Biological Journal of the Linnean Society

Forsdyke, Donald R

doi: 10.1093/biolinnean/bly035pmid: N/A

Abstract Mechanisms initiating a sympatric branching process that can lead to new species are broadly classified as chromosomal and genic. Chromosomal mechanisms are supported by breeding studies involving exchanges of individual chromosomes or their segments between mouse subspecies. There are also studies of the rapidly mutating mouse PR domain containing 9 (Prdm9) gene, which encodes PRDM9, a protein targeting meiotic recombination hotspots. When PRDM9 is bound to duplex DNA symmetrically with equal strength, the repair of mutations in one parental strand, based on information on the allelic strand (conversion), would seem to be unbiased in discriminating between strands. Therefore, mismatches detected between pairing paternal and maternal DNA strands (heteroduplexes) undergo unbiased conversions (to homoduplexes). This leaves uncertainty on whether a mutation has been corrected or compounded. However, a hypothetical tagging of mismatch regions, so that both strands are epigenetically marked as uncertain, would make it possible, over numerous generations, for mutations to be corrected (biased conversions) whenever asymmetry is detected. Thus, variation would decrease, and members of a species could remain within its bounds. Intriguingly, new experimental studies show that, when chromosomally interpreted, PRDM9 also works through asymmetrical epigenetic labelling to confine members to species bounds. To the extent that the experimentally observed and hypothetical anti-speciation asymmetries can be related, chromosomal mechanisms are further supported. INTRODUCTION Hypotheses on the initiation of a sympatric branching process that can lead to new species are broadly classified as chromosomal and genic (Forsdyke, 2001, 2010; Nei & Nozawa, 2011; Nevo, 2012). Agreement is sought as to which initiation mechanisms are genuinely, rather than hypothetically, capable of originating species, and which are most likely to have operated in the general case (Kliman, Rogers & Noor, 2001; Forsdyke, 2004; Johannesson, 2010). Some chromosomal hypotheses invoke sequence disparities between parental chromosomes so that meiotic pairing of ‘homologous’ chromosomes fails within the gonad of their offspring (hybrid). Thus, there can be no exchange of DNA segments (recombination), and the production of gametes ceases (hybrid sterility). Arising from studies of the PR domain containing 9 gene (Prdm9), there is now ‘fresh evidence for a genetic connection between recombination and hybrid sterility’, suggesting ‘the intriguing possibility that recombination and speciation are mechanistically coupled’ (Payseur, 2016). However, the ‘speciation genes’ commonly invoked to explain this have become increasingly elusive. Specific genes affecting species initiation have been proposed for the fruit fly (Mallet, 2006). Although the existence of similar genes in other species is doubted (Schartl, 2008; Louis, 2009; Kao, Schwartz & Sherlock, 2010), the Prdm9 gene that is expressed in early germ cell maturation (Hayashi, Yoshida & Matsui, 2005; Mihola et al., 2009) is now thought a likely ‘first mammalian candidate for a speciation gene’ (Flachs et al., 2012). More definitely, it is ‘the only mammalian speciation gene yet identified’ (Davies et al., 2016). However, studies to be reviewed here suggest that its major role is to encode a genome maintenance protein that works to retain a line of organisms within species bounds. Thus, it can also be regarded as an inhibitor of speciation – an anti-speciation gene – that opposes speciation when initiated chromosomally (Brand & Presgraves, 2016; Reese & Forsdyke, 2016). Whatever the outcome, ‘cracking the curious case of PRDM9 promises to provide important insights into large swaths of biology, from human genetics to speciation’ (Ségurel, Leffler & Przeworski, 2011). This review has three parts. The first considers the roles of hybrid sterility as a primary reproductive barrier and of the Prdm9 protein product, PRDM9, as an adder of epigenetic methylation marks to histones. This designates the chromosomal location of meiotic recombination hotspots where there can be an asymmetric conversion of information in one parental genome to that of the other. However, like some other genes whose products target DNA, Prdm9 is on a mutational ‘treadmill’, being forced to change rapidly to keep pace with DNA sequence changes. Its DNA target ‘calls the tune’, yet in a paradoxical way. Hotspots appear self-destructive in that a given PRDM9 protein faces, generation after generation, diminishing target availability, thus ‘obliging’ its gene to mutate rapidly in response to an as-yet-unexplained evolutionary pressure for the designation of new targets. The second part of the review deals with the new light cast on chromosomal speciation by experimental modifications of PRDM9 that confers on it the ability to tune call, so shifting hotspot locations. Finally, the possibility is considered that these observations relate to a postulated asymmetry requirement for the epigenetic marking of sequences whose accuracy is in doubt (Reese & Forsdyke, 2016). Prdm9 AND RECOMBINATION HOTSPOTS Histone modifications by Prdm9 Of the barriers with a potential to initiate the reproductive isolation needed for sympatric divergence of one species into two, the hybrid sterility resulting from defective meiosis can play only a primary role (Fig. 1). Any barrier can be primary, but there is an onus upon those believing another barrier (e.g. hybrid inviability) to be primary to show that it had not been preceded by a period, however brief, of hybrid sterility (Forsdyke, 2017a, b, 2018). Regarding mechanisms, although controversial there is now growing evidence that nucleic acid ‘pair-first’ (rather than ‘break-first’) models accord well with meiotic recombination as described for many organisms (Forsdyke, 1996, 2017c; Gladyshev & Kleckner, 2014, 2016, 2017; Chapman et al., 2017). Subversion of such pairings by sequence disparity would create conditions favouring speciation. Figure 1. View largeDownload slide Of three sets of reproductive barriers leading to sympatric species initiation, the hybrid sterility barrier can play only a primary role. Given that the germinal cycle (gamete, zygote, embryo, meiotic adult gonad, gamete, etc.) is recursive, any point, be it before or after union of gametes to form a zygote, will serve to mediate primary arrest of the cycle (points 1, 2 and 3 in A). The arrow indicates clockwise progression, but numbering is anticlockwise. When 3 is primary, the cycle arrests and downstream events (2 and 1) cannot occur in that individual. When 2 is primary, then downstream event (1) cannot occur in that individual. When 1 is primary, it cannot affect 3 and 2 in the same individual because collaboration with a partner generates a new cycle (B). Thus, it is normal for the cycle to interrupt when the hybrid sterility barrier (1) is absent. If prezygotic barriers (3) are absent, gametes from pairing partners (dashed lines) meet as a new individual (a new cycle). If hybrid inviability barriers (2) are absent in that new individual, then gonadal gametogenesis follows, provided that hybrid sterility barriers (1) are also absent. The cycle then resumes in another new individual, and so on. In many extant species, 3 is the reproductive isolating barrier we identify today (e.g. mouse cannot copulate with elephant). In others, 2 is the identified barrier. Neither identification excludes the possibility of an earlier role for 1 as a fundamental barrier acting at the time of an initial divergence into two species (still present in certain viruses; Forsdyke, 1996). Thus, the temporal sequence of successive barriers may sometimes be 1, 2, 3. Figure 1. View largeDownload slide Of three sets of reproductive barriers leading to sympatric species initiation, the hybrid sterility barrier can play only a primary role. Given that the germinal cycle (gamete, zygote, embryo, meiotic adult gonad, gamete, etc.) is recursive, any point, be it before or after union of gametes to form a zygote, will serve to mediate primary arrest of the cycle (points 1, 2 and 3 in A). The arrow indicates clockwise progression, but numbering is anticlockwise. When 3 is primary, the cycle arrests and downstream events (2 and 1) cannot occur in that individual. When 2 is primary, then downstream event (1) cannot occur in that individual. When 1 is primary, it cannot affect 3 and 2 in the same individual because collaboration with a partner generates a new cycle (B). Thus, it is normal for the cycle to interrupt when the hybrid sterility barrier (1) is absent. If prezygotic barriers (3) are absent, gametes from pairing partners (dashed lines) meet as a new individual (a new cycle). If hybrid inviability barriers (2) are absent in that new individual, then gonadal gametogenesis follows, provided that hybrid sterility barriers (1) are also absent. The cycle then resumes in another new individual, and so on. In many extant species, 3 is the reproductive isolating barrier we identify today (e.g. mouse cannot copulate with elephant). In others, 2 is the identified barrier. Neither identification excludes the possibility of an earlier role for 1 as a fundamental barrier acting at the time of an initial divergence into two species (still present in certain viruses; Forsdyke, 1996). Thus, the temporal sequence of successive barriers may sometimes be 1, 2, 3. Furthermore, altough cellular nucleic acids are closely associated with proteins that could affect recombination-related nucleic acid interactions, many aspects of DNA pair-first mechanisms can be reproduced in vitro in the absence of proteins (Danilowicz et al., 2009; Forsdyke, 2016, p. 180). Assuming nucleic acid evolution to have preceded that of proteins (Reanney, 1979; Poole & Logan, 2005), one can envisage the kinetics of recombination – once entirely dependent on nucleic acid chemistry (Forsdyke, 2007) – to have been affected by later-evolving proteins. Thus, in practice, pair-first mechanisms cannot really be first. There is need for a prior modification of proteins, such as the histones around which DNA is organized as nucleosome complexes. Modification of such nucleoprotein (chromatin) complexes to improve access to DNA is one of the functions of PRDM9, the zinc finger part of which recognizes with high specificity a short DNA sequence motif at the nucleosome surface, so defining a 1–2 kb recombination ‘hotspot’ region. This recognition activates the PR/SET domain of the protein that catalyses the local addition of methyl groups to certain lysine residues of one of the histone proteins (H3). Trimethylation of H3 generates a ‘nucleosome-depleted region’ where recombination is facilitated (i.e. homologous strands can pair, there is branch migration to increase the area scanned, and the enzyme causing double-strand breaks, SPO11, can be recruited). Conversion restrains diversity The idea that meiosis, apart from generating diversity by shuffling parental genomes, also has the potential to restrain diversity, has a long history. Montgomery (1901) proposed that meiosis would ‘rejuvenate’ the colourfully staining material (chromatin) in disparate chromosomes: When the spermatozoon conjugates with the ovum there is a mixture of cytoplasm with cytoplasm, of karyolymph with karyolymph, … but there is no intermixture of chromatin, for the chromosomes … remain more separated from one another than at any other stage …. But after this beginning stage of the germinal cycle, … in the synapsis stage … we find a positive attraction between paternal and maternal chromosomes. The reason for the final union of these chromosomes is obvious; it is evidently to produce a rejuvenation of the chromosomes. From this standpoint the conjugation of the chromosomes in the synapsis stage may be considered the final step in the process of conjugation of the germ cells. It is a process that effects the rejuvenation of the chromosomes; such rejuvenation could not be produced unless chromosome of different parentage joined together, and there would be no apparent reason for chromosomes of like parentage to unite. The idea that the rejuvenation could involve exchange of linear information between disparate chromosomes was outlined by Winge (1917) and later elaborated in terms of DNA mismatch repair (Bernstein & Bernstein, 1991; Forsdyke, 2001, 2007; Gorelick & Heng, 2010; Chapman et al., 2017; Bernstein, Bernstein & Michod, 2018). If we assume rejuvenation to imply a restoration (‘correction’) of DNA sequences that have strayed from the species norm (i.e. they are to be returned closer to their formerly ‘pristine’ base composition and order), then it follows that a lack of such rejuvenation would allow DNA differences to accumulate, so possibly leading to reproductive isolation and speciation. Rejuvenation would restrain diversification, and hence speciation. Fundamental to this are mechanisms of meiotic recombination. Meiosis within a parental gonad generally provides for the unbiased distribution of grandpaternal and grandmaternal genomic information among offspring. Sometimes detected as a failure of characters to follow a Mendelian pattern of distribution, the asymmetrical conversion of information from one genome into that of the other has been called ‘gene conversion’. But, in principle, the asymmetry would also apply to non-genic sequences. Given two formerly homologous segments, A and B, which have come to differ slightly (i.e. are now heterologous), information in A can be converted into that of B, or vice versa. Thus, regarding individual offspring, diversity is diminished. However, the direction of the conversion may be generally unbiased in that in one individual A may have converted to B, and in another B may have converted to A. Thus, at the population level, in the absence of conversion bias there will be no change in the frequencies of A and B. Diversity is not diminished. On the other hand, rather than let natural selection be the arbiter of frequencies should one segment prove more advantageous than the other, if there were sufficient information, a parent could ‘place an educated bet’ on the conversion direction to adopt. For example, if it were ‘known’ that A was the species norm and/or that B had experienced a recent mutation, then there could be an appropriate conversion bias. Given that most mutations are detrimental, resulting in negative selection in the line of organisms within which they occur, the best bet would be to convert B into A. If all parents were to bet in a similar manner then diversity within the population would be diminished. Intriguingly, as will be considered later, various studies show that a sequence motif recognized by PRDM9 (so defining a recombination hotspot) is directionally converted in mice heterozygous for that hotspot in favour of the sequence less strongly recognized by that protein (meiotic drive). I also consider later the possibility of ‘educated bets’ in the light of studies of incipient speciation in mice. DNA inconstancy drives the protein treadmill Proteins often act as enzymes catalysing changes in other molecules (their specific substrates), which may be micromolecules (e.g. glucose) or macromolecules (e.g. DNA). During the course of evolution, there are changes in proteins as a consequence of mutations in the genes that encode those proteins. Their substrates do not usually participate in this directly. For example, as organisms evolve, specific enzymes may change and so improve the utilization of glucose. That is natural selection. But glucose itself has no say in this. It does not change to lighten or impede the task of the enzymes. For some enzymes, however, DNA is a specific substrate. DNA both encodes the enzymes (locally in the regions of their genes) and is either their local or general substrate (target). However, DNA, by virtue of base changes, can change its character. Thus, it has the power to lighten or impede the tasks of the enzymes that act on it. Furthermore, the constancy of glucose means that the enzymes of glucose metabolism can count on glucose being the same from generation to generation. The limited constancy of DNA means that some of the proteins that react with it are on an evolutionary treadmill. They are ‘red queens’ that must change as their substrate changes (Lesecque et al., 2014). Some proteins target nucleic acids generically by virtue of their unchanging phosphate–ribose structures, whereas others are not indifferent to the composition or order of bases. Therefore, as nucleic acid bases change, the proteins that target nucleic acids more specifically may need to adapt functionally to ensure optimal binding to their substrate. This adaptive treadmill implies that, as lines diverge, the mutation rate in nucleic acid recognition proteins must be greater than in most other proteins (Paz et al., 2006). If its genes encoding nucleic acid recognition proteins do not rapidly mutate, an organism may not survive. Here the DNA ‘dog’ may wag the protein ‘tail’, but sometimes the converse may apply. Changes in proteins encoded by certain ‘mutator’ genes can exert a genome-wide effect on base composition (Cox & Yanofsky 1969). Likewise, a mutated Prdm9 gene can, in one fell swoop, exert a genome-wide effect on the nature and location of target sites that focus recombination events (see later). Paradoxically, however, over evolutionary time it is usually these target sites that ‘call the tune’, resulting in the positive selection of Prdm9 mutants: ‘PRDM9 has been under selective pressure to switch to new targets. However, the reasons for this selective pressure remain mysterious’ (Lesecque et al., 2014). The hotspot conversion paradox The genomic locations of recombination events are usually not randomly distributed. In microorganisms, there are strand-specific, mainly gene-located, ‘crossover hotspot instigator’ (Chi) sequences that are recognized by enzymes with a role in recombination (Lao & Forsdyke, 2000). Multicellular organisms also have sequence hotspots where recombination initiates most frequently. Such hotspots may relate to genic boundaries (e.g. transcription start sites), either as a primary feature (Forsdyke, 2011a) or as a default option when sites (motifs) recognized by PRDM9 are not operative (Brick et al., 2012). In organisms possessing a functional Prdm9 gene (e.g. primates, rodents), the primary hotspot option is often extra-genic or within introns. These DNA regions have more potential than exons to extrude the kissing-loop structures that may be of importance for recombination (Forsdyke, 1996, 2007). Chi being a conserved sequence in bacteria, it was thought initially that there would be a conserved, specifically targeted binding sequence for PRDM9. Much attention was paid to a degenerate 13-base consensus motif that was often found in human hotspots. However, although recombination initiation hotspots may endure for many generations, eventually they disappear (erode) owing to an unexplained directional conversion bias that accompanies recombination (Fig. 2). Thus, Wahls & Davidson (2011) observed that ‘hotspots seed their own destruction’; hence, to maintain recombination, fresh sites with different sequence characteristics must emerge. Focusing on yeasts, which do not have the Prdm9 gene, they noted that ‘within the genome is a collection of hotspot-active DNA sites and a reservoir of “cryptic” DNA sequence motifs that can be rendered active by as little as a single-base-pair substitution’. Under this ‘equilibrium dynamic model’, the substitution refers to a DNA target motif, not to the enzyme that recognizes that motif. However, the possibility was entertained that for organisms with the Prdm9 gene, novel motifs could be so designated by its PRDM9 product if the gene were appropriately mutated (‘Prdm9 shift model’). Accordingly, the widely distributed hotspot sequences that had eroded could be viewed as having ‘called the tune’, and the gene had ‘responded’ by mutation (i.e. the mutation conferred some selective advantage). The latter view is now supported by extensive studies, such as those on the infertility of the male offspring of crosses between mouse subspecies carried out in the Forejt laboratory (see later). Figure 2. View largeDownload slide Cyclic erosion of PRDM9 binding sites during repeated within-species crosses. Homologous Prdm9 genes are shown as white rectangles flanked by horizontal lines (genomes; red and blue). At the top, curved arrows indicate potential PRDM9 targets (X) at various genomic locations. On the left, parental gametes unite to form normal F1 hybrids with symmetrical target sites. These hybrids, in turn, form parental-type gametes (the choice of sex is arbitrary here). On the right, a mutation has changed a target site in one parental gamete (loss of X). Despite the local asymmetry, both normal and mutant sites are recognized at meiosis by PRDM9, and chromosomal pairing and recombination occur, albeit perhaps less efficiently. However, there is a conversion bias in favour of the less strongly recognized mutant parent genome. Although, through the generations, members of the majority population continue to produce unmutated target sites (scheme on the left), recurrent crosses with members of the growing mutant population (the cycle on the right; thick lines) erode this lead. Thus, the Prdm9 gene, in its X-recognizing form, becomes progressively irrelevant. Figure 2. View largeDownload slide Cyclic erosion of PRDM9 binding sites during repeated within-species crosses. Homologous Prdm9 genes are shown as white rectangles flanked by horizontal lines (genomes; red and blue). At the top, curved arrows indicate potential PRDM9 targets (X) at various genomic locations. On the left, parental gametes unite to form normal F1 hybrids with symmetrical target sites. These hybrids, in turn, form parental-type gametes (the choice of sex is arbitrary here). On the right, a mutation has changed a target site in one parental gamete (loss of X). Despite the local asymmetry, both normal and mutant sites are recognized at meiosis by PRDM9, and chromosomal pairing and recombination occur, albeit perhaps less efficiently. However, there is a conversion bias in favour of the less strongly recognized mutant parent genome. Although, through the generations, members of the majority population continue to produce unmutated target sites (scheme on the left), recurrent crosses with members of the growing mutant population (the cycle on the right; thick lines) erode this lead. Thus, the Prdm9 gene, in its X-recognizing form, becomes progressively irrelevant. When comparing diverging sequences, it is argued that that which is conserved between the sequences is evolutionarily important, whereas that which is less conserved is less important; indeed, that is why the ‘hand of evolution’ may have ‘chosen’ to discard it. Thus, when confronting the ‘hotspot conversion paradox’ (Fig. 2), Boulton, Myers & Redfield (1997) noted that because ‘biased gene conversion is a typical consequence of recombination at hotspots’, then ‘the sites thought to initiate crossing over cannot be maintained by the benefits of the events they cause’. In other words, change itself seemed to have an adaptive value. Therefore, there are circumstances in which non-conservation seems to have a greater adaptive value than conservation. Likewise, proteins such as PRDM9 are themselves transient, displaying extensive variation over time. Their genes are on a treadmill, having to mutate to keep up with the constantly changing hotspot landscape. Thus, there is high within-species variation (e.g. polymorphism) and even greater between-species (e.g. human–mouse) variation in the DNA recognition regions (zinc fingers) of Prdm9 genes. Grey et al. (2011) note: The PRDM9 gene is well-conserved among metazoans, however the domain encoding the zinc finger array experiences an accelerated evolution in several lineages, including rodents and primates. This accelerated evolution is restricted to codons responsible for the DNA-binding specificity of PRDM9 zinc fingers, which appear to have been subjected to positive selection. Prdm9 AND SPECIATION Subspecies as incipient species Breeding studies from the Forejt laboratory, involving exchanges of individual chromosomes or chromosome segments between members of two mouse subspecies [Mus musculus musculus (Mmm) and Mus musculus domesticus (Mmd)], support chromosomal, rather than genic, disparity as a basis for species initiation (Bhattacharyya et al., 2013, 2014; Gregorova et al., 2018). The subspecies had not been designated as ‘species’ because the reproductive isolation that defines species was incomplete in that a degree of fertility was still evident among female offspring (Haldane’s rule). Without geographical isolation (i.e. in sympatry) there remains a potential for interbreeding and gene flow. Thus, the subspecies were incipient species with a potential eventually to attain full species status (Mérot et al., 2017; Foote, 2018). Two important strains within the subspecies are PWD (from Mmm) and B6 (from Mmd). They are estimated to have diverged from a common mouse ancestor ~0.5 Mya and, although their DNA sequences are closely identical, they differ in their Prdm9 genes and the multiple specific targets of the PRDM9 proteins encoded by those genes. The latter targets can be represented (see Fig. 3, top) as Xs (for PWD) and black dots (for B6). These symbols represent the target sequence motifs as they now exist and, largely because of past erosions, they are likely to differ from versions that were active at earlier times (closer to the time of the initial PWD–B6 divergence). Figure 3. View largeDownload slide Hybrid sterility induced by pervasive reciprocal asymmetry of PRDM9 target sites when members of mouse subspecies are crossed (left), and its rescue in a transgene ‘humanized’ strain (right). At the top, subspecies-specific Prdm9 genes are shown as rectangles (PWD strain, white; B6 strain, black). Corresponding potential genomic targets (Xs and black dots) are indicated by curve arrows. On the left, gametes unite to form F1 hybrids that develop and grow normally. However, owing to a pervasive asymmetrical meiotic conversion bias that impairs gamete formation, they are sterile (i.e. Prdm9 acting as a potential ‘speciation gene’). On the right, the B6 version of the Prdm9 gene has been transgenically modified (‘humanized’) such that the zinc fingers no longer recognize B6 targets (black dots), but do recognize sequences elsewhere in the genome (striped dots) that happen to resemble closely the target of the human PRDM9 gene. When the humanized B6 mouse is crossed with a normal PWD mouse, meiosis proceeds normally owing to the symmetry between targets (striped dots). This overcomes the effects of pervasive asymmetries related to the presence of the PWD version of Prdm9, and the mice are fertile (i.e. Prdm9 acting as a potential ‘anti-speciation gene’). Figure 3. View largeDownload slide Hybrid sterility induced by pervasive reciprocal asymmetry of PRDM9 target sites when members of mouse subspecies are crossed (left), and its rescue in a transgene ‘humanized’ strain (right). At the top, subspecies-specific Prdm9 genes are shown as rectangles (PWD strain, white; B6 strain, black). Corresponding potential genomic targets (Xs and black dots) are indicated by curve arrows. On the left, gametes unite to form F1 hybrids that develop and grow normally. However, owing to a pervasive asymmetrical meiotic conversion bias that impairs gamete formation, they are sterile (i.e. Prdm9 acting as a potential ‘speciation gene’). On the right, the B6 version of the Prdm9 gene has been transgenically modified (‘humanized’) such that the zinc fingers no longer recognize B6 targets (black dots), but do recognize sequences elsewhere in the genome (striped dots) that happen to resemble closely the target of the human PRDM9 gene. When the humanized B6 mouse is crossed with a normal PWD mouse, meiosis proceeds normally owing to the symmetry between targets (striped dots). This overcomes the effects of pervasive asymmetries related to the presence of the PWD version of Prdm9, and the mice are fertile (i.e. Prdm9 acting as a potential ‘anti-speciation gene’). Erosion of target motifs A consequence of the erosion of specific target sites (Fig. 2) is that today’s PWD mice encode PRDM9 proteins with less affinity for their target sites on the PWD genome than for target sites at the corresponding positions on the B6 genome (the latter are not symbolized in Fig. 3, top right). In other words, today’s PWD-encoded PRDM9 proteins would, if given the opportunity, display stronger binding to a B6 genome than to a PWD genome. Likewise, today’s B6-encoded PRDM9 proteins would, if given the opportunity, display stronger binding to a PWD genome than to a B6 genome. That opportunity arises when members of the two subspecies are crossed. For a within-subspecies cross (Fig. 2), the decreased affinity does not disturb the general symmetry of binding of PRDM9 proteins. As long as appreciable affinity is retained, there should be fertile offspring. However, when there is a between-subspecies cross (Fig. 3) the binding is generally asymmetrical, and the story is very different. When a PWD female is crossed with a B6 male, all male offspring are sterile (Forsdyke, 2000; Delph & Demuth, 2016). A likely mechanism for this is shown at the left of Figure 3. As a result of the prior erosion, on the genome donated by the PWD gamete the target sites for the PRDM9 encoded by the PWD version of Prdm9 are weaker than the same target sites on the genome donated by the B6 gamete. Likewise, on the genome donated by the B6 gamete the target sites for the PRDM9 encoded by the B6 version of Prdm9 are weaker than the same target sites on the genome donated by the PWD gamete. The widespread asymmetry is somehow sufficient to impede meiotic pairing, and the individual is sterile. Humanized targets restore fertility For within-subspecies crosses, binding site affinity progressively decreases over evolutionary time (Fig. 2), with the potential to lead to sterility. Thus, individuals with Prdm9 mutations affecting the PRDM9 zinc fingers (so designating fresh binding sites) should have a selective advantage. Indeed, experimental exchanges of mouse Prdm9 genes indicate restoration of fertility (Mihola et al., 2009). An extreme version of this was to insert a human zinc finger DNA recognition region into a mouse Prdm9 gene (Fig. 3, right). Given both the great length of their genomes and the separation of humans and mice from their common ancestor some 150 Mya, it was likely that there would be some symmetrical, non-eroded potential target sequences in the mouse genome (indicated by striped dots in Fig. 3). These would be targeted, through its protein product, by the transgenically modified Prdm9 gene. B6 targets (black dots in Fig. 3, right) were denied a corresponding B6 Prdm9-encoded protein, but the pervasive asymmetry of PWD targets should have been recognized by the corresponding PWD Prdm9-encoded protein. The finding that the cross was fertile indicated a primary role for PRDM9 zinc finger mutations in preserving within-species fertility (Davies et al., 2016). Thus, the eroding multiple target sites in a genome ‘call the tune’. The Prdm9 gene must ‘respond’ to maintain fertility. This maintenance is in keeping with the observation of Oliver et al. (2009) for humans that ‘allelic variations at the DNA-binding positions of human PRDM9 zinc fingers show significant association with decreased risk of infertility’. Davies et al. (2016) concluded that ‘the full fertility of humanized mice implies there are unlikely to be any specific essential PRDM9 binding sites’. EPIGENETIC MARKING OF SUSPECT SEQUENCES An anti-speciation gene From generation to generation, growing DNA sequence diversity between two intraspecies groups makes speciation more likely (Harvey et al., 2017). Conversely, anything that facilitates the transgenerational uniformity of genomic information should make speciation less likely. As mentioned above, in many quarters Prdm9 is considered a ‘speciation gene’, indicating its direct positive involvement in the speciation process, such as by increasing the sequence diversity that can lead to hybrid sterility. Indeed, PRDM9 can work to induce sterility as shown in Figure 3 (left side). However, concerning this positive involvement, Davies et al. (2016) remark that ‘the rapid evolution of the zinc finger array of PRDM9 implies an unexpected transience of this direct role’. In keeping with this, Figure 3 (right side) raises the possibility that the development of sterility would be forestalled by zinc finger mutations (of which the humanized version is an extreme example) that would work to reduce variation and produce fertile offspring. The sterility could indeed be evolutionarily transient. Supporting this view, human PRDM9 zinc finger mutations have been found to be associated with protection against hybrid sterility (Irie et al., 2009; Oliver et al., 2009). Thus, PRDM9 can be viewed as normally promoting recombination, hence providing opportunities for repairing mutations and so impeding the sequence divergence that can lead to speciation. It opposes the hybrid sterility that is ‘characterized by failure of pairing (synapsis) of homologous chromosomes and an arrested meiotic prophase owing to lack of repair of recombination intermediates’ (Davies et al., 2016). In this light, Prdm9 is best regarded as an anti-speciation gene in that it facilitates the correction of meiotic mismatches (heterozygosities). There is now diminishing support for the view that adverse genic interactions (epistasis), referred to as Dobzhansky–Muller incompatibilities (leading to hybrid inviabilities), are as fundamental to speciation as meiotic chromosomal mismatches (Kao et al., 2010; Forsdyke, 2011b). Such interactions between fruit fly lines are likely to have evolved late in the speciation process (Mallet, 2006; Liénard, Araripe & Hartl, 2016; Fuller et al., 2017). A tale of two asymmetries Yet none of this easily explains the ‘hotspot paradox’ (Boulton et al., 1997), namely the erosion of PRDM9 target sites when asymmetrically disposed between parental genomes (Fig. 2). For some reason, the Prdm9 gene is on a treadmill being constantly ‘obliged’, through mutation, to reinvent itself, so encoding a novel PRDM9 protein product that can summon up fresh symmetrical target sites. These sites then become epigenetically marked through histone methylation. In another context, the epigenetic symmetry/asymmetry issue also emerged with a hypothetical mechanism for transgenerational error correction through biased meiotic conversion (Reese & Forsdyke, 2016; Forsdyke, 2016, p. 332). Here, a means by which a repair process could achieve a gradual (multigenerational) bias against mutations was envisioned. When meiotic branch migration encounters a point mutation, a base pair mismatch arises and is repaired, so either the maternal or the paternal base is converted to the complement of its opposing base in the heteroduplex. In the absence of any reason to expect a systematic bias in this repair process, chance alone would seem to determine the fate of a mutation. However, it is supposed that where a meiotic strand exchange first encounters a base pair mismatch (indicating a likely mutation) both strands are epigenetically marked to designate uncertainty. Such a mark, if retained within the new organism’s germ cell line, will convey the message ‘there is a near 50% chance that this allele is a mutation’. If this epigenetic tag could promote subsequent meiotic strand exchange and invite the next meiotic partner (less likely to harbour a mutation) to serve as the donor strand in the subsequent mismatch repair, this could ensure that, over many generations, biased conversions will remove mutations. Such a process, by reducing the rate of mutation-driven genetic divergence within breeding populations, would act as a force against speciation. Figure 4 outlines the proposed tagging principle, which has some resemblance to the epigenetic tagging of damaged DNA in cell lines (Kafer et al., 2016). When meiotic conversion is randomly biased, a mutation in one of the parental genomes can be either repaired or compounded (Fig. 4, outcomes 1 and 3). However, this uncertainty in the repair process can be registered by adding an epigenetic tag symmetrically to both strands in the repaired region (Fig. 4, outcomes 2 and 4). Subsequently, the tagged gametes can be compared with normal gametes provided by partners from the main, non-mutant population. It would be an ‘educated bet’ that the conversion should result in replacement of mutant with normal sequence. Therefore, there could be a selection pressure for the conversion to be from untagged to tagged DNA strands. With PRDM9, the conversion from donor to recipient is also asymmetrical (Fig. 2). Could this experimentally observed asymmetry and the hypothetical asymmetries, both of which may oppose speciation, be related? Could the two asymmetries be functionally linked with retention of appropriate directionalities? Indeed, could one be the evolutionary raison d’être of the other? These questions remain with us, as does the problem of the confinement of Prdm9 to certain taxa. Figure 4. View largeDownload slide Marking as ‘suspect’ a sequence whose accuracy is in doubt facilitates remedy (albeit delayed) by directional conversion bias. Resulting from a normal cross (top left), healthy homozygous F1 hybrids are representative of the species, members of which produce normal gametes (curved arrow). However, a mutation (X) in one parent (top right) may produce a heterozygous F1 hybrid. If the mutation is such that mutant and non-mutant strands cannot be distinguished, meiotic conversion bias will occur randomly, with four possible outcomes (1, 3, 2 and 4). 1: bias in favour of the unmutated strand corrects the mutation. 3: bias in favour of the mutated strand compounds the mutation. 2 and 4: in the region of the heterozygosity, conversion is accompanied by epigenetic tagging as ‘suspect’ (*). This affects both parental strands, irrespective of whether the mutation has been corrected (2) or compounded (4). In the middle, the tagged parents produce tagged gametes (the sex here is arbitrary). At the bottom, crosses of tagged parents with members of the general population produce unilaterally tagged hybrids. Directionally biased meiotic conversion in favour of non-tagged DNA strands removes both the mutation (if present) and the tag. Figure 4. View largeDownload slide Marking as ‘suspect’ a sequence whose accuracy is in doubt facilitates remedy (albeit delayed) by directional conversion bias. Resulting from a normal cross (top left), healthy homozygous F1 hybrids are representative of the species, members of which produce normal gametes (curved arrow). However, a mutation (X) in one parent (top right) may produce a heterozygous F1 hybrid. If the mutation is such that mutant and non-mutant strands cannot be distinguished, meiotic conversion bias will occur randomly, with four possible outcomes (1, 3, 2 and 4). 1: bias in favour of the unmutated strand corrects the mutation. 3: bias in favour of the mutated strand compounds the mutation. 2 and 4: in the region of the heterozygosity, conversion is accompanied by epigenetic tagging as ‘suspect’ (*). This affects both parental strands, irrespective of whether the mutation has been corrected (2) or compounded (4). In the middle, the tagged parents produce tagged gametes (the sex here is arbitrary). At the bottom, crosses of tagged parents with members of the general population produce unilaterally tagged hybrids. Directionally biased meiotic conversion in favour of non-tagged DNA strands removes both the mutation (if present) and the tag. Prdm9 cannot cover all potholes Given the great length of genome sequence in need of screening for errors, it would seem reasonable to suppose that recombination ‘hotspots might massively increase search efficiency by directing homology search to PRDM9 binding sites’ (Davies et al., 2016). Given this supposed advantage, why should hotspots erode? Some 25000–50000 hotspots were detected in the human genome (Myers et al., 2005). However, only a small percentage of potential PRDM9 binding sites are used in any one meiotic cell (Balcova et al., 2016; Davies et al., 2016). Thus, at any point in time it is not the number of hotspots that is limiting: ‘Not all PRDM9 binding sites become hotspots, and the reasons for this remain unclear’ (Altemose et al., 2017). Increasing PRDM9 dosage experimentally appears to increase hotspot usage, suggesting that the availability of PRDM9 can be rate limiting (Flachs et al., 2012; Davies et al., 2016). One unexplained implication of this is that an allelic pair of homologous hotspots could have PRDM9 molecules asymmetrically bound, not because of the erosion effect (Fig. 2), but because there was insufficient PRDM9. Recent work suggesting that their proper functioning requires aggregation (multimerization) of individual PRDM9 molecules (as homo- or heteromeric complexes) complicates this further (Baker et al., 2015; Altemose et al., 2017). However, another implication is that, in normal circumstances, widely dispersed, epigenetically tagged heterologies (e.g. polymorphisms) may need to persist for many generations if they are to come under the surveillance of PRDM9. Failing their removal from the population by natural selection (or drift), the hotspot erosion process (Fig. 2) would serve to expose cryptic PRDM9 targets and hence, over the generations, bring different genomic segments under surveillance, so ultimately achieving repair (Fig. 3). Then there is the pothole repair analogy. If a town has roads with many potholes and few repair vans, it makes sense to direct those vans to the most potholed roads. Thus, simply speeding up the pace of genome-wide surveillance might not be enough. It would seem better if close-knit collectives of epigenetically marked suspect sites could somehow (i.e. by being designated as hotspots) be recognized as needing more urgent treatment than isolated sites. Consistent with this, Arbeithuber et al. (2015) note that ‘regions in close vicinity to these PRDM9 binding sites also showed a significant enrichment of polymorphisms in humans’. Although some polymorphisms are stable (perhaps having escaped tagging), it seems possible that a local collective of polymorphisms that are enriched for epigenetic ‘suspect’ tags (Fig. 4) might somehow be able to attract some of the limited number of PRDM9 molecules to an otherwise cryptic hotspot in its proximity (Fig. 5). A hypertagged region could influence hotspot placement. Alternatively, such a collective of tagged heterologies might be able to trigger local hotspot mutation to increase affinity for PRDM9 or to send a long-range signal to the Prdm9 gene that it is time to mutate its zinc fingers so as to reconfigure its global control. Whatever the ‘pothole’ search mechanism, this underlines the value of focusing recombination to hotspots, rather than randomly. Figure 5. View largeDownload slide Hypothetical recruitment of a fresh PRDM9 target site by a chromosome sequence that has been epigenetically marked (*) to bias meiotic conversion. Details are as in Figure 2. At the top, the curved grey arrow indicates a presumptive target site with the same sequence as those that have been targeted in previous generations (curved black arrows). The small grey arrow indicates epigenetic sequence modification prompted by some aspect of the asterisk-marked region. At the bottom, the fresh target site is recognized. Given that the PRDM9 concentration is limiting, one of the previous target sites is vacated. The suspect sequence with a mutation (X) has been directionally converted and the tag eliminated. Figure 5. View largeDownload slide Hypothetical recruitment of a fresh PRDM9 target site by a chromosome sequence that has been epigenetically marked (*) to bias meiotic conversion. Details are as in Figure 2. At the top, the curved grey arrow indicates a presumptive target site with the same sequence as those that have been targeted in previous generations (curved black arrows). The small grey arrow indicates epigenetic sequence modification prompted by some aspect of the asterisk-marked region. At the bottom, the fresh target site is recognized. Given that the PRDM9 concentration is limiting, one of the previous target sites is vacated. The suspect sequence with a mutation (X) has been directionally converted and the tag eliminated. Some of the multitude of recombination-related proteins so far discovered (very important but not mentioned here) could also be involved. Noting that an increase in PRDM9 dosage only ‘partially rescues hybrid sterility of PWD x B6 F1 males’, Balcova et al. (2016) distinguish the identification of hotspot targets from the rate of events subsequent to that identification, which implicate a gene. Their results are interpreted as ‘strongly indicating an independent control of global crossover rate variation and genomic crossover placement’. Given the current pace of research in this area, a role for gene products could soon emerge (Gregorova et al., 2018). CONCLUSIONS Speciation is a possible outcome of increasing variation between individuals in a species. Although variation can be diminished, usually in local genome regions, by selective sweeps, there are also numerous internal mechanisms operating genome-wide to maintain DNA sequence integrity (Bernstein & Bernstein, 1991). This could account for the ‘mysterious’ selective pressure on Prdm9 (Lesecque et al., 2014). The quicker Prdm9 can mutate to facilitate symmetrical recognition of new hotspot sequences, the quicker and more comprehensive can be the repair of heterozygosities, and the less likely the triggering of a speciation event. Given that this now appears to be the main role of PRDM9, it follows that Prdm9 is best regarded as an anti-speciation gene (Brand & Presgraves, 2016; Reese & Forsdyke, 2016). The case for this, which includes explaining the hotspot conversion paradox, is made here in the context of a chromosomal basis for hybrid sterility. To this extent, the chromosomal hypothesis is supported over genic alternatives as governing the initiation of sympatric speciation in the general case. Of course, numerous genes participate in the orchestration of meiosis, and malfunction of only one can produce hybrid sterility (Forsdyke, 2010). However, in general, it is malfunction of the most distinctive feature of meiosis – the sequence-dependent pairing of homologous chromosomes – that is likely to be decisive in sparking divergence into new species, at least in a sympatric context. ACKNOWLEDGEMENTS Virgil Reese, Root Gorelick, Rosemary Redfield and five anonymous referees gave valuable advice. Queen’s University hosts my biohistory web pages (http://post.queensu.ca/~forsdyke/evolutio.htm or https://archive-it.org/collections/7641). The bioRxiv preprint server has an early version of this review (Forsdyke, 2017b). REFERENCES Altemose N, Noor N, Bitoun E, Tumian A, Imbeault M, Chapman JR, Aricescu AR, Myers SR. 2017. A map of human PRDM9 binding provides evidence for novel behaviors of PRDM9 and other zinc-finger proteins in meiosis. eLife  6: e28383. Google Scholar CrossRef Search ADS   Arbeithuber B, Betancourt AJ, Ebner T, Tiemann-Boege I. 2015. Crossovers are associated with mutation and biased gene conversion at recombination hotspots. Proceedings of the National Academy of Sciences of the United States of America  112: 2109– 2114. Google Scholar CrossRef Search ADS   Baker CL, Petkova P, Walker M, Flachs P, Mihola O, Trachtulec Z, Petkov PM, Paigen K. 2015. Multimer formation explains allelic suppression of PRDM9 recombination hotspots. PLoS Genetics  11: e1005512. Google Scholar CrossRef Search ADS   Balcova M, Faltusova B, Gergelits V, Bhattacharyya T, Mihola O, Trachtulec Z, Knopf C, Fotopulosova V, Chvatalova I, Gregorova S, Forejt J. 2016. Hybrid sterility locus on chromosome X controls meiotic recombination rate in mouse. PLoS Genetics  12: e1005906. Google Scholar CrossRef Search ADS   Bernstein C, Bernstein H. 1991. Aging, sex and DNA repair . San Diego: Academic Press. Bernstein H, Bernstein C, Michod RE. 2018. Sex in microbial pathogens. Infection, Genetics and Evolution  57: 8– 25. Google Scholar CrossRef Search ADS   Bhattacharyya T, Gregorova S, Mihola O, Anger M, Sebestova J, Denny P, Simecek P, Forejt J. 2013. Mechanistic basis of infertility of mouse intersubspecific hybrids. Proceedings of the National Academy of Sciences of the United States of America  110: E468– E477. Google Scholar CrossRef Search ADS   Bhattacharyya T, Reifova R, Gregorova S, Simecek P, Gergelits V, Mistrik M, Martincova I, Pialek J, Forejt J. 2014. X chromosome control of meiotic chromosome synapsis in mouse inter-subspecific hybrids. PLoS Genetics  10: e1004088. Google Scholar CrossRef Search ADS   Boulton A, Myers RS, Redfield RJ. 1997. The hotspot conversion paradox and the evolution of meiotic recombination. Proceedings of the National Academy of Sciences of the United States of America  94: 8058– 8063. Google Scholar CrossRef Search ADS   Brand CA, Presgraves DC. 2016. Evolution: on the origin of symmetry, synapsis, and species. Current Biology  26: 325– 328. Google Scholar CrossRef Search ADS   Brick K, Smagulova F, Khil P, Camerini-Otero RD, Petukhova GV. 2012. Genetic recombination is directed away from functional genomic elements in mice. Nature  485: 642– 645. Google Scholar CrossRef Search ADS   Chapman KM, Wilkey MM, Potter KE, Waldman BC, Waldman AS. 2017. High homology is not required at the site of strand invasion during recombinational double-strand break repair in mammalian chromsomes. DNA Repair  60: 1– 8. Google Scholar CrossRef Search ADS   Cox EC, Yanofsky C. 1969. Mutator gene studies in Escherichia coli. Journal of Bacteriology  100: 390– 397. Danilowicz C, Lee CH, Kim K, Hatch K, Coljee VW, Kleckner N, Prentiss M. 2009. Single molecule detection of direct, homologous, DNA/DNA pairing. Proceedings of the National Academy of Sciences of the United States of America  106: 19824– 19829. Google Scholar CrossRef Search ADS   Davies B, Hatton E, Altemose N, Hussin JG, Pratto F, Zhang G, Hinch AG, Moralli D, Biggs D, Diaz R, Preece C, Li R, Bitoun E, Brick K, Green CM, Camerini-Otero RD, Myers SR, Donnelly P. 2016. Re-engineering the zinc fingers of PRDM9 reverses hybrid sterility in mice. Nature  530: 171– 176. Google Scholar CrossRef Search ADS   Delph LF, Demuth JP. 2016. Haldane’s rule: genetic bases and their empirical support. Journal of Heredity  107: 383– 391. Google Scholar CrossRef Search ADS   Flachs P, Mihola O, Šimeček P, Gregorová S, Schimenti JC, Matsui Y, Baudat F, de Massy B, Piálek J, Forejt J, Trachtulec Z. 2012. Interallelic and intergenic incompatibilities of the Prdm9 (Hst1) gene in mouse hybrid sterility. PLoS Genetics  8: e1003044. Google Scholar CrossRef Search ADS   Foote AD. 2018. Sympatric speciation in the genomic era. Trends in Ecology & Evolution  33: 86– 95. Google Scholar CrossRef Search ADS   Forsdyke DR. 1996. Different biological species “broadcast” their DNAs at different (G+C)% “wavelengths”. Journal of Theoretical Biology  178: 405– 417. Google Scholar CrossRef Search ADS   Forsdyke DR. 2000. Haldane’s rule: hybrid sterility affects the heterogametic sex first because sexual differentiation is on the path to species differentiation. Journal of Theoretical Biology  204: 443– 452. Google Scholar CrossRef Search ADS   Forsdyke DR. 2001. The origin of species, revisited . Montreal: McGill-Queen’s University Press. Forsdyke DR. 2004. Chromosomal speciation: a reply. Journal of Theoretical Biology  230: 189– 196. Google Scholar CrossRef Search ADS   Forsdyke DR. 2007. Molecular sex: the importance of base composition rather than homology when nucleic acids hybridize. Journal of Theoretical Biology  249: 325– 330. Google Scholar CrossRef Search ADS   Forsdyke DR. 2010. George Romanes, William Bateson, and Darwin’s ‘weak point’. Notes and Records of the Royal Society of London  64: 139– 154. Google Scholar CrossRef Search ADS   Forsdyke DR. 2011a. The selfish gene revisited: reconciliation of Williams-Dawkins and conventional definitions. Biological Theory  5: 246– 255. Google Scholar CrossRef Search ADS   Forsdyke DR. 2011b. The B in BDM. William Bateson did not advocate a genic speciation theory. Heredity  106: 202. Google Scholar CrossRef Search ADS   Forsdyke DR. 2016. Evolutionary Bioinformatics , 3rd edn. New York: Springer. Google Scholar CrossRef Search ADS   Forsdyke DR. 2017a. Base composition, speciation, and why the mitochondrial barcode precisely classifies. Biological Theory  12: 157– 168. Google Scholar CrossRef Search ADS   Forsdyke DR. 2017b. The chromosomal basis of species initiation: Prdm9 as an anti-speciation gene. BioRxiv . doi: https://doi.org/10.1101/170860 Forsdyke DR. 2017c. Speciation: Goldschmidt’s chromosomal heresy, once supported by Gould and Dawkins, is again reinstated. Biological Theory  12: 4– 12. Google Scholar CrossRef Search ADS   Forsdyke DR. 2018. Hybrid sterility can only be primary when acting as a reproductive barrier for sympatric speciation. arXiv : 1802.01996v2. Available at: https://arxiv.org/ftp/arxiv/papers/1802/1802.01996.pdf Fuller ZL, Leonard CJ, Young RE, Schaeffer SW, Phadnis N. 2017. The role of chromosomal inversions in speciation. BioRxiv . doi: 10.1101/211771. Gladyshev E, Kleckner N. 2014. Direct recognition of homology between double helices of DNA in Neurospora crassa. Nature Communications  5: 3509. Google Scholar CrossRef Search ADS   Gladyshev E, Kleckner N. 2016. Recombination-independent recognition of DNA homology for repeat-induced point mutation (RIP) is modulated by the underlying nucleotide sequence. PLoS Genetics  12: e1006015. Google Scholar CrossRef Search ADS   Gladyshev E, Kleckner N. 2017. Recombination-independent recognition of DNA homology for repeat-induced point mutation. Current Genetics  63: 389– 400. Google Scholar CrossRef Search ADS   Gorelick R, Heng HHQ. 2010. Sex reduces genetic variation: a multidisciplinary review. Evolution  65: 1088– 1098. Google Scholar CrossRef Search ADS   Gregorova S, Gergelits V, Irena Chvatalova I, Bhattacharyya T, Valiskova B, Fotopulosova V, Jansa P, Wiatrowska D, Forejt J. 2018. Modulation of Prdm9-controlled meiotic chromosome asynapsis overrides hybrid sterility inmice. eLife 7: e34282. Grey C, Barthès P, Chauveau-Le Friec G, Langa F, Baudat F, de Massy B. 2011. Mouse PRDM9 DNA-binding specificity determines sites of histone H3 lysine 4 trimethylation for initiation of meiotic recombination. PLoS Biology  9: e1001176. Google Scholar CrossRef Search ADS   Harvey MG, Seeholzer GF, Smith BT, Rabosky DL, Cuervo AM, Brumfield RT. 2017. Positive association between population genetic differentiation and speciation rates in New World birds. Proceedings of the National Academy of Sciences of the United States of America  114: 6328– 6333. Google Scholar CrossRef Search ADS   Hayashi K, Yoshida K, Matsui Y. 2005. A histone H3 methyltransferase controls epigenetic events required for meiotic prophase. Nature  438: 374– 378. Google Scholar CrossRef Search ADS   Irie S, Tsujimura A, Miyagawa Y, Ueda T, Matsuoka Y, Matsui Y, Okuyama A, Nishimune Y, Tanaka H. 2009. Single nucleotide polymorphisms of the PRDM9 (MEISETZ) gene in patients with nonobstructive azoospermia. Journal of Andrology  30: 426– 431. Google Scholar CrossRef Search ADS   Johannesson K. 2010. Are we analyzing speciation without prejudice? Annals of the New York Academy of Sciences  1206: 143– 149. Google Scholar CrossRef Search ADS   Kafer GR, Li X, Horii T, Suetake I, Tajima S, Hatada I, Carlton PM. 2016. 5-Hydroxymethylcytosine marks sites of DNA damage and promotes genome stability. Cell Reports  14: 1283– 1292. Google Scholar CrossRef Search ADS   Kao KC, Schwartz K, Sherlock G. 2010. A genome-wide analysis reveals no nuclear Dobzhansky-Muller pairs of determinants of speciation between S. cerevisiae and S. paradoxus, but suggests more complex incompatibilities. PLoS Genetics  6: e1001038. Google Scholar CrossRef Search ADS   Kliman RM, Rogers BT, Noor MAF. 2001. Differences in (G+C) content between species: a commentary on Forsdyke’s “chromosomal viewpoint” of speciation. Journal of Theoretical Biology  209: 131– 140. Google Scholar CrossRef Search ADS   Lao PJ, Forsdyke DR. 2000. Crossover hotspot instigator (Chi) sequences in Escherichia coli occupy distinct recombination/transcription islands. Gene  243: 47– 57. Google Scholar CrossRef Search ADS   Lesecque Y, Glémin S, Lartillot N, Mouchiroud D, Duret L. 2014. The red queen model of recombination hotspots evolution in the light of archaic and modern human genomes. PLoS Genetics  10: e1004790. Google Scholar CrossRef Search ADS   Liénard MA, Araripe LO, Hartl DL. 2016. Neighboring genes for DNA-binding proteins rescue male sterility in Drosophila hybrids. Proceedings of the National Academy of Sciences of the United States of America  113: E4200– E4207. Google Scholar CrossRef Search ADS   Louis EJ. 2009. Origins of reproductive isolation. Nature  457: 549– 560. Google Scholar CrossRef Search ADS   Mallet J. 2006. What does Drosophila genetics tell us about speciation? Trends in Ecology & Evolution  21: 386– 393. Google Scholar CrossRef Search ADS   Mérot C, Salazar C, Merrill RM, Jiggins CD, Joron M. 2017. What shapes the continuum of reproductive isolation? Lessons from Heliconius butterflies. Proceedings of the Royal Society B: Biological Sciences  284: 20170335. Google Scholar CrossRef Search ADS   Mihola O, Trachtulec Z, Vicek C, Schimenti JC, Forejt J. 2009. A mouse speciation gene encodes a meiotic histone H3 methyltransferase. Science  323: 373– 375. Google Scholar CrossRef Search ADS   Montgomery TJ. 1901. A study of the chromosomes of the germ cells of metazoa. Transactions of the American Philosophical Society  20: 154– 236. Google Scholar CrossRef Search ADS   Myers S, Bottolo L, Freeman C, McVean G, Donnelly P. 2005. A fine-scale map of recombination rates and hotspots across the human genome. Science  310: 321– 324. Google Scholar CrossRef Search ADS   Nei M, Nozawa M. 2011. Roles of mutation and selection in speciation: from Hugo de Vries to the modern genomic era. Genome Biology and Evolution  3: 812– 829. Google Scholar CrossRef Search ADS   Nevo E. 2012. Speciation: chromosomal mechanisms. In Encyclopedia of life sciences . Chichester, UK: John Wiley. doi: 10.1002/9780470015902.a0001757.pub3. Oliver PL, Goodstadt L, Bayes JJ, Birtle Z, Roach KC, Phadnis N, Beatson SA, Lunter G, Malik HS, Ponting CP. 2009. Accelerated evolution of the Prdm9 speciation gene across diverse metazoan taxa. PLoS Genetics  5: e1000753. Google Scholar CrossRef Search ADS   Payseur BA. 2016. Genetic links between recombination and speciation. PLoS Genetics  12: e1006066. Google Scholar CrossRef Search ADS   Paz A, Kirzhner V, Nevo E, Korol A. 2006. Coevolution of DNA-interacting proteins and genome ‘dialect’. Molecular Biology and Evolution  23: 56– 64. Google Scholar CrossRef Search ADS   Poole AM, Logan DT. 2005. Modern mRNA proofreading and repair: clues that the last universal common ancestor possessed an RNA genome. Molecular Biology and Evolution  22: 1444– 1455. Google Scholar CrossRef Search ADS   Reanney DC. 1979. RNA splicing and polynucleotide evolution. Nature  277: 598– 600. Google Scholar CrossRef Search ADS   Reese VR, Forsdyke DR. 2016. Meiotic pairing inadequacies at the levels of X chromosome, gene, or base: epigenetic tagging for transgenerational error-correction guided by a future homologous duplex. Biological Theory  11: 150– 157. Google Scholar CrossRef Search ADS   Schartl M. 2008. Evolution of Xmrk: an oncogene, but also a speciation gene? BioEssays  30: 822– 832. Google Scholar CrossRef Search ADS   Ségurel L, Leffler EM, Przeworski M. 2011. The case of fickle fingers: how the PRDM9 zinc finger protein specifies meiotic recombination hotspots in humans. PLoS Biology  9: e1001211. Google Scholar CrossRef Search ADS   Wahls WP, Davidson MK. 2011. DNA sequence-mediated, evolutionarily rapid redistribution of meiotic recombination hotspots. Genetics  189: 685– 694. Google Scholar CrossRef Search ADS   Winge Ö. 1917. The chromosomes, their number and general importance. Comptes-Rendus des Travaux due Laboratoire Carlsberg  13: 131– 275. © 2018 The Linnean Society of London, Biological Journal of the Linnean Society This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)