Hybridization generates novel genomic combinations by merging divergent genomes, each of which has been uniquely refined by natural selection and stochastic processes through evolutionary time. This rapid blend of divergent DNA results in a myriad of genomic consequences and often leads to genomic instability in Saccharomyces yeasts (Dunn et al., 2013; Morard et al., 2020; Steensels et al., 2021) and other organisms (Dion-Côté and Barbash, 2017; Gibeaux et al., 2018). Recent years have seen considerable progress in identifying and describing genomic features of hybrids. For instance, it was recently found that genetic outcomes of hybridization are surprisingly repeatable in crosses between closely related species (Langdon et al., 2021; Moran et al., 2021). However, there are many open questions about the predictability of hybrid genomic architecture, the prevalence of genomic instabilities, and their evolutionary relevance. Can we predict hybrid genomes from parental genomic, demographic, or ecological features at all?

Some outcomes of hybridization are well-known. Hybridization between genetically distant yeast species usually causes significant genomic dysfunction. As a result, only a small fraction (<1%) of the spores of interspecific Saccharomyces F1 hybrids are viable and fertile. This lucky subset however can reproduce asexually and/or sexually and may be evolutionarily successful, in some cases even establish new lineages that are reproductively isolated from the parents. Reproductive isolation of intra- and interspecific hybrids can be mediated by ecology, e.g., through adaptation to a new niche (Leducq et al., 2016) or by endogenous factors, e.g., through allopolyploidy or chromosomal rearrangements preventing backcrosses (Charron et al., 2014). Whole genome duplication (WGD) has been shown to help circumvent severe fitness loss in the F2 hybrid generation (Marsit et al., 2021). However the F1 spores of allotetraploid hybrids are not able to mate with each other or a parent because their heterozygosity at the mating type locus renders them sterile (Pfliegler et al., 2012). Recently, two genetic mechanisms have been shown to restore fertility in normally sterile crosses, giving hybrids a route to meiotic recombination and adaptation. This includes return-to-growth (RTG) in intraspecific Saccharomyces cerevisiae crosses (Mozzachiodi et al., 2020), and massive loss of heterozygosity (LOH) (D'Angiolo et al., 2020).

But before hybrid genomes are stable and established, they often undergo wholesale, rapid changes (Sipiczki, 2008, 2018; Pfliegler et al., 2012; Lopandic, 2018; Steensels et al., 2021). Due to the failure of homeologeous chromosomes to pair, and antirecombination preventing regular chromosomal segregation in hybrid meiosis (Hunter et al., 1996; Rogers et al., 2018; Bozdag et al., 2021), the gametes (spores) of F1 hybrids are often aneuploid (reviewed in Gilchrist and Stelkens, 2019). Aneuploidy can cause the dysregulation of gene expression, which is often detrimental to the cell (Pavelka et al., 2010a,b; Sheltzer et al., 2012; Dürrbaum and Storchová, 2016). Thus, most extra chromosomes are short-lived and shed in consecutive mitotic divisions. However, some aneuploidies have been shown to be well tolerated in wild strains of S. cerevisiae through gene dosage compensation (Gasch et al., 2016; Tsai and Nelliat, 2019; Scopel et al., 2021) and some specific disomies, trisomies or segmental aneuploidies can in fact provide quick adaptive solutions, especially when cells face environmental stress (e.g., heat or antibiotics) (Rancati et al., 2008; Selmecki et al., 2009; Chen et al., 2012; Yona et al., 2012; Morard et al., 2019). The higher the relative fitness effect of a specific aneuploidy, the more likely this aneuploidy will remain a more permanent feature of the hybrid genome, but so far these effects remain unquantified in interspecific yeast hybrids.

Here, for the first time, we test for patterns of chromosomal mis-segregation across a diverse set of hybrid genomes with different species ancestries, evolutionary histories, and ecological niches. We predict aneuploidies and their adaptive relevance to vary with the parental genomic background of the hybrid cross and with environmental conditions. Specifically, we expect more genetically divergent parents to generate more aneuploid offspring. We also predict that the more skewed the parental genomic contributions are in a hybrid genome, the fewer aneuploidies it will carry, because overall sequence similarity increases, mitigating the risk for chromosome missegregation. Consequently, we predict “older” introgressed genomes to be more euploid and stable overall, as they had more time for postzygotic genomic processes after hybridization.

Due to ongoing advances in sequencing technologies and an increasing appreciation for the industrial and evolutionary potential of yeast hybrids, a large number of hybrid genomes have been recently sequenced. Here, we employed cutting-edge bioinformatic tools and pipelines (sppIDer, AAF, MuLoYDH, BWA, FreeBayes, Control-FREEC) and custom scripts to analyze the genomic architecture of all 204 interspecific Saccharomyces hybrids published to date (Figure 1). This collection contains hybrids isolated from fermentation environments (beer, lager, wine, and whiskey), but also from semi-wild, human-associated environments (olives, fruit, forest soils, and clinical and laboratory settings) (Supplementary Data Sheet). Most hybrids have two parental species, but some genomes contain contributions from three or even four different species. In most cases, but not all, the species identity of the parents is known. Our analysis includes early generation hybrids with largely equal genomic content from both parent species, but the large majority of hybrids are older and have more complex evolutionary histories, i.e., they are the result of postzygotic genomic processes including LOH, recurrent miss-segregation of chromosomes in meiosis and mitosis, and repeated backcrossing with one of the parent species. Although the hybrids in this study are not the direct F1 or F2 offspring of interspecific crosses, they are derived from this “true” hybrid offspring, resulting in genetically admixed hybrid segregants and are referred to as hybrids throughout this study.

Our study provides a comprehensive analysis of the prevalence of aneuploidy, similarities in the genomic composition, and mtDNA inheritance patterns across this diverse set of hybrid genomes. We highlight genomic patterns that apply to hybrids across all species and ecological backgrounds, but also report interesting cross- and environment-specific idiosyncrasies. Our results are an important step in understanding the factors shaping divergent hybrid genomes.

Hybridization between divergent species generates novel genomic combinations and offers insights into the genomic architecture underlying fitness. Interspecific hybridization has been shown to have severe genomic consequences (Smukowski Heil et al., 2017; Zhang et al., 2020), often resulting in highly unstable genomes characterized by instability (Dunn et al., 2013; D'Angiolo et al., 2020; Morard et al., 2020; Steensels et al., 2021). Leveraging high-throughput sequencing data from 204 interspecific hybrids within the Saccharomyces genus, we offer a novel perspective on patterns of genetic ancestry, mtDNA inheritance, aneuploidy and LOH. We tailored and integrated existing bioinformatic pipelines (such as sppIDer and MuLoYDH) with other bioinformatic tools (such as AAF) and custom Python scripts, offering the unique opportunity to expand our knowledge of hybrid genomic architecture and facilitate the comparison of the genomic ancestry of hybrids with diverse genetic backgrounds.

The hybrid genomes analyzed here are the result of vastly divergent crosses, made from parent yeast species as genetically distant as humans and mice (Shen et al., 2018). For this study, we were constrained by the high-throughput sequencing datasets available for interspecific hybrids (Figure 1). The data is not uniformly distributed among the possible hybrid crosses within Saccharomyces. We recognize that this can limit and potentially bias interpretations across genetic ancestries due to varying sample sizes. Here, we have focused on general patterns and therefore hope to limit speculation in crosses with limited available data. Interestingly, we found that as genetic distance increased (from ~10 to ~22% nucleotide differences genome-wide), the number of shared sequences in hybrid genomes also increased (Figure 2C). Thus, more genetically divergent parents resulted in genomically more similar hybrids. The reasons for this are yet unclear but this might be the result of stronger negative epistasis at larger genomic distances constraining the number of viable genomic combinations in hybrids. This is interesting with respect to the evolution of reproductive isolation in this species group. Recent work suggests that a large contributor to inviability and sterility in homoploid yeast hybrids is the failure of homeologeous chromosomes to pair and antirecombination, causing chromosomal missegregation during meiosis (Ono and Greig, 2019; Bozdag et al., 2021). When antirecombination was suppressed in S. cerevisiae × S. paradoxus hybrids, viability of hybrid gametes increased 70-fold (Bozdag et al., 2021). Another process that can lead to increased rates of aneuploidy upon hybridization is the gradual loss of chromosomes through both meiosis and mitosis (Sipiczki, 2018, 2019). The remaining contributions to reproductive isolation between Saccharomyces yeast species are likely combinations of Dobzhansky-Muller incompatibilities (DMIs), i.e., complex negative epistatic interactions involving multiple loci with weak fitness effects, which are difficult to detect statistically (Li et al., 2013) and have so far only been conclusively demonstrated to cause mortality in a large set of intraspecific S. cerevisiae crosses (Hou et al., 2014). The patterns found here, of increased genomic similarity in more divergent crosses, are consistent with complex DMIs restricting hybrid genomic architecture.

Aneuploidy, the loss or gain of a chromosome, is a common result of hybridization. In S. cerevisiae × S. paradoxus hybrids, chromosomal non-disjunction leaves a substantial portion of F1 spores (the recombined gametic products of hybrid meiosis) with at least one aneuploidy (Boynton et al., 2018; Rogers et al., 2018). Here, we found that across all hybrid genomes, aneuploidy was very frequent with 44% of genomes having at least one significant deviation in sequencing read depth. The majority of these aneuploidies were chromosomal gains (36%), fewer were chromosome losses (15%). In a recent study on 237 laboratory S. cerevisiae × S. paradoxus F2 hybrids, we found that 88% of all genomes contained aneuploidies (Zhang et al., 2020). This supports the notion that early hybrid generations contain more aberrant chromosome numbers than later generation hybrids that are more stabilized, as is the case for many of the genomes included in this study. This result also shows that the prevalence of aneuploidy is more than twice as high in hybrid meiosis outcomes compared to regular meiosis within S. cerevisiae, where aneuploidy was recently reported to affect 21% of all crosses, with variation between clades (Scopel et al., 2021).

We found that aneuploidy prevalence strongly depended on the genetic background of the hybrid cross. Among cross types, aneuploidy ranged from none of the genomes containing aneuploidies to 54% of the genomes having aneuploidies (Figures 6C,D). Against our expectation, there was no relationship between aneuploidy and genetic distance between parental species. But interestingly, aneuploidy was most frequent when genomic contributions from the two parental species approached equal shares in hybrid genomes (Figures 6A,B). When hybrid genomes harbored genomic content mostly from a single species, with only minor introgressions from the other species, read depth variance was consistently low. Equal genomic contributions are indicative of the hybrid's evolutionary history, suggesting the hybridization event was relatively recent, with limited time for stabilizing postzygotic processes and backcrossing. Older hybrids with a more complex history likely benefit from genome stabilization, allowing them to return to euploidy, which is reflected in our data.

Another interesting pattern is that aneuploidy in yeast is more common in smaller chromosomes (Kumaran et al., 2013; Zhu et al., 2016; Peter et al., 2018; Gilchrist and Stelkens, 2019). We found the same relationship between chromosome size and aneuploidy (measured as mean read depth change) in this large set of interspecific yeast hybrids (Figure 7A). As chromosome size increased, the rate of aneuploidies decreased, likely due to the higher number of genes with important functions on larger chromosomes (Scopel et al., 2021). Smaller chromosomes tend to have fewer genes resulting in a smaller chance of dosage sensitivity. Other factors may also play a role in this, e.g., aneuploidy of the largest yeast chromosome 4 has been shown to cause a delay for entry into the cell cycle (Torres et al., 2007), likely due to the increased biological burden of protein synthesis of the extra chromosome copy. Together, this results in selection against aneuploidy of large chromosomes. In agreement with this, we found that larger chromosomes (including chromosome 4) were less likely to have gains or losses in these hybrid genomes.

Aneuploidy has been viewed as a double-edged sword with potential for adaptation (especially to stressful environments) but also fitness costs (Tsai and Nelliat, 2019). So far, aneuploidy has only been demonstrated to cause variation in fitness and adaptation in non-hybrid yeast crosses (e.g., Yona et al., 2012). For example, Scopel et al. (2021) report that whether a specific aneuploidy was neutral, detrimental or beneficial depended on the genetic background of crosses involving different strains of S. cerevisiae. It is reasonable to expect the same to be true in hybrids. Indeed, several industrial hybrid strains are stable aneuploids (Dunn et al., 2005; Peris et al., 2012; Okuno et al., 2016) and experimental evolution has recovered aneuploidies in different interspecific hybrids (Peris et al., 2017, 2020; Gallone et al., 2019; Langdon et al., 2019). In fact, due to the much higher frequency of chromosome missegregation in hybrid meiosis, we predict aneuploidy to feature more prominently in the adaptive evolutionary history of hybrid strains, simply because there are more occasions to recruit extra copies of genes located on aneuploid chromosomes for adaptation (Gilchrist and Stelkens, 2019).

Our results on mitochondrial DNA inheritance patterns confirm previous findings that hybrids usually retain mitochondria from the majority nuclear genome parent. For instance, it has been shown that S. cerevisiae × S. kudriavzevii hybrids with S. kudriavzevii mtDNA do not tolerate large losses of S. kudriavzevii nuclear genomic content, likely due to important species-specific interactions of proteins encoded in nuclear and mitochondrial genes (Peris et al., 2012, 2018). However, we saw remarkable exceptions to this rule, e.g., in S. cerevisiae × S. eubayanus hybrids that contained almost exclusively S. eubayanus mitochondria, even in genomes with a majority S. cerevisiae nuclear genome (Supplementary Figure 3B). These hybrids have been found in low or moderate temperature environments within the Northern European limit of the grape vine distribution, and mostly in low temperature lager fermentation environments (Figure 1), suggesting that these hybrids benefit from cold tolerance mitochondrial genes of the non-cerevisiae parent (Baker et al., 2019; Li et al., 2019). Only four hybrids within this cross inherited the S. cerevisiae mtDNA and three of these had a nuclear genome that was predominantly inherited from S. cerevisiae (>70%). In S. cerevisiae × S. uvarum hybrids it has been shown that mitochondrial inheritance patterns are dominated by one species and that the dominant mitochondrial parent is strain and cross specific (Verspohl et al., 2018). Therefore, mitochondria from certain S. cerevisiae strains were universally inherited, regardless of which S. uvarum strain it was crossed with. Similarly, S. uvarum strains were identified exhibiting dominant inheritance pattern independent of the S. cerevisiae mating partner. This suggests interactions between mtDNA and nuclear DNA or potentially be influenced by the mtDNA itself. Additionally, mitochondrial retention in S. cerevisiae × S. uvarum hybrids was recently shown to be heavily influenced by the effects of mitochondrial genes on nuclear expression and fitness, which in turn often depend on environmental factors (Hewitt et al., 2020). For instance, in rich media at cold temperatures, S. uvarum mitochondria were retained, whereas S. cerevisiae mitochondria were retained on non-fermentable carbon sources regardless of temperature. Thus, both intrinsic genetic (co-evolutionary constraints between mitochondrial and nuclear genes), and environment-dependent adaptive advantages may explain the patterns we observe here.

In the aftermath of faulty chromosomal segregation in yeast hybrid meiosis, various mechanisms can lead to the LOH, including the loss of entire chromosomes of one of the parental subgenomes (hemizygosity), base mismatch repair during homeologous recombination of heterozygous sites and break-induced replication (D'Angiolo et al., 2020). LOH events have recently been shown to occur at a consistently higher rate than mutations indicating a vital role in genome evolution (Dutta et al., 2021). Here, we found in a subset of ecologically diverse S. cerevisiae × S. paradoxus hybrids, that LOH was common, at approximately 30% of genomic regions. The majority of these regions were homozygous for S. cerevisiae genetic markers. It is interesting that these five hybrids share common LOH regions, despite being isolated from different ecological niches (olives, soil and wine). There are vast regions on chromosome 12 primarily homozygous for S. cerevisiae markers, whereas chromosome 5 was largely left heterozygous. Although our analysis here is quite limited in sample size (five genomes only), the patterns we see may be indicative of intrinsic genetic incompatibilities between divergent genes located in these regions. LOH has been shown to allow hybrids to overcome reproductive isolation and enables introgression between vastly divergent yeast species (D'Angiolo et al., 2020). Alternatively, LOH can be viewed as maximizing fitness potential within the hybrid genome. There is evidence for LOH to play a major role in the fitness and adaptation of yeast both in naturally occurring hybrids and interspecific crosses generated and evolved in the laboratory (Smukowski Heil et al., 2017; Lancaster et al., 2019). However, the small sample size examined here does not allow for strong conclusions about these complex interactions.

In summary, our analysis provides evidence for increased similarity of hybrid genomes with increasing parental sequence divergence. This is interesting, considering that the more divergent two parental genomes are, the more can potentially go wrong during hybrid meiosis. On the other hand, more sequence differences lead to less efficient chromosome pairing in meiosis, which in turn leads to fewer meiotic divisions and less recombination and overall, potentially leading to more similar hybrid offspring. It is also important to note that all the hybrid genomes included in this study are viable outcomes of hybridization that had time to stabilize their genomes through postzygotic processes since the hybridization event. Thus, the patterns we observe may be caused by stronger negative epistasis at larger genomic divergence, putting constraints on hybridization outcomes. We also provide the first evidence for higher genome stability (less aneuploidy) the more the parental genomic contributions deviated from a 1:1 ratio in the hybrid genome. However, some questions remain. For instance, do hybrid genomes contain different levels of structural variation than their parents, such as copy number variation, translocations or inversions? Since all sequence data available for hybrid yeast genomes to date is short read data, analysis of structural variants has been limited. In addition, most genomic data so far comes from only two yeast species (S. cerevisiae and S. paradoxus). This is expected to change soon with high quality long read assemblies becoming more affordable. Another open question is about the role of transposable elements (TEs) in hybrid yeast adaptation and speciation. Hybridization has been shown to not increase the rate of TE mobilization in natural and laboratory interspecific crosses (Hénault et al., 2020; Smukowski Heil et al., 2021). But TE copy numbers in hybrids were strongly dependent on the specific genotypes used for the cross (Hénault et al., 2020) and interestingly, species-specific mitochondrial inheritance changed transposition rate by an order of magnitude (Smukowski Heil et al., 2021). Scrutinizing patterns and predictors of genomic instability in hybrids as we have done here, can potentially advance our understanding of the role of hybridization in adaptation to environmental stress, the evolution of drug resistance, and cell line disorders.

Publicly available datasets were analyzed in this study. The datasets analyzed for this study can be found in the European Nucleotide Archive (https://www.ebi.ac.uk/ena/browser/home). The accession numbers for each hybrid genome are shown in Supplementary Data Sheet 1.

DB and RS conceptualized and designed the study and wrote the manuscript. DB performed data analysis and data visualization. DP provided assistance with sppIDer scripts and performed mitochondrial genome assembly of the Holarctic S. eubayanus strain. All authors read, commented, and approved the final manuscript.

This work was supported by the Swedish Research Council (2017-04963 to RS), the Knut and Alice Wallenberg Foundation (2017.0163 to RS), and the Wenner-Gren Foundations (UPD2018-0196, UPD2019-0110 to DB). Work performed at the National Genomics Infrastructure (NGI)/Uppsala Genome Center (project SNIC 2019/8-23) has been funded by RFI/VR and Science for Life Laboratory, Sweden. DP is a postdoctoral researcher funded by the Research Council of Norway grant no. RCN 274337 and a senior researcher supported by the Valencian International University (VIU).

DP declares receiving royalties from VIU based on publication productivity. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.