Enforcement of Postzygotic Species Boundaries in the Fungal Kingdom

SUMMARY

Understanding the molecular basis of speciation is a primary goal in evolutionary biology. The formation of the postzygotic reproductive isolation that causes hybrid dysfunction, thereby reducing gene flow between diverging populations, is crucial for speciation. Using various advanced approaches, including chromosome replacement, hybrid introgression and transcriptomics, population genomics, and experimental evolution, scientists have revealed multiple mechanisms involved in postzygotic barriers in the fungal kingdom. These results illuminate both unique and general features of fungal speciation. Our review summarizes experiments on fungi exploring how Dobzhansky-Muller incompatibility, killer meiotic drive, chromosome rearrangements, and antirecombination contribute to postzygotic reproductive isolation. We also discuss possible evolutionary forces underlying different reproductive isolation mechanisms and the potential roles of the evolutionary arms race under the Red Queen hypothesis and epigenetic divergence in speciation.

INTRODUCTION

Described by Darwin as “the mystery of mysteries,” how a new species arises in nature is a long-standing question in evolutionary biology. Over the past few decades, scientists have dissected the genetic basis of hybrid incompatibility in several model organisms, shedding light on the molecular mechanisms underlying reproductive isolation. When combined with ecological and population genomics data, this information can advance our understanding of the general factors driving speciation. The fungal kingdom contains species found in diverse ecological niches (1). Moreover, fungi often establish close relationships with plants and animals (2–5). The ever-increasing number of fungal species that have been investigated at the population and genomic levels represents a wonderful resource for scientists to study speciation.

Speciation is an evolutionary process during which a population diverges and becomes genetically distinct from other populations (6). In the past, morphological features and growth profiles were often used to classify fungal species (7). With conceptual and technical advances, the biological species concept (BSC) and the phylogenetic species concept (PSC) have been integrated to update the relationships between many previously defined species (7, 8). The BSC is the most accepted concept for defining species in eukaryotic organisms, but it requires the measurement of reproductive isolation between different species. Some fungi are observed to reproduce only asexually or cannot be mated under laboratory conditions, which complicates the use of the BSC. On the other hand, the PSC uses gene or genome sequence similarity to classify species, which can be applied even to unculturable isolates (7). Moreover, several studies have shown a good correlation between sequence divergence and the species defined by the BSC (9, 10), further supporting the use of the PSC to complement the BSC in mycology. Currently, many fungal species are defined by the PSC, and the levels of reproductive isolation remain undetermined in some clades.

The fungal kingdom constitutes both multicellular and unicellular species. Multicellular fungi are composed of hyphal filaments that can bunch together into a structure called the mycelium. During the sexual cycle, hyphae from two individuals fuse to form a mycelium that contains haploid nuclei of both parents (a stage called plasmogamy). The dikaryotic state may persist for many generations before nuclear fusion (a stage called karyogamy), depending on the species. Once a diploid zygote has formed, meiosis progresses to produce either four haploid nuclei or four haploid spores. Yeasts are unicellular fungi that have evolved in distinct phylogenetic lineages. Two haploid yeast cells of different mating types can fuse to produce a diploid zygote. Zygote cells can enter the meiotic cell cycle to produce haploid spores or may continue to divide asexually. Several yeast species have been studied extensively in the laboratory, as they are model organisms widely adopted for genetic and molecular biology experiments.

Fungi possess several unique characteristics in terms of their cell physiology and life cycle, allowing them to speciate via mechanisms distinct from those in animals and plants. First, asexual reproduction is extensive among fungi, potentially facilitating clonal evolution and even incipient speciation (11, 12). Under natural selection, asexual reproduction fosters genetic homogeneity within subpopulations sharing similar ecologies but allows genetic diversity to accumulate over time between subpopulations of different niches. For example, it was estimated that one outcrossing event happens every ~50,000 generations in the budding yeast Saccharomyces cerevisiae. Therefore, after an outcrossing event, the accumulation of loss-of-heterozygosity (LOH) events might lead to rapid differential genome evolution during vegetative growth (13). Even in similar environments, clonal evolution can lead to diverse populations by means of selfish genetic-element-driven evolution or genetic drift (14, 15). Moreover, changes in ploidy or chromosome number are commonly observed in fungi when forced to adapt to environmental fluctuations (16, 17). Asexual reproduction allows such adapted clones to expand and form distinct populations. Polyploidy and aneuploidy facilitate chromosomal rearrangements and genome evolution (18–21), which can also lead to reproductive isolation and speciation. Many fungal species establish symbiotic relationships with plants and animals, and these host-symbiont interactions are important drivers of genetic diversity in both fungi and their hosts (4, 22). The dynamics of these host-symbiont interactions (such as arms races between hosts and parasites) and related traits (such as host specificity, symbiont transmission rate, and reproductive mode of the fungal symbionts) often contribute to reproductive isolation and, ultimately, speciation rates (23, 24). In this article, we review the known molecular mechanisms underlying postzygotic isolation in fungi and discuss the possible evolutionary forces involved in this process. Since most genetic data are derived from a few well-studied species, the information that we present is inevitably biased. However, we incorporate current knowledge into more generalized models when we interpret population and genomics data for other species.

MOLECULAR MECHANISMS UNDERLYING POSTZYGOTIC ISOLATION

Dobzhansky-Muller Incompatibility

Dobzhansky-Muller incompatibility is the best-known mechanism contributing to reproductive isolation between different species. It is induced by interacting components that cannot function properly when alleles from different species are mixed in hybrids (25–27). If the incompatibility has a dominant effect, it leads to F1 hybrid lethality or sterility. In contrast, recessive incompatibilities affect only the haploid F1 gamete or diploid F2 progeny with homozygous alleles (Fig. 1). When studying genetic incompatibility, scientists often search for interacting genes (or other components such as centromeric DNA) harbored by the nuclear genome (nuclear-nuclear incompatibility). However, recent studies have indicated that interactions between nuclear and mitochondrial genomes (mitochondrial-nuclear incompatibility) can also play a vital role in reproductive isolation (28).

FIG 1.

Reproductive isolation caused by Dobzhansky-Muller incompatibility. A and B represent two interacting components that diverge among populations during evolution. A and B can be encoded by nuclear or mitochondrial genomes, and proper A-B interactions are important for cell physiology. If the incompatibility has a dominant effect, it will lead to F1 hybrid lethality or sterility (represented by dashed circles). Recessive incompatibilities affect only the haploid F1 gamete or diploid F2 progeny with homozygous alleles.

Yeast geneticists noted quite early that strong postzygotic isolation exists between the budding yeast S. cerevisiae and its close relatives (29). Although F1 hybrid diploids of budding yeast grow normally under laboratory conditions, their haploid gametes (spores) display low viability, indicative of hybrid sterility. Subsequent studies have shown that F1 sterility is caused primarily by mismatch repair-dependent antirecombination during meiosis I (discussed in detail below) rather than strong dominant incompatibility (30–32). Nevertheless, multiple cases of recessive incompatibilities eliciting different effects have been revealed. A genome-wide analysis of hybrids between S. cerevisiae and Saccharomyces paradoxus indicated that only complex incompatibilities involving multiple loci with mild effects exist between these two nuclear genomes (33). However, the resolution of that analysis was not sufficiently high to identify the incompatible loci. More recently, it was proposed that multiprotein complexes may form the basis of complex incompatibility (34). Since the formation of a functional protein complex involves interactions between different subunits, compromising individual interactions may yield only weak effects. However, when the assembly of an entire complex is disrupted due to failures in multiple interactions, the effects can be synergistic, as predicted under the complex incompatibility model (35). Many cellular processes and regulatory pathways are executed by protein complexes (36), but it remains to be determined if incompatible complex subunits have a general impact on reproductive isolation during speciation.

The first case of strong recessive incompatibility in yeast was revealed by a systematic screen using chromosome replacement lines in which individual S. cerevisiae chromosomes were replaced by homologs from Saccharomyces uvarum (previously known as S. bayanus) (37). The resulting incompatibility is caused by the AEP2 gene located on S. bayanus chromosome 13 and the OLI1 gene located on the S. cerevisiae mitochondrial genome (mitochondrial DNA [mtDNA]), leading to mitochondrial dysfunction and sporulation defects. In wild-type cells, Aep2 binds to the 5′ untranslated region (UTR) of OLI1 mRNA to facilitate the translation of Oli1, a subunit of the ATP synthase complex (38). The 5′ UTR of OLI1 mRNA has significantly diverged between these two species, resulting in the failure of S. bayanus Aep2 to interact with S. cerevisiae OLI1 mRNA. Thus, the AEP2-OLI1 gene pair represents two-locus Dobzhansky-Muller mitochondrial-nuclear incompatibility.

In a follow-up study, two other strong mitochondrial-nuclear incompatibilities were discovered among three yeast species, i.e., S. cerevisiae, S. paradoxus, and S. bayanus (39). One of them, the MRS1-COX1 gene pair, again involves a protein-RNA interaction. The Mrs1 protein is required for the splicing of specific introns of mitochondrial COX1 (40). The loss of an ancient COX1 intron from the S. cerevisiae lineage means that S. cerevisiae Mrs1 cannot splice this ancient intron possessed by the S. paradoxus and S. bayanus mitochondrial genomes, leading to respiratory defects in hybrid gametes (39). Interestingly, these strong mitochondrial-nuclear incompatibilities appeared at different evolutionary time points that correlate with the divergence of different yeast lineages, implying that mitochondrial-nuclear incompatibilities might be among the first genetic barriers to arise during yeast speciation.

Why is mitochondrial-nuclear incompatibility so commonly observed among the yeast species closely related to S. cerevisiae (also known as the Saccharomyces sensu stricto yeasts)? Several features of the yeast mitochondrial genome, including its high mutation rate, dynamic genomic structure, and small effective population size, make it a promising system for population-specific interactions to evolve between the mitochondrial and nuclear genomes. The extensiveness of asexual reproductive lifestyles in yeasts allows mtDNA to accumulate mutations easily in individual lineages. Once mtDNA has changed, the nuclear genome needs to adapt due to the intimate interactions between these two genomes. If mitochondrial-nuclear coevolution is driven in different directions in two populations, it may eventually lead to hybrid incompatibility and speciation (41). A recent study shows that nucleus-encoded pentatricopeptide-repeat-containing (PPR) proteins display considerable binding flexibility, which allows them to coevolve rapidly with their substrates (42). PPR proteins belong to the largest RNA-binding protein family in eukaryotes, which function primarily in mitochondria or chloroplasts (43). Most yeast PPR proteins are involved in stabilizing and/or regulating the translation of mitochondrion-borne RNAs (44). In Saccharomyces sensu stricto yeasts, about two-thirds of PPRs have evolved different levels of incompatibility between species (42), explaining why many observed mitochondrial-nuclear incompatibilities involve protein-RNA interactions. The highly evolvable nature of PPR proteins enables cells to accelerate compensatory adaptations to the altered RNA substrates encoded by mtDNA. It is worth mentioning that in plants, such PPR proteins are often involved in restoring the fertility caused by mitochondrion-associated cytoplasmic male sterility (CMS) (45). This further suggests the crucial role of PPR proteins in mitochondrial-nuclear coevolution.

The mitochondrial-nuclear coevolution model is further supported by studies on S. cerevisiae populations. When endogenous mtDNA of S. cerevisiae cells was replaced by mtDNA from different natural populations, the cybrid cells often exhibited reduced fitness under various conditions (46, 47). These results suggest that coevolution between mitochondrial and nuclear genomes occurs quickly and frequently, even at the population level. The next key question is the primary driving force underlying such coevolution, i.e., whether it is due simply to genetic drift or caused by environmental adaptation or intergenomic conflicts (41).

Killer Meiotic Drivers in Fungi

Meiotic drive systems can function as a genetic barrier to reduce gene flow between distinct populations. They are well known in plants and animals, but a growing body of evidence indicates that meiotic drives may also be common in fungi (14, 15). True meiotic drivers and killer meiotic drivers are two general classes of meiotic drive systems (48). In organisms where all meiotic products do not differentiate into oocytes, true meiotic drivers favor their own transmission to the progeny by manipulating the meiotic system so that driver carriers have a greater chance of becoming oocytes than nondriver carriers. In contrast, killer meiotic drivers behave like genetic parasites that destroy sibling gametes lacking the driver allele. Killer meiotic drivers have been observed in a wide range of eukaryotes and are important contributors to the causes of infertility, genome evolution, and speciation (48). In fungi, all meiotic products become haploid spores, and no specialized oocyte is formed, so killer meiotic drivers are common. Below, we discuss a few well-characterized killer meiotic drive systems identified in Podospora, Neurospora, and Schizosaccharomyces species. The killer meiotic drivers observed in species of the plant-pathogenic fungal genera Bipolaris and Fusarium are not included here since much less is known about them (49).

Podospora anserina het-s system.

In the dung-inhabiting ascomycete fungus Podospora anserina, spore-killing meiotic drive factors often result in the abortion of two of the four spores in asci. At least two distinct types of killer meiotic drives have been identified in P. anserina, i.e., the het-s and Spok systems (50, 51). The [Het-s] cytoplasmic genetic element was initially identified in 1952 and was recognized as a prion nearly half a century later (52). During the sexual cycle of P. anserina, the male gamete contributes very little cytoplasm to the heterokaryon, so the cytoplasm of a zygote is essentially maternally inherited. In the diploids generated from a cross between [Het-S] males and [Het-s] females, the maternally inherited cytoplasmic HET-s prions (the killers) are transmitted to all meiotic products (spores). However, het-S allele-inheriting spores will also produce the HET-S protein (the target), which can be converted by HET-s prions into an oligomeric integral membrane protein, resulting in the death of het-S spores due to the destabilization of their cell membrane (53). The het-s system requires an interaction between the killer [Het-s] prion and the target HET-S protein for spore killing, representing a typical example of the “killer-target drive system” (Fig. 2A). A similar killer-target drive system is also observed in the fruit fly Drosophila melanogaster (54), suggesting that it is a common strategy for meiotic drivers.

FIG 2.

Reproductive isolation caused by two types of killer meiotic drives. (A) Model for killer-target meiotic drive systems. CT, cytoplasmic transmission. (B) Model for poison-antidote meiotic drive systems. The poison and antidote can be encoded by a single genetic locus or two genetic loci.

Despite its strong impact on reproductive isolation, the het-s system is speculated to have evolved from systems for allorecognition, i.e., the ability to discriminate self from nonself (55). Like other fungi, Podospora spp. can perform hyphal anastomosis, by which vegetative hyphae fuse upon encountering each other. This process facilitates nutrient acquisition and colony establishment between cooperating colonies (56). However, it also increases the risk of spreading mycoviruses or other parasites between different populations (57). The het loci function in allorecognition by controlling the viability of heterokaryons, termed heterokaryon incompatibility. This cell death reaction is triggered when genetically distinct hyphae fuse, thereby reducing the potential for parasite transmission between populations (55). In P. anserina, several het loci have been shown to exert pleiotropic effects on hybrid sterility or inviability (58). It is important to know the extent to which these loci contribute to reproductive isolation in natural habitats. By combining field observations, population genomics data, and computer simulations, a recent study elegantly addressed this issue and showed that the antagonistic interactions of two unlinked loci, het-r and het-v, allow coexisting P. anserina populations to maintain reproductive isolation in the wild (59). If the situation remains unchanged, these populations may accumulate further divergence and eventually become different species.

P. anserina Spok system.

In addition to the het-s system, Podospora spp. host another multigene family of poison-antidote meiotic drivers, the Spok spore killers (60). In heterozygous diploids, the Spok genes act as autonomous elements that kill nearly all of the spores without inheriting the killer gene. Various Spok genes (Spok1, Spok2, Spok3, and Spok4) have been discovered in different isolates of Podospora. Spok1 was first discovered in Podospora comata, a sibling species of P. anserina, whereas the homologous gene Spok2 was found at high population frequencies among P. anserina strains (60). In contrast, Spok3 and Spok4 occur at low to intermediate frequencies and are always associated with a large genomic region termed the Spok block (51). This Spok block can be found at different chromosomal locations in different individuals, but only 1 copy exists per genome. Location switching of the Spok block is probably mediated by a giant tyrosine-recombinase-mobilized DNA transposon, Enterprise (61). The number of Spok genes in the Spok block (which carries Spok3, Spok4, or both) varies among isolates. Combined with the effects of variable genomic locations, these divergences may explain the different degrees of spore-killing phenotypes observed in various crosses.

Some Spok genes can both drive and suppress other drivers, leading to complicated epistatic relationships among Spok genes. For instance, Spok1 and Spok4 confer mutual resistance to each other (51). Moreover, Spok1 can act as a factor in resistance to Spok2 but not vice versa (60). Although Spok genes are localized at different genomic positions in different P. anserina strains, the sequences of Spok genes are highly conserved in P. anserina. The predicted enzymatic functions (nucleases and kinases) of SPOK proteins probably constrain their evolutionary divergence (51, 60) since both the killing and resistance functions are attributed to these two separate functional domains (51).

Neurospora intermedia Sk-1, Sk-2, and Sk-3.

Wild populations of Neurospora sitophila and N. intermedia carry spore killers (Sk) similar to other poison-antidote meiotic drivers. Spore killer 1 (Sk-1) was the first spore killer found in Neurospora about 40 years ago (62). However, it was not until recently that a single gene, Spk-1, was identified to carry both poison and antidote functions, like the Spok system in Podospora (63). In contrast, the killer and resistance factors are encoded by multiple genes in the Spore killer 2 (Sk-2) and Spore killer 3 (Sk-3) systems. They are complex meiotic drive elements mapping to a large 30-cM region surrounding the centromere of chromosome 3, where recombination is suppressed between killer and sensitive strains (64). Rfk-1, a gene required for Sk-2-based spore killing, has been mapped to the right edge of the recombination suppression region (65, 66), whereas the resistant-to-spore-killer (rsk) gene has been identified on its left flank (67). Distinct alleles of the rsk gene serve as antidotes for Sk-2 and Sk-3 drivers. However, the Sk-2 rsk allele does not protect against the Sk-3 poison and vice versa (62). The rsk gene encodes a fungus-specific protein without recognizable motifs (67), and the molecular mechanisms underlying how rsk prevents the spore destruction caused by Sk driver poisons have not yet been fully elucidated.

A dense set of inversions interspersed with transposable elements has been identified in both the Sk-2 and Sk-3 haplotypes, likely contributing to the suppression of recombination (68). Interestingly, although Sk-2 and Sk-3 are located in the same chromosomal region, phylogenetic analysis indicates that they originated independently or at least have a long separate evolutionary history (68). The similar patterns of tandem inversions and dense repeat clusters in Sk-2 and Sk-3 strains likely reflect the convergent genetic architecture necessary to generate tight linkages between genes causing meiotic drive and resistance.

Schizosaccharomyces pombe wtf genes.

The Schizosaccharomyces pombe wtf gene family contains several active meiotic drive genes. Like other single‐gene drivers (such as the P. anserina Spok system), they are each self‐sufficient in terms of executing meiotic drive and resistance. Two proteins, a poison and an antidote, can be generated from a wtf gene via alternative transcriptional and translational start sites. A short transcript encoding the gamete-killing poison is first expressed prior to the meiotic divisions, so all four spores generated by a heterozygote contain the poison. However, a long transcript encoding an antidote is produced after spore individualization, and this protein largely remains within the wtf allele-carrying spores. This coordinated expression of poison and antidote proteins results in biased wtf transmission to viable spores (69, 70) (Fig. 2B). Due to the shared sequences between the poison and antidote proteins, mutations that make novel poisons simultaneously generate compatible antidotes, thereby rendering the system highly evolvable. In a study that surveyed 22 other wtf-like genes for meiotic drive phenotypes, eight new drivers sharing only 30 to 90% protein identity with other known drivers were identified. Even though the sequences of these wtf genes are highly divergent, they are generally driven into >85% of gametes when heterozygous (71).

How does a given Wtf poison kill a spore in which it resides, and how does an antidote neutralize the poison? Wtf proteins contain multiple predicted transmembrane domains (72). Poisoned spores are often misshapen and lose their membrane integrity, suggesting that Wtf poisons may kill the cells by oligomerizing to form a pore in vital membranes as spore development progresses (69, 70). A recent study showed that the Wtf4 poison protein forms dispersed toxic aggregates that trigger nuclear condensation and fragmentation. However, the Wtf antidote protein can coassemble into the aggregate with the Wtf4 poison and promote its trafficking to vacuole-associated sites (73). Moreover, ectopically expressed Wtf poison and antidote proteins cause similar phenotypes in vegetative S. cerevisiae cells, indicating that S. pombe-specific or meiosis-specific pathways are not required for the killing and neutralization effects of wtf genes (73). The protein aggregation model offers a possible explanation as to why highly diverse wtf genes can result in the same phenotype since the self-aggregating feature is probably less evolutionarily constrained than maintaining a specific enzymatic activity or interaction partner.

Evolutionary implications of killer meiotic drivers in speciation.

Killer meiotic drivers are usually regarded as selfish genetic elements that sacrifice the benefit of their host to enhance their own transmission in the population. Since there is strong selective pressure for them to exert their effect quickly, meiotic driver-induced reproductive isolation can be established rapidly between individual populations. In S. pombe, reproductive isolation is commonly observed between isolates with very low genetic divergence (>99.5% DNA sequence identity) (74), which may be explained by the high copy numbers of wtf genes in their genomes. However, once a driver has become fixed in a population, the selective pressure to maintain the driver’s killing function is alleviated. Depending on its cost (i.e., the cost of producing a poison that is constantly neutralized by its antidote), a driver may degenerate and completely lose its function. Nevertheless, how killer meiotic drivers contribute directly to speciation remains unclear, representing a fertile ground for further study.

In addition to their driver-killing effects, deleterious mutations linked to single-gene drivers (such as the P. anserina Spok system) have a high likelihood of being fixed in a population together with the driver. Such events can lead to suppressor coevolution (i.e., of the deleterious mutation), thereby creating another reproductive barrier. In multiple-gene drivers (such as the Sk system of Neurospora), recombination suppression often occurs in the chromosomal region between killing and resistance genes to enhance their linkage, which can result in chromosomal rearrangements and the accumulation of mutations. These changes can all contribute to reproductive isolation among different populations, even after the drivers have lost their functions.

Chromosomal Rearrangements

Chromosome rearrangements can have substantial impacts on hybrid sterility and viability. During meiosis I, homologous chromosomes pair, recombine, and then segregate. If two homologous chromosomes are not collinear due to chromosomal rearrangements, this may lead to the unbalanced segregation of the genes on the translocated arms. The functions and numbers of the missegregated genes will determine the severity of such events (Fig. 3). Moreover, the effect can be multiplied if there are multiple translocated chromosomes. A well-documented example is the S. paradoxus-Saccharomyces cariocanus hybrid. These two closely related yeast species have highly similar genome sequences but differ in four large reciprocal chromosomal translocations. The spore viability of the hybrid is drastically reduced to ~6% of that of the individual pure species, i.e., close to the value expected from the unbalanced segregation of essential genes on the translocated arms (75, 76). It was suggested previously that chromosomal rearrangements are not the primary mechanism underlying the reproductive isolation of the Saccharomyces sensu stricto group given that chromosome translocation events are not correlated with the sequence-based phylogeny (75). However, recent population genomics data have shown that in natural isolates, chromosomal rearrangements are far more frequent than previously believed (77, 78), indicating that they play a general role in blocking gene flow among yeast populations.

FIG 3.

Reproductive isolation caused by chromosomal rearrangements. Chromosomal translocations are used as an example. If the translocation contains the essential gene (represented by dark- and light-gray blocks), some meiotic products will lack this essential gene and die (represented by dashed circles).

A recent genomics analysis of pathogenic Cryptococcus species revealed that multiple chromosomal inversions and translocations exist among three closely related species (with ~6% sequence divergence), indicating that chromosomal rearrangements might contribute to reproductive isolation in the genus Cryptococcus (79). By reconstituting such rearrangements in a single species, Cryptococcus neoformans, it was revealed that chromosomal translocations can indeed create a strong reproductive barrier (80).

How does a rearranged chromosome become fixed in a local population? Many yeast species are homothallic, meaning that haploid cells harboring a rearranged chromosome can switch their mating type and mate to generate diploid cells homozygous for rearranged chromosomes. Moreover, asexual reproduction is common in the fungal life cycle, allowing cells with rearranged chromosomes to establish clonal populations. Therefore, rearranged chromosomes may have a higher likelihood of becoming fixed in fungal species than in other eukaryotes, even without endowing strong beneficial effects. In addition to genetic drift, rearranged chromosomes can also spread in the population due to adaptive benefits. Adaptive chromosomal rearrangements are frequently observed in natural yeast strains and other fungi (81–84). In a study by Colson and colleagues, chromosome translocations similar to those detected in Saccharomyces mikatae were recreated in S. cerevisiae before measuring cell fitness (85). The translocated strains consistently outcompeted the reference strain (lacking translocation) under various conditions, implying that the chromosomal rearrangements detected in S. mikatae might have become fixed in the population due to adaptive benefits. Meiotic drive represents another potential mechanism for fixing chromosomal rearrangements in a population (86, 87). As described above for Neurospora, the Sk system represses recombination between the killer and resistance genes by generating chromosomal rearrangements, and such events have evolved multiple times, as reported for Sk-2 and Sk-3 (68). In Schizosaccharomyces, Zanders and colleagues also observed that a chromosomal translocation links a pair of driving loci, further contributing to hybrid infertility (88).

Mismatch Repair-Dependent Antirecombination

Mismatch repair-dependent antirecombination is the best-studied mechanism of reproductive isolation in the Saccharomyces sensu stricto yeasts. Mismatch repair systems correct DNA mismatches generated from DNA polymerase misincorporation errors or DNA damage. The mismatch repair machinery has been found to suppress recombination between homologous regions with certain levels of sequence divergence (termed antirecombination) (89). Thus, it can interfere with meiotic recombination and cause chromosome missegregation in hybrid cells formed between different species, leading to hybrid sterility (Fig. 4). By deleting PMS1 or MSH2, which are involved in the mismatch repair system, fertility increased ~7- to 10-fold in hybrids of S. cerevisiae and its sister species S. paradoxus (30). In a follow-up study, hybrid fertility was further enhanced to nearly nonhybrid levels by repressing the SGS1 and MSH2 genes (32). In these mutant cells, improved fertility is correlated with increased meiotic recombination, supporting the prediction of the antirecombination model. Similarly, disruption of the mismatch repair system increased the spore viability of hybrid cells formed between diverged S. cerevisiae strains (90).

FIG 4.

Reproductive isolation caused by mismatch repair-dependent antirecombination. When two parental genomes have high sequence divergence, mismatch repair-dependent antirecombination interferes with meiotic recombination in hybrid cells, leading to chromosome missegregation. The spores missing essential chromosomes are inviable (represented by dashed circles).

The dominant effect of antirecombination suggests that it is a major cause of F1 hybrid sterility in Saccharomyces sensu stricto yeasts. However, antirecombination as a reproductive isolation barrier works only when species accumulate sufficient sequence divergence throughout the whole genome, meaning that the yeast species need to be separated for a long time to accumulate such divergence. Geographic or ecological isolation is probably not the only factor underlying such separation since these species are often found in the same locations or ecological niches, and outcrossing can be further facilitated through insect vectors (91, 92). Other above-mentioned reproductive isolation barriers are likely to contribute to reducing gene flow during the incipient stage of speciation, allowing populations to diverge further. Interestingly, in animals, hybrid sterility caused by Prdm9 incompatibility also involves meiotic recombination and chromosomal pairing (93). This provides an excellent example of how genetic incompatibility interacts with the chromosomal level of sequence divergence to establish reproductive isolation between populations.

CURRENT APPROACHES FOR IDENTIFYING THE MECHANISMS INVOLVED IN POSTZYGOTIC ISOLATION

Chromosome Replacement and Introgression

Several genetic tools have been established in budding yeast that allow scientists to exchange individual chromosomes or mitochondrial DNA between different species (94). As described above, chromosome replacement lines have been constructed between multiple yeast species to identify genes involved in strong or mild mitochondrial-nuclear incompatibilities (37, 39, 42, 95). These replacement lines also represent good material for dissecting transcriptional misregulation or unstable protein complex formation involving multiple incompatible loci.

Hybrid introgression is often used in other species to map incompatible genetic loci, but it cannot be efficiently applied to different yeast species due to sequence divergence-induced antirecombination. Nevertheless, genetic mapping approaches similar to introgression have been used to analyze different populations of the same species to reveal the early onset of reproductive isolation and incipient speciation. A study using a cross between the S288C and SK1 strains of S. cerevisiae identified mild two-locus incompatibility in the DNA mismatch repair pathway (96). Mismatch repair is initiated by binding MutS-homologous proteins to a base-base mismatch or an insertion-deletion loop, followed by the ATP-dependent recruitment of MutL-homologous proteins to begin the repair process (reviewed in reference 97). In S. cerevisiae, the Mlh1-Pms1 heterodimer is the major MutL-homologous complex for postreplicative mismatch repair (98). The observed incompatibility involves S288C-derived MLH1 and SK1-derived PMS1, leading to elevated mutation rates and reduced spore viability. Further domain-swapping experiments revealed that single-amino-acid changes in each protein, Mlh1 D761 from S288C and Pms1 K822 from SK1, are sufficient to cause the incompatibility (96).

In a large-scale study, Hou and colleagues crossed 27 natural isolates of S. cerevisiae and examined their F1 haploid progeny under different growth conditions (99). Approximately 24% of the tested crosses revealed condition-specific genetic incompatibility. Moreover, most cases represented complex incompatibility involving more than two genetic loci. These data suggest that complex incompatibilities may play a crucial role in incipient speciation. The same study also identified condition-specific two-locus incompatibility caused by a TAA nonsense mutation in the COX15 gene and the UAA (ochre) suppressor SUP7 in a clinical strain, YJM421 (77, 99). Since COX15 encodes a mitochondrial inner membrane protein required for heme A synthesis (100, 101), the hybrid progeny exhibit respiratory deficiency if they inherit solely the mutant cox15 allele but not the suppressor SUP7. These results suggest that negative epistatic interactions between alleles are probably quite common in natural environments. The recent discovery that the repression of SGS1 and MSH2 restores meiotic recombination in a highly sequence-divergent hybrid can enable scientists to construct introgression lines between diverged species (32), thereby accelerating the discovery of genetic incompatibilities among different species.

Population Genomics

Speciation is an evolutionary process that begins with local genomic barriers to the exchange of genes associated with adaptation, intrinsic incompatibility, selfish elements, or assortative mating. These barriers then spread until complete reproductive isolation is established (6). Theoretically, a population genomics approach can reveal regions of reduced gene flow by detecting loci with greater differentiation than expected from the average across many loci. Given advances in genome sequencing technologies, scientists can now compare the genomes of many natural populations and search for the genomic regions involved in speciation.

Over the past 2 decades, populations of several yeast species have been whole-genome sequenced. In 2009, Liti and colleagues compared the genomes of 36 S. cerevisiae and 35 S. paradoxus strains isolated from various geographic locations and ecological niches (102). They observed a much faster decay of linkage disequilibrium in S. cerevisiae than in S. paradoxus, revealing the impact of human domestication on S. cerevisiae evolution. That study also uncovered that genomic segments from many S. cerevisiae mosaic strains are not related to any known pure lineages, implying the existence of as-yet-undiscovered lineages or lineages that are already extinct. A follow-up study surveyed 1,011 S. cerevisiae strains from diverse environments across five continents to maximize the breadth of their ecological and geographical origins (103). The resulting genomic data showed that strains collected from Taiwan, mainland China, and other regions of East Asia exhibit the highest genome sequence diversity, supporting that the Far East is the geographic origin of this species. That conclusion is further evidenced by two other studies showing that the genetic diversity of Chinese and Taiwanese strains is higher than that of strains from other parts of the world (77, 104). Interestingly, polyploidy and aneuploidy, which are often observed in domesticated yeast strains, are less common in wild populations, indicating that they may represent a transient state of adaptation or the outcome of human selection. Moreover, a genome-wide association study (GWAS) showed that copy number variants have greater impacts on the phenotypic landscape than single-nucleotide polymorphisms (SNPs) (103).

Although population genomics studies of S. cerevisiae have identified several genetic loci associated with specific environmental responses or potentially adaptive phenotypes, so far, none of them have been demonstrated to be involved in reproductive isolation between populations or species. Thus, further experiments are needed to test their roles in speciation. In addition, combining population data with species pangenomes should allow scientists to analyze and identify potential local genomic barriers to gene exchange. Interestingly, chromosomal rearrangements are often observed in various S. cerevisiae and S. paradoxus populations, and they likely contribute to reproductive isolation among these populations (75, 77, 78, 81, 105, 106).

The fission yeast S. pombe is another species for which natural populations have been widely investigated. To date, a few hundred natural strains isolated from various locations and ecological niches have been whole-genome sequenced (107–109). One study examined the patterns of polymorphisms and identified a region close to the end of chromosome 3 that displayed an extremely high level of divergence among strains (108). The authors of that study postulated that this diverged region may be responsible for reproductive isolation, but that hypothesis has not yet been experimentally tested. Like budding yeast, population genomics of S. pombe revealed that chromosomal rearrangements are common in S. pombe populations (107–109). Jeffares and colleagues have shown that hybrid spore viability is inversely correlated with differences in chromosomal rearrangement and the numbers of SNPs, indicating that both genetic features contribute to reproductive isolation among different strains (109). In another study, various chromosomal rearrangements observed in natural isolates were engineered in one strain background to compare their phenotypic effects. Several of these rearrangements resulted in improved fitness under various growth conditions, suggesting that they are potentially adaptive (110).

Neurospora species represent the best-studied model filamentous fungi for genomics and molecular biology research. One study analyzed 48 natural isolates of Neurospora crassa derived from two recently diverged populations, one from subtropical Louisiana and another from the tropical Caribbean basin, revealing two genomic islands of extreme divergence (111). Interestingly, the subtropical Louisiana populations grow better at low temperatures, and two genes in the observed divergence islands have functions related to cold responses. This outcome indicates that the genes in the genomic islands may confer a local selective advantage that can ultimately give rise to condition-specific genetic incompatibility, as observed in S. cerevisiae (99).

The recent growth of population genomics has provided perspectives and information revolutionizing research on speciation in many fungi, including nonmodel fungal species (112–116). However, there is a paucity of experimental studies that explicitly distinguish between alternative hypotheses pertaining to the process of speciation. For example, it is unclear if adaptive traits arise from the tight physical linkage of multiple genetic changes or a single genetic change with pleiotropic effects (117). It is also challenging to distinguish mutation-order speciation from ecological speciation (118). Consequently, technologies such as RNA interference (RNAi) and gene editing (e.g., CRISPR) that interfere with or alter gene function should be developed further to facilitate genetic manipulations, representing promising approaches to discovering the origins of fungal diversity and answering seminal evolutionary questions.

Experimental Evolution

Laboratory evolution experiments allow direct real-time tests of speciation theories and are an excellent complement to other approaches (119). The integration of genomic tools in an experimental speciation framework has advanced our understanding of how reproductive isolation evolves. Using S. cerevisiae, Dettman and colleagues evolved 12 populations derived from a single diploid progenitor under two selective conditions, high salinity (S) and low-glucose minimal medium (M). After hundreds of generations, the evolved cells exhibited improved fitness under selective conditions (120). However, S/M hybrids exhibited a reduced rate of mitotic reproduction and a diminished meiotic reproduction efficiency. That outcome validates the ecological speciation hypothesis whereby adaptation to altered growth environments can lead to reproductive isolation (121). By whole-genome sequencing of three evolved clones, the authors identified six evolved alleles contributing to the adaptive phenotypes (122). Among these alleles, negative epistasis between evolved MKT1 in the high-salt environment and PMA1 in the low-glucose environment strongly reduced the fitness of S/M hybrids in the low-glucose environment. Mkt1 is a major regulator of the mRNAs involved in mitochondrial function and P bodies (123). PMA1 encodes an essential ATP-driven proton pump responsible for maintaining pH homeostasis and transmembrane potential (124). The defects of S/M hybrids are caused mainly by altered intracellular pH arising from the negative interaction between evolved PMA1 and MKT1 (125). These results also provide an excellent example of how Dobzhansky-Muller incompatibility evolves when two populations adapt to different environments.

A similar approach was used on the filamentous fungus Neurospora to understand the effects of initial genetic diversity and divergent adaptation on the evolution of reproductive isolation (126). Low-diversity and high-diversity founding populations were generated from intraspecific (N. crassa × N. crassa) and interspecific (N. crassa × Neurospora intermedia) crosses. Parallel cultures were then propagated in two selective environments, high salinity and low temperature. After evolution, it was observed that divergent adaptation resulted in greater reproductive isolation than parallel adaptation (which had evolved under the same conditions). Intriguingly, altered reproductive success was observed for both prezygotic isolation (perithecial production) and postzygotic isolation (progeny viability). Moreover, the effects of divergent adaptation on reproductive isolation were more pronounced for populations with higher initial genetic variation. Genetic analysis revealed that the observed incompatibility occurs in a two-locus, two-allele model with asymmetric negative epistasis, which is consistent with the Dobzhansky-Muller model of genetic incompatibilities. Laboratory evolution experiments with both S. cerevisiae and N. crassa provide direct evidence that reproductive isolation can evolve quickly among populations that adapt to different growth environments, as suggested by the ecological speciation model (121).

Interestingly, negative epistasis was also observed among yeast populations that had been experimentally evolved under the same selective conditions (127). When the genomes of five evolved clones isolated from glucose-limited cultures were sequenced, adaptive mutations were found to have evolved multiple times independently in two loci, i.e., MTH1 and HXT6/HXT7. However, when mutations in these two loci were combined, the double mutants exhibited lower fitness than wild-type cells, indicative of incompatibility between the evolved MTH1 and HXT6/HXT7 alleles. Negative epistatic interactions were also observed in another evolution experiment. When Ono and colleagues analyzed pairwise interactions of the first-step mutations that independently evolved in a fungicide-containing environment, one-third of them exhibited negative epistasis (128). Together, these results provide evidence that even the same selective force can generate adaptive mutations that are incompatible with each other, supporting the idea of mutation-order speciation (121).

Similar to nuclear-nuclear incompatibility, the evolution of mitochondrial-nuclear incompatibility has been observed in an evolution experiment in which yeast cells were challenged with mitochondrial stress to select for improved mitochondrial DNA stability (129). Although nuclear mutations were identified to be the primary contributor to evolved phenotypes, some evolved strains showed incompatibility with the ancestral mitochondrial DNA, implying coevolution between the mitochondrial and nuclear genomes.

Other than genetic incompatibility, genomic rearrangements have often been observed in yeast populations that have evolved under various selective conditions (130–133). One such study used reconstitution experiments to show that genome restructuring in an evolved clone isolated under prolonged starvation conditions indeed led to reproductive isolation (134).

Hybrid Transcriptomics

It is generally accepted that variation in gene expression shapes the phenotypic and molecular diversity of organisms. Organisms frequently alter how their genes are regulated in order to adapt to divergent ecological niches (135–137). However, gene misregulation in hybrid cells can occur as a by-product of regulatory divergence, compromising hybrid fitness and impeding gene flow between two diverse populations. Comparative transcriptomics has been used to study animal hybrids, and such analyses have revealed potential cases of gene misregulation causing hybrid dysfunction (138–140). Applying similar approaches to fungal hybrids may prove fruitful.

Changes in gene expression can occur via mutations in regulatory sequences proximal to the target gene body (cis effects) or from variations elsewhere in the genome that impact the upstream propagating regulators (trans effects) (141). Distinguishing between the relative contributions of cis and trans effects can be achieved using a viable interspecific hybrid in which two parental genomes are exposed to the same trans-acting environment. Consequently, differences in expression between the orthologous genes observed in hybrids are assumed to result from cis effects, whereas interspecific differences lost in hybrids are attributed to trans effects (142). The cis and trans effects on gene expression divergence have been dissected in intraspecific and interspecific yeast hybrids. Consistent with observations from Drosophila (143), cis variation contributes to the majority of expression divergence between species. In contrast, intraspecific or condition-specific expression differences are often caused by trans changes (144–147). These results allow us to understand not only the evolutionary trajectories of regulatory divergence but also the potential physiological impacts of misregulated gene expression on hybrid cells.

Interestingly, rather than exhibiting changes in gene expression levels, an S. cerevisiae-S. paradoxus hybrid presented earlier meiotic gene expression than either parental species (148). Although the causal factors and cellular effects of such misregulation were not identified, the results indicate that temporal dynamics should be explicitly considered when evaluating gene expression. There have been few hybrid transcriptomic studies in fungi other than budding yeasts. However, in a study of a natural diploid hybrid of the genus Epichloë (149), the authors observed that although most homeologous genes retained their parental expression patterns, ~1,200 genes exhibited expression bias or expression reversal (meaning that the homeolog expression patterns are entirely different from those of the parents), but the possible physiological effects of these misregulated genes were not analyzed further.

Transcriptional rewiring is an extensive form of regulatory divergence in which orthologous transcription factors (TFs) regulate different target genes in different species (150, 151). Such events would endow strong incompatibility to the diverged organisms. Accumulating evidence supports that transcriptional rewiring events are not as rare as previously thought. Although most examples of such rewiring have been observed between distantly related species, it has been reported that the binding targets of certain yeast TFs can be rapidly changed. One study used chromatin immunoprecipitation and DNA microarray (ChIP-on-chip) experiments to examine the binding of two TFs, Ste12 and Tec1, in three closely related yeast species, S. cerevisiae, S. mikatae, and S. bayanus (152). Only 20% of gene promoters retained binding of these TFs in all three species. Although it is unclear how many TFs share similar dynamics in terms of their binding sites, such rapid divergence in binding sites is likely to fuel regulatory divergence in closely related species. Moreover, another recent study showed that the altered binding preferences of TFs are caused primarily by variations outside the DNA-binding domains (153). To generate an overall picture of switched binding preferences, a systematic approach must be adopted that combines both experiments and bioinformatics to detect changes in orthologous TFs.

LOOKING BACK AND MOVING FORWARD

Possible Driving Forces of Speciation

Traditionally, speciation has been thought to arise from adaptation to diverse ecological niches or genetic drift due to geographic isolation. However, evidence from studies of fungal speciation indicates that divergent adaptation and prolonged geographic isolation may not be necessary in some cases. Experimental evolution data suggest that if organisms acquire different initial adaptive mutations even under the same environmental conditions, they may undertake completely different evolutionary trajectories at later stages (127, 128). Therefore, incompatible mutations can accumulate in different populations, leading to speciation, as proposed under the “mutation-order speciation” model (121). Further studies of the fitness landscape are required to advance our understanding of mutation-order speciation and its contribution to how species evolve and diverge. The “conflictual speciation” model proposes that speciation can be driven by the arms race between two coevolving partners, with reproductive isolation evolving as a by-product of antagonistic selection among genomic elements having divergent fitness interests (154). As predicted by the Red Queen hypothesis (155), an arms race is a zero-sum game, meaning that the average relative fitness of evolved cells remains constant, even though the evolved mutations are adaptive. Certain forms of genomic conflict implicated in the evolution of reproductive isolation have been observed in fungi, including selfish elements and host genomes, mitochondrial and nuclear genomes, and fungus-host interactions.

Selfish elements are genetic segments that can enhance their own transmission at the expense of other genes in the genome. Examples include transposable elements, meiotic drivers, and postsegregation killers that lead to hybrid dysfunction. A similar scenario arises for mitochondrial and nuclear genomes within a cell. Selfish mtDNA that replicates more efficiently than wild-type mtDNA can spread in the cytosol despite compromised respiratory function. The arms race between mtDNA and host nuclear genomes can accelerate cytonuclear coevolution. However, when two populations harboring coadapted mitochondrial and nuclear mutations encounter each other and generate hybrids, mismatched mitochondria and nuclei may cause hybrid breakdown. Coevolutionary dynamics evidencing Red Queen arms races have also been observed for parasitic fungus-host interactions. For instance, immune or behavioral defense mechanisms in hosts play roles in maintaining the genetic diversity of parasitic fungi (4, 156, 157). Even though molecular studies of how fungus-host interactions contribute to fungal speciation are sparse, recent advances in genetic and genomic tools will greatly facilitate research into this topic in the future.

What Are the Roles of Epigenetic Factors in Speciation?

Genomes are vulnerable to selfish genetic elements, and epigenetic mechanisms are central to preserving genome integrity and defending the genome against selfish DNA elements (158, 159). Thus, epigenetic factors may play a crucial role in escalating arms races between selfish elements and host genomes. However, altered epigenetic systems can also have a global impact on transcriptional regulation or rewiring, giving rise to another layer of complexity. Once a transcriptional network diverges, its target genes are likely misregulated in hybrid cells, resulting in fitness defects. Similarly, selfish elements can trigger a broad range of genetic variations, including novel gene function or regulation, as well as chromosomal rearrangements, which cumulatively may contribute to the genomic divergence between individual populations. Consequently, if these epigenetic factors do not exhibit strong incompatibility among populations, evolved reproductive isolation may be attributed to misregulated gene expression or chromosomal rearrangements without an awareness of the contribution of epigenetic factors. Unraveling the impact of epigenetic factors on the entire process would require a detailed analysis of how selfish elements are repressed at different evolutionary stages.

Apart from genome defense, recent studies have shown that epigenetic variation, like genetic variation, can act as a substrate for natural selection (160). Epigenetic processes rapidly modify the state of gene expression and genomic structures, enhancing phenotypic plasticity in response to the changing environment. For example, stress-induced centromeric histone eviction enables adaptation via aneuploidy in Candida albicans (161). Similarly, stress-induced prions enable heritable stress resistance by activating subtelomeric domains in S. cerevisiae (162). Stress-induced changes in the chromatin state that mediate programmed evolvability have been identified in several fungal pathogens (163, 164). Once beneficial epigenetic modifications become heritable (epigenetic assimilation), they can increase in frequency in a population (165). This process can help populations, especially clonal ones, survive under life-threatening environmental stress conditions. Epigenetic modifications can also facilitate mutational assimilation, resulting in novel morphology, behavior, or physiology and, ultimately, speciation (166). This hypothesis suggests that when individual populations diverge, epigenetic divergence occurs prior to any genetic differentiation. Although several studies have investigated the role of natural epigenetic variation in plant and animal speciation (167–169), similar studies on fungi are surprisingly scarce (159, 170).

CONCLUDING REMARKS

It is an exciting time to study the driving forces and mechanisms underlying fungal speciation. The field is poised to grow considerably in the coming years, as high-throughput sequencing greatly facilitates the detection of the genetic components driving the process of reproductive isolation. Moreover, improvements in genome-editing technology will boost our ability to identify and characterize the molecular mechanisms responsible for divergence, especially in fungal systems that had previously been genetically intractable. These advances are paramount to addressing the many remaining unanswered questions. New epigenomic tools might also open up new avenues for interpreting if the epigenetic machinery can operate as a potential first step in fungal speciation. These cutting-edge technologies can further complement experimental speciation to enable scientists to conduct direct real-time tests of speciation theories (171).