Fertility Cost (or Sometimes a Lack of It) in Relation to Heterozygosity for Robertsonian Rearrangements in Mammals: A Review Open Access

To evolutionary biologists, chromosomal rearrangements in diploid organisms are of particular interest because of their impact on meiosis in the heterozygous state. The first division of meiosis is tuned to structurally equal homologs with pairing, crossing over, and segregation of the bivalent. But in heterozygotes for chromosomal rearrangements, the homologs differ and the heteromorphic configuration that forms during prophase I have to change the way it pairs, recombines and segregates compared with a normal bivalent. This can affect the exchange of genes between the homologs, the equality of transmission of the homologs and the fertility of the heterozygote.

These meiotic traits of heterozygotes for chromosomal rearrangements may have an impact on the fate of the rearrangement in the population, i.e., whether it is likely to be lost from the population, become fixed or remain as a polymorphism in a population. If a rearrangement becomes fixed in a population, then the meiotic behavior of heterozygotes may become a factor in maintaining adaptive differences and/or contributing to reproductive isolation from other populations.

Robertsonian (Rb) rearrangements are one type of chromosomal rearrangement found in various diploid organisms, including mammals, and are the focus of this review (Table 1). Rb rearrangements were first described over a century ago in Orthoptera by WRB Robertson [1], after whom they are named. Rb fusions (also known as centric fusions and Rb translocations) involve the joining of two acrocentric chromosomes at their centromeres to form a metacentric chromosome (Fig. 1a). For simplicity, the general terms acrocentric, meaning a chromosome with a centromere at or near one end of the chromosome, and metacentric, meaning a chromosome with an internal centromere, are used in preference to more restrictive terms (e.g., telocentric with a centromere exactly at the end of a chromosome or submetacentric if the centromere is not precisely in the middle of the chromosome). An Rb fission is the reverse process to an Rb fusion, with the generation of two acrocentrics from a metacentric by splitting of the chromosome at the centromere (Fig. 1b). A third type of Rb rearrangement is a whole-arm reciprocal translocation (WART) [2, 3]. WARTs at their simplest (as shown in Fig. 1c) involve the swapping of chromosome arms between a metacentric and an acrocentric or between two metacentrics, such that a new type of metacentric is generated. Thus, Rb rearrangements in some way alter the composition of acrocentric and metacentric chromosomes in a karyotype, but do not change the number of chromosome arms, i.e., the nombre fondamental (NF) [4] remains the same. The breakages involved in Rb rearrangements are most likely to occur in repetitive DNA at the centromere and, therefore, in general, probably cause no direct phenotypic effect. The high frequency of occurrence of Rb rearrangements in mammals over evolutionary time and the apparent phenotypic normality of carriers of recently arisen Rb rearrangements indicates that this is a reasonable assumption. The possibility of a direct impact of Rb mutations on phenotype is not considered further in this review, but see [5, 6].

Based on decades of cytogenetics on mammals, Rb rearrangements are probably the most commonly recorded change to the karyotype [91]. Because they involve whole chromosome arms, they are easily detected. More recently, genomics has been used to describe structural variation in the genome, which includes not only the chromosomal rearrangements that can be detected by classical cytogenetics but also much finer-scale changes [92]. Genomic studies suggest that inversions (ranging from small to large) are the most frequent structural variation in the mammalian genome, but because they have been much more difficult to detect than Rb rearrangements by classical cytogenetics, there are fewer empirical data on their impacts, although there have been many theoretical studies on the importance of inversions to adaptation and speciation.

The subject of this review is the fertility of heterozygotes for Rb rearrangements in mammals, i.e., a consideration of how their abnormal meiosis can directly and obviously explain any infertility (reduction in numbers of offspring) or sterility (zero offspring). We consider in particular how the infertility (or lack of it) may affect, first, the probability of fixation of Rb rearrangements and, second, the probability that chromosomal forms become distinct species. Most discussion of levels of infertility in Rb heterozygotes has been directed toward these two aspects, and less toward the existence or otherwise of balanced polymorphism in mammals, although this latter topic is discussed by [93, 94].

Following [2], we distinguish two broad types of Rb heterozygotes: simple and complex (Fig. 2). For particular chromosome arm combinations, simple Rb heterozygotes have a metacentric and homologous acrocentrics, this condition arising through either Rb fusion or fission. This leads to formation of a chain-of-three (CIII) configuration at prophase I, also known as a trivalent (Fig. 2a). An individual may be heterozygous for multiple Rb rearrangements and would be expected to form multiple trivalent configurations at prophase I (Fig. 2b). In the case of a complex Rb heterozygote, the karyotype includes metacentrics which share an arm in common (i.e., they show monobrachial homology). There may be just two metacentrics with monobrachial homology and these heterozygotes would produce a chain-of-four configuration (CIV) at prophase I (Fig. 2c). However, complex heterozygotes may have more than two metacentrics showing monobrachial homology and, depending on which combinations of metacentrics show monobrachial homology, a longer chain configuration could be formed at prophase I, or a ring configuration, or multiple non-bivalent configurations with at least one composed of two or more metacentrics (Fig. 2d–f). Complex heterozygotes can arise when there have been multiple Rb fusions and/or one or more WARTs.

This article is about the fertility of Rb heterozygotes as occur within species or can be generated by captive crosses within species or between closely related species. Rb variation at these taxonomic levels typically involves autosomes; this will be assumed unless stated otherwise.

Two taxa stand out as exceptional in terms of the number of studies on the fertility of both simple and complex Rb heterozygotes. These are the western house mouse Mus musculus domesticus and the common shrew Sorex araneus. Both taxa are abundant and easily collected and show exceptional Rb variation in nature, with numerous discrete populations distinguished by differing sets of metacentrics and acrocentrics (known as Rb races). Our knowledge of fertility in Rb heterozygotes has been augmented by other species of mammals which, like the shrews and mice, have been studied in natural populations or involving crosses of wild-derived individuals in captivity. There have also been studies of humans and of domestic animals.

The most obvious way to measure fertility costs associated with heterozygosity for Rb rearrangements in mammals is to compare litter sizes (or other measure of number of offspring) in Rb heterozygotes versus homozygotes. This has been done by making appropriate crosses in captivity (e.g., [26]) or studying pregnancies in nature (e.g., [95]). Even more valuable is to measure long term, or even lifetime, reproductive output (e.g., [19]).

When there is a partial reduction in litter size in Rb heterozygotes versus homozygotes, the infertility can often be attributed to anaphase I nondisjunction (e.g., [39]). In Rb heterozygotes, the expected heteromorphic configurations at meiosis I contain more elements than a bivalent (Fig. 2), and it can be anticipated that this deviation from the norm makes the configurations more prone to errors in segregation at anaphase I, the more so the greater the number of elements in the configurations. For configurations of three or more chromosomes in Rb heterozygotes, the “correct” segregation is alternate (every other chromosome in a configuration going to the same pole), but, instead, segregation may be adjacent and this is what is termed “nondisjunction” (Fig. 3). Thus, considering an individual that is a simple heterozygote for the metacentric 1.2, a meiotic trivalent (CIII) configuration is expected (Fig. 2a). Alternate segregation would result in 1.2 going to one pole at anaphase I and the homologous acrocentrics 1 and 2 going to the other pole, leading to chromosomally balanced gametes (Fig. 3). Nondisjunction would be represented by 1.2 and 1 going to one pole and 2 to the other, or 1.2 and 2 going to one pole and 1 to the other (i.e., adjacent segregation), or all three chromosomes going to one pole (i.e., 3:0 segregation, see [87])(Fig. 3). We use the term nondisjunction for any error of segregation, including situations where chromosomes fail to form chiasmata and therefore segregate independently [96]. When Rb heterozygous mammals show nondisjunction, the aneuploid gametes produced fertilize as readily as euploid gametes [97], such that aneuploid zygotes are produced which generally die around the time of implantation (monosomics) or during gestation (trisomics) [98].

Nondisjunction frequencies for particular categories of Rb heterozygotes can be determined by chromosome counts of metaphase II cells in both males and females (e.g., [21]). Nondisjunction frequencies can also be estimated from typing embryos and fetuses and detecting aneuploidy that way (e.g., [27, 28]). Metaphase II counts are much easier to make for males than females and it is possible to score substantial numbers of cells for each distinct karyotype, if there are several individuals with that karyotype (e.g., [7]). It is difficult to identify which chromosome is aneuploid in a metaphase II spread using a conventional stain, although FISH/chromosome painting may be applied to identify particular chromosomes allowing an estimate of nondisjunction frequencies for those (e.g., [30]). FISH can also be applied to detect aneuploidy for particular chromosomes in decondensed sperm (SpermFISH), allowing thousands of gametes per individual male to be typed (e.g., [74]).

As well as the difficulties associated with chromosome segregation at anaphase I, the configurations of three or more chromosomes in Rb heterozygotes can display abnormalities earlier in meiosis I, triggering germ cell death. In both male and female mammals germ cell death can be associated with asynapsis of chromosomal regions that normally would be synapsed, and in males it can also occur if the normally asynapsed XY sex bivalent shows unusual synapsis [99, 100]. Germ cells modify asynapsed regions to make them transcriptionally silenced, while synapsed regions are transcriptionally active [101]; if regions that are typically transcriptionally active become silenced, or if regions that are normally silenced become active – such inappropriate gene expression may be cell lethal. Again, the unusualness of the heteromorphic configuration in Rb heterozygotes is likely the basis for errors in pairing [102], particularly near the mutation site (the centromere) where late pairing and incomplete homology may promote asynapsis, deviating from the typical synapsed state of autosomes at pachytene (Fig. 4). Interestingly, if a chromosome region shows non-homologous synapsis (rather than asynapsis) that appears to protect it from cell death as if it were showing straight pairing [99] (Fig. 4). Synaptonemal complex analysis of pachytene cells allows an assessment of straight pairing and non-homologous synapsis (e.g., [103]) and also asynapsis and meiotic silencing of unpaired chromatin (MSUC)(e.g., [15]). The presence of asynapsis and associated MSUC are traits associated with cell death, although this contention is not without controversy [34]. Pachytene studies can include an assessment of nuclear architecture; determining, for example, if the heteromorphic configuration is brought into close contact with the sex bivalent and thereby potentially disrupting the asynaptic state of that [37, 104]. It is also possible to assess the level of germ cell death directly, comparing heterozygotes with the standard, homozygote, condition. In females, this can be done by histology of the ovary, counting the number of follicles (e.g., [45]). In males, the presence of germ cells on slides made from a testis suspension (for metaphase II counts or for synaptonemal complexes, e.g., [61]) gives an indication of the amount of germ cell death. More precision for males comes from measuring testis mass [105] or epididymal sperm counts [106] and from testis histology where the ratio of primary spermatocytes to spermatids can provide a valuable quantitative measure (the expectation is 1:4 [107]). Germ cell death in females may cause infertility by reducing reproductive lifespan [99]; litter sizes earlier in their reproductive life maybe normal. In males, it may require a major reduction in sperm to reduce reproductive output at any particular time (10% or less of normal levels [106]). Some types of Rb heterozygotes in a particular species may always show a complete interruption to spermatogenesis – thereby always being completely sterile. Other types of heterozygote with some sperm production may show normal or partial fertility in some individuals but sterility in others whose sperm levels are consistently below the threshold (e.g., [72]).

As well as estimating infertility (nondisjunction and germ cell death, and their consequences – reduced reproductive output) directly, as described above, there are indirect ways of measuring infertility. The Rb heterozygotes in some species of mammals, including house mice and common shrews, are particularly found in areas of contact and hybridization between Rb races, known as “hybrid zones” [108, 109]. Any infertility associated with the Rb heterozygotes will reduce their fitness. Hybrid zones where the hybrids show unfitness are known as “tension zones” and the width of the tension zone is determined by the level of hybrid fitness (the greater the unfitness, the narrower the zone) and dispersal (the greater the dispersal, the wider the zone) [110]. If, as seems reasonable, the major cause of a reduced fitness in Rb heterozygotes is their infertility, then comparing the width of hybrid zones characterized by different types of Rb heterozygote provides an indirect estimate of fertility in those heterozygotes (e.g., [111]).

Table 1 provides more than 150 entries on fertility of Rb heterozygotes in mammals. The majority of these relate to single studies reporting fertility on a particular category of Rb heterozygote. Sometimes the comparison with homozygotes is incorporated into the entry, in other cases a separate entry is made for homozygotes. Where two or more studies are of a similar nature and can be conveniently included in a single entry, that is done. There are also some instances where data for a particular species have previously been reviewed, and the entry refers to what is written in that review. The precision of the data varies from study to study. Where possible, sample sizes are given.

Below we will summarize the findings reported in Table 1. However, for scholars interested in the details of fertility in Rb heterozygotes among different species of mammal, we strongly recommend spending time perusing the table in detail. The data are heterogeneous in all sorts of respects, and it is through careful line-by-line inspection that one gets the best “feel” for the generalities.

In our summary, we first consider broad categories of heterozygotes, making comparisons among different species. We then have a section on “case-by-case variation” where we investigate how infertility may reflect generalities that transcend those different categories. We will not continually refer to Table 1: that is where to look for references to species-specific work being reported.

Starting with findings in the wild common shrews [95], near-normal fertility has increasingly been seen as usual for naturally occurring simple Rb heterozygotes forming one or a few trivalents (typically, one to three trivalents: “1–3 CIII heterozygotes”) in mammals. In the early 1980s it was surprising that Rb rearrangements, so commonly distinguishing species of mammals, should have such a mild effect on fertility, given assumptions held by some evolutionary biologists that closely related species are reproductively isolated due to unfitness associated with heterozygosity for such rearrangements (see e.g., [112]). In the case of the common shrew, there have been numerous confirmatory studies to show that, compared with homozygotes, there is little loss of fertility in 1–3 CIII heterozygotes. These studies have examined litter sizes (fetal counts), anaphase I nondisjunction in males (metaphase II counts), anaphase I nondisjunction in males and females (fetal karyotypes), synaptonemal complexes in males and germ cell death in females (histological analysis) and males (histological analysis and sperm counts). There have been slight signs of infertility in these simple heterozygotes, e.g., some fetal products of trivalent nondisjunction [8] and somewhat lower follicle counts in simple heterozygotes than homozygotes. However, the average fertility loss is very small. This marginal infertility is supported by measured chromosome cline widths in hybrid zones which indicate that the selection against heterozygotes forming one trivalent maybe less than 1% [111]. Other species with a natural (or close to natural) genome where simple heterozygotes forming one (sometimes more than one) trivalent have similar fertility to homozygotes include the earth-colored mouse, South African pouched mouse, Brazilian marsh rat, Chinese striped hamster, Middle East blind mole-rat, goitered gazelle, impala and red brocket deer. Domestic sheep also show this trait.

Considering other species examined with a natural or near-natural genome, in some instances fertility in 1–3 CIII heterozygotes is perceptibly reduced compared with homozygotes, but rarely exceptionally so. The best-represented species in Table 1, the house mouse, shows a mixed picture in terms of infertility. Several studies show a common shrew-like result of low nondisjunction in 1–3 CIII heterozygotes, less than 4% in males, but others show values over 10%. Litter sizes, where they have been measured, have been similar to homozygotes. Germ cell death tends to be slightly higher in male 1–3 CIII heterozygotes than homozygotes, and sperm counts tend to be slightly lower. Unfortunately, the data for female 1–3 CIII heterozygotes house mice are limited.

Large Japanese field mice also show a nondisjunction frequency about 10% higher in male CIII heterozygotes than homozygotes. For Molina’s grass mouse, the nondisjunction frequency in CIII heterozygotes is about 10% higher than one type of homozygote (metacentric), but about 20% lower than the other (acrocentric). For black rats, 1–2 CIII heterozygotes have reduced litter sizes and in some cases are semisterile. Zaisan mole voles that are 1–2 CIII heterozygotes can be fertile or sterile.

Other data relate to humans and domestic and farmed species. In humans there can be extreme infertility in CIII heterozygotes, high germ cell death leading to oligospermic and even sterile males (although other individuals heterozygous for the same chromosomes are not infertile in this way) [72]. Nondisjunction frequencies can be high in humans that are CIII heterozygotes, particularly in females (with documented values up to 60%). The recorded pregnancy risks of aneuploid fetuses (products of nondisjunction) do not match the more extreme estimates of nondisjunction from gametes, but there is a substantial risk of trisomy 21 (15% in the second trimester) for females heterozygous for the Rb metacentric 14;21. Female cattle and pigs that are CIII heterozygotes also show reduced reproductive output and/or relatively high nondisjunction. The data on arctic fox are mixed as to whether litter sizes are reduced in CIII heterozygotes or not. There can also be substantially enhanced nondisjunction in CIII heterozygotes of the house mouse when wild-derived Rb metacentrics are introduced into laboratory mouse strains, as already noted by [39].

Overall, 1–3 CIII heterozygotes often have near-normal fertility. However, principally in humans and domestic species, and particularly in females, there can be high nondisjunction. Sterility due to germ cell death is sporadic in 1–3 CIII heterozygotes. The data for humans are exceptional in that they reveal individuals that are sterile even when the norm is milder infertility – thereby showing the full spread of responses to Rb heterozygosity for a particular metacentric. For other species sample sizes are smaller and levels of recorded infertility generally refer to population averages; this refers to all types of heterozygotes, not just those in this section.

The best studied species for the fertility of simple Rb heterozygotes forming many trivalents is the house mouse. In general, the more extreme the heterozygosity in terms of number of trivalents, the greater the infertility (Fig. 5). In the house mouse it is possible to get simple Rb heterozygotes forming nine trivalents (9 CIII heterozygotes). These are produced by crosses between wild or wild-derived individuals of a 2n = 22 Rb race (with 9 metacentrics) and either laboratory mice with the standard 2n = 40 all-acrocentric karyotype or wild (or wild-derived) mice with 2n = 40. Highly heterozygous hybrids are produced in nature when 2n = 22 and 2n = 40 mice make contact, and in some cases 2n = 31 individuals (F₁ hybrids) are generated in the laboratory by crossing the two chromosomal forms that would hybridize in nature, and those F₁s are 9 CIII heterozygotes. Fertility can be low in these heterozygotes, e.g., anaphase I nondisjunction frequencies of over 70% in females, litter sizes of 0–1 compared with the norm of 4–5, and some cases of complete male sterility. In other instances, the fertility reduction is not so great (litter size of 63% of normal).

Interestingly, the examples of simple Rb heterozygotes forming many trivalents in other species also show a mild impact to fertility. Thus, South African pouched mice with heterozygotes forming up seven trivalents show little reduction in litter size, and testis mass and seminiferous tubule diameters similar to homozygotes. Zaisan mole voles that were 10 CIII heterozygotes also showed only slightly reduced litter sizes and no obvious disruption to spermatogenesis. Crosses between the Zaisan and northern mole voles also generated 10 CIII heterozygotes. However, in this case the heterozygotes were sterile, producing no offspring and the males producing no sperm. This extreme phenotype no doubt reflects to some extent the genic differences between the species; whether the chromosomal heterozygosity contributes to the sterility is unknown. The same also applies to male interspecific hybrids of lemurs (common brown X black) that are 8 CIII heterozygotes and sometimes (usually?) sterile. However, one out of six individuals did produce offspring. Hybrids between common brown and gray-headed lemurs that are 3–6 CIII heterozygotes are fertile with spermatogenesis not obviously different from the pure species (i.e., homozygotes).

The fertility data reviewed above for simple Rb heterozygotes forming one or a few trivalents indicate that a small difference in chromosome number between species is unlikely to be the sole cause of their reproductive isolation. The data that have accumulated for complex Rb heterozygotes allows us to examine further the potential of chromosomal rearrangements in promoting speciation. In response to new data on complex Rb heterozygotes in the house mouse in the 1970s and 1980s, the monobrachial speciation model was developed by Baker & Bickham [113], itself building off Capanna’s earlier emblematic speciation model [114]. The monobrachial speciation model posits that different Rb fusions may become fixed in different populations of a species, that fixation being possible because simple Rb heterozygotes forming one trivalent are reasonably fertile (which we now know to be true: see above). If metacentrics in those different populations (Rb races) share a single arm in common it means that on contact and hybridization the F₁s produced will be complex heterozygotes. In the simplest case a CIV configuration will be formed in such F₁s (Fig. 2c). It is an explicit expectation in Baker & Bickham’s model that CIV heterozygotes will be sterile. In fact, in the common shrew, the pig, the Brazilian marsh rat, the Zaisan mole vole and the house mouse the reduction in fertility has been found to be much less extreme. In both the common shrew and the Brazilian marsh rat anaphase I nondisjunction in heterozygous males is close to that of homozygotes and litter sizes also are not greatly affected by the CIV configuration. Germ cell death is slightly elevated in male shrew CIV heterozygotes relative to homozygotes but certainly plenty of mature sperm are observed. Zaisan mole vole CIV complex heterozygotes likewise have been found to have plentiful sperm, although reduced litter size. There is a degree of infertility afflicting CIV heterozygotes in house mice, but again not indicating a propensity for sterility. Anaphase I nondisjunction is ca. 17% (compared with a norm of 0–3% for homozygotes) and germ cell death is ca. 50% (compared with a norm of less than 30% for homozygotes – even homozygotes show some germ cell death between the spermatocyte and spermatid stages of spermatogenesis). In the Talas tuco-tuco, there is also greater asynapsis at pachytene in CIV heterozygotes relative to simple heterozygotes – but again individuals are expected to be reasonably fertile. The fact that CIV heterozygotes typically are not greatly infertile, and the same can be said for RIV heterozygotes (see below), is important because these are the types of heterozygote produced when a WART arises in a population. Thus, infertility may not be an excessive hindrance to fixation of new WARTs, which indicates that it is realistic to include them in chromosomal phylogenies [115, 116].

Complex heterozygotes in the common shrew producing a chain-of-five (CV) or a ring-of-four (RIV) meiotic configuration also do not show massive infertility. Male anaphase I nondisjunction is slightly elevated versus homozygotes but still less than 12% (nondisjunction in homozygotes is about 2% and about 4% in CIII heterozygotes [7]). However, the combination of enhanced male and female anaphase I nondisjunction and germ cell death in CV heterozygotes causes a sufficient reduction in fitness to promote selection for acrocentric chromosomes in hybrid zones where they occur [117]. Presumably average litter sizes are smaller and/or frequency of sterile males is increased and/or female reproductive lifespan is reduced. The presence of an “acrocentric peak” in such hybrid zones [108] means that the unfit CV heterozygotes are unlikely to be produced because that requires the presence of the relevant metacentrics, which are lacking because of the dominance of acrocentrics in the hybrid zone. Modeling supports the verbal argument that an acrocentric peak may be favored by natural selection when Rb races produce F₁ hybrids that are CV heterozygotes [118]. Once again house mouse CV heterozygotes show greater infertility (anaphase I nondisjunction, germ cell death) than in the common shrew, and sometimes similar and sometimes more extreme than recorded in mouse CIV heterozygotes. The infertility suffered by CV heterozygotes in house mice from Valtellina, northern Italy [21], was important because it apparently was sufficient to promote reinforcement [119]. In this case selection against the complex heterozygotes has seemingly favored assortative mating, such that Rb races in contact no longer hybridize. Thus, natural selection appears to have operated in different ways to avoid production of the same type of unfit complex heterozygote in shrews and mice.

Complex heterozygotes with short meiotic chains can be produced in other species than common shrews and house mice. Crosses between long-haired and dusky rats generates a hybrid with a CV configuration and three trivalents. In this case both males and females can be fertile, although there is a substantial reduction of litter size (by ca. 70%). Given that this is a cross between species and the fact that multiple heterozygosities are involved, it is difficult to tease out the contribution of the meiotic CV configuration to the infertility; genic factors and compounding effects of multiple non-bivalent configurations could be involved. Substantially reduced breeding success (very few litters, small numbers of offspring when there are litters) and much pachytene arrest in male spermatogenesis is also seen in CV + 2 CIII heterozygotes that are interspecific hybrids between Mongolian and Eversmann’s hamsters. For lemurs, fertility has been examined in male CIV or CV heterozygotes (also forming multiple trivalents) that are hybrid products of multiple species. One of the hybrids examined produced very few sperm and was undoubtedly sterile, but others showed plentiful sperm and had offspring. In Zaisan mole rats CV heterozygotes showed similar degrees of infertility to CIV heterozygotes.

The data on complex heterozygotes forming long meiotic configurations is limited to the house mouse, common shrew and lemurs. House mice with very long configurations (CXI, CXV, CXVI, CXVII, CXIX, RXVI, RXVIII) or multiple moderate length configurations (CVII + CVIII, 2 CIX) show great infertility. All these examples of meiotic chains are accompanied by male sterility (complete arrest of spermatogenesis in every instance), but some female CXV and CXVII heterozygotes have been shown to produce offspring (with very small litter sizes). Both male and female complex heterozygote mice with very long ring configurations have signs of highly reduced fertility but there is the possibility that some individuals may produce young (again very few). Heterozygous mice with rather shorter chains or rings (CVI, RVI) were also not typically male sterile (they show pachytene asynapsis and greater germ cell death than homozygotes, but not to an exceptional level) and female RVI heterozygotes have a similar number of follicles as homozygotes. In lemurs, interspecific hybrid males examined that had long chains (CXI, CXIII) were apparently sterile, but there is the complication of additional configurations (CIII, RVI) present, as well as potential genic contributions to sterility, given that the species being crossed may have diverged genetically in other ways than karyotypically. An interspecific hybrid male lemur that was a CVI + CIV + CIII heterozygote also had very few sperm and was likely sterile.

In the common shrew, long meiotic chains or multiple shorter chain configurations (CVII, CVIII, CIX, CX, CXI, CVI + CV, CVII + CIV) do not typically appear to be associated with male sterility. Measured germ cell death is moderate rather than severe (not exceeding 40%, compared to ca. 10% in homozygotes), morphologically normal sperm are seen on slides, and any measured anaphase I nondisjunction is also not extreme. This also applies to heterozygotes with CIV + RIV, CV + RIV, RIV + RIV configurations. The width of hybrid zones in the common shrew also indicates that the unfitness of F₁s with long meiotic configurations is not exceptional. The hybrid zones at the contact of the Novosibirsk and Tomsk Rb races and the Moscow and Seliger Rb races are most informative. The width of the former hybrid zone suggests selection against the F₁s (CIX heterozygotes) of 1–10% and for the latter hybrid zone selection against the F₁s (CXI heterozygotes) of 8–90%; the range depending on what dispersal value is used [111]. Neither of these hybrid zones have the complications of a geographical barrier.

For our own, well-studied species, the frequency of anaphase I nondisjunction for individuals of normal karyotype is well known to be higher in females than males [120], and available data indicate that the same distinction applies to human Rb heterozygotes (Table 1) [38, 71, 74]. Does that mean, for mammals in general, that female Rb heterozygotes will tend to suffer greater anaphase I nondisjunction than males? In house mice and common shrews, there are indications from Table 1 that 1–3 CIII heterozygotes do show higher nondisjunction in females (for the house mouse see also Fig. 5b extending to heterozygotes with a higher number of CIII configurations). More detailed confirmation comes from the extensive studies of wild-derived metacentrics introduced into laboratory mouse genomes (coupled with studies of Rb fusions that have arisen in laboratory mice) where “considerably higher meiotic non-disjunction rates occur in female rather than male carriers of the same Rb metacentric” [39] (see also [38, 55]). Considering house mice with other Rb karyotypes, Chatti et al. [35] do a particularly detailed comparison of 9 CIII heterozygotes comparing litter size and number of litters of house mice over 1 year, and males do have higher fertility.

Rb heterozygotes of both sexes have a tendency for increased germ cell death compared with homozygotes. For males, it may require a substantial drop in sperm output to cause sterility [106]. However, average values of germ cell death for a particular heterozygote category may fail to capture the impact of individual variation among the males. If there is an increase in germ cell death in Rb heterozygotes there will likely be an increase in numbers of sterile males because the shift in the distribution of germ cell death values will increase the number of individuals that cross the sterility threshold. This should be borne in mind in considering the data in Table 1 and in making an assessment of unfitness associated with particular categories of heterozygote (see discussion above on the selective response to the unfitness of CV heterozygotes in common shrews and house mice).

When there is an increased germ cell death in female Rb heterozygotes relative to homozygotes, it will reduce their reproductive lifespan [99]. Thus, in both males and females, even moderately increased germ cell death may reduce fitness – it does not require complete sterility. But in terms of likelihood of complete sterility associated with Rb heterozygosity, there is a clear sex difference. As seen very clearly in house mice, male complex heterozygotes forming long-chain configurations are always sterile. However, females can still have litters, as documented particularly carefully by Grize et al. [42] for CXV heterozygotes.

Above we separate our descriptions of simple Rb heterozygotes forming few meiotic trivalents from those forming many trivalents, and our descriptions of complex Rb heterozygotes forming short versus long meiotic configurations. Thus, for the whole range of species considered, there is a tendency for greater infertility the greater the number of meiotic heteromorphic configurations and the longer those configurations, as already noted with fewer data [2]. These relationships can also be seen within species for both the common shrew and house mouse (Table 1) and are supported by studies of hybrid zones in both species. Chromosome cline widths tend to be narrower when the F₁s have a larger number or greater size of heteromorphic configurations [108, 109].

Studies on the house mouse reveal another important feature. While males with long meiotic chain configurations are inevitably sterile with complete arrest of spermatogenesis, males with equally large ring configurations have much more normal testis histology and do produce some sperm [38, 39, 43]. Thus, male complex heterozygotes forming large meiotic rings have at least the potential of offspring, although 7 males in one study did not yield offspring [36], presumably because most of the surviving sperm were aneuploid due to anaphase I nondisjunction. There are indications that females with large meiotic ring configurations may also have higher fertility than females with long-chain configurations, with the latter suffering a higher frequency of anaphase I nondisjunction as well as greater germ cell death (Table 1).

For the common shrew, it is possible to compare germ cell death in Rb heterozygotes forming short meiotic chains and rings (Table 1). Four studies of CIV and CV heterozygotes showed 23.5–29.0% germ cell death (based on a total of 19 males), while two studies of RIV heterozygotes showed 18.5–24.0% germ cell death (based on a total of 10 males). Once again, the ring-forming heterozygotes have higher fertility.

It is important to note these findings showing greater infertility in chain-forming complex heterozygotes than ring-forming complex heterozygotes with an equivalent length of configuration, i.e., an equivalent level of Rb heterozygosity. It strongly indicates that the infertility that is observed in complex heterozygotes is a property of the chromosomes themselves rather than the result of some independently accumulated genic difference between the Rb races being crossed to generate the complex heterozygotes. The genic differences between Rb races may well correlate with the number of metacentric differences, maybe accumulated in allopatry, but it is happenstance whether F₁s between those Rb races should form meiotic rings or chains. The greater likelihood of male sterility associated with a long meiotic chain than a ring in house mice presumably reflects the presence of terminal acrocentrics which is the only difference between the ring and chain configurations (Fig. 2). Centromeric asynapsis and association with the XY bivalent has been noted for the terminal acrocentrics in long-chain configurations [40, 43]. Asynapsis and MSUC occurs elsewhere in chain and ring configurations [11, 15, 43], and the negative effects associated with that presumably accumulate the longer the configuration, but the additional very strong effect of asynapsis of the terminal acrocentrics may cause consistent cell death and complete arrest of spermatogenesis, as seen in house mice with long meiotic chain configurations. Johannisson & Winking [43] particularly emphasize that it is not just asynapsis of the acrocentrics that is important, it is the high level of association of the asynapsed acrocentrics of long-chain configurations with XY bivalents that is critical (see also data for lemurs [69, 70]). An association of the heteromorphic configuration and the XY bivalent is completely absent in males with long ring configurations (again supported by lemurs). Thus, the disruption of gene silencing of the XY in many pachytenes, through the interaction with the long-chain configuration, could explain the greater germ cell death in males with long chains than males with long rings, and explain the greater germ cell death in males with long chains than females with long chains.

Different metacentrics in the same species may show different anaphase I nondisjunction frequencies when present as CIII heterozygotes. This has been particularly well demonstrated in the house mouse [39]. Such variation between metacentrics is not particularly surprising and could have a variety of different explanations, although more work is needed to fully understand the causes. Winking et al. [28] carried out an elegant study in the house mouse, constructing simple Rb heterozygotes for two different metacentrics and demonstrating very different anaphase I nondisjunction frequencies for the two trivalents. This demonstrates that two distinct trivalent configurations in the same cell may interact with the spindle in very different ways, leading to differences in accuracy of segregation. This difference in interaction with the spindle potentially includes differences between the trivalents in whether they always form the necessary crossovers – if chromosome arms in a trivalent do not form a crossover the acrocentric will segregate independently from the metacentric. Interestingly, it has been demonstrated that the different chromosome arms in a trivalent may have a different incidence of anaphase I nondisjunction [27].

For CIII heterozygotes in house mice, there is generally lower anaphase I nondisjunction in individuals collected from natural populations than when wild-derived Rb metacentrics have been introduced into a laboratory mouse genetic background [24, 39]. Incompatibilities between the wild-derived Rb metacentric and the laboratory mouse genome into which it is introduced could promote meiotic perturbation (nondisjunction) through disturbed kinetochore-spindle interactions or inadequate chiasmata-formation, as suggested by the autonomy of different metacentrics in the same cell shown by Winking et al. [28]. Additionally, or alternatively, it can be proposed that there is an impact of hybridity at the cellular level, with a relationship between germ cell death and anaphase I nondisjunction [121]. Increased germ cell death could occur through direct physiological effects associated with hybrid incompatibilities, and nondisjunction of the trivalent could be a sublethal response at the cellular level [121]. This could reflect, for instance, perturbation of the tightly coordinated meiotic progression [122, 123].

Mole voles provide an interesting case study relating to hybridity and germ cell death. Male mole voles that are 10 CIII heterozygotes may show relatively unperturbed spermatogenesis when they are the product of a cross within a species, but show spermatogenic arrest when they are the product of a between species cross (Table 1 [62]). This spermatogenic arrest could be a direct physiological response of interspecies hybridity, but there is also greater asynapsis in the interspecific hybrids than the intraspecific hybrid. Therefore, the interspecies hybridity could, to some extent, be inducing germ cell death by perturbing the meiotic pairing behavior of the chromosomes.

The interplay of chromosomal properties and hybrid genomes is also evident in the common shrew. In chromosomal hybrid zones within the species, microsatellite data show little interruption to gene flow, while equivalent hybrid zones (in terms of the types of Rb heterozygote present) between species (common shrew and Valais shrew) show considerable interruption to gene flow [124]. That interruption to gene flow likely reflects hybrid unfitness through infertility. Interestingly, one of the two hybrid zones between the common shrew and Valais shrew has a more extreme chromosomal difference than the other and shows the greater interruption to gene flow [124], indicating that the genic perturbation may be operating (at least partially) through perturbed behavior of the chromosomes, in a similar way as suggested for wild/laboratory house mouse hybrids.

In both pigs and cattle, CIII heterozygotes suffer sufficient infertility to cause a financial cost to the farming industry [79, 89]. This impact of Rb metacentrics has led to interest in the mutational process that underlies their origin and the development of simple ways to type individuals, e.g., for the 1;29 Rb metacentric in cattle [125‒127]. It is unknown why the infertility of these domesticates should be higher than, say, the common shrew. However, the population genetics of livestock is very different from a wild-living small mammal. Where they occur in common shrews, CIII heterozygotes are exposed to intense natural selection over many generations, which may favor a reduction in both nondisjunction and germ cell death (this can also apply to other forms of Rb heterozygote occurring in the species, see below). Rb metacentrics in pigs and cattle are not exposed to these sorts of selection pressures. When Rb fusions arise sporadically in humans, they likewise are not exposed to “shrew-like” selection pressures. Nondisjunction frequencies in CIII heterozygotes are much higher in both male and female humans compared to common shrews (Table 1). This higher nondisjunction results in levels of aneuploidy that causes considerable medical concern [71]. Interestingly, although some studies of 1–3 CIII heterozygotes in wild house mice have demonstrated shrew-like nondisjunction frequencies (<5% in males), others (including those with large sample sizes) have shown much higher frequencies (>10% in males)(Table 1). This may be a consequence of the “unnaturalness” of the population genetics of the house mouse. Population dynamics and dispersal in the house mouse reflects their close association with humans, with great fluctuations in population size and occurrence of both very short and very long dispersal distances [128, 129]. Thus, mouse Rb heterozygotes found in nature may sometimes be produced on contact of populations that are genetically rather distinctive because one population has dispersed from a distant location, or because of independent population dynamics in the two populations. Genetic distinctiveness of even nearby mouse populations has been documented, including between populations distinguished by Rb metacentrics (e.g., [130]). Such genetic distinctiveness could contribute to higher infertility in mouse Rb heterozygotes in a comparable way to hybridity of laboratory and wild mice. Rb heterozygotes generated in the laboratory may also represent crosses between natural populations with quite different population histories. Additionally, it is worth noting that some house mouse populations show intense inbreeding [131], which may also impact fertility of Rb heterozygotes, as seen in mole voles [59].

Finally, it should be emphasized that there are various entries in Table 1 where Rb heterozygotes have been generated by crosses between species or between major genetic forms within species. Infertility in the Rb heterozygotes in these cases will be the result of some combination of the effects of chromosomal heterozygosity and the effects of genic hybridity. This includes the entries for the black rat, three species interspecific hybrids of lemurs, crosses between subspecies of pig and the following interspecific crosses: long-haired × dusky rat, Mongolian × Eversmann’s hamsters, Zaisan × northern mole vole, common brown × gray-headed lemur, common brown × black lemur, gray-headed × black lemur, crowned × black lemur, gray-headed × collared brown lemur, sheep × argali.

The substantial datasets for house mouse and common shrew allow a comparison of fertility of equivalent types of Rb heterozygote (Table 1). We have already discussed 1–3 CIII heterozygotes, where sometimes house mice and common shrews show a similar low fertility cost, and other times house mice have a higher level of infertility. Considering other types of Rb heterozygote, all the indications are that common shrews show higher fertility than house mice. In particular, no heterozygous category of male shrew has been found to be typically sterile (even those with the most extreme meiotic chains: CIX, CX, CXI, CVII + CIV, CVI + CV). In contrast, all complex heterozygous male mice with long chains have been found to have total meiotic arrest and complete sterility (including CVII + CVIII heterozygotes, CXI heterozygotes and heterozygotes with longer configurations). Gropp et al. [38] go further than this to state: “complete sterility … is constantly observed in all heterozygous males with longer chains [than CV],” though this may refer to mice with a partially laboratory mouse genetic background for chains of a moderate length. This difference in tendency for sterility between mice and shrews may relate, in part, to the opportunities for natural selection discussed above. As can be seen from Table 1, the house mouse complex heterozygotes with long chains that have been studied were generated by crosses in the laboratory, often between individuals from populations that are geographically distant, while the common shrew complex heterozygotes analyzed were usually collected from nature. Natural selection has apparently responded to the infertility of short-chain complex heterozygotes in the common shrew by selection for acrocentrics in multiple hybrid zones [108, 117]. However, if F₁ hybrid shrews are complex heterozygotes with long meiotic chains, there is less indication of selection for acrocentrics, and grounds for why this is the case [108]. So, under these circumstances, long-chain complex heterozygotes may be generated in substantial numbers over many generations in any particular hybrid zone and, therefore, selection may be expected to reduce any infertility in these heterozygotes (e.g., reducing the extent of germ cell death). This exposure to natural selection may explain the higher fertility of complex heterozygotes forming meiotic long chains studied in the common shrew relative to those studied in the house mouse (which have not be exposed to natural selection).

However, when comparing the three best studied mammalian systems with regards to fertility in Rb heterozygotes (common shrew, house mouse, human), it may not just be population traits or level of recent exposure to natural selection that are important to explain the stark differences in Rb heterozygote fertility between species. There may be species-specific aspects of the chromosomes, kinetochore, spindle, chiasma-formation and progression of the meiotic process that favor/disfavor regular behavior of the heteromorphic configurations in Rb heterozygotes. Even for normal karyotypes, humans are particularly prone to anaphase I nondisjunction, particularly females [120], and inadequacy in crossover formation may have an important role [132]. The error-prone meiosis in humans could reflect relaxed selection, but it could also relate to other factors. Given the tendency for anaphase I nondisjunction in humans of normal karyotype, infertility in Rb heterozygotes is perhaps not a surprise. It is also worth noting from Table 1 that percentages of germ cell death determined from spermatocyte:spermatid ratios tend to be higher for homozygous mice than homozygous shrews, hinting at differences in the baseline level of that trait between species, as well as the differences in frequency of anaphase I nondisjunction already discussed.

In terms of the relatively high fertility in the common shrew, one interesting feature of the species is that males standardly have a meiotic trivalent configuration for the sex chromosomes (XY₁Y₂) [7], and therefore have to function in the constant presence of a multivalent. The X in the common shrew is a compound chromosome composed of the original X and an autosome; the mutation creating this chromosome occurred well before the origin of the common shrew [133]. Thus, selection will have had a long period to act on the meiotic behavior of this trivalent; it is found to be very orderly in present-day common shrews [7]. Having this trivalent already part of the normal meiotic condition in the common shrew does this mean that meiocytes in shrews are somehow less prone to catastrophic effects of multivalents on chromosomal pairing and segregation?

There are other factors that can explain variation in Rb heterozygote fertility among the entries in Table 1. Clearly use of different methods, or the same methods by different researchers, is going to contribute to such variation. For instance, using different methods, the level of infertility associated with Rb heterozygotes in humans is very different according to whether it is estimated by study of gametes or study of second trimester fetuses (Table 1). The arctic fox provides an example where approximately the same methods are used by different researchers, but different assessments on the fertility of Rb heterozygotes are reached (Table 1). Variation in sample sizes between studies may be part of the explanation for this. The choice of individuals used for fertility studies also has an impact. For instance, Vasco et al. [32] showed that the level of asynapsis and germ cell death decreased with age in 8 CIII heterozygotes; showing that age of animal can make a difference to results obtained.

There are some examples where researchers have studied individuals heterozygous for both Rb rearrangements and another type of chromosomal rearrangement, allowing a comparison between the two. Galindo et al. [90] studied male hybrids among chromosomal forms of the red brocket deer. These hybrids were both Rb heterozygotes and tandem fusion heterozygotes, and spermFISH analysis was used to compare the frequencies of anaphase I nondisjunction associated with each type of heteromorphic configuration. For Rb heterozygosity it was about 2% and for tandem fusion heterozygosity it was about 30%. For chromosomal rearrangements that arise sporadically in livestock, reciprocal translocations are associated with greater infertility than Rb rearrangements [79, 134, 135]. A single trivalent configuration in simple Rb heterozygotes is more likely to show regular behavior than the heteromorphic configurations in heterozygotes for tandem fusions and reciprocal translocations [93]. Compared with these other chromosomal rearrangements, Rb rearrangements are mild in their impacts, and may rather easily get fixed; hence, it is not surprising how often they contribute to the difference in karyotype among closely related species. Inversions, likely the most common structural variants in mammals, may have even less fertility cost (e.g., [79]).

The title of this section is a straw man. It is obvious that infertility is not the only thing worthy of study in considering origin, evolution and significance of Rb variation. Over the years there has been a particular interest in infertility because it is an intriguing feature that is relatively easy to measure (though not without its difficulties). This section is a reminder about the importance of other factors in explaining Rb variation.

In considering fixation of Rb fusions and fissions, there has been an interest in the fertility of CIII heterozygotes because high infertility would make fixation difficult. The finding that CIII heterozygotes in many species have near normal fertility lessens that issue. But fixation is more complicated than just measuring infertility of heterozygotes. It also depends on mutation rate, which has rarely been estimated (see [136]). In terms of the evolutionary forces bringing about the fixation, genetic drift can be viewed as the default and was assumed in many of the early theoretical models (e.g., [137]). More recently, arguments have been made for selective advantage of Rb rearrangements [138] and meiotic drive in their favor [139, 140], the latter supported by data on the house mouse.

In considering the role of Rb rearrangements in reproductive isolation and therefore potentially making a causal link between karyotypic difference and species diversification, the extent of infertility in Rb heterozygotes is not the only factor of importance to consider. There has long been known to be recombination suppression around the chromosomal breakpoints (the centromere) in CIII heterozygotes [141]. This is of importance because it can allow any genic differences between two chromosomal forms to be maintained between those forms on hybridization. Recombination suppression can also allow a build-up of reproductive isolation between the two forms despite interbreeding [142]. A reduction in gene flow around the centromere of Rb metacentrics has been demonstrated in hybrid zones of Rb races of the house mouse, but it is difficult to distinguish between recombination suppression and hybrid infertility as the cause of this reduction – both would cause the same signal of reduced gene flow around the centromere, the mutation site [143, 144].

Different species of mammal very often differ by chromosome number, including closely related species, and that difference is primarily caused by Rb fusions or fissions [145]. Therefore, it may be suggested that somehow Rb rearrangements can promote speciation [146]. Closely related species of mammal can differ by a single or few Rb rearrangements, or by many. As an example of a small difference in karyotype, the ocelot and related species (2n = 36) differ from most cats (2n = 38) by a single Rb fusion [147]. Where species differ by one or a few Rb rearrangements, the data in Table 1 indicate that the unfitness due to their structural difference alone is unlikely to be sufficient to cause reproductive isolation. That does not mean that Rb rearrangements are uninvolved in speciation – it just means that infertility due purely to chromosomal behavior is not sufficient to cause reproductive isolation. One or a few Rb rearrangements could help promote speciation if genic differences between the diverging populations act in concert with the chromosomal difference enhancing the aberrant behavior of the heteromorphic configuration, thereby causing greater infertility, as has been demonstrated in the house mouse [39]. Or genic incompatibilities could add to unfitness in other ways. One or a few Rb rearrangements could also help promote speciation if recombination suppression around the centromeres allow the chromosomes to remain genetically divergent and build-up further genetic differences during contact between the chromosomal forms (i.e., Rb races) [142].

Sometimes sister species differ by multiple Rb rearrangements. Once again, from the fertility data, it is very likely that the hybrids would not be completely sterile based on their structural difference alone. This applies both to simple heterozygotes with many trivalents and to complex heterozygotes even with very long configurations. Earlier in this article we showed that a premise of the monobrachial speciation model [113] – that CIV heterozygotes are sterile – does not hold. Even in the house mouse, the archetypal species for the monobrachial speciation model, complex heterozygotes with very long meiotic chains (CXV) are not completely sterile (females can produce some offspring) [42]. However, F₁ hybrids do not have to be completely sterile for the chromosomes to be the driving force for speciation. If they are sufficiently infertile, they could still promote speciation by reinforcement. This has been suggested in house mice in Italy for complex heterozygotes with a relatively short meiotic CV chromosomes [119]. It is interesting that the same type of heterozygote may also promote a different outcome, despeciation [108]. Thus, in a common shrew chromosomal hybrid zone between Rb races in Britain [117], it appears that the infertility of CV heterozygotes as F₁ hybrids creates a selection pressure in favor of acrocentric chromosomes, generating an acrocentric peak in the hybrid zone, thereby reducing the chance to produce the unfit CV heterozygotes. Therefore, in this case, selection is creating a situation where speciation is less rather than more likely.

Not only are the selective responses to infertility of complex heterozygotes of relevance to speciation and despeciation, such selective responses can also explain instances of reticulate evolution among Rb races in the common shrew and house mouse, relating also to the infertility shown by multiple simple heterozygotes [148].

Returning to sister species that differ by multiple Rb rearrangements, sometimes these do differ by a substantial number of Rb rearrangements. Although proof is still lacking, these are candidates for reproductive isolation promoted by infertility of the F₁ hybrids due to Rb heterozygosity. As an example, the European and American beavers differ by metacentrics with monobrachial homology such that an F₁ hybrid between them would be a CVII + CIX + 2CIII heterozygote [149]. Such a hybrid would likely have very low fertility on chromosomal grounds.

Compared with simple heterozygotes for one or a few trivalents, the greater infertility of either simple heterozygotes with many trivalents or complex heterozygotes may mean that F₁ hybrids with those characteristics in synergy with genic incompatibilities may be more likely to promote reproductive isolation between hybridizing Rb races. This is easiest to envisage for situations where the F₁ hybrids are complex heterozygotes with a single long configuration. Under these circumstances, there is just a single type of hybrid with low fertility that could be enhanced by genic incompatibilities of the hybridizing Rb races, leading to complete reproductive isolation, or speciation by reinforcement may be feasible even without the synergy of genic incompatibilities. Reproductive isolation could be attained while the two Rb races are in contact with each other and hybridizing. When F₁s have multiple heteromorphic configurations – multiple trivalents or multiple longer configurations or a mixture of the two – then the chromosomal clines in the hybrid zone may become staggered on contact of the Rb races, meaning that the F₁ karyotype is rarely or never observed and the hybrid karyotypes would show less extreme infertility on chromosomal grounds [148]. These differences in impact of different types of F₁ hybrids relate to what happens after contact of the different Rb races. The accumulation of genic differences between different Rb races in allopatry, and their synergy with the chromosomal differences, may together create tighter, unstaggered hybrid zones if the two forms do come into contact [150]. This is apparently the case on contact of the common and the Valais shrew [124].

How to view the contribution of infertility of Rb heterozygotes to speciation in mammals? It appears that only in the case of Rb races producing complex heterozygotes might there be sufficient hybrid unfitness for chromosomal difference alone to promote reproductive isolation, e.g., through reinforcement. However, more generally, the contribution of the infertility of Rb heterozygotes to reproductive isolation goes hand-in-hand with other genetic markers that show underdominance or otherwise reduce gene flow between two diverging Rb races within a species. What the work on infertility in Rb heterozygotes has done is to alert researchers to possible ways that Rb rearrangements may contribute to the speciation process. Studies with genomics will increasingly help elucidate this contribution [92]. This is illustrated by the work of Jónsson et al. [151] based on the entire genomes of all contemporary equids, which showed past gene flow between species despite massive difference in karyotype attributable to Rb rearrangements (2n = 32–64). This suggests that infertility due to karyotypic difference has not in itself been sufficient to stop successful interbreeding between these taxa, further emphasizing that infertility of Rb heterozygotes is one potential step in the speciation process, but not the whole story.

Figure 6 provides a pictorial summary of the main trends in fertility of Rb heterozygotes, as reviewed here.

Our review builds on the work of many researchers and we sincerely acknowledge that contribution. We apologize for any oversight in representing that work. Oxana Kolomiets is one of those whose work touched on fertility costs in Rb heterozygotes, and we particularly pay tribute to her here. One of us (J.B.S.) had the honor to co-author two articles with Oxana and treasures his interactions with her. We also wish to salute the journal Cytogenetics/Cytogenetics and Cell Genetics/Cytogenetic and Genome Research for publishing so many of the articles that have reported on the fitness of Rb heterozygotes in mammals.

J.B.S. is on the editorial board of Cytogenetics and Genome Research. The authors have no other conflicts of interest to declare.

This work did not require funding.

J.B.S. contributed to the study conception and design and wrote the original draft. Visualization was performed by J.J.H. Writing – review and editing – was performed by J.J.H. and J.B.S. All authors read and approved the final manuscript.