Mechanisms of Transmission Ratio Distortion at Hybrid Sterility Loci Within and Between Mimulus Species.

Hybrid incompatibilities are a common correlate of genomic divergence and a potentially important contributor to reproductive isolation. However, we do not yet have a detailed understanding of how hybrid incompatibility loci function and evolve within their native species, or why they are dysfunctional in hybrids. Here, we explore these issues for a well-studied, two-locus hybrid incompatibility between hybrid male sterility 1 (hms1) and hybrid male sterility 2 (hms2) in the closely related yellow monkeyflower species Mimulus guttatus and M. nasutus By performing reciprocal backcrosses with introgression lines (ILs), we find evidence for gametic expression of the hms1-hms2 incompatibility. Surprisingly, however, hybrid transmission ratios at hms1 do not reflect this incompatibility, suggesting that additional mechanisms counteract the effects of gametic sterility. Indeed, our backcross experiment shows hybrid transmission bias toward M. guttatus through both pollen and ovules, an effect that is particularly strong when hms2 is homozygous for M. nasutus alleles. In contrast, we find little evidence for hms1 transmission bias in crosses within M. guttatus, providing no indication of selfish evolution at this locus. Although we do not yet have sufficient genetic resolution to determine if hybrid sterility and transmission ratio distortion (TRD) map to the same loci, our preliminary fine-mapping uncovers a genetically independent hybrid lethality system involving at least two loci linked to hms1 This fine-scale dissection of TRD at hms1 and hms2 provides insight into genomic differentiation between closely related Mimulus species and reveals multiple mechanisms of hybrid dysfunction.

Hybrid incompatibilities are a common outcome of genomic divergence among closely related species. Across diverse taxa, a number of genes for hybrid inviability and sterility have been identified [see Presgraves (2010), Maheshwari and Barbash (2011), Sweigart and Willis (2012), Ouyang and Zhang (2013)], but we still know very little about how such genes function and initially evolve within their native species. One possibility is that the initial mutations are selectively “neutral” and become fixed by random genetic drift. Alternatively, the mutations might increase in frequency because they benefit the native species for reasons that are incidental to their role in reproductive isolation, for example by promoting ecological adaptation (Schluter and Conte 2009). Yet another possibility is that hybrid incompatibilities arise through recurrent bouts of intragenomic conflict within species (Frank 1991; Hurst and Pomiankowski 1991). In this last scenario, selfish genetic elements (e.g., transposons, meiotic drivers, and “gamete killers”) manipulate host reproduction to bias their own transmission. Because these actions are often detrimental to host fitness, there is then selective pressure for compensatory mutations or suppressors to neutralize the effects of selfish evolution (Burt and Trivers 2006).

The idea that intragenomic conflict involving segregation distorters might be a major source of hybrid incompatibilities has resurged in recent years (Johnson 2010; McDermott and Noor 2010; Presgraves 2010; Crespi and Nosil 2013), largely due to influential studies in Drosophila that have mapped hybrid segregation distortion and hybrid sterility to the same genomic locations (Tao ; Phadnis and Orr 2009; Zhang ). Similarly, in plants, classic and recent crossing studies have revealed gamete killers that affect both transmission ratios and fertility; at these loci, one parental allele causes the abortion of gametes carrying the other allele [e.g., tobacco: (Cameron and Moav 1957), wheat: (Loegering and Sears 1963), tomato: (Rick 1966), rice: (Sano 1990; Long ; Yang ), and Arabidopsis: (Simon )]. Although suggestive of a causal link between selfish genetic elements and hybrid incompatibilities, few studies have proven a history of segregation distortion within species. Thus, in most cases, an alternative possibility is that segregation distortion acts exclusively in hybrid genetic backgrounds, and is a consequence rather than a cause of the incompatibility.

In seed plants, hybrid incompatibilities can act in either the diploid sporophyte or the haploid gametophyte, two stages of the life cycle that are controlled by different sets of genes and subject to distinct evolutionary forces (Walbot and Evans 2003; Gossmann , 2016). Unlike in animal systems, which have very little haploid gene expression in sperm or egg cells (Braun ; Barreau ), thousands of genes are expressed in plant gametophytes (i.e., pollen and embryo sacs in angiosperms) (Wuest ; Rutley and Twell 2015). As a result, hybrid sterility in plants can be caused by genetic incompatibilities that affect the haploid gametophytes or the diploid sporophytic tissues surrounding the gametes (e.g., tapetum for pollen and ovule cells for the embryo sac). Of these two possibilities, the former appears to be much more common among the ∼50 hybrid sterility loci that have been identified between subspecies of Asian cultivated rice, Oryza sativa ssp. japonica and O. sativa ssp. indica (Morishima ; Ouyang and Zhang 2013). A large number of gametic incompatibilities have also been shown to contribute to TRD in crosses between populations of Arabidopsis lyrata (Leppala ). This bias toward gametic incompatibilities might be due to differences in the number of mutations that affect the two classes of hybrid sterility and/or to the fact that recessive alleles are exposed in the haploid gametophyte (similar to genes on heteromorphic sex chromosomes). Additionally, rates of evolution might be accelerated for gametophytic genes due to sex-specific selection (Gossmann ). It is also possible that intragenomic conflict is more common in the gametophyte; any selfish genetic element that can disable gametes carrying the alternative allele will have a direct impact on its own transmission.

Of the handful of plant hybrid sterility genes that have been cloned, all are in rice, most are gametic, and many appear to have evolved via neutral processes. The two most straightforward examples involve pollen defects caused by loss-of-function alleles at duplicate genes (Mizuta ; Yamagata ), consistent with a model of divergent resolution via degenerative mutations and genetic drift (Werth and Windham 1991; Lynch and Force 2000). The remaining six cases all involve gamete killers (Long ; Kubo , 2016a,b; Yang ; Yu ), which might be taken as evidence for pervasive selfish evolution within rice species. However, molecular characterization of these hybrid sterility systems has provided little support for this scenario. For example, the S5 locus causes female sterility in japonica-indica hybrids when gametes carry an incompatible combination of “killer” and “protector” alleles at three, tightly linked genes (Yang ). The two domesticated subspecies carry null alleles in distinct components of the killer–protector system. Because both derived haplotypes are perfectly compatible with the ancestral haplotype, it seems unlikely that they entailed fitness costs. Although it is conceivable that intragenomic conflict played a role in the initial formation of the S5 haplotype (i.e., the ancestral killer/protector combination might represent a resolved conflict), it does not seem to be the cause of the current reproductive barrier between japonica and indica. Similarly, at the Sa locus, which causes japonica-indica hybrid male sterility, patterns of molecular variation and the prevalence of neutral alleles that are compatible in all crosses suggest that hybrid dysfunction may have evolved unopposed by natural selection (Long ; Sweigart and Willis 2012). A key feature of these gamete killers is that they are caused by two or more tightly linked, epistatic genes (Long ; Yang ; Kubo 2013, 2016a). Adding to the complexity, some of them require additional, unlinked loci that act sporophytically (Kubo , 2016a,b). Taken together, these studies suggest that hybrid sterility in rice is polygenic and might evolve without significant fitness costs within species. However, it is not yet clear if these themes are generalizable to other plant systems.

Here, we investigate patterns of TRD associated with a two-locus hybrid sterility system between the closely related monkeyflower species, Mimulus guttatus and M. nasutus. Previously, we fine-mapped the two incompatibility loci—hms1 and hms2—to small nuclear genomic regions of ∼60 kb each on chromosomes 6 and 13 (Sweigart and Flagel 2015). We also discovered evidence that the hms1 incompatibility allele is involved in a partial selective sweep within a single population of M. guttatus, but the underlying cause of the sweep is unknown (Sweigart and Flagel 2015). Additionally, because the hms1 sterility allele is embedded in a nearly invariant, 320 kb haplotype, it is not yet clear whether hms1 or a linked locus is the target of the sweep. This polymorphic hybrid sterility system provides a unique opportunity to test directly whether selfish evolution within species can lead to incompatibilities between species.

Previously, in crosses between M. guttatus and M. nasutus, we observed TRD of genotypes at both hms1 and hms2 (Sweigart ; Sweigart and Flagel 2015), but the causes have remained unexplored. Additionally, these previous studies did not test directly whether the hms1-hms2 incompatibility acts in the gametophyte or sporophyte, although patterns of F2 hybrid sterility seemed to suggest the latter. Results from these studies suggested that the incompatibility acts in the diploid sporophyte, with the M. guttatus allele at hms1 acting dominantly in combination with recessive M. nasutus alleles at hms2 to cause nearly complete male sterility and partial female sterility (Sweigart ). Consistent with this genetic model, pollen viability is ∼20% in F2 hybrids that are heterozygous for hms1 and homozygous for M. nasutus alleles at hms2 (hms1GN; hms2NN), much lower than the 50% expected for a strictly gametic hybrid incompatibility (with hms1G; hms2N causing dysfunction). Moreover, because a gametic hybrid incompatibility should cause transmission bias at both interacting loci, we would expect a deficit of M. guttatus alleles at hms1 equal to that of M. nasutus alleles at hms2. Although F2 hybrids do indeed show a deficit of M. nasutus alleles at hms2, allelic transmission at hms1 follows the Mendelian expectation (Sweigart ).

In the current study, we used ILs and a reciprocal backcross design to distinguish among at least four possibilities for TRD in genomic regions linked to hms1 and hms2: (1) distortion through male gametes due to pollen competition and/or pollen sterility, (2) distortion through female gametes due to female meiotic drive (e.g., Fishman and Saunders 2008) and/or ovule sterility, (3) TRD through both male and female gametes due to an incompatibility that affects both gametophytes (e.g., Kubo ), and (4) distortion caused by selection against zygotes. In a series of crossing experiments, we investigated the mechanism of TRD at hms1 and hms2 and addressed the following specific questions. Is hybrid transmission bias at hms1 and/or hms2 a simple byproduct of gametic hybrid sterility? Is there evidence for hybrid transmission bias at these loci independent of gamete sterility? Are hybrid sterility and TRD genetically separable? Does TRD at hms1 occur within M. guttatus? Our results provide insight into the mechanisms of hybrid sterility and transmission distortion, and into the evolutionary dynamics of incompatibility alleles within species.

Materials and Methods

Study system and plant lines

The M. guttatus species complex is a group of phenotypically diverse wildflowers with abundant natural populations throughout much of western North America. In this study, we focus on M. guttatus and M. nasutus, two members of the complex that diverged roughly 200,000 yr ago (Brandvain ). These species occupy a partially overlapping range, and are primarily differentiated by mating system. M. guttatus is predominantly outcrossing with showy, insect-pollinated flowers, whereas M. nasutus is highly self-fertilizing with reduced flowers. In geographic regions where the two species cooccur, they are partially reproductively isolated by differences in floral morphology, flowering phenology, and pollen-pistil interactions (Diaz and Macnair 1999; Martin and Willis 2007; Fishman ). Hybrid incompatibilities are also common, but variable (Vickery 1978; Christie and Macnair 1987; Sweigart ; Case and Willis 2008; Martin and Willis 2010). Despite these barriers to interspecific gene flow, sympatric populations display evidence of genome-wide introgression (Sweigart and Willis 2003; Brandvain ; Kenney and Sweigart 2016).

Previous work identified two nuclear incompatibility loci, hms1 and hms2, which cause nearly complete male sterility and partial female sterility in a fraction of F2 hybrids between an inbred line of M. guttatus from Iron Mountain, Oregon (IM62), and a naturally inbred M. nasutus line from Sherar’s Falls, Oregon (SF5) (Sweigart ). In 2015, Sweigart and Flagel generated a large SF5-IM62 F2 mapping population (N = 5487) to fine-map hms1 and hms2 to regions of ∼60 kb on chromosome 6 and chromosome 13, respectively. Hybrids carrying at least one incompatible M. guttatus allele at hms1 in combination with two incompatible M. nasutus alleles at hms2 display extreme male sterility (i.e., 0–5% pollen viability) and partial female sterility (Sweigart ). Furthermore, the hms1 locus is polymorphic within the Iron Mountain population (Sweigart ) and several inbred lines derived from that site are known to carry compatible alleles that do not cause hybrid sterility when crossed to M. nasutus (Sweigart and Flagel 2015). In experimental crosses to test for TRD at hms1 within M. guttatus, we used a compatible line called IM767. In total, three inbred lines were used in different crossing schemes to test for TRD within and between species (see below). SF5 is compatible at hms1 and incompatible at hms2, IM62 is incompatible at hms1 and compatible at hms2, and IM767 is compatible at hms1 and hms2.

All plants were grown in the greenhouse at the University of Georgia. For all crosses, seeds were planted into 96-cell flats containing Fafard 3B potting mix (Sun Gro Horticulture, Agawam, MA), stratified for 7 d at 4°, and then placed in a greenhouse with supplemental lights set to 16 hr days. Plants were bottom-watered daily and temperatures were maintained at 24° during the day and 16° at night.

IL crossing design to investigate mechanisms of TRD between M. guttatus and M. nasutus

Previously, two reciprocal nearly isogenic line (NIL) populations carrying M. nasutus (SF5) or M. guttatus (IM62) introgressions in the opposite genetic background were generated (Fishman and Willis 2005). Briefly, a single SF5 × IM62 F1 and IM62 × SF5 F1 individual each served as the initial seed parent then underwent four generations of backcrossing to create a BN4 NIL population (SF5 × IM62 F1, M. recurrent parent) and a BG4 NIL population (IM62 × SF5 F1, M. recurrent parent). Within the BN4 and BG4 populations, each NIL carries a unique complement of heterozygous introgressions in a genome that is expected to be 93.75% homozygous for the recurrent parent’s alleles. To determine the genomic locations of the heterozygous introgressed regions, the NILs were genotyped at microsatellite and gene-based markers distributed throughout the genome (L. Fishman, unpublished). We selected three NILs with introgressions spanning hms1 or hms2 for further genetic analyses. Against a largely M. guttatus background, the BG4.476 NIL is heterozygous for an introgression that includes hms1 and ∼78% of the physical distance along chromosome 6. The BG4.149 line is heterozygous for an introgression that spans ∼71% of chromosome 13 and includes hms2. Against a M. nasutus background, the BN4.62 line is heterozygous for ∼75% of chromosome 13, including hms2. In addition to these NILs, we used an hms1 IL, RSB4, created after four generations of recurrent selection for hybrid sterility with backcrossing to M. nasutus, starting from a sterile SF5-IM62 BC1 individual (Sweigart ); the heterozygous introgression spans ∼50% of chromosome 6.

To characterize TRD between M. guttatus and M. nasutus, we used a multi-step crossing scheme, starting with the NILs and RSB4 (described above), to create a set of lines carrying specific two-locus genotypes at hms1 and hms2. First, to generate ILs that carry heterozygous alleles at both hms1 and hms2 in an otherwise M. guttatus or M. nasutus genetic background, we crossed BG4.476 to BG4.149, and BN4.62 to RSB4. From those progeny, we identified hms1-hms2 double heterozygotes by genotyping with markers that flank hms1 (M8 and M24) and hms2 (M51 and MgSTS193), as described previously (Sweigart and Flagel 2015). Next, to generate individuals that carry various two-locus combinations at hms1 and hms2, we self-fertilized doubly heterozygous ILs from each genetic background (i.e., IL-G and IL-N = M. guttatus and M. nasutus backgrounds, respectively). These crosses are expected to yield nine different two-locus genotypes each (typical of an F2), five of which are heterozygous at hms1 and/or hms2 (Figure 1). Surprisingly, one of the relevant IL-N hms1-hms2 genotypes was not recovered (hms1GG; hms2GN, see Figure 1); the hms1-introgression could not be made homozygous for M. guttatus alleles against an M. nasutus genetic background (see Results). We assessed male fertility (i.e., pollen viability) for the nine experimental IL genotypes (five for IL-G and four for IL-N) as described previously (Sweigart , 2007).

Figure 1

Crossing design for backcross experiment using introgression lines (ILs). For each genotype, two chromosome pairs are shown (one with hms1 and one with hms2). We constructed two sets of ILs with heterozygous introgressions at both hms1 and hms2; the IL-G has an M. guttatus genetic background (gray shading) and the IL-N has an M. nasutus genetic background (white). These doubly heterozygous ILs were self-fertilized to generate progeny with two-locus genotypes that are heterozygous at hms1 and/or hms2. These five progeny types were then reciprocally backcrossed to M. guttatus and M. nasutus. G = M. guttatus allele (gray); N = M. nasutus allele (white). TRD, transmission ration distortion.

To test the effect of hms1 genotype on transmission at hms2 and vice versa, we reciprocally backcrossed each of the nine ILs to both M. guttatus (IM62) and M. nasutus (SF5) (Figure 1). Thus, for each IL, we generated four reciprocal backcross populations allowing us to dissect sex-specific TRD. For each IL, two of the backcrosses used the emasculated IL as the seed parent in crosses to IM62 and SF5 lines (i.e., IL-IM62 and IL-SF5) and two used the IL as the pollen parent in crosses to emasculated IM62 and SF5 plants (i.e., IM62-IL, and SF-IL). If hms distortion occurs through pollen (due to pollen competition or a gametic incompatibility), we expect TRD in one or both of the backcrosses using the IL as the paternal parent, but not as the maternal parent. If, instead, female meiotic drive and/or a female gametic incompatibility occurs at these hms loci, we would expect to see TRD in both backcrosses with the IL as the seed parent, but not with the IL as the pollen parent. Finally, if TRD is caused by the loss of diploid zygotes (or seedlings), it should be apparent in both reciprocal crosses to the same recurrent parent (i.e., regardless of the sex of the IL). For all crosses, the female parent was emasculated 1–2 d before hand-pollination to prevent self-pollination. Sample sizes for the progeny classes ranged from 33 to 215 individuals (average N = 136).

For each hms locus, we performed factorial ANOVAs in Jmp Pro 13.0 to examine if genotype ratios were affected by four factors: (1) IL genetic background, (2) IL genotype at the interacting hms locus, (3) backcross direction, and (4) identity of the recurrent parent.

Crossing design to examine TRD within M. guttatus

To determine whether TRD at the polymorphic hms1 incompatibility locus occurs between incompatible and compatible alleles from the Iron Mountain population of M. guttatus, we generated reciprocal F2 and backcrossed populations with IM62 and IM767. We previously determined that the IM767 inbred line carries a compatible allele at hms1 (i.e., one that does not carry the 320 kb haplotype or cause sterility in combination with SF5 alleles at hms2). The IM62 and IM767 inbred lines were intercrossed reciprocally and a single F1 hybrid from each was self-fertilized to form reciprocal F2 populations (IM62 × IM767: N = 267 and IM767 × IM62: N = 315). To identify putative female- and male-specific sources of TRD, and to distinguish between meiotic/gametic mechanisms vs. zygotic selection, we generated reciprocal backcrosses with IM62 and IM767. We used a single F1 hybrid (IM62 × 767; maternal parent listed first) to generate four backcross populations to the recurrent parents (F1-IM62 BC1, IM62-F1 BC1, F1-IM767 BC1, and IM767-F1 BC1). Two of these backcrosses used the emasculated F1 as the seed parent and two used the F1 as the pollen donor in crosses to the emasculated recurrent parents.

We also wanted to examine the effect of M. nasutus hms2 alleles on patterns of within-M. guttatus TRD at hms1. We wondered if having M. nasutus alleles at hms2 has the potential to unleash severe distortion at hms1, even in an otherwise M. guttatus genetic background. To address this question, we intercrossed IM767 with a BG4-NIL (BG4.275) that is heterozygous for an SF5 introgression spanning ∼36% of chromosome 13 including hms2 (in an IM62 genetic background; Supplemental Material, Figure S2). We self-fertilized two of the resulting F1s to generate F2 hybrids segregating for SF5 alleles at hms2 against an IM62-IM767 F2-like genetic background. We then genotyped at hms-linked markers (M183 for hms1 and MgSTS193 for hms2) to identify IM62-IM767 hms1 heterozygotes in combination with three different hms2 genotypes: (1) IM62 homozygotes, (2) IM767 homozygotes, or (3) SF5 homozygotes. Using each of these three genotypic classes, we performed reciprocal backcrosses to IM767 (Figure S2).

Assessment of TRD

To examine patterns of TRD at the hms1 and hms2 loci, we collected leaf tissue from individual plants and isolated genomic DNA using a rapid extraction protocol (Cheung et al. 1993) modified for 96-well format. To infer the hms1 and hms2 genotypes of hybrid progeny generated from crosses between IM62 and SF5, we determined genotypes at a multiplexed set of fluorescently labeled markers that flank hms1 (M8 and M24) and hms2 (MgSTS193 and M51) following amplification protocols used previously (Sweigart , 2017). We excluded individuals with crossovers between either pair of flanking markers; based on expected frequency of double crossovers between flanking markers, genotyping error rates for hms1 and hms2 were each < 1%. For experimental crosses involving IM62 and IM767, only one tightly linked marker was used to infer genotype at hms1 (M183). Based on expected crossovers between hms1 and M183, the genotyping error rate was < 1%. All fluorescently labeled marker products were run on an ABI 3730 at the University of Georgia Genomics Facility. Genotypes were scored automatically using GeneMarker (SoftGenetics), with additional hand scoring when necessary. We used χ2 tests with two degrees of freedom to determine if hms-linked genotypes were significantly distorted.

Data availability

All plant lines are available upon request. Genotype data for fine-mapping TRD at hms1 and hms2 are provided in Table S1.

Results

TRD in M. nasutus-M. guttatus F2 hybrids

As part of previous efforts to fine-map Mimulus hybrid incompatibility loci (Sweigart and Flagel 2015), we generated a large M. nasutus-M. guttatus F2 hybrid mapping population (N = 5487) and genotyped all individuals at gene-based markers flanking hms1 (M8 and M24) and hms2 (M51 and MgSTS193). As previously reported (Sweigart ; Sweigart and Flagel 2015), we observed significant TRD in F2 genotypes at both hybrid sterility loci (Table 1). At hms1, we observed a significant excess of heterozygotes, but allelic transmission did not differ from the Mendelian expectation. The observed genotype ratios at hms1 also differed significantly from the expectation given the random union of two gametes with the observed allele frequencies. At hms2, we observed an excess of M. guttatus homozygotes and a deficit of M. nasutus homozygous genotypes, as well as a significant bias toward M. guttatus alleles. However, genotype ratios at hms2 do not differ from what is expected given the observed allele frequencies. Taken together, these patterns suggest TRD at hms1 might be driven primarily by zygotic selection, whereas hms2 appears to be influenced primarily by selection among gametes.

Table 1

Genotype and allele frequencies at hms1 and hms2 in an M. nasutus-M. guttatus F2 population (N = 5487)

	Allele Frequencya	Genotype Frequencyb
Locus	O	O	E
hms1	0.49:0.51	0.22:0.55:0.23****	0.24:0.50:0.26
hms2	0.62:0.38****	0.38:0.48:0.14	0.38:0.47:0.14

P < 0.0001 based on χ2 tests of observed frequencies vs. the Mendelian expectation with 2 d.f. for genotypes and 1 d.f. for allele frequencies. O, observed; E, expected.

O allele frequencies are reported as M. guttatus:M. nasutus (G:N). At hms2, but not hms1, allele frequencies significantly differ from the Mendelian expectation (0.5:0.5).

O and expected E genotype frequencies are reported as M. guttatus homozygotes:heterozygotes:M. nasutus homozygotes (GG:GN:NN). Expected genotype frequencies shown are calculated from the random union of gametes with the observed frequencies. At hms1, genotypes differ significantly (P < 0.0001) from both the Mendelian expectation (0.25:0.5:0.25) and from the expectation given the random union of gametes with the observed allele frequencies. At hms2, genotypes differ significantly (P < 0.0001) from the Mendelian expectation but not from the expectation given the random union of gametes with the observed allele frequencies.

When considered together, the two-locus genotypes at hms1 and hms2 differ significantly from the Mendelian expectation (χ2 = 389.372, d.f. = 8, P < 0.0001, N = 5487). Although the two-locus genotypes are also significantly different from the expectation given the observed allele frequencies at hms1 and hms2 shown in Table 1 (χ2 = 71.626, d.f. = 8, P < 0.0001), the values are much more closely aligned (Table 2). Particularly notable is the deficit of two genotypic classes (hms1GG; hms2NN and hms1NN; hms2GG) and the excess of two others (hms1GG; hms2GG and hms1NN; hms2NN; Table 2). This pattern of two-locus disequilibrium follows the expectation for gametic action of hms1-2 sterility (i.e., with hms1G; hms2N gametes tending to be sterile). However, the observed F2 transmission ratios at hms1 and hms2 cannot be entirely explained by hms1G; hms2N gametic sterility (Table S2). This phenomenon, whether acting through one or both parents, would be expected to reduce the transmission of M. guttatus alleles at hms1, in the same way that it reduces M. nasutus alleles at hms2. However, there is no indication of allelic transmission bias at hms1 in the F2 hybrids. Taken together, these results suggest that gametic expression of the hms1-hms2 incompatibility is important, but not the sole contributor, to patterns of TRD in F2 hybrids.

Table 2

Observed and expected genotype frequencies at hms1 and hms2 in F2 hybrids and IL F2 hybrids

	F2 (5487)a			IL-G F2 (167)b		IL-N F2 (200)c
Genotype hms1; hms2	E: Mendelian	E: O Allele Freq	O	O	E: Backcross	O	E: Backcross	E: hms1GG = Lethal
GG; GG	0.0625	0.093	0.099	0.066	0.107	0	0.119	0
GG; GN	0.1250	0.115	0.100	0.114	0.106	0	0.069	0
GG; NN	0.0625	0.035	0.022	0.006	0.200	0	0.090	0
GN; GG	0.1250	0.191	0.208	0.174	0.176	0.185	0.193	0.241
GN; GN	0.2500	0.236	0.268	0.234	0.249	0.300	0.249	0.310
GN; NN	0.1250	0.073	0.071	0.054	0.078	0.075	0.058	0.073
NN; GG	0.0625	0.098	0.070	0.102	0.072	0.085	0.077	0.096
NN; GN	0.1250	0.121	0.117	0.180	0.133	0.225	0.151	0.188
NN; NN	0.0625	0.037	0.047	0.072	0.061	0.130	0.0740	0.092

E, expected; O, observed; Freq, frequency; IL, introgression line.

F2 genotype counts significantly differ from the Mendelian expectation (χ2 = 389.372, d.f. = 8, P < 0.0001) and from what is expected for the random union of gametes given the observed allele frequencies (see Table 1) and independent assortment at hms1 and hms2 (χ2 = 71.626, d.f. = 8, P < 0.0001).

IL-G F2 genotype counts significantly differ from the Mendelian expectation (χ2 = 18.7910, d.f. = 8, P = 0.0160), but not from what is expected based on allelic transmission in the IL backcrosses (see Table 4, χ2 = 5.9730, d.f. = 8, P = 0.6502).

IL-N F2 genotypes significantly differ from the Mendelian expectation (χ2 = 86.4090, d.f. = 8, P < 0.0001) and from what is expected based on allelic transmission in the IL backcrosses (see Table 4, χ2 = 62.0370, d.f. = 8, P < 0.0001), but not from what is expected from the IL backcrosses + hms1GG homozygote death (χ2 = 3.5950, d.f. = 5, P = 0.6090).

Table 4

Two-locus transmission ratios at hms1 and hms2 in backcross progeny of doubly heterozygous ILs

			hms1;hms2a
♀b	♂b	Nc	G;G	G;N	N;G	N;N	P
IL-G	G	293	0.31	0.20	0.24	0.25
IL-G	N	232	0.28	0.24	0.25	0.22
IL-N	G	354	0.30	0.13	0.30	0.28	***
IL-N	N	320	0.43	0.15	0.22	0.19	****
Average			0.33	0.18	0.25	0.24
G	IL-G	298	0.32	0.05	0.35	0.28	****
N	IL-G	450	0.40	0.13	0.24	0.23	****
G	IL-N	104	0.34	0.03	0.30	0.34	****
N	IL-N	212	0.32	0.10	0.30	0.29	***
Average			0.34	0.08	0.30	0.28

P < 0.05, ** P < 0.01, *** P < 0.005, and **** P < 0.0001 based on χ2 tests of observed frequencies vs. the Mendelian expectation. G, M. guttatus background; N, M. nasutus background; IL, introgression line.

Two-locus allelic combination at hms1 and hms2 inherited from IL parent. G = M. guttatus allele; N = M. nasutus allele.

Backcrosses using ILs (M. guttatus background = IL-G; M. nasutus background = IL-N) to the IM62 line of M. guttatus (G) and the SF5 line of M. nasutus (N). ♀ indicates the maternal parent and ♂ indicates the paternal parent.

Number of progeny assessed.

M. nasutus-M. guttatus IL crosses reveal multiple causes of F2 distortion

To investigate several possible causes of F2 TRD at hms1 and hms2, we performed a crossing experiment using the IL-Gs and IL-Ns. In this crossing design (Figure 1), individuals with one of several possible two-locus hms1-hms2 genotypes, in each of the IL genetic backgrounds, were crossed reciprocally to M. guttatus (IM62) and M. nasutus (SF5). By scoring hms1 and hms2 genotypes in the progeny of these crosses, we were able to examine the effects of several factors, including parental genotype, genetic background, and cross direction, on transmission ratios at the two-hybrid sterility loci. Of the 36 crosses performed, 12 showed significant TRD at hms1 and/or hms2 (Table 3; note that two crosses were unsuccessful due to hybrid male sterility). For both hms1 and hms2, parental genotype at one locus has a strong effect on allelic transmission at the other (hms1 affects hms2: F = 37.69, P < 0.0001 and hms2 affects hms1: F = 7.80, P = 0.004; Figure S1). For hms2, cross direction is also important, with stronger TRD occurring through pollen (F = 72.33, P < 0.0001). Neither the genetic background nor the identity of the recurrent parent significantly affected transmission ratios at hms1 or hms2 (results not shown).

Table 3

Allelic transmission ratios at hms1 and hms2 in IL-backcross progeny

♀a	♂a	hms1; hms2b	Nc	hms1 %Gd	hms2 %Ge
IL-G	G	GN; GG	101	0.56
		GN; NN	171	0.60
		GG; GN	163		0.53
		NN; GN	158		0.47
		GN; GN	293	0.46	0.54
IL-G	N	GN; GG	189	0.55
		GN; NN	119	0.64*
		GG; GN	49		0.53
		NN; GN	132		0.50
		GN; GN	232	0.52	0.54
G	IL-G	GN; GG	382	0.55
		GN; NN	No seeds	–
		GG; GN	120		0.86****
		NN; GN	187		0.50
		GN; GN	298	0.37***	0.67****
N	IL-G	GN; GG	636	0.62****
		GN; NN	No seeds	–
		GG; GN	158		0.90****
		NN; GN	187		0.52
		GN; GN	450	0.53	0.64****
IL-N	G	GN; GG	266	0.44
		GN; NN	593	0.48
		GG; GN	N/a		–
		NN; GN	325		0.55
		GN; GN	354	0.42*	0.59*
IL-N	N	GN; GG	211	0.48
		GN; NN	317	0.52
		GG; GN	N/a		–
		NN; GN	43		0.54
		GN; GN	320	0.58*	0.66****
G	IL-N	GN; GG	113	0.46
		GN; NN	85	0.71**
		GG; GN	N/a		–
		NN; GN	250		0.53
		GN; GN	104	0.37*	0.64*
N	IL-N	GN; GG	177	0.51
		GN; NN	194	0.72****
		GG; GN	N/a		–
		NN; GN	188		0.57
		GN; GN	212	0.42	0.61*

P < 0.05, ** P < 0.01, *** P < 0.005, and **** P < 0.0001 based on χ2 tests of observed frequencies vs. the Mendelian expectation. G, M. guttatus background; N, M. nasutus background; IL, introgression line; N/a, not applicable.

Backcrosses using ILs (M. guttatus background = IL-G; M. nasutus background = IL-N) to the IM62 line of M. guttatus (G) and the SF5 line of M. nasutus (N). ♀ indicates the maternal parent and ♂ indicates the paternal parent.

Two-locus genotype for hms1 and hms2. GG = M. guttatus homozygote; GN = heterozygote; and NN = M. nasutus homozygote.

Number of progeny assessed. Two crosses were unsuccessful (labeled “no seeds”) because the IL-G with the genotype hms1GN; hms2NN was completely male sterile. The IL-N with the genotype hms1GG; hms2GN could not be generated (see text) and is labeled “n/a.”

Percent M. guttatus (G) alleles at hms1 transmitted to progeny from heterozygous IL parent.

Percent M. guttatus (G) alleles at hms2 transmitted to progeny from heterozygous IL parent.

The pattern of TRD at hms2 follows what is expected if hybrid sterility acts through gametes. For example, if pollen grains are inviable when they carry M. guttatus alleles at hms1 in combination with M. nasutus alleles at hms2, the effect of hms1 paternal genotype on TRD at hms2 should be additive. Indeed, progeny from males that carry one or two M. guttatus alleles at hms1 show a 28 or 76% undertransmission of M. nasutus alleles at hms2 relative to the Mendelian expectation (Figure S1). Consistent with the action of a gametic incompatibility, backcross progeny of doubly heterozygous IL parents (i.e., hms1GN; hms2GN) are much less likely to come from gametes with an M. guttatus allele at hms1 in combination with an M. nasutus allele at hms2 (Table 4). In these crosses, the hms1G; hms2N gamete type is undertransmitted through both sexes, though the effect is stronger through males. Undertransmission is also more severe in crosses to IM62 (M. guttatus) and against the IL-N genetic background (Table S3).

If the hms1-hms2 incompatibility acts through gametes, we might expect patterns of pollen viability to predict rates of TRD through males. To examine this possibility, we measured pollen viability in various two-locus genotypes of the IL-Gs and IL-Ns (Table 5). In general, patterns of male fertility and TRD are indeed related. For example, pollen viability is 64% in IL-Gs that are hms1GG; hms2GN. For this genotype, if we assume equal transmission of M. guttatus and M. nasutus alleles into pollen and attribute all sterility to hms1G; hms2N, then the M. guttatus allele at hms2 should be present in 78% of progeny when this individual is used as the paternal parent in a cross (which is close to the observed frequency of 86%, Table 3). Similarly, for IL-Gs that are hms1GN; hms2GN, if we assume that all hms1G; hms2N gametes are inviable (and divide the remaining 7% sterility equally among the other three two-locus genotypes), we expect M. guttatus allele frequencies of 33 and 66% at hms1 and hms2, respectively. These values are very similar to what we observe when this IL-G genotype is backcrossed to M. guttatus (37 and 67%, Table 3).

Table 5

Pollen viability for various hms1-2 IL genotypes

Genetic Background	hms1; hms2	Na	PVb
IL-G	GG; GN	5	0.64 (0.04)
	NN; GN	16	0.79 (0.04)
	GN; GN	16	0.67 (0.06)
	GN; GG	12	0.71 (0.06)
	GN; NN	3	0.18 (0.17)
IL-N	NN; GN	15	0.88 (0.02)
	GN; GN	14	0.81 (0.03)
	GN; GG	13	0.85 (0.02)
	GN; NN	18	0.09 (0.01)

IL, introgression line; G, M. guttatus background; N, M. nasutus background; PV, pollen viability.

Number of individuals scored.

Pollen viability given as the proportion viable pollen grains per flower (for a haphazard sample of 100). PV is the average of two flowers and the number in parentheses is the SE.

At hms1, TRD is more complex. On the one hand, M. guttatus alleles at hms1 are undertransmitted due to the hms1G; hms2N gametic sterility discussed above (Table S3). On the other hand, in many of the IL-backcrosses, M. guttatus alleles at hms1 are overrepresented among the progeny (Table 2). This effect is most pronounced when the IL parent is heterozygous at hms1 and homozygous for M. nasutus alleles at hms2 (Figure S1; note that this genotype is not completely sterile so crosses can still be performed). Remarkably, this direction of TRD is exactly the opposite of what is expected if hms1 transmission is primarily influenced by the hms1G; hms2N gametic incompatibility. Moreover, pollen viability in IL-Gs and IL-Ns with the genotype hms1GN; hms2NN is much lower than the 50% expected for gametic expression of hybrid male sterility (Table 5), consistent with overtransmission of M. guttatus hms1 alleles into pollen. Note that if these two TRD mechanisms—hms1G; hms2N gamete sterility and overtransmission of M. guttatus hms1 alleles—counteract each other in F1 hybrids and in doubly heterozygous ILs, it could explain why their progeny carry hms1 alleles in roughly Mendelian proportions (Figure S1 and Table 2). Consistent with this idea, backcross progeny of doubly heterozygous ILs are most often products of the hms1G; hms2G gamete type (Table 4).

Additionally, a genetically distinct hybrid incompatibility appears to affect transmission of hms1 against an M. nasutus genetic background. Self-fertilization of a doubly heterozygous IL-N individual produces no M. guttatus homozygotes at the hms1 locus (Table 2), a genotype expected to appear in a quarter of the progeny (IL-N F2 N = 200, expected frequency = 50). When instead this same doubly heterozygous IL-N genotype is crossed to IM62 (in either direction), progeny homozygous for M. guttatus alleles at hms1 are recovered (Table S4). Note that selfing the doubly heterozygous IL-N produces offspring with isogenic M. nasutus genetic backgrounds, whereas the backcross to IM62 results in progeny with genetic backgrounds that are F1-like. Taken together, these results suggest that the hms1 region is involved in yet another hybrid incompatibility. This one causes lethality in hybrids that are homozygous for M. guttatus alleles at hms1-linked loci and homozygous for M. nasutus alleles at one or more unlinked loci. Given the large size of the hms1-containing IL (representing 50% of chromosome 6), it seems likely that additional genetic loci contribute to hybrid lethality, rather than hms1 itself.

By scoring genotype frequencies in the progeny of reciprocal backcrosses involving the doubly heterozygous ILs (hms1GN; hms2GN), it is possible to track which two-locus hms1-2 meiotic products are transmitted through pollen and ovules. If we use these observed two-locus gametic allele frequencies (instead of assuming equal proportions of the four two-locus gamete types) to calculate expected genotype frequencies in the selfed progeny of doubly heterozygous ILs (i.e., IL-F2 populations), the resulting values do not significantly differ from observed proportions (Table 2 and Table 4). To fully account for observed genotype frequencies in the IL-N F2, it is also necessary to assume complete lethality of M. guttatus homozygotes at hms1 (Table 2; note that this hybrid lethality is not reflected in IL backcross allele frequencies because progeny do not carry the requisite M. nasutus genetic background for expression of the incompatibility).

In summary, we have identified at least three sources of hms1-hms2 TRD in M. nasutus-M. guttatus F2 hybrids: (1) undertransmission of pollen and, to a lesser extent, ovules that carry an M. guttatus allele at hms1 in combination with an M. nasutus allele at hms2, presumably due to gametic inviability; (2) overtransmission of M. guttatus alleles at hms1, an effect that occurs through males and females, and does not depend on genetic background; and (3) hybrid lethality in individuals homozygous for M. guttatus alleles at hms1 (and linked genomic regions) in combination with M. nasutus homozygosity at one or more unlinked loci.

Fine-mapping TRD

In previous (Sweigart and Flagel 2015) and ongoing efforts to fine-map hms1 and hms2, we identified a small subset of SF5-IM62 F2 hybrids that were recombinant for one or both sets of hms-flanking markers. With the goal of genetically mapping TRD in both regions, we self-fertilized these recombinants to generate F3 progeny and examined genotype frequencies at both sets of flanking markers (Figure 2 and Figure 3). We reasoned that TRD in the F3 progeny should only be observable if the causal locus is heterozygous in the F2 parent. If, instead, the TRD-causing locus is homozygous (for either M. guttatus or M. nasutus alleles), loci in the adjacent heterozygous region should segregate in a Mendelian fashion.

Figure 2

Genetic dissection of hms2-linked TRD in Mimulus. A physical map of ∼4.5 Mb surrounding the hms2 region is shown, including the positions of genetic markers (indicated with triangles along the top). F2 recombinants are shown with horizontal bars representing genotypes in the genomic region linked to hms2 and squares indicate genotypes at hms1 and across the genetic background (white = M. nasutus homozygote, gray = heterozygote, and black = M. guttatus homozygote). Deviation from the Mendelian expectation (0.25) of M. nasutus homozygotes (NN) in the F3 progeny is given. N indicates the number of F3 progeny scored from each individual. * P < 0.05, ** P < 0.01, *** P < 0.005, and **** P < 0.0001 based on χ2 tests of observed frequencies vs. the Mendelian expectation. TRD, transmission ration distortion.

Figure 3

Genetic dissection of hms1-linked TRD in Mimulus. A physical map of 15 Mb surrounding the hms1 region is shown, including the positions of genetic markers (indicated with triangles along the top) and the 320 kb hms1 haplotype (shown as a solid black bar with dotted lines extending downward). F2 recombinants are shown with horizontal bars representing genotypes in the genomic region linked to hms1, and squares indicate genotypes at hms2 and across the genetic background (white = M. nasutus homozygote, gray = heterozygote, and black = M. guttatus homozygote). Deviation from the Mendelian expectation (0.25) of M. nasutus homozygotes (NN) in the F3 progeny is given for the top group of 11 F2 recombinants. Deviation from the Mendelian expectation (0.25) of M. guttatus homozygotes (GG) in the F3 progeny is given for the bottom group of six F2 recombinants and the doubly heterozygous ILs. N indicates the number of F3 progeny scored from each individual. *** P < 0.005 and **** P < 0.0001, based on χ2 tests of observed frequencies vs. the Mendelian expectation. ILs, introgression lines; TRD, transmission ration distortion.

As in the IL crosses, patterns of hms2-linked TRD were consistent with the action of hms1G; hms2N gametic sterility. In this genomic region, the most extreme TRD occurred in the two F3 families that descended from F2 hybrids with the hms1GG; hms2GN genotype (Figure 2). Despite this general support for hms1-hms2 gametic sterility, hms2-linked TRD could not be unambiguously mapped to a particular genomic region (no interval in Figure 2 is perfectly associated with presence/absence of TRD). Presumably, genetic background in these F2 hybrids can mask TRD associated with hms1G; hms2N gametic sterility (e.g., 28_22) or mimic it (e.g., 02_66).

At hms1, the two contributors to TRD were decoupled in F2 recombinants, with M. guttatus homozygotes overrepresented in some F3 families and underrepresented in others (Figure 3). As with the IL experiments, the most significant overtransmission of M. guttatus alleles at hms1 appears in the progeny of F2 hybrids that are homozygous for M. nasutus alleles at hms2 (Figure 3, first two F2s). This TRD phenotype maps to an 800 kb region that includes hms1, but we have too few recombinants to determine if the hybrid TRD phenotype is genetically separable from hybrid sterility. For a distinct set of hms1 F2 recombinants, we observed a severe deficit of M. guttatus homozygotes among their F3 progeny (Figure 3, last six F2 individuals), consistent with the expression of hybrid lethality as seen in the IL experiments. This TRD phenotype maps to at least two independent loci in the hms1 region and is not affected by hms2 genotype, suggesting a distinct genetic basis for this hybrid incompatibility.

TRD at hms1 within M. guttatus

To investigate whether hms1-linked TRD is a strictly hybrid phenomenon or also occurs within M. guttatus, we generated reciprocal F2 progeny between IM62 and IM767. These two inbred lines carry distinct alleles at hms1 and show very different patterns of variation in the surrounding genomic region. The IM62 line carries an incompatible, hybrid sterility-causing hms1 allele embedded within a distinctive 320 kb haplotype, whereas IM767 carries a compatible (i.e., nonsterility causing) allele at hms1 and typical levels of nucleotide variation in the region (Sweigart and Flagel 2015). Because genotype frequencies at hms1 did not differ significantly between reciprocal F2 populations (data not shown), we pooled data from both directions of the cross. We observed modest but significant TRD at hms1 with an excess of IM62 homozygotes (frequency of IM62 homozygotes to heterozygotes to IM767 homozygotes: expected 0.25:0.5:0.25, observed 00.27:0.54:0.19, χ2 = 6.479, d.f. = 2, P = 0.0027, N = 582). However, the bias in allelic transmission toward IM62 was not significant (frequency of IM62:IM767 alleles: expected 0.5:0.5, observed 0.54:0.46, χ2 = 0.151, d.f. = 1, P < 0.151, N = 582) and genotype frequencies did not significantly differ from the expectation given the allele frequencies (χ2 = 2.025, d.f. = 2, P = 2.025, N = 582). To further investigate the mechanism of hms1-linked TRD, we performed reciprocal backcrosses using IM62 and IM767. However, unlike in the IM62-IM767 F2 hybrids, all four backcross populations exhibited nearly perfect Mendelian ratios (expected 0.50:0.50; F1 × IM62 = 0.50:0.50, N = 279; F1 × IM767 = 0.50:0.50, N = 281; IM62 × F1 = 0.51:0.49, N = 189; and IM767 × F1 = 0.49:0.51, N = 188). These results suggest that there is little to no transmission bias favoring the hms1 incompatibility allele or the associated 320 kb haplotype within the Iron Mountain population.

Finally, we wanted to investigate if the presence of M. nasutus alleles at hms2 increases the transmission bias of IM62 at hms1, even in an otherwise M. guttatus genetic background. To address this question, we examined genotype frequencies in the reciprocal backcross progeny of individuals that were heterozygous IM62/IM767 at hms1 and segregating for an M. nasutus introgression at hms2 (against an otherwise IM62-IM767 F2 genetic background; Figure S2). Indeed, extreme TRD at hms1 (i.e., bias toward the IM62 allele > 70%) was only observed in the backcross progeny of one individual (08_60) that was also homozygous for M. nasutus alleles at hms2 (Table 6). These results suggest that overtransmission of the IM62 allele at hms1, which appears to require M. nasutus alleles at hms2, may occur exclusively in hybrids.

Table 6

Transmission of IM62 vs. IM767 at hms1 varies depending on hms2 genotype

hms2 Genotype	F2 IDa	%IM62b
		F2 Male	F2 Female
IM62	02_02	0.58 (74)	0.55 (64)
	02_46	0.48 (121)	0.43 (28)
	06_31	0.55 (179)	0.29 (41)
	06_70	0.55 (123)	0.50 (116)
	06_96	0.41 (46)	–
	Combined	0.53 (543)	0.47 (249)
IM767	02_17	0.45 (53)	0.56 (122)
	02_48	0.54 (79)	0.49 (84)
	02_68	0.56 (39)	0.49 (141)
	06_39	0.55 (107)	–
	Combined	0.53 (278)	0.51 (347)
SF	08_60	0.77 (104)****	0.73 (75)***
	12_09	0.50 (111)	0.54 (41)
	Combined	0.62 (215)**	0.66 (116)*

P < 0.05, ** P < 0.01, *** P < 0.005, **** P < 0.0001 based on χ2 tests of observed genotype frequencies vs. the Mendelian expectation. ID, identifier.

Individual IDs for F2 progeny from BG4275 to IM767 crosses. At hms1, all F2 individuals used were heterozygous for IM62 and IM767 alleles; at hms2, individuals used were homozygous for IM62, IM767, or SF alleles (see text for details).

Percent IM62 alleles at hms1 transmitted to progeny from IM62 to IM767 heterozygous parent. Value given in parentheses is the number of progeny assessed.

Discussion

TRD is commonly observed among hybrid offspring of recently diverged species, but the evolutionary significance is not always clear. In this study, we identified multiple contributors to hybrid TRD in genomic regions linked to two Mimulus hybrid sterility loci hms1 and hms2, revealing a fine-scale complexity reminiscent of several previously characterized hybrid incompatibilities (Davis and Wu 1996; Long ; Yang ; Kubo ). We have discovered that hybrid transmission bias is caused, in part, by gametic action of the hms1-hms2 incompatibility itself. However, the effects of the gametic hybrid sterility are partially obscured by an opposing (and currently unknown) mechanism that results in overtransmission of the M. guttatus hms1 incompatibility allele in certain hybrid genetic backgrounds. In addition, our genetic analyses uncovered an independent hybrid lethality system with at least two incompatibility loci tightly linked to hms1. Strikingly, we found no evidence of biased transmission of the hms1 incompatibility allele within M. guttatus, providing little support for selfish evolution as the cause of a recent, partial sweep at hms1 (Sweigart and Flagel 2015). Instead, it appears that TRD at hms1 and hms2 might occur exclusively in hybrids.

Gametic action of hms1-hms2 hybrid incompatibility

Our finding that the hms1G; hms2N gamete type is severely undertransmitted in six of the eight backcrosses involving doubly heterozygous ILs (hms1GN; hms2GN) is strong evidence of gametic action of the incompatibility. This result runs counter to our previous interpretation of the finding that pollen viability is reduced from the F1 to F2 generation, which seemed to suggest a diploid (sporophytic) genetic basis for the hms1-hms2 incompatibility (Sweigart ). In general, for a hybrid incompatibility that affects the gametophyte, sterility is expected to be less severe in the F2 generation due to the inviability of recombinant F1 gametes and regeneration of parental combinations. However, in this case, it appears that removal of hms1G; hms2N F1 gametes is somewhat balanced by overtransmission of M. guttatus alleles at hms1. Moreover, incomplete penetrance of F1 hybrid gametic sterility (i.e., some hms1G; hms2N gametes do contribute to the F2 generation, see Table 4) produces a small fraction of F2 hybrids that are completely sterile because they are homozygous for incompatible alleles (i.e., hms1GG; hms2NN).

As an independent line of evidence for gametic expression of the hms1-hms2 incompatibility, it is apparently difficult to introgress M. nasutus hms2 alleles into an M. guttatus genetic background. In the BG4-NIL population (i.e., fourth-generation NILs that carry SF5 introgressions in an IM62 genetic background; see Materials and Methods from this study and Fishman and Willis 2005), only 2.8% of individuals (5/175) are heterozygous at MgSTS45, a marker ∼2 cM from hms2 (L. Fishman, unpublished results). This level of distortion is notable: of the 194 markers genotyped in this BG4 population, only four of them show lower heterozygosity and three of those map near a meiotic drive locus that strongly favors the M. guttatus allele (Fishman and Saunders 2008). In the BN4-NIL population (i.e., fourth-generation NILs that carry IM62 introgressions in an SF5 genetic background; see Materials and Methods), heterozygous introgressions at MgSTS45 are much more common, occurring in 10% of individuals (18 of 181). This result is not unexpected given that M. guttatus alleles at hms2 are perfectly compatible with M. nasutus alleles at hms1.

Unlike in animals, hybrid incompatibilities in plants are often gametic (Morishima ; Koide ; Leppala ). Based on his studies of hybrid sterility between the indica and japonica varieties of O. sativa, Oka (1974) first suggested that defects in pollen development might be caused by loss-of-function alleles at duplicate genes (Oka 1974). Indeed, two cases of this duplicate gametic lethal model have now been demonstrated at the molecular level (Mizuta ; Yamagata ). For Mimulus hms1 and hms2, there is no evidence that gene duplicates are involved (Sweigart and Flagel 2015), but a similar pattern of hybrid sterility is expected to result from a two-locus hybrid incompatibility between any genes expressed in the gametophyte. Additionally, the fact that the hms1-hms2 incompatibility seems to affect both the male and female gametophyte (the hms1G; hms2N gamete type is undertransmitted through both sexes) is consistent with our finding that these loci contribute to both hybrid male sterility and hybrid female sterility (Sweigart ). Gametic hybrid incompatibilities that affect the fertility of both sexes have also been discovered in tomato, rice, and Arabidopsis (Rick 1966; Koide ; Leppala ), though they are apparently less common than those that act in only one sex (Morishima ; Koide )

Additional sources of TRD

Our fine-scale dissection of TRD at hms1 and hms2 provides insight into genomic differentiation between closely related Mimulus species and reveals a potentially complex genetic basis for hybrid dysfunction. In other systems, fine-mapping has often revealed multiple, tightly linked hybrid incompatibility loci that show independent effects (Wu and Davis 1993; Kubo ; Simon ) or epistasis (Long ; Yang ; Kubo ). In one particularly complex example from indica and japonica, fine-mapping revealed two tightly linked genes involved in independent two-locus pollen killer systems (Kubo ). Because of this tight linkage, pollen killing had initially appeared to be caused by a single, three-locus interaction (Kubo et al. 2008). Remarkably, both of these pollen killer systems involve interactions between sporophytic and gametophytic genes, as well as additional modifier loci (Kubo ). The picture emerging from such studies is one of hybrid sterility regulated by multiple, interconnected molecular networks, potentially involving many genes.

A key question for hms1 and hms2 is whether the same genes cause the gametic incompatibility and transmission bias of M. guttatus at hms1. The latter is particularly strong when hms2 is homozygous for M. nasutus alleles (Figure S1 and Table 3), suggesting that it might be caused by an interaction between the two loci. Additionally, the presence of hms2NN also appeared to unleash severe hms1 TRD in one of the two IM62-IM767 F2 populations in which it was present (Table 6), suggesting that hms2 might be necessary but not sufficient for hms1 TRD. On the other hand, overtransmission of hms1G does not seem to absolutely require hms2NN (e.g., we observed 62% transmission of hms1G in M. nasutus × IL-GGN;GG, Table 3), which might argue against its direct involvement. Indeed, for the IL-Gs, there is a bias toward hms1G in all backcross populations except those involving doubly heterozygous IL parents (i.e., hms1GN; hms2GN), which, because they express the hms1G; hms2N gametic inviability, might obscure additional sources of hms1 TRD. Going forward, additional rounds of high-resolution fine-mapping will be needed to pinpoint the causal genes and determine if Mimulus hybrid sterility and TRD are genetically separable. Such efforts in rice have been successful in disentangling the complex phenotypic effects of linked hybrid sterility loci (e.g., Kubo ).

Identifying the molecular genetic basis of hms1 TRD might also provide insight into its mechanisms. Because the bias toward M. guttatus alleles at hms1 occurs through both males and females, the simplest single explanation is a gamete-killing system that affects pollen and seeds. Alternatively, it is possible that independent mechanisms (and genetic loci) cause sex-specific TRD, such as pollen competition in males (e.g., Fishman ) and meiotic drive in females (e.g., Fishman and Saunders 2008). Whatever the cause, overtransmission of hms1G is apparently exacerbated by M. nasutus alleles at hms2 to the point of overwhelming the effects of the hms1G; hms2N gametic incompatibility. Indeed, the direction of TRD in the backcross progeny of hms1GN; hms2NN ILs is counterintuitive: because of the hms1G; hms2N gametic incompatibility, one expects transmission bias to be toward M. nasutus alleles. Instead, we observed exactly the opposite, namely, strong transmission bias toward M. guttatus at hms1. This finding might help explain < 50% of pollen inviability in ILs with the genotype hms1GN; hms2NN. If hms1G alleles are highly overrepresented in pollen of such individuals due to gamete killing or some other mechanism, the gametic incompatibility will be expressed more often than expected under Mendelian inheritance. However, to explain the bias toward M. guttatus alleles in the backcross progeny, the gamete-killing phenotype has to be stronger than the gametic incompatibility. In other words, some fraction of hms1G; hms2N gametes must survive, and in greater numbers than hms1N; hms2N gametes, to form zygotes. Clarifying the role of hms2 in hms1 TRD, and whether it acts through the diploid sporophyte or haploid gametophyte, will be an important step toward understanding the mechanistic basis of hybrid distortion.

Surprisingly, our crossing experiments revealed at least two additional hybrid incompatibility loci linked to hms1. These loci, which contribute to TRD in the IL-Ns, appear to cause hybrid inviability and involve recessive alleles from both Mimulus species; against an M. nasutus genetic background, the hms1 region cannot be made homozygous for M. guttatus alleles. The precise locations of these hybrid lethality loci are not yet known (Figure 3), but both potentially overlap with the 320 kb haplotype associated with the hms1 incompatibility allele (Sweigart and Flagel 2015). This nearly invariant haplotype, which includes 30 genes, has recently risen to intermediate frequency in the Iron Mountain population of M. guttatus. The fact that multiple hybrid incompatibility loci are associated with this sweeping haplotype suggests that natural selection within a single population might have profound consequences for reproductive isolation between Mimulus species.

Implications for the evolution of hybrid sterility in Mimulus

An emerging theme in speciation genetics is that selfish evolution within species might be a major driver of hybrid incompatibilities. Decades of genetic analysis have provided a detailed mechanistic understanding of classic segregation distorters within Drosophila and mouse species (see Presgraves 2008), and more recent studies have shown that hybrid sterility and hybrid TRD can be caused by the same genes (Phadnis and Orr 2009; Zhang ). However, very few studies have directly linked these two ends of the spectrum, testing whether incompatibility alleles act as selfish genetic elements within species. In one recent exception, Case showed population genomic evidence for coevolution between a selfish cytoplasmic male sterility gene and a nuclear restorer of fertility (Rf locus) within the Iron Mountain population of M. guttatus (Case ). These same two loci also cause hybrid male sterility between M. guttatus and M. nasutus, suggesting that intragenomic conflict within Iron Mountain contributes to interspecific reproductive barriers.

Direct evidence for selfish evolution is missing from all of the hybrid gamete eliminators that have been cloned in rice (Long ; Kubo , 2016a,b; Yang ; Yu ). In most of these hybrid sterility systems, patterns of molecular variation at the causal genes in japonica, indica, and their wild ancestor O. rufipogon suggest that hybrid incompatibility alleles may never have expressed their killing phenotypes within species [e.g., Long ; Yang ; also see Sweigart and Willis (2012­)]. In plants, it is also important to consider that even if gamete eliminators do arise within species and evolve selfishly to bias their own transmission, they might do so without any cost to individual fitness (Rick 1966). Especially for pollen killers, a sufficient number of viable pollen grains might still remain to fertilize all available ovules. Under a scenario of selfish evolution with no fitness costs, there is no conflict and, thus, no mechanism for generating hybrid incompatibilities.

Despite evidence for a recent selective sweep of the hms1-associated haplotype in the Iron Mountain population (Sweigart and Flagel 2015), our crossing experiments suggest there is no transmission bias favoring the IM62 hms1 incompatibility allele. One caveat to this finding is that TRD at hms1 might vary in different genetic backgrounds; even if there is no transmission bias between the IM62 and IM767 hms1 alleles, TRD might occur in other heterozygous combinations. Alternatively, Iron Mountain individuals, including IM62 and IM762, might carry suppressors at hms2. However, given the recentness of the hms1-associated sweep (i.e., ∼63 generations old; Sweigart and Flagel 2015), it seems unlikely that there has been sufficient time for a suppressor to evolve. Instead, M. guttatus from Iron Mountain and elsewhere may carry a “permissive” allele at hms2 that allowed the evolution of the IM62 hms1 variant without it expressing any transmission bias or sterility. Consistent with this idea, the incompatibility allele at hms2 seems to be specific to M. nasutus (Sweigart ), indicating this species likely carries the derived allele. Thus, instead of being driven by selfish evolution within M. guttatus, it appears that TRD at hms1 is limited only to hybrids. These findings leave open the possibility that hms1 evolution within Iron Mountain may have been driven by ecological adaptation. Further molecular characterization of these hybrid incompatibility loci and direct investigations of the fitness effects of alternative alleles at hms1 will be important steps toward identifying the evolutionary causes of this reproductive barrier.

Supplementary Material

Supplemental material is available online at www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300148/-/DC1.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.