HETEROZYGOSITY AND ITS UNEXPECTED CORRELATIONS WITH HYBRID STERILITY

Authors


Abstract

In general, heterozygosity is considered to be advantageous, primarily because it masks the effects of deleterious recessive alleles. However, there is usually a reduction in fitness in individuals that are heterozygous due to the pairing of two species (heterospecific). Because the parental alleles arose along separate evolutionary paths, they may not function properly when brought together within an individual. The formation of these unfit interspecies hybrids is one of the mechanisms that maintains species isolation. Interestingly, it has been observed that later-generation individuals resulting from a backcross to one parent are more often sterile than those resulting from a backcross to the other parent, but the mechanism underlying this trend is unknown. Here, I show that one direction of backcross produces offspring with more heterospecific genome, and that this is correlated with the directionality seen in backcross hybrid sterility. Therefore, the directionality in sterility is likely due to the different amounts of heterospecific genome present in the two backcrosses. Surprisingly, in spite of the potential fitness consequences, I also find that interspecies laboratory backcrosses in general yield an excess of heterospecific individuals, and that this trend is consistent across multiple taxa.

When an individual is homozygous at multiple loci, fitness is often reduced (Falconer and Mackay 1996; Charlesworth and Charlesworth 1999). High levels of homozygosity typically result from inbreeding, which produces homozygotes for deleterious recessive alleles. A homozygous individual will also have reduced genetic variation in subsequent offspring, and therefore a reduced ability to respond successfully to selection. Numerous studies have found advantages to being heterozygous (Ferreira and Amos 2006; Gemmell and Slate 2006; Hoffman et al. 2007; Rubenstein 2007), and there is evidence that females of some species will show preference in mate choice for novel males (Hughes et al. 1999; Masters et al. 2003), which are more likely to result in heterozygous offspring.

In contrast to the benefits of heterozygosity within a species, there are usually deleterious effects of being heterozygous when the two alleles come from two different species (but see Pfennig 2007). Indeed, one of the primary mechanisms that isolates species from one another is the reduced fitness of interspecies hybrids. For example, if the two parents are genetically adapted to different environmental niches, and if the alleles for this adaptation act additively, interspecies heterozygosity (heterospecificity) may create a midparent phenotype not suited to the ecological niche of either parent species (Rice and Hostert 1993; Schluter 1998). In a similar fashion, intermediate behavioral phenotypes or mating cues can be maladaptive or impair the ability of a hybrid to find a mate (Mavárez et al. 2006).

There may also be reduced interspecies hybrid fitness through dysfunction due to hybrid sterility or hybrid inviability. This hybrid dysfunction is thought to arise when alleles that interact epistatically diverge along separate paths in the two species. When these divergent alleles are brought together in the heterospecific hybrid they are no longer compatible and cause some level of reduced fitness. These deleterious epistatic interactions, known as “Dobzhansky–Muller incompatibilities” (Dobzhansky 1936; Muller 1940), are therefore thought to result from alleles that function normally within their respective species, but interact negatively when brought together in a hybrid.

The presence of hybrid sterility or inviability is more often seen in the heterogametic (e.g., XY or ZW) sex, a trend known as “Haldane's Rule” (Haldane 1922). The genes that cause hybrid dysfunction must act normally within the genetic background of their “own” species, because pure-species individuals are fertile and viable; only when the genes from the two species are placed together does dysfunction arise. Because this dysfunction more strongly affects the heterogametic sex, it is theorized to be caused by an interaction between a recessive factor on the X (or Z) chromosome and a dominant factor on an autosome (Muller 1940; Orr and Turelli 1996). These interactions would be more likely to affect the heterogametic sex because there is only one allele of an X-linked (or Z-linked) locus in a heterogametic individual: recessive X-linked loci can be “exposed” in a heterogametic hybrid but are “masked” in a homogametic hybrid. Although the dominance theory is the leading hypothesis used to explain hybrid dysfunction, very few interacting pairs of genes have been identified (Brideau et al. 2006; Lee et al. 2008). The lack of identification of autosomal loci has primarily been attributed to the notion that there are multiple autosomal loci that likely have complex epistatic interactions (Tao et al. 2003; Coyne and Orr 2004), and that the degree of dominance may be modified by these epistatic interactions (Chang and Noor 2010).

Many studies have used quantitative trait locus (QTL) mapping to attempt to locate regions of the genome that contribute to interspecies hybrid sterility, as well as to map other traits that vary between two species. In QTL mapping, the phenotype of recombinant individuals is measured and compared to their genotype. To circumvent the sterility barrier to creating recombinant individuals, the fertile F1 progeny (e.g., females) can be crossed back to pure-species individuals to create a “backcross” population for mapping (e.g., “Backcross B” in Fig. 1; Coyne and Orr 1989, 1997).

Figure 1.

Creation of backcross hybrids. Species B (gray) and C (white) are crossed to produce a fertile F1 individual (usually female). These females can be crossed to males from either parental species to produce backcross offspring. One chromosome homolog comes from the parental species and the other is a mix of the two parental genomes due to recombination in the F1. Homologous chromosomes are represented as paired bars; only one set of autosomes is shown. The Y chromosome is represented as a bent bar, and the chromosome it is paired with is the X.

In addition to the recessive-dominant interactions found in F1 individuals, additional types of interactions are possible in these backcross offspring due to the presence of recombinant chromosomes (Figs. 1 and 2A, B). The relatively random nature of recombination means that each individual can have a different number and combination of interacting factors, which is thought to give rise to the observed range of sterility from one backcross individual to the next. Because both parental genomes are represented on the recombinant X chromosome, both directions of recessive-dominant interaction are now possible, for example species C-X interacting with B-autosome and the reciprocal B-X interacting with C-autosome (Fig. 2B). Because there are regions of the autosomes that are homospecific in a backcross individual, deleterious recessive–recessive interactions are also possible. Although overall there are a greater number of possible genetic interactions in backcross individuals, it should be noted that there are fewer possible recessive–dominant interactions of the original type found in the F1 generation because many X chromosome loci in the backcross offspring are now homospecific (compare Fig. 2A,B).

Figure 2.

Potential genetic interactions in backcross individuals containing different amounts of heterospecificity. For simplicity, only interactions present in backcross to Species B are shown (see Fig. 1). Homologous chromosomes are represented as paired bars; only one set of autosomes (A) is shown. The Y chromosome is represented as a bent bar, and the chromosome it is paired with is the X. Gray bars represent genome coming from Species B (B); white bars represent genome from Species C (C). The description of the number of interactors compared to the hypothetical illustrative average (B) are described on the right.

Backcross hybrids are often mapped in both directions of cross, as seen in the data presented in Table 1. The offspring resulting from the backcross to Species B (Backcross B) can be compared to those from backcross to Species C (Backcross C; Fig. 1). Interestingly, the level of hybrid sterility is usually not equal in these two directions of backcross: Backcross B, for example, consistently produces fewer fertile offspring than those produced from Backcross C. The cause of this directionality is not known, but closer examination of this anomaly may provide insight into the genetic mechanism of hybrid sterility.

Table 1.  Comparison of backcrosses and their relative sterility.
GenusSpecies B1Species CF1 cross (female×male)Second Cross2Reference3
  1. 1The backcross to (B) produces more sterile offspring than the backcross to (C).

  2. 2Individuals are from a backcross (BC) to the listed parent, except the Iris dataset where F2 data were sorted by similarity to one parent or the other to create simulated BC groups. The studies (1) and (2) on Anopheles were on female sterility; all others refer to male sterility. The BC direction that produces more sterile offspring is marked by a superscript “MS” for “more sterile.”

  3. 3If multiple citations, those with an “*” provided sterility data and those without provided genotype data.

DrosophilaD. mauritianaD. simulanssim×mauBC mauritiana MS(Zeng et al. 2000)
BC simulans 
 D. sechelliaD. simulanssim×sechBC sechellia 1 MS(Orgogozo et al. 2006)
  BC simulans 1   
    BC sechellia 2 MS(Macdonald and Goldstein 1999)
    BC simulans 2 
D. santomeaD. yakubayak×santBC santomea MS(Moehring et al. 2006a*, b)
BC yakuba
 D. pseudoobscuraD. persimilispseu×perBC pseudoobscura MS
BC persimilis
(Noor et al. 2001)
AnophelesA. arabiensisA. gambiaegam×arabBC arabiensis 1 MS(Slotman et al. 2005)
BC gambiae 1
BC arabiensis 2 MS(Slotman et al. 2004)
BC gambiae 2
 A. gambiaeA. arabiensisgam×arabBC gambiae 3 MS(Slotman et al. 2004)
    BC arabiensis 3 
OryzaO. sativa japonicaO. sativa indicaind×japonF2 japonica MS(Yan et al. 2003; Uga et al. 2007)
F2 indica
IrisI. brevicaulisI. fulvabrev×fulvaBC brevicaulis MS(M. Arnold, pers. comm.;
    BC fulvaBouck et al. 2005)

Here, I investigate the potential role of different types of genetic interactions in hybrid sterility by examining interspecies backcross genotype data compiled from pre-existing studies on a variety of species. I first sought to identify the prevalent type of genomic interactors contributing to hybrid sterility by comparing the amount of heterospecificity on the X chromosome versus autosomes in backcross individuals and its correlation with sterility. The expectation is that individuals with greater heterospecificity on the X and less heterospecificity on the autosomes (Fig. 2C) will have the greatest observed sterility because this combination provides the greatest number of recessive–dominant and recessive–recessive interactions. The amount of genome available for possible recessive–recessive interactions should drive this trend more than the possibility of recessive–dominant interactions because most loci affecting sterility appear to act recessively (Masly and Presgraves 2007). If the sterility of backcross individuals is instead primarily affected by the same recessive–dominant interactions governing the sterility of F1 individuals, then I would expect a large effect of the heterospecificity of the X chromosome but no effect of the autosomes in at least one backcross direction (because one homolog is entirely of one species and contains all of the dominant interactors; Fig. 2D).

I then examined the types of genetic interactions driving the differential sterility of the two directions of backcross. These interactions cause one direction of backcross to suffer greater sterility than the other direction of backcross. If this trend is driven by a difference in strength and/or quantity of interactors present in one species over the other, then I would expect to see opposite correlations in the two directions of backcross. For example, if Species B (Fig. 1) has more or stronger interactors, I would see the greatest levels of sterility whenever there is more Species B DNA present in the genome. In Backcross B individuals this occurs when there is high heterospecificity on the X and low heterospecificity on the autosomes (Fig. 2C); in Backcross C individuals this occurs when there is low heterospecificity on X and high heterospecificity on the autosomes.

Interestingly, the data did not support any of the scenarios outlined above. Indeed, the only consistent trend was that the greater the level of heterospecificity, the greater the level of sterility, regardless of the marker location (although there is a stronger effect of the X chromosome) or the direction of backcross. This correlation can be extended to the observed directionality in backcross hybrid sterility: individuals produced from one direction of backcross had a greater level of heterospecificity, and that direction also demonstrated the greater level of hybrid sterility. Thus the directionality in backcross hybrids can potentially be explained by a differential amount of heterospecificity. Surprisingly, an additional trend was found in backcross hybrids: an excessive amount of overall heterospecificity above expected, observed across multiple taxa.

Methods

Data were compiled from existing literature (see Table 1). The criteria for inclusion in this study were: (1) crosses had to be between species; (2) crosses to both parental species producing backcross individuals, or an F2 population, had to be genotyped at multiple markers throughout the genome; and (3) the relative sterility of the two backcross directions had to be known. Every available study that met these criteria was included in this analysis. Ten datasets met these criteria: five for Drosophila (fruitflies), three for Anopheles (mosquitoes), one for Oryza (rice), and one for Iris (iris). Two of the Drosophila datasets are for the same species pair (D. simulans×D. sechellia), and the three Anopheles datasets are for the same species pair (A. gambiae and A. arabiensis).

Marker genotype data were used to determine the amount of heterospecificity within an individual's genome. Although the type of molecular marker varied (e.g., microsatellite, insertion/deletion, etc.), all of the studies presented here only genotyped a molecular marker if it had species-specific alleles. In recombinant offspring, therefore, an individual that is homospecific at a particular molecular marker would show a homozygous genotype whereas an individual that is heterospecific at that marker would show a heterozygous genotype.

Results

HIGH X CHROMOSOME HETEROSPECIFICITY IS CORRELATED WITH STERILITY

The greatest number of negative interspecies genetic interactions, and hence the highest degree of sterility, should be found in individuals with a high level of heterospecificity on the X chromosome and a low level of heterospecificity on the autosomes (Fig. 2C). Due to the way sterility is measured, we should observe a negative correlation between the amount of heterospecificity on the X (higher values = more heterospecificity) and the degree of sterility (lower fertility values = more sterile), and a positive correlation between the amount of heterospecificity on the autosomes and sterility.

According to the data, individuals that are sterile have a significantly greater percentage of heterospecific X chromosome than individuals that are fertile, and the correlation between heterospecificity and sterility is strongly negative, as expected (Table 2). In other words, the greater proportion of heterospecific genome present on the X, the more likely an individual will be sterile. However, the opposite trend is not consistently observed for the autosomes. Only half of the datasets showed a significantly lower percentage of autosome heterospecificity in sterile individuals compared to fertile individuals, and the correlation between heterospecificity and sterility was inconsistent and weak (Table 2 and Fig. 3). Thus, although the X chromosome strongly follows the pattern we would expect to induce the greatest levels of sterility, the autosomes do not consistently demonstrate the expected correlation (data fit example Fig. 2D), at least for the datasets available for this comparison, which were all in Drosophila.

Table 2.  Heterospecificity and its correlation with sterility scores.
BackcrossN1Individual sterility scoresSignificant markers
X chromosomeAutosomesX chromosomeAutosomes
% Het2 (St:NSt) Correlation3% Het2 (St:NSt) Correlation3% Het4 (St:NSt)% Het4 (St:NSt)
  1. 1N= total number of individuals scored. The amount of heterospecificity of each individual was calculated based on genotype data, and subsequently compared to the sterility score for that individual.

  2. 2Percent heterospecificity for N individuals that are scored as sterile compared to not sterile (St:NSt). P-values from a t-test denoting significant differences between the amount of heterospecificity in sterile versus nonsterile individuals are listed as*P<0.05,**P<0.005, and***P<0.0001. Only genotypes for which there is a sterility score are used; all datasets are Drosophila species.

  3. 3Pearson's rank correlation coefficient between individual heterospecificity and sterility scores. A negative correlation indicates that a higher level of heterospecificity is associated with a decrease in fertility (=greater sterility). P-values denoting significant correlations are listed as*P<0.05,**P<0.005, and***P<0.0001.

  4. 4Percent heterospecificity for markers that were found to be significantly associated with sterility compared to those that were not (St:NSt) via QTL mapping; “none”=no markers were in this dataset. Only genotypes for which there is a fertility score are used. None of the values were significantly different by a Wilcoxon rank-sum test.

  5. 5 Only a single marker was in this dataset.

BC sech 2 20060.1: 34.8***−0.30***54.2: 60.8* 0.15*56.55:54.6 57.3:55.5
BC sim 2 20060.5: 13.6***−0.70***58.2: 54.0−0.0939.8:38.3 none:56.3
BC sant 55070.7: 31.0***−0.46***48.7: 54.9** 0.0351.6:49.4 none:52.5
BC yak 55084.0: 34.2***−0.73***47.1: 56.9*** 0.16**41.5:40.1554.0:53.1
BC pseu114755.2: 28.3***−0.34***54.3: 47.4***−0.17***44.8:none51.4:51.7
BC per104448.7: 26.1***−0.44***51.1: 49.1−0.07*41.2:none56.15:49.8
Figure 3.

Difference in % heterospecificity in sterile versus fertile samples. The difference in percent heterospecificity for individuals scored as sterile versus fertile for their X chromosome (solid black squares) and autosomes (solid gray triangles). The difference in percent heterospecificity for markers determined by QTL mapping to be associated with sterility versus not associated with sterility for the X chromosome (open black squares) and autosomes (open gray triangles). The horizontal dashed line is at zero, where steriles have just as much heterospecificity as fertiles; larger values indicate that steriles have a higher percent heterospecificity than fertiles. Data of this nature were only available for the Drosophila species datasets, and are presented in the same order as in Table 3.

It is possible that the correlation with hybrid sterility may only be present for those markers that are found through QTL mapping to have a significant association with hybrid sterility compared to those that are not, thus explaining the unexpected lack of an autosomal correlation when comparing individual sterility scores. In this case, one would expect the effect to be significant when comparing markers affiliated with sterility via QTL mapping to those that are not if the genotype origin is important. Interestingly, however, this trend is not statistically significant when comparing markers that were found to be correlated with sterility to those that were not (Table 2 and Fig. 3). Although the individual marker effects were not statistically significant, there was a significant overall effect among the entire dataset for markers affiliated with sterility to have a greater amount of heterospecificity than those that were not associated with sterility, regardless of chromosome origin (Wilcoxon signed-rank test; W- = 1, P<0.02). In other words, the X chromosome and autosomal markers that had a significant association with hybrid sterility were more likely to be heterospecific than nonsignificant markers. This indicates that autosomal recessive interactors, which can only have an effect when homospecific, are not likely the key players in backcross hybrid sterility, at least for those markers that achieved statistical significance through QTL mapping. Although these results are dependent upon studies in Drosophila, the conclusions may apply more broadly to other taxa.

INCREASED HETEROSPECIFICITY IS CORRELATED WITH THE DIRECTIONALITY OF HYBRID STERILITY

When F1s are crossed back to the two different parental species, there is a difference in the amount of sterile offspring produced from the two directions of backcross. If this effect is due to differences in the strength of interactors present in the two parental species, I would expect to observe opposite effects of heterospecificity on sterility: heterospecifics would be more likely to be sterile in one direction of backcross, and homospecifics would be more likely to be sterile in the other direction of backcross. This trend was not observed for any of the datasets as heterospecifics were more likely to be sterile for both directions of backcross (Table 2).

It is possible that an aspect of heterospecificity itself, rather than particular trends of genetic interactors, drives the degree of sterility. As demonstrated in the previous section, the level of heterospecificity is highly correlated with an individual's sterility score, with individuals that are more heterospecific (particularly on the X chromosome; Fig. 3) having a greater level of sterility. In the case of the differential sterility in backcrosses, if a greater amount of heterospecificity is present in the more sterile direction of backcross compared to the less-sterile direction, the degree of heterospecificity could also explain the directionality of backcross hybrid sterility. This is indeed what is observed (Fig. 4 and Table 3). The backcross direction producing more sterile offspring has a significantly greater amount of heterospecificity than the direction that produces less-sterile offspring in all but one dataset (Wilcoxon rank sign test; W- = 2; P<0.01). The one dataset that does not follow this trend (sant-yak) may be due to the way in which it was scored: although the average sterility score is indeed lower for D. santomea (indicating less fertility), only backcrosses to D. yakuba create the most severe phenotype of complete sterility, which is the absolute lowest score. Therefore, if this dataset was scored on the presence or absence of complete steriles, then it too would fit the trend. Although the directionality of sterility and heterospecificity might have separate explanations, it is likely that they are connected. Thus it seems that the directionality in the degree of backcross sterility can be explained by a directionality in the amount of heterospecificity.

Figure 4.

Directional increase in heterospecificity. The difference between percentage heterospecificity in the backcross producing more sterile offspring (BC to Species B, listed on top) and the backcross producing less-sterile offspring (BC to Species C, bottom), sorted from smallest to highest difference in overall heterospecificity (black diamonds). The average percent heterozygosity for the X chromosome is shown as gray squares; for the autosomes is shown as gray triangles. Positive values indicate more heterospecificity in BC-B; negative values indicate more heterospecificity in BC-C. The dashed line is at zero, where BC-B and BC-C have the same amount of heterospecificity. Separate X and autosome data were not available for Iris and Oryza. The difference for gam3/arab3 X chromosome (44%) is outside the visual range of the graph.

Table 3.  Amount of heterospecificity in backcross individuals.
BackcrossN1Percent heterospecificityMarkers >50%:total5
Total2XAuto.Highest3Lowest3TotalXAuto.
  1. 1N is the total number of genotypes. These genotype values were scored as homospecific or heterospecific and then used to determine if the amount of heterospecificity deviates from expected.

  2. 2Marker heterospecificity values that are significantly different from the expected 50% by a chi-squared test are denoted with *P<0.05, **P<0.005, and ***P<0.0001. This is not calculated for the separate X and autosome measures as subdividing the dataset reduced the power to be able to detect significance, and several datasets have only a single marker in one of those categories. N/A=data not available.

  3. 3If markers with the highest or lowest heterospecificity are present on the X chromosome, are designated with a superscript “X”; for japon, ind, brev, and fulva it is not known if the highest or lowest was on the X.

  4. 4The outlier value of BC arab3 was not included in this calculation.

  5. 5The ratio of the number of markers with heterospecificity >50% to the total number assayed. For the averages, the ratio is followed by the% of markers that have >50% heterospecificity. N/A=data not available.

  6. 6As the lowest value for BC arab3 is an outlier, the second lowest value is shown in parenthesis.

  7. Note that this table compares levels of heterospecificity between the more sterile and less-sterile backcross, but does not differentiate between the heterospecificity levels in sterile versus fertile individuals within each of those datasets.

BC mau20,90156.4***50.757.365.746.7X 40:45 5:6 35:39
BC sim20,10656.1***58.455.770.8X49.9 44:45 6:6 38:39
BC sech 1123056.0**64.054.668.0X27.0 12:14 2:2 10:12
BC sim 110,35949.949.449.972.132.3X 13:25 1:2 12:23
BC sech 2760055.4***54.855.666.041.0 35:38 8:8 27:30
BC sim 2760052.6**38.756.366.035.5X 27:38 0:8 27:30
BC sant34,64151.5***46.453.957.944.5X 21:32 0:10 21:22
BC yak34,65151.8***51.452.059.145.7 21:32 8:10 13:22
BC pseu28,72949.943.151.155.439.2X  7:13 0:2  7:11
BC per14,08949.542.351.056.641.1X 12:16 0:2 12:14
BC arab 1113158.2**48.563.064.9N/AN/AN/AN/A
BC gam 184053.651.854.666.7N/AN/AN/AN/A
BC arab 2951361.3***47.362.064.956.1 20:21 0:1 20:20
BC gam 2903057.1***52.057.466.749.8 19:21 1:1 18:20
BC gam 3903057.1***53.757.262.151.2 21:21 1:1 20:20
BC arab 3951356.5*** 9.758.964.19.7X (54.6)6 20:21 0:1 20:20
F2 japon909649.1N/AN/A56.838.8N/AN/AN/A
F2 ind900848.1*N/AN/A56.540.7N/AN/AN/A
BC brev47,38059.0***N/AN/A∼70∼30N/AN/AN/A
BC fulva39,72031.0***N/AN/A∼65∼30N/AN/AN/A
Total Average322,853 51.2***48.7454.663.841.9312:382 (81.7%)32:60 (53.3%)280:322 (87.0%)
More sterile BC168,497 54.9***47.355.163.442.0156:184 (84.8%)16:30 (53.3%)140:154 (90.9%)
Less-sterile BC154,356 47.1***50.1454.064.241.8156:198 (78.8%)16:30 (53.3%)140:168 (83.3%)

Because the original comparison found that sterility was highly correlated with the amount of heterospecificity on the X chromosome, but not the autosomes, I then examined whether the same trend was observed with the directionality in heterospecificity; in other words, whether the directionality in heterospecificity was specific to the X chromosome. Only four of the eight available datasets had higher average heterospecificity for the X chromosome in the backcross direction producing more sterile offspring, and the overall average heterospecificity was lower in the backcross that produces more sterile offspring than in the backcross that produces less-sterile offspring (47.5% vs. 50.1%, respectively; Fig. 4 and Table 3), which is opposite of what is expected. This suggests that the X chromosome is not contributing to the directional trend in heterospecificity, and there may in fact be less heterospecificity on the X chromosome in the more sterile backcross. In contrast, six of the eight datasets had a higher average heterospecificity in the more sterile direction of backcross for the autosomes, and the heterospecificity across the entire dataset was higher in the more sterile backcross (55.1%) than the less-sterile backcross (54.0%), suggesting that the autosomes may be the primary contributors to the greater level of heterospecificity in the more sterile backcross (data fit example Fig. 2E). The comparison of X chromosome and autosomal effects is not entirely consistent, however, and has a bias in sampling because there are far fewer markers sampled on the X compared to the autosomes. If additional data could show that this trend is not due to bias, however, then the sterility directionality may be due to an overall difference in heterospecificity across the entire genome, independent of a particular chromosome.

THERE IS EXCESS HETEROSPECIFICITY IN BACKCROSS DATASETS

Although one direction of backcross has a greater number of heterospecifics than the other, both directions of backcross have an excess of heterospecifics compared to expected (Fig. 5). A backcross individual has an equal chance of being heterospecific or homospecific at any particular locus, and thus one would expect to see approximately half of the genotypes be heterospecific and half of them be homospecific. What is observed, however, is that approximately 80% of the markers that are genotyped have greater than expected heterospecificity (right columns, Table 3). Again, the X chromosome is likely not contributing to this trend as approximately 50% of those genotypes are greater than expected (and half are less than expected), as would be predicted by chance. The trend is instead driven by the autosome genotypes, of which 87% show greater than expected heterospecificity (data fit example Fig. 2E). The overall excess of heterospecifics is not merely driven by linkage, as the pattern of excess was not clustered in groups of genetically linked markers. Again, however, there is the caveat of sampling bias as there are more genotype data for autosomes than for the X chromosome.

Figure 5.

Excess heterospecificity in backcross hybrids. Sorted from smallest to highest difference in total percent heterospecificity (black diamonds). The average percent heterozygosity for the X chromosome is shown as gray squares; for the autosomes is shown as gray triangles. The horizontal dashed line is the expected value of 50% heterozygosity. Separate X and autosome data were not available for Iris and Oryza. The value for the arab3 X chromosome (9.7%) is outside the visual range of the graph.

Discussion

The mixed genome of interspecies hybrids can cause dysfunction through a variety of possible negative genetic interactions. Although any single interaction may potentially cause dysfunction within an individual, when these effects are averaged over multiple individuals we can create a map of the various combinations throughout the genome that influence hybrid sterility. The genomic combination that should create the greatest number of negative interactors in a population of backcross individuals (Fig. 1) would be to have a large number of both recessive–dominant and recessive–recessive interactions (Fig. 2C). In this example, individuals with a large amount of heterospecificity on the X chromosome and a low amount of heterospecificity on the autosomes would be more likely to be sterile. The data presented here do not fit this model.

In interspecies studies that use backcross datasets, three trends have emerged. The first is that the amount of heterospecificity of the X chromosome is highly correlated with an individual's sterility, whereas the autosome heterospecificity is only weakly correlated with sterility, and in conflicting directions (Table 2 and Fig. 3; similar to Fig. 2D). Additionally, there was a strong correlation between the overall genotype of the X chromosome and sterility, but only a weak correlation among markers found to be significant via QTL mapping and sterility. The generally weak and inconsistent correlation between autosomes and sterility, regardless of the method of comparison, suggests that individual loci may have opposing effects on the autosomes or that there is variable and inconsistent strength and prevalence of recessive–recessive and recessive–dominant interactors on the autosomes. Alternatively, the autosomes may simply not be major contributors to interspecies hybrid sterility. It is also possible that overall genome homology, rather than individual loci, has the strongest influence on the degree of hybrid sterility.

The most likely explanation as to why the autosomal genotype is less relevant to the final sterility phenotype, however, is that the same recessive–dominant interactions that are presumed to cause F1 hybrid sterility are also the major factors affecting backcross hybrid sterility. Backcross individuals have one entire set of chromosomes from the backcross parent (Fig. 1), and thus an entire complement of any dominant autosomal interactors that may reside in the genome. In this scenario, the presence of any recessive X chromosome interactor would render an individual sterile, explaining the strong correlation of X genotype with backcross hybrid sterility. If this is true, then the disproportionately larger number of recessive autosomal loci that have been found (Masly and Presgraves 2007) may contribute less overall to hybrid sterility than the less-abundant dominant loci. Because this comparison was only able to be completed for Drosophila, data from other taxonomic groups would allow for an assessment of whether this is a more universal trend. It is also possible that my failure to detect a consistent effect of the autosomes is due to the inviability of certain genotypes, preventing those individuals from being scored. Although this cannot be ruled out, the dominance of these interactors and their effects would have to follow an inconsistent pattern to create the inconsistent pattern observed for sterility.

The second trend is that the amount of heterospecificity is highly correlated with the directionality of backcross sterility (Fig. 4). In other words, the backcross direction that produces offspring that have a greater amount of heterospecificity is also the direction of backcross that produces a greater proportion of sterile offspring, suggesting that differential levels of heterospecificity may cause the directionality of sterility observed in backcross hybrids. Unlike the data presented above, there was not a strong effect of the X chromosome driving this directionality, and it instead appears to be driven by the heterospecificty of the autosomes. This complement increases the number of recessive–dominant interactions that are possible, whereas decreasing the number of recessive–recessive interactions that are possible (the scenario depicted in Fig. 2E). As seen in the correlations with sterility, it again appears that recessive–recessive interactions may not play as strong of a role as recessive–dominant interactions. Although these comparisons cover a broad range of taxa, from flies and mosquitoes to plants, it is still a very small sample of organisms with which to make this comparison. Additionally, the X chromosome data contain far fewer genotypes than the autosomal data (Table 3), and thus a wider variety of taxa with more extensive marker data is needed to resolve this discrepancy.

It is possible that the directionality in hybrid sterility is correlated with, but not caused by, the directionality in heterospecificity. One possibility is that the directionality is caused by cytoplasmic-nuclear genome interactions due to an incompatibility between the mitochondrial and nuclear genomes. The backcross to the species that provided the male for the F1 cross would then always be less fertile. This is not true for three of the 10 species pairs examined here (pseu × per, gam × arab3, brev × fulva), and it was shown that cytoplasmic effects did not contribute to hybrid sterility in an additional two of the 10 species pairs (mau × sim, sech × sim; Zeng and Singh 1993). Therefore, cytoplasmic-nuclear genome interactions are unlikely to contribute to the overall trend. Maternal effects might also explain the directionality of sterility. This would act in a similar fashion to the cytoplasmic-nuclear interactions, but with the interaction coming from factors that the mother contributes early in development. If the maternal factors in the F1 female are more like those from the parental female in the cross, then one would assume that increased sterility results from an abnormal interaction between the maternal factors and the backcross genome. The data assembled here do not provide evidence either for or against this as a contributing factor, and it requires further testing.

Lastly, there is a trend of a greater amount of heterospecificity in these backcross populations than expected, irrespective of the direction of backcross, as the majority of markers that were genotyped had heterospecificity levels greater than the expected 50% (Table 3 and Fig. 5). Heterozygosity can be deleterious when there are negative genetic interactions, such as in interspecies hybrids, and there should potentially be a deficiency of heterospecifics in interspecies datasets due to the presence of interacting factors affecting viability. It is surprising, therefore, that there are more heterospecific genotypes in the offspring than expected by random inheritance. This effect is not evenly distributed in the genome, as the autosomes have a far greater excess of heterospecific markers than the X chromosome (as in Fig. 2E). This measured excess is unlikely to be due to human error, as these studies were carried out in multiple laboratories, and scoring error is more likely to result in the failure to score an allele than the addition of alleles, which would reduce the number of scored heterospecifics rather than increase the number.

Because most of these studies have used inbred or long-standing laboratory stocks, the observed excess heterospecificity may be attributed to the masking of deleterious recessive alleles that accumulated in these laboratory stocks. In other words, when they are hybridized with another species, heterosis (hybrid vigor) allows for a greater amount of heterospecificity than expected if individuals that are homospecific have a lower rate of survival. The assumption would be that the benefits of heterospecificity due to heterosis are greater than the deleterious effects of interspecies genetic dysfunction. Although inbred populations can indeed benefit from a general increase in fitness due to heterosis, it is unlikely that heterosis alone accounts for the observed trend of increased heterospecificity. Because most studies cited here scored young adults and all living adults were scored, heterosis would have to significantly affect development to adulthood or short-term viability (rate of death of young individuals) in order for it to account for the number of heterospecifics in the scored populations. In other words, backcross individuals that are more homospecific (pure-species) would have to die before reaching adulthood whereas heterospecific (interspecies) individuals do not. Loci that prevent survivorship to adulthood when homospecific would rapidly be eliminated from a (homospecific) laboratory stock. Because most pure-species stocks are able to be maintained, and trends of early-age die-off or failures in development have not been reported within the species examined here, heterosis is unlikely to be the cause of the observed increase in heterospecificity. However, further studies measuring both hatch rates and sterility would be able to determine if heterosis is a contributing factor.

An alternative explanation for increased heterospecificity, and one that is intriguing to consider, is that there is a molecular mechanism to promote increased heterozygosity in offspring. Novel males have an increased mating success, either through female preference for novel males (Hughes et al. 1999; Masters et al. 2003) or through novel sperm having increased success at fertilization when present in a mix of sperm (Tregenza and Wedell 2002; Pemberton et al. 2003). The increased success of novel males likely evolved due to the resulting heterozygous offspring having a fitness advantage (Ferreira and Amos 2006; Gemmell and Slate 2006; Hoffman et al. 2007; Rubenstein 2007). Separate studies have demonstrated the existence of interspecies meiotic drive that can affect sex ratios (Tao et al. 2001; Orr and Irving 2005). Additionally, sperm have been shown to swim collectively to reach the egg, and are sensitive enough to tell relatedness of fellow sperm (Fisher and Hoekstra 2010). It seems possible, therefore, that sperm and/or eggs may be sensitive enough to determine the relatedness of their fertilization partner and could selectively increase the chances of fertilizations that would produce heterozygous offspring, thus increasing their fitness. Unlike meiotic drive, in which a particular chromosome has an increased chance of being present in the next generation, this “heterozygote drive” would promote any chromosome that increases heterozygosity, and thus would vary depending on which gametes are paired. Like many drive mechanisms, additional genetic measures might evolve to counter the drive, causing it to be “exposed” only when the countermeasures are absent, as in interspecies hybrids that have evolved along separate trajectories. If the mechanism varies in its effectiveness when interspecies genomes are mixed in different proportions, that is in reciprocal backcross progeny, it could also explain the observed presence of greater heterospecificity in one direction of backcross over the other. Although this is purely conjecture, the possibility is raised to draw attention to an additional level of selection that is plausible, yet unexplored.


Associate Editor: T. Chapman

ACKNOWLEDGMENTS

I am deeply grateful to all of the individuals who contributed datasets for this analysis. I also want to thank A. Johnson for assistance with the appropriate statistics, and G. Barker, M. Noor, D. Garfield, V. McNiven, J. Pardy, J. Vincent, and two anonymous reviewers for discussions and helpful comments. This work was funded by an NIH National Research Service Award, an NSERC Discovery grant and a Canada Research Chair to AJM.

Ancillary