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.

is that the initial mutations are selectively neutral and become fixed by random genetic 42 drift. Alternatively, the mutations might increase in frequency because they benefit the 43 native species for reasons that are incidental to their role in reproductive isolation -by 44 promoting ecological adaptation, for example (Schluter and Conte 2009). Yet another 45 possibility is that hybrid incompatibilities arise through recurrent bouts of intragenomic 46 conflict within species (Frank 1991;Hurst and Pomiankowski 1991). In this last scenario, 47 selfish genetic elements (e.g., transposons, meiotic drivers, gamete killers) manipulate 48 host reproduction to bias their own transmission. Because these actions are often 49 detrimental to host fitness, there is then selective pressure for compensatory mutations or 50 suppressors to neutralize the effects of selfish evolution (Burt and Trivers 2006). 51 The idea that intragenomic conflict involving segregation distorters might be a 52 major source of hybrid incompatibilities has resurged in recent years (Johnson 2010;53 McDermott and Noor 2010; Presgraves 2010; Crespi and Nosil 2013), largely due to 54 influential studies in Drosophila that have mapped hybrid segregation distortion and 55 hybrid sterility to the same genomic locations (Tao et al. 2001; Phadnis and Orr 2009b; 56 10 in one or both of the backcrosses using the IL as the paternal parent, but not as the 251 maternal parent. If, instead, female meiotic drive and/or a female gametic incompatibility 252 occurs at these hms loci, we would expect to see TRD in both backcrosses with the IL as 253 the seed parent, but not with the IL as the pollen parent. Finally, if TRD is caused by the 254 loss of diploid zygotes (or seedlings), it should be apparent in both reciprocal crosses to 255 the same recurrent parent (i.e., regardless of the gender of the IL). For all crosses, the 256 female parent was emasculated 1-2 days before hand-pollination to prevent self- To determine whether transmission ratio distortion at the polymorphic hms1 266 incompatibility locus occurs between incompatible and compatible alleles from the Iron 267 Mountain population of M. guttatus, we generated reciprocal F 2 and backcrossed 268 populations with IM62 and IM767. We previously determined that the IM767 inbred line 269 carries a compatible allele at hms1 (i.e., one that does not carry the 320-kb haplotype or 270 cause sterility in combination with SF5 alleles at hms2). The IM62 and IM767 inbred 271 lines were intercrossed reciprocally and a single F 1 hybrid from each was self-fertilized to 272 form reciprocal F 2 populations (IM62 x IM767: N = 267; IM767 x IM62: N = 315). To 273 identify putative female-and male-specific sources of TRD, and to distinguish between 274 meiotic/gametic mechanisms versus zygotic selection, we generated reciprocal 275 backcrosses with IM62 and IM767. We used a single F 1 hybrid (IM62 x 767; maternal 276 parent listed first) to generate four backcross populations to the recurrent parents (F 1 -277 IM62 BC 1 , IM62-F 1 BC 1 , F 1 -IM767 BC 1 , IM767-F 1 BC 1 ). Two of these backcrosses used 278 the emasculated F 1 as the seed parent and two used the F 1 as the pollen donor in crosses 279 to the emasculated recurrent parents. 280 11 We also wanted to examine the effect of M. nasutus hms2 alleles on patterns of 281 within-M. guttatus TRD at hms1. We wondered if having M. nasutus alleles at hms2 has 282 the potential to unleash severe distortion at hms1, even in an otherwise M. guttatus 283 genetic background. To address this question, we intercrossed IM767 with a BG 4 -NIL 284 (BG 4 .275) that is heterozygous for an SF5 introgression spanning ~36% of chromosome 285 13 including hms2 (in an IM62 genetic background; Figure S2). We self-fertilized two of 286 the resulting F 1 s to generate F 2 hybrids segregating for SF5 alleles at hms2 against an 287 IM62-IM767 F 2 -like genetic background. We then genotyped at hms-linked markers 288 (M183 for hms1 and MgSTS193 for hms2) to identify IM62-IM767 hms1 heterozygotes 289 in combination with three different hms2 genotypes: 1) IM62 homozygotes, 2) IM767 290 homozygotes, or 3) SF5 homozygotes. Using each of these three genotypic classes, we 291 performed reciprocal backcrosses to IM767 ( Figure S2). 292 293

Assessment of transmission ratio distortion 294
To examine patterns of TRD at the hms1 and hms2 loci, we collected leaf tissue from 295 individual plants and isolated genomic DNA using a rapid extraction protocol (Cheung et 296 al. 1993) modified for 96-well format. To infer the hms1 and hms2 genotypes of hybrid 297 progeny generated from crosses between IM62 and SF5, we determined genotypes at a 298 multiplexed set of fluorescently labeled markers that flank hms1 (M8 and M24) and hms2 299 (MgSTS193 and M51) following amplification protocols used previously (Sweigart et al. markers, genotyping error rates for hms1 and hms2 were each < 1%. For experimental 303 crosses involving IM62 and IM767, only one tightly linked marker was used to infer 304 genotype at hms1 (M183). Based on expected crossovers between hms1 and M183, the 305 genotyping error rate was < 1%. All fluorescently labeled marker products were run on 306 an ABI 3730 at the University of Georgia Genomics Facility. Genotypes were scored 307 automatically using GeneMarker (SoftGenetics), with additional hand scoring when 308 necessary. We used chi-square tests with two degrees of freedom to determine if hms-309 linked genotypes were significantly distorted.  (Table 1). At hms1, we observed a significant 320 excess of heterozygotes, but allelic transmission did not differ from the Mendelian 321 expectation. The observed genotype ratios at hms1 also differed significantly from the 322 expectation given the random union of two gametes with the observed allele frequencies. frequencies. Taken together, these patterns suggest TRD at hms1 might be driven 327 primarily by zygotic selection, whereas hms2 appears to be influenced primarily by 328 selection among gametes. 329 When considered together, the two-locus genotypes at hms1 and hms2 differ 330 significantly from the Mendelian expectation (X 2 = 389.372, d.f. = 8, P < 0.0001, N = 331 5487). Although the two-locus genotypes are also significantly different from the 332 expectation given the observed allele frequencies at hms1 and hms2 shown in Table 1 (X 2 333 = 71.626, d.f. = 8, P <0.0001), the values are much more closely aligned (Table 2). 334 Particularly notable is the deficit of two genotypic classes (hms1 GG ; hms2 NN and hms1 NN ; 335 hms2 GG ) and the excess of two others (hms1 GG ; hms2 GG and hms1 NN ; hms2 NN ; Table 2). 336 This pattern of two-locus disequilibrium follows the expectation for gametic action of 337 hms1-2 sterility (i.e., with hms1 G ; hms2 N gametes tending to be sterile). However, the 338 observed F 2 transmission ratios at hms1 and hms2 cannot be entirely explained by hms1 G ; 339 hms2 N gametic sterility (Table S1). This phenomenon, whether acting through one or 340 both parents, would be expected to reduce the transmission of M. guttatus alleles at hms1, 341 in the same way that it reduces M. nasutus alleles at hms2. However, there is no 342 13 indication of allelic transmission bias at hms1 in the F 2 hybrids. Taken together, these 343 results suggest that gametic expression of the hms1-hms2 incompatibility is important, 344 but not the sole contributor, to patterns of transmission ratio distortion in F 2 hybrids. 345 346

M. nasutus-M. guttatus IL crosses reveal multiple causes of F 2 distortion 347
To investigate several possible causes of F 2 transmission ratio distortion at hms1 and 348 hms2, we performed a crossing experiment using the IL-G and IL-Ns. In this crossing 349 design (Figure 1), individuals with one of several possible two-locus hms1-hms2 350 genotypes -in each of the IL genetic backgrounds -were crossed reciprocally to M. 351 guttatus (IM62) and M. nasutus (SF5). By scoring hms1 and hms2 genotypes in the 352 progeny of these crosses, we were able to examine the effects of several factors, 353 including parental genotype, genetic background, and cross direction, on transmission 354 ratios at the two hybrid sterility loci. Of the 36 crosses performed, 12 showed significant 355 transmission ratio distortion at hms1 and/or hms2 (Table 3; note that two crosses were 356 unsuccessful due to hybrid male sterility). For both hms1 and hms2, parental genotype at 357 one locus has a strong effect on allelic transmission at the other (hms1 affects hms2: F = 358 37.69, P < 0.0001; hms2 affects hms1: F = 7.80, P = 0.004; Figure S1). For hms2, cross 359 direction is also important, with stronger TRD occurring through pollen (F = 72.33, P < 360 0.0001). Neither the genetic background nor the identity of the recurrent parent 361 significantly affected transmission ratios at hms1 or hms2 (results not shown).  (Table 4). In 371 these crosses, the hms1 G ; hms2 N gamete type is under-transmitted through both sexes, 372 14 though the effect is stronger through males. Under-transmission is also more severe in 373 crosses to IM62 (M. guttatus) and against the IL-N genetic background (Table S2). 374 If the hms1-hms2 incompatibility acts through gametes, we might expect patterns 375 of pollen viability to predict rates of transmission ratio distortion through males. To 376 examine this possibility, we measured pollen viability in various two-locus genotypes of 377 the IL-G and IL-Ns (Table 5) guttatus allele at hms2 should be present in 78% of progeny when this individual is used 382 as the paternal parent in a cross (which is close to the observed frequency of 86%, Table  383 3). Similarly, for IL-Gs that are hms1 GN ; hms2 GN , if we assume that all hms1 G ; hms2 N 384 gametes are inviable (and divide the remaining 7% sterility equally among the other three 385 two-locus genotypes), we expect M. guttatus allele frequencies of 33% and 66% at hms1 386 and hms2, respectively. These values are very similar to what we observe when this IL-G 387 genotype is backcrossed to M. guttatus (37% and 67%, Table 3). 388 At hms1, TRD is more complex. On the one hand, M. guttatus alleles at hms1 are 389 under-transmitted due to the hms1 G ; hms2 N gametic sterility discussed above (Table S2). 390 On the other hand, in many of the IL-backcrosses, M. guttatus alleles at hms1 are 391 overrepresented among the progeny (Tables 2 and 2.1). This effect is most pronounced 392 when the IL parent is heterozygous at hms1 and homozygous for M. nasutus alleles at 393 hms2 ( Figure S1; note that this genotype is not completely sterile so crosses can still be 394 performed). Remarkably, this direction of TRD is exactly the opposite of what is 395 expected if hms1 transmission is primarily influenced by the hms1 G ; hms2 N gametic 396 incompatibility. Moreover, pollen viability in IL-G and IL-Ns with the genotype hms1 GN ; 397 hms2 NN is much lower than the 50% expected for gametic expression of hybrid male 398 sterility (  Figure S1). Consistent with this idea, backcross 403 15 progeny of doubly heterozygous ILs are most often products of the hms1 G ; hms2 G gamete 404 type (Table 4). 405 Additionally, a genetically distinct hybrid incompatibility appears to affect 406 transmission of hms1 against an M. nasutus genetic background. Self-fertilization of a 407 doubly heterozygous IL-N individual produces no M. guttatus homozygotes at the hms1 408 locus (Table 2) Table 2, Table 4). 425 To fully account for observed genotype frequencies in the IL-N F 2 , it is also necessary to 426 assume complete lethality of M. guttatus homozygotes at hms1 (Table 2; note that this 427 hybrid lethality is not reflected in IL backcross allele frequencies because progeny do not 428 carry the requisite M. nasutus genetic background for expression of the incompatibility). 429 In summary, we have identified at least three sources of hms1-hms2 TRD in M. In previous (Sweigart and Flagel 2015) and ongoing efforts to fine-map hms1 and hms2, 440 we identified a small subset of SF5-IM62 F 2 hybrids that were recombinant for one or 441 both sets of hms flanking markers. With the goal of genetically mapping TRD in both 442 regions, we self-fertilized these recombinants to generate F 3 progeny and examined 443 genotype frequencies at both sets of flanking markers (Figures 3 and 4). We reasoned that 444 TRD in the F 3 progeny should only be observable if the causal locus is heterozygous in 445 the F 2 parent. If, instead, the TRD-causing locus is homozygous (for either M. guttatus or 446 M. nasutus alleles), loci in the adjacent heterozygous region should segregate in a 447

Mendelian fashion. 448
As in the IL crosses, patterns of hms2-linked TRD were consistent with the action 449 of hms1 G ; hms2 N gametic sterility. In this genomic region, the most extreme TRD 450 occurred in the two F 3 families that descended from F 2 hybrids with the hms1 GG ; hms2 GN 451 genotype (Figure 2). Despite this general support for hms1-hms2 gametic sterility, hms2-452 linked TRD could not be unambiguously mapped to a particular genomic region (no 453 interval in Figure 2 is perfectly associated with presence/absence of TRD). Presumably, 454 genetic background in these F 2 hybrids can mask TRD associated with hms1 G ; hms2 N 455 gametic sterility (e.g., 28_22) or mimic it (e.g., 02_66). 456 At hms1, the two contributors to TRD were decoupled in F 2 recombinants with M. To investigate whether hms1-linked TRD is a strictly hybrid phenomenon or also occurs 471 within M. guttatus, we generated reciprocal F 2 progeny between IM62 and IM767. These  Figure S2). Indeed, extreme TRD at hms1 (i.e., bias 499 toward the IM62 allele > 70%) was only observed in the backcross progeny of one 500 individual (08_60) that was also homozygous for M. nasutus alleles at hms2 (Table 6). 501 These results suggest that over-transmission of the IM62 allele at hms1, which appears to 502 require M. nasutus alleles at hms2, may occur exclusively in hybrids. interpretation of the finding that pollen viability is reduced from the F 1 to F 2 generation, 528 which seemed to suggest a diploid (sporophytic) genetic basis for the hms1-hms2 529 incompatibility (Sweigart et al. 2006). In general, for a hybrid incompatibility that affects 530 the gametophyte, sterility is expected to be less severe in the F 2 generation due to the 531 inviability of recombinant F 1 gametes and regeneration of parental combinations. 532 However, in this case, it appears that removal of hms1 G ; hms2 N F 1 gametes is somewhat 533 balanced by over-transmission of M. guttatus alleles at hms1. Moreover, incomplete 534 penetrance of F 1 hybrid gametic sterility (i.e., some hms1 G ; hms2 N gametes do contribute 535 to the F 2 generation, see Table 4) produces a small fraction of F 2 hybrids that are 536 completely sterile because they are homozygous for incompatible alleles (i.e., hms1 GG ; 537 hms2 NN ). to result from a two-locus hybrid incompatibility between any genes expressed in the 560 gametophyte. Additionally, the fact that the hms1-hms2 incompatibility seems to affect 561 both the male and female gametophyte (the hms1 G ; hms2 N gamete type is under-562 transmitted through both sexes) is consistent with our finding that these loci contribute to 563 both hybrid male sterility and hybrid female sterility (Sweigart et al. 2006). Gametic 564 hybrid incompatibilities that affect the fertility of both sexes have also been discovered in A key question for hms1 and hms2 is whether the same genes cause the gametic 584 incompatibility and transmission bias of M. guttatus at hms1. The latter is particularly 585 strong when hms2 is homozygous for M. nasutus alleles (Table 3, Figure S1), suggesting 586 it might be caused by an interaction between the two loci. Additionally, the presence of 587 hms2 NN also appeared to unleash severe hms1 TRD in one of the two IM62-IM767 F 2 588 populations in which it was present (Table 6), suggesting hms2 might be necessary but 589 21 not sufficient for hms1 TRD. On the other hand, over-transmission of hms1 G does not 590 seem to absolutely require hms2 NN (e.g., we observed 62% transmission of hms1 G in M. 591 nasutus x IL-G GN;GG , Table 3), which might argue against its direct involvement. Indeed, 592 for the IL-Gs, there is a bias toward hms1 G in all backcross populations except those 593 involving doubly heterozygous IL parents (i.e., hms1 GN ; hms2 GN ), which, because they 594 express the hms1 G ; hms2 N gametic inviability, might obscure additional sources of hms1 595 TRD. Going forward, additional rounds of high-resolution fine-mapping will be needed 596 to pinpoint the causal genes and determine if Mimulus hybrid sterility and TRD are 597 genetically separable. Such efforts in rice have been successful in disentangling the 598 complex phenotypic effects of linked hybrid sterility loci (e.g., (Kubo et al. 2016a). 599 Identifying the molecular genetic basis of hms1 TRD might also provide insight 600 into its mechanisms. Because the bias toward M. guttatus alleles at hms1 occurs through 601 both males and females, the simplest single explanation is a gamete killing system that 602 affects pollen and seeds. Alternatively, it is possible that independent mechanisms (and 603 genetic loci) cause sex-specific TRD, such as pollen competition in males (e.g., (Fishman 604 et al. 2008)) and meiotic drive in females (e.g., (Fishman and Saunders 2008). Whatever 605 the cause, over-transmission of hms1 G is apparently exacerbated by M. nasutus alleles at 606 hms2 to the point of overwhelming the effects of the hms1 G ; hms2 N gametic 607 incompatibility. Indeed, the direction of TRD in the backcross progeny of hms1 GN ; 608 hms2 NN ILs is counterintuitive: because of the hms1 G ; hms2 N gametic incompatibility, 609 one expects transmission bias to be toward M. nasutus alleles. Instead, we observed 610 exactly the opposite, namely, strong transmission bias toward M. guttatus at hms1. This 611 finding might help explain < 50% pollen inviability in ILs with the genotype hms1 GN ; 612 hms2 NN . If hms1 G alleles are highly overrepresented in pollen of such individuals due to 613 gamete killing or some other mechanism, the gametic incompatibility will be expressed 614 more often than expected under Mendelian inheritance. However, to explain the bias 615 toward M. guttatus alleles in the backcross progeny, the gamete killing phenotype has to 616 be stronger than the gametic incompatibility. In other words, some fraction of hms1 G ; 617 hms2 N gametes must survive -and in greater numbers than hms1 N ; hms2 N gametes -to 618 form zygotes. Clarifying the role of hms2 in hms1 TRD, and whether it acts through the 619 22 diploid sporophyte or haploid gametophyte, will be an important step toward 620 understanding the mechanistic basis of hybrid distortion. 621 Surprisingly, our crossing experiments revealed at least two additional hybrid 622 incompatibility loci linked to hms1. These loci, which contribute to TRD in the IL-Ns, 623 appear to cause hybrid inviability and involve recessive alleles from both Mimulus 624 species: against an M. nasutus genetic background, the hms1 region cannot be made 625 homozygous for M. guttatus alleles. The precise locations of these hybrid lethality loci 626 are not yet known (Figure 3), but both potentially overlap with the 320-kb haplotype 627 associated with the hms1 incompatibility allele (Sweigart and Flagel 2015 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.
**** P < 0.0001 based on X 2 tests of observed frequencies versus the Mendelian expectation with 2 d. f. for genotypes and 1 d. f. for allele frequencies.    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 standard error. 2 Percent IM62 alleles at hms1 transmitted to progeny from IM62-IM767 heterozygous parent. Value given in parentheses is the number of progeny assessed.