to saturation in YPD.
Vigorous vortexing was performed for 3-4 minutes, followed by 5 minutes of centrifugation at maximum speed. The top aqueous layer was transferred to 1 mL 100 % EtOH. After an inversion mixture, DNA was precipitated at -80 °C for at least 10-15 minutes. A second centrifugation at maximum speed was performed, and the supernatant was discarded.
We washed the pellet with 700 µl of 70 % EtOH, and we centrifuged again to remove any residue or trace of the supernatant. The pellet was dried and resuspended in 100 µl of EB at 50-60 °C for 30 minutes. To remove RNA, we incubated the solution with 0.5 µl of 10 mg/ml RNase A for 30 minutes at 37 °C. DNA was quantified with a Qubit 2.0 Fluorometer (ThermoFisher Scientific). For PCR-RFLP, resulting PCR products were digested with a restriction enzyme or a combination of multiple restriction enzyme assays able to discriminate among Saccharomyces species (New England Biolabs, Ipswich, MA).
An extended PCR-RFLP pattern, developed in previous publications (11,20) and this study, are detailed in Table S5. Undigested PCR products were visualized on a 1.5 % agarose gel, while digested PCR products were visualized on a 3 % agarose gel.
Adaptive laboratory evolution (ALE) and colony selection Two of the three six-species hybrids (during construction, the third lost S. cerevisiae chromosome IV, where Scheffersomyces (Pichia) stipitis xylose utilization genes had been inserted) were evolved in triplicate at room temperature in tubes with two independent media conditions: 3.0 mL YPD or 3.0 mL YPX (1 % yeast extract, 2 % peptone, and 2 % xylose). Three to five days of fermentation were performed to allow cells to consume the sugars, and an aliquot of each replicate was transferred at of 0.1 optical density at 600 nm (OD600) to a fresh medium until it reached approximately 80 generations. A colony from each independent ALE experiment, regardless of whether they were evolved in glucose or xylose, was selected on YPX plates (1 % yeast extract, 2 % peptone, 2 % xylose, and 2 % agar).

Genome sequencing and chromosome composition
Genomic DNA (gDNA) samples from the diploidized S. cerevisiae strain and the two-, three-, and six-species hybrids were submitted to the DOE Joint Genome Institute for paired-end Illumina sequencing. Libraries were constructed according to the manufacturer's instructions. Sequencing of the flow cell was performed on an Illumina MiSeq using MiSeq Reagent kits, following a 2x150 nucleotide (nt) indexed run recipe.
Curated raw reads were submitted to the SRA database as Bioproject PRJNA476226 (Table S6).
Genomic characterization was performed with sppIDer (17). Our combined nuclear reference genome was built with the genome assemblies of S. cerevisiae GLBRCY22-3 (21), which is a close relative of the biofuel reference strain used here; S. paradoxus CBS432; S. arboricola CBS10644 (22); S. mikatae IFO1815; S. kudriavzevii ZP591; S. uvarum CBS7001 (23); and Saccharomyces eubayanus FM1318 (24). Our combined mitochondrial reference genome was built with the mitochondrial assemblies of the aforementioned strains (22, 24, 25), except for CBS7001, whose mitochondrial genome is still not completely assembled (26). Instead, we used the mitochondrial genome of a close relative, S. uvarum CBS395 (25). Raw Illumina paired-end reads and the combined reference genomes were the input data of sppIDer, which is essentially a wrapper that runs published tools to map the short reads to the combined reference genomes and creates several colorful and visually intuitive outputs (17). Here, we show depth of coverage plots from those species contributing genomes.
The percentage of each Saccharomyces nuclear genome retained in the hybrid was calculated as follows: where Pspp, is the percentage for one of the parental species; Ct, is the number of windows with a coverage mean value above 2; Ws, is the window size; and Gs is the reference genome size for that parental species. This calculation yielded a good approximation of contribution of each species to the genome of the hybrid. For each strain, the number of chromosomes, and the ploidy were estimated from the sppIDer plots. The number of chromosomal aberrations was based on the number of gains, losses, or unbalanced translocations detected in the sppIDer plots. One chromosomal gain, loss, or unbalanced translocation was counted as one aberration. Aberrations observed in one hybrid and maintained in the offspring of subsequent crosses were not counted again; only new aberrations for each cross were reported in the aberration plot ( Figure 1B).

Microtiter plate growth curves
We compared the growth kinetics of the S. cerevisiae reference strain yHRW135, the ancestors of the two six-species hybrids retaining the chromosome IV (yHRWh39, yHRWh56), and populations of the evolved hybrids. Growth was tested in YPD and YPX at room temperature. Strains or populations were pre-cultured in 3.0 mL YPD or YPX, depending of the medium tested. After pre-culture, 10 µl of saturated culture was inoculated into a 96-well plate (Nunc, Roskilde, Denmark) containing 240 µl of identical medium as the pre-culture. Spaces between the wells in the plates were filled with sterile H2O to maintain the humidity of the plates. The reference strain was cross-inoculated in all conditions; for example, yHRW135 pre-cultured in YPX was tested in both YPD and in YPX.
To monitor the growth of strains and populations in the different media conditions, we inoculated 96-well plates and placed them in a BMG FLUOstar Omega at 20 °C.
Absorbance at 595 nm was monitored every 15 min for 2 days. Background absorbance was subtracted from the average of three negative controls containing the uninoculated medium being tested. Kinetic parameters for each condition were calculated in GCAT v6. 3 (27). Median and standard deviations from three independent biological replicates were calculated in R (28) (Table S2). For each medium condition, parameters were normalized against the reference strain, yHRW135, which had been pre-cultured and grown in the medium tested.
Culture wheel growth curves Strains isolated from single colonies from evolved hybrids, ancestral hybrids, and the reference strain (yHRW135) were pre-cultured in YPX, and inoculated at an initial OD600 of 0.1 into 3 mL glass tubes containing YPX. Growth was monitored by measuring OD600.
Kinetic parameters were calculated as above. Median and standard deviations from six independent biological replicates were calculated as above. These experimental conditions most closely matched the conditions in which the strains were evolved, and they are reported in Figure 1C and Table S1.

Statistical analysis
Data analysis and plots were performed in R (28). A linear model of regressions were added to the plot in Figure 1B using the geom_smooth option in the R package ggplot2, and r 2 and significance of regression were calculated with summary(lm(y ~ x)), where x was the number of species, and y was the number of observed aberrations.
The impact of mitochondrial inheritance ( Figure S5B) in the retention of the nuclear genome of those hybrids involving S. cerevisiae was tested using a multifactor ANOVA in R, using summary(aov(P ~ M * C)), where P is the percentage retained of the S. cerevisiae nuclear genome; M is the mitochondrial genome, which was encoded as a binary character (either as the S. cerevisiae mitochondrial genome or that of another species); and C is the type of strain (i.e. classified as parental S. cerevisiae; two-, three-, four-, or ancestral six-species hybrid; and evolved six-species hybrid). t-tests for significant differences between frequency of chromosome gains and losses and Wilcoxon rank sum tests for significant differences in the kinetic parameters shown in Figure 1C and S6, respectively, were performed in R.

Supplementary Text
High instability of chromosome III Industrial conditions offer a niche to interspecific hybrids due to the stressful conditions HyPr exploits the rare-mating system by heterologously expressing the HO gene from differentially marked plasmids (11). We expected that Ho would cut one copy of the heterozygous MAT locus of the diploid strain and use homology repair to convert the locus from heterozygous to homozygous, presumably over a small gene conversion patch using either a homologous chromosome or a silent mating cassette as a template. Although

Heteroplasmic state
During interspecific hybridization, hybrids can inherit one of the two parental mitotypes or a recombinant version (44). In general, one of the parental mitotypes was quickly fixed during the generation of our hybrids here, except for: the allotetraploid S. kudriavzevii x S. mikatae yHRWh4, the allotetraploid S. cerevisiae x S. uvarum yHRWh10, and the sixspecies hybrids yHRWh36, which were all heteroplasmic ( Figure S4).
These three exceptions were all part of the same crossing scheme, and we offer a set of related possible explanations. The presence of selfish elements, such as homing endonucleases, could explain why multiple mitotypes were retained in yHRWh4. In this case, a portion of Smik-COX1, a gene with a high number of introns invaded by homing endonuclease genes (45), seems to have been introduced into the Skud mitochondrial genome (mtDNA) (Figure S4A,C). The most intriguing result occurred while generating the yHRWh10 hybrid, which retained most of both parental mtDNAs (S. cerevisiae and S. uvarum) and remained in a heteroplasmic state ( Figure S4A). We recently demonstrated that, during formation of S. cerevisiae x S. uvarum hybrids, the frequency of petite generation (e.g. strains with incomplete or no mitochondrial genome) was higher when the hybrid inherited a S. uvarum mtDNA (26). Introgression of the F-SceIII homing endonuclease gene into a Suva-mtDNA restored normal mitochondrial retention (26). In yHRWh10, we did not detect the presence of F-SceIII, which might have influenced the loss of mtDNA in its descendants. For example, the six-species hybrid yHRWh36 retained only a small region of Sarb-mtDNA and Suva-mtDNA ( Figure S4C). The role of F-SceIII in the stabilization of the mitochondrial genome is still unclear, but it may be playing an important role in our crosses. In another recent study, mtDNA inheritance was also demonstrated to be dominated by one parental strain due to nuclear-mitochondrial interactions, rather than occurring stochastically (46).
The loss of mitochondrial genomes in particular hybrid combinations, and the unusually high or low coverage ( Figure S4) might suggest that interactions between nuclear-encoded mitochondrial proteins with the mtDNA are unbalanced. In such cases, one model proposes that an oligomeric circular mtDNA form precedes ρstrain formation (47). The Illumina read coverages for some regions of the mtDNAs were surprisingly varied in some hybrids, such as yHRWh8, yHRWh13, and most of the hybrids used for yHRWh36 generation ( Figure S4). Formation of mtDNA concatemers are regulated by levels of Din7p and Mhr1p (47); when hybrids form, these protein levels might be misregulated due to different cis-and trans-regulation, which could increase or decrease the copy numbers of specific mitochondrial regions. Technical artifacts from Illumina sequencing also cannot be excluded, and this phenomenon and process merits further study.  The hybridization steps necessary to generate the six-species hybrids yHRWh36, yHRWh39, and yHRWh56 are 5 represented in panels A), B) and C), respectively. Yeast cells are represented in gray, and chromosomes are colored 6 according to the Saccharomyces species. The strain names of our lab's copy of some strains (Table S2) are displayed in 7 parentheses below the original culture collection strains. Drug-marker resistance is indicated above yeast cells. Systematic 8 crosses are highlighted with arrows to form a pedigree. The black lightning bolt symbol represents the doxycycline shock to 9 promote MATing-type switching or loss to facilitate hybridization. was inferred based on the mitosppIDer output ( Figure S4). Presence, absence, or incomplete mitochondrial genomes 16 are colored in yellow, white, and black, respectively. Synthetic hybrids were ordered as they were used to generate the next 17 hybrid ( Figure S1). Diploidized GLRBCY101 (yHRW134) and yHRWh4 are shown multiple times because of their use in 18 multiple crossing schemes. Note that considerable karyotypic diversity continued to be generated during 80 generations of 19 ALE, but each evolved strain is easily recognized as more similar to its ancestral six-species hybrid. that portion of the genome. Panels were ordered to represent synthetic hybrid data based on the order they were used to 77 generate the next hybrid ( Figure S2). sppIDer produces multiple plots (17) used for generating the individual six-species hybrids ( Figure S2). sppIDer produces multiple plots (17), but here we show 93 the log2 of the average coverage of ~44 bp windows normalized to the mitochondrial genome-wide average coverage. When 94 a synthetic hybrid is formed between ρ + parental strains, a heteroplasmic state can be maintained for several generations, 95 but eventually, a parental or recombinant mitochondrial genome is generally fixed (44). In some hybrids, this heteroplasmic 96 state persisted, and the names of hybrids are colored in a gradient according to the detected mitochondrial genomes; these 97 colors are also displayed in Figure S5 A) The percentage of each Saccharomyces species genome is stacked in the bar plot for each synthetic hybrid. Synthetic 103 hybrids are displayed in the order in which they were used ( Figure S1). yHRWh4 is shown multiple times because of its use 104 in two crossing schemes. We did not expect 100% genome contribution for each Saccharomyces species, even for recently 105 created hybrids, because some genomic regions (e.g. repeats) are not unambiguously detectable with Illumina sequencing 106 data. Percentage bars are colored according to each species' contribution. The strain names are colored based on the 107 mitochondrial inheritance inferred from mitosppIDer output ( Figure S4). Synthetic hybrids with two or more detected 108 mitochondrial genomes or regions are shown as gradient color based on the species' contributions. B) The nuclear 109 composition of the S. cerevisiae parent, synthetic hybrids, and evolved hybrids is plotted according to mitochondrial 110 inheritance. Hybrids with or without the S. cerevisiae mitochondrial genome are colored in red and light blue, respectively. 111 112 Figure S6 Growth kinetics for ancestral and evolved six-species hybrids in a 114

microplate. 115
Growth measured as area under the curve (AUC) for the ancestor and evolved six-116 species hybrids, after normalization to the S. cerevisiae reference strain, yHRW135 (full 117 microplate kinetic parameters are reported in Table S2). Different shapes indicate the 118 media in which the synthetic six-species hybrids were evolved. Colors differentiate parent 119 strains, while each data point represents an evolved replicate population.