Introgression and Parental Conflict Shape Repeated Occurrences of Postzygotic Isolation in Mimulus

Summary

Postzygotic reproductive isolation is often thought to accumulate as a byproduct of neutral divergence. Yet, it frequently evolves rapidly, in line with non-neutral evolution. A major driver of intrinsic postzygotic reproductive barriers are intragenomic conflicts, such as conflict between maternal and paternal interests in resource allocation to offspring (i.e. parental conflict). Parental conflict may underlie hybrid seed inviability, a common and rapidly evolving reproductive barrier in angiosperms. Nevertheless, in closely related, hybridizing species, it remains unclear how intragenomic conflicts and introgression interact to determine the fate of incompatibility alleles in nature. Here, we explore repeated incidences of hybrid seed inviability in a rising model: the Mimulus guttatus species complex. Using an extensive, range-wide crossing survey, we discover patterns of hybrid seed inviability within the widespread M. guttatus that are better described by geography than phylogeny. These patterns of reproductive isolation transgress M. guttatus species boundaries, as geographically-proximate but phylogenetically-distant species also exhibit similar patterns of hybrid seed inviability with allopatric populations of M. guttatus. We find strong support that patterns of reproductive isolation are consistent with parental conflict. Lastly, we provide evidence that introgression may underlie shared patterns of hybrid seed inviability between two species within this complex. Such introgression could have led to cascading reproductive isolation with other closely related species, creating a complex landscape of incompatibility. Overall, this work suggests that parental conflict and introgression can interact to shape the rapid and repeated evolution of strong reproductive isolation in the wild.

Graphical Abstract

eTOC Blurb

Frayer, Soliman et al. demonstrate that range-wide patterns of hybrid seed inviability between populations of a wildflower species complex are consistent with parental conflict. They further show that in at least one species pair, introgression may underlie these shared patterns of reproductive isolation.

Introduction

The evolution of intrinsic postzygotic isolation is often thought of as a slow accumulation of neutral differences^1–3, yet these barriers can evolve rapidly and dynamically throughout divergence, suggesting a non-neutral evolutionary history^4–8. Intrinsic incompatibilities are common among diverse taxa^4,9–14, indicating that these barriers are fundamental in speciation. Moreover, their study can reveal how closely related taxa diverge in basic developmental processes, such as gametogenesis or offspring development^{9,10,13,15–19}.

Hybrid incompatibility can evolve rapidly via intragenomic conflicts, such as parental conflict^20–27. Parental conflict arises when unequal relatedness among offspring in a brood leads to opposing parental optima for resource allocation^20,21,28,29. Under this theory, selection should favor paternal alleles that draw more resources from the ovule-parent to their offspring, while maternal alleles arise to combat that excess via resource repression^28,29. This parental “tug of war” can lead to an evolutionary arms race wherein sets of maternal and paternal alleles become coadapted within a given population or species³⁰. Parental conflict manifests most significantly in taxa with specialized tissues that facilitate nutrient exchange from ovule-parents to offspring, namely the endosperm in angiosperms and the placenta in eutherians²⁶. Mechanistically, this could be achieved through genomic imprinting, a specialized form of gene regulation that has independently evolved in angiosperms and eutherians, wherein genes exhibit parent-of-origin biased expression²¹. Parental conflict theory posits that within lineages, maternally- and paternally-acting alleles should be evenly matched, resulting in typical offspring development. However, crosses among lineages with different histories of conflict can reveal selfish alleles with such parent-of-origin effects, leading to dysfunctional offspring development and potentially death^23,25,31,32. Several testable predictions arise from parental conflict theory: First, reciprocal hybrids should exhibit parent-of-origin specific growth abnormalities. Namely, F₁s sired by a lineage that has experienced a longer or more intense history of conflict should exhibit an overgrowth phenotype known as “paternal excess”, while the reciprocal cross should exhibit an undergrowth phenotype (i.e. “maternal excess”)^21,30. We refer to these as high- and low-conflict lineages, and note that these definitions align with high- and low-endosperm balance number, a similar conceptual framework used in crop breeding^33,34. Second, if parent-of-origin growth effects cause hybrid inviability, size asymmetries should correlate with inviability³⁵. Third, incompatibility should involve tissues that facilitate nutrient exchange among maternal parents and offspring²⁶. Last, crosses that restore the balance of maternally- and paternally-acting alleles (i.e. crosses between lineages with similar histories of conflict) should be compatible, even if their histories of conflict are evolutionarily independent³⁵. This creates a final testable prediction that the outcome of crosses should be predictable based on inferred histories of conflict.

Hybrid seed inviability is common across angiosperms^10,35–47. Overgrowth phenotypes often involve delayed cell division or cellularization in the endosperm, both of which tend to be lethal^{35,37,40–42,46,48,48}. Conversely, undergrowth phenotypes result from accelerated endosperm development^{12,35,37,40,42,46}. Furthermore, hybrid seed inviability can evolve rapidly²⁵ including in taxa as recently diverged as 10,000 years¹². The observation that hybrid seed inviability has evolved rapidly and independently across diverse angiosperms, with consistent patterns of dysfunctional seed development, indicates that it is highly repeatable at the phenotypic level. However, it remains unclear if this repeatability extends to the genetic level, and what the source of any shared genetic variation might be. These questions have fundamental implications for evolution, revealing how much of the genome is susceptible to genetic conflict and how quickly these alleles can arise and spread.

Introgression can promote the spread of alleles among populations^49,50. While typically thought to hinder speciation, introgression can contribute to reproductive isolation by introducing novel ecologically-mediated reproductive barriers^51,52. Introgression from one species into another can also enable reproductive isolation with a third species^53,54. Theory based on meiotic drivers predicts that once a selfish-element invades a new population, it should spread^55,56, and empirical work on introgressed selfish-elements supports this prediction^55–57. However, it is unclear under what conditions an intrinsic incompatibility that has evolved via intragenomic conflicts could persist as a reproductive barrier. Furthermore, if and how theory developed for meiotic drivers translates to an allele evolving via parental conflict remains unknown.

Here, we leverage the Mimulus guttatus species complex to understand the repeated occurrence of hybrid seed inviability. We identify several instances of hybrid seed inviability across geographically and/or phylogenetically distinct taxa. We test whether these patterns of crossability conform to the prediction of parental conflict. We then assess whether the genetic basis of hybrid seed inviability between two incidences is shared, and whether they have a history of introgression. Our results support a scenario in which introgression of selfish elements led to the repeated evolution of reproductive isolation between lineages.

Results

Repeated occurrence of hybrid seed inviability

The Mimulus guttatus species complex is a recent radiation of monkeyflowers endemic to western North America^58,59 (Figure 1A,B; crown age estimate: 674kya⁴⁴). This species complex includes several uniquely named species, as well as the widespread M. guttatus, which is paraphyletic and comprises two major clades⁶⁰ (‘Northern’ and ‘Southern’; Figure 1A). We first performed a range-wide crossing survey of 23 populations of M. guttatus, including both the Northern and Southern clades, in a nearly full diallel design (426/506 unique directions, 1,742 crosses), and identified strong and asymmetric patterns of hybrid seed inviability between geographically distinct populations of M. guttatus (Figure 1B,C; Table S1). Crosses between populations from the Sierra mountains/foothills and coastal or northern parts of the range were largely inviable, producing flat seeds that did not germinate when the Sierran populations served as the sire. In contrast, the reciprocal crosses produced viable seeds with high germination rates (Figure 1C; Figure S1; Sierran x Northern M. guttatus: χ²=1427, DF=3, p<0.0001; Sierran x Coastal M. guttatus: χ²=2724, DF=3, p<0.0001). There was no seed inviability in crosses between populations of M. guttatus from the north and coast (herein “Northern” and “Coastal”) or within the Sierran lines—these crosses formed plump, viable seeds with high germination rates (Figure S1; Northern x Coastal M. guttatus: χ²=6.88, DF=3, p=0.08). Despite broad consistency, we found some population-specificity in the magnitude of hybrid seed inviability. Specifically, two Northern M. guttatus accessions exhibited more viable seeds with Sierran lines and fewer with other Northern lines (i.e. CRES and TOK; Figure 1C).

Figure 1. Hybrid seed inviability segregates within the widespread M. guttatus species complex.

A) A maximum likelihood phylogeny of several lineages within the M. guttatus species complex with lines used in the crossing survey denoted by colored shapes at the tips. Inferences of the history of conflict that a subset of these taxa have experienced are denoted by the inner greyscale ring. These inferences are based on the current work, as well as that of others³⁵. The outer colored ring denotes named species or lineages of M. guttatus and M. decorus that are inferred to have different histories of conflict. B) The geographic distribution of the lines used in the crossing survey. Samples are colored according to species/lineage. C) Proportion of viable seeds for pairwise crosses survey between each of the 25 populations. Grey boxes denote incomplete crosses. See also Figure S1 and Table S1.

The discovery that populations of M. guttatus from the Sierra mountains and foothills exhibited a consistent crossing barrier with populations from elsewhere in the range is surprising, because the Sierran M. guttatus themselves do not form a monophyletic group. Instead, they span both major clades on our phylogeny (Figure 1A), with some individuals being more closely related to Northern M. guttatus and others being more closely related to Southern M. guttatus (which includes Coastal M. guttatus). Crossability was more strongly predicted by geographic distance than phylogenetic distance, when considering only crosses involving this Sierran lineage (geographic distance: F=44.7, DF=1, p<0.00001, phylogenetic distance: F=5.9, DF=1, p=0.01). This finding highlights a novel incidence of hybrid seed inviability segregating within geographically distinct populations of M. guttatus, albeit with some population-specificity in the magnitude of incompatibility.

Given the geographic pattern of incompatibility, we next tested whether two additional Sierran endemic species (M. glaucescens and M. corallinus) also exhibited this incompatibility. We crossed two populations of M. glaucescens to 21 populations of M. guttatus from our original crossing survey. Both Sierran M. guttatus and M. glaucescens behaved similarly in crosses with either Northern or Coastal M. guttatus, although seed inviability was more severe in crosses sired by M. glaucescens (Figure 1C). While crosses between Sierran M. guttatus and M. glaucescens produced viable seeds, there was a small but significant decline in viable seed production when M. glaucescens sired the cross (~6–7%; χ²=36.4, DF=3, p<0.001; Figure S1). We then crossed another, more distantly related Sierran endemic, M. corallinus to a single population of Northern M. guttatus. We again found strong, asymmetric hybrid seed inviability, wherein seeds were inviable when M. corallinus sired these crosses but were viable when M. corallinus served as the dam (Figure 2B; χ²=68, DF=3, p<0.0001). Mimulus glaucescens and M. corallinus are distantly related in the species complex, with M. glaucescens belonging to the Northern clade and M. corallinus belonging to the Southern clade (Figure 1A), reinforcing that this novel incidence of hybrid seed inviability is not defined by the phylogeny, but by geography.

Figure 2. Evidence for parental conflict.

Crosses of Sierran M. guttatus (SG), M. corallinus (C), and Southern M. decorus (SD) to Northern M. guttatus (NG; columns). A-C) The proportion of viable seeds for reciprocal crosses. D-F) The seed width for all reciprocal crosses. G-I) The result of germination assays (solid outline) and embryo rescued 8 days after pollination (8 DAP; dotted outline) for reciprocal crosses. In plots A-F, letters indicate groupings of crosses with significant differences. In plots G-I, pairs where embryo rescue significantly differs from the control are denoted by an asterisk. See also Figure S2.

These patterns of hybrid seed inviability mirror those previously described in the M. guttatus species complex involving a phylogenetically-distinct and allopatric species: Southern M. decorus³⁵ (Figure 1A). Crosses between Southern M. decorus and a subset of 9 Northern and Coastal M. guttatus lines used in the crossing survey above reveal a broadly consistent pattern, although the pattern here is more symmetric than previously reported³⁵ (Figure 2A–C). In all cases, hybrid seeds are more likely to be inviable when M. corallinus, M. glaucescens, Sierran M. guttatus, or Southern M. decorus serve as the sire in crosses with Coastal or Northern M. guttatus. Although lines of Coastal and Northern M. guttatus did vary in the severity of incompatibility (Figure 1C, Figure S1), patterns of crossability between any given accession of M. guttatus with Sierran M. guttatus or Southern M. decorus were highly correlated (i.e. lines that exhibit more symmetric patterns of incompatibility with Sierran M. guttatus also show more symmetric patterns of incompatibility with Southern M. decorus; r²=0.79, DF=13, p=0.0005; Figure S2). We next sought to test whether patterns of repeated hybrid seed inviability conformed to the predictions of parental conflict.

The role of parental conflict in hybrid seed inviability

Parental conflict predicts continuous antagonistic selection between maternal- and paternal-interests in resource allocation to offspring²¹. To investigate whether this novel incidence of hybrid seed inviability may be caused by parental conflict, we next tested the four predictions outlined in the introduction. First, under parental conflict, we predict that F₁ seeds should differ in size^20,21,61. Consistent with this prediction, all incompatible crosses exhibited parent-of-origin growth effects in reciprocal F₁ seeds, wherein seeds sired by M. corallinus, M. glaucescens, Sierran M. guttatus, or Southern M. decorus were larger than the reciprocal cross, indicative of a paternal excess phenotype (Figure 2D–F; Figure S1).

Second, if parent of origin effects on growth were causing hybrid seeds to die, then the extent of asymmetry in reciprocal F₁ seed size should be correlated with seed viability³⁵. Indeed, we find that the level of asymmetry in F₁ seed size is highly correlated with asymmetries in F₁ seed viability (r²=−0.712, DF=168, p<0.0001) and germination proclivity (r²=−0.559, DF=168, p<0.0001).

Third, under parental conflict, disrupted seed development should manifest in the endosperm²⁰. Therefore, F₁ viability should be rescued if embryos are grown on a nutrient-rich media. We performed embryo rescues for a subset of crosses between Northern M. guttatus and each of Sierran M. guttatus, M. corallinus, and Southern M. decorus. While we qualitatively found that germination increased in rescued seeds in four out of six crosses (Figure 2 G–I), only the M. corallinus x Northern M. guttatus cross was statistically significant (Figure 2H). Despite a lack of statistical significance, we were able to rescue crosses that otherwise do not germinate. Moreover, several within-species embryo rescues showed reduced germination (Figure 2G–I), suggesting that the rescue process itself can be damaging. Therefore, while embryo rescues do not perfectly restore germination, the endosperm likely does play an important role in this incompatibility.

Lastly, parental conflict theory dictates that the outcome of crosses should be predictable based on the histories of conflict, even if phylogenetically distant. Based on the observation of paternal excess when Sierran M. guttatus, M. glaucescens, or M. corallinus were sires, we infer that they have experienced a stronger history of conflict than Coastal or Northern lineages of M. guttatus. We tested this inference by crossing these lineages with each other, as well as with a species that had previously been identified as having a stronger history of conflict (Southern M. decorus). We predict that these species should all be compatible with one another, despite not being closely related (Figure 1A). In line with this hypothesis, crosses among all high-conflict lineages (i.e. Southern M. decorus, M. corallinus, M. glaucescens, and Sierran M. guttatus) yielded viable seeds with no parent-of-origin size differences (Figure 3A,B; Figure S3). Conversely, crosses between high-conflict lineages (i.e. M. corallinus and Sierran M. guttatus) and a species that has previously been shown to exhibit the lowest history of conflict of all taxa described herein (i.e. Northern M. decorus³⁵), exhibit reciprocal and nearly complete inviability (Figure 3C,D). This suggests that high-conflict lineages behave in predictable ways: inviability is most severe between lineages with the most disparate histories of conflict, and viability is recovered between lineages with more similar histories of conflict, despite varying divergence times. However, this prediction may also be true if incompatibility loci have introgressed, a possibility we explore below.

Figure 3. Cross compatibility is predictable based on inferred histories of conflict.

High-conflict lineage Southern M. decorus is reciprocally crossed to A) Sierran M. guttatus and C) M. corallinus, and no hybrid seed inviability is observed. The low-conflict lineage, Northern M. decorus, is reciprocally crossed to B) Sierran M. guttatus and D) M. corallinus, and nearly complete hybrid seed inviability is observed. Notably, it has been previously shown that the closely related Southern and Northern M. decorus have complete hybrid seed inviability when crossed to each other³⁵. Letters denote cross directions with significant differences. See also Figure S3.

Repeated incidences of hybrid seed inviability have a shared genetic basis

We next sought to understand if the shared patterns of cross incompatibility were caused by a shared genetic basis. To test this hypothesis, we used a complementation test, where two lines exhibiting a similar phenotype are hybridized, and the presence or absence of phenotypic segregation in F₂ hybrids can reveal whether the genetic basis of the focal trait is shared. Using this logic, we created an F₂ population between two lineages that both exhibit complete hybrid seed inviability when acting as the sire in a cross with Northern M. guttatus—M. corallinus and Southern M. decorus (Figure 2). We crossed a single inbred line of Northern M. guttatus (IM62) to M. corallinus, Southern M. decorus, their reciprocal F₁s, and F₂s, with Northern M. guttatus acting as the dam in all crosses (Figure 4A). If the genetic basis of hybrid seed inviability was largely independent, F₂s should segregate for alleles associated with hybrid seed inviability, and thus some fraction of F₂s would be at least partially compatible with Northern M. guttatus. By contrast, if the genetic basis of hybrid seed inviability was shared, then no genetic variation would segregate among an F₂ population and all F₂ individuals should be fully incompatible with Northern M. guttatus. Out of 761 crosses (>41,000 seeds, 170 unique F₂ individuals), we did not recover a single viable hybrid seed, strongly suggesting that the genetic basis of hybrid seed inviability between Northern M. guttatus and each of M. corallinus and Southern M. decorus is shared (Figure 4A; Figure S4).

Figure 4. The shared genetic basis of hybrid seed inviability between M. corallinus and Southern M. decorus may be due to a history of introgression.

A) A diagram of the crossing design for the complementation test. B) The complementation test supports the shared genetic basis of hybrid seed inviability between Northern M. guttatus and each of M. corallinus and Southern M. decorus. C) PCA of Northern M. decorus, Southern M. decorus, and M. corallinus. D) PCAdmix results for these three lineages. K=2 was the best supported ancestry grouping E) HyDe estimates that 18.2% of the genome of Southern M. decorus comes from M. corallinus. See also Figure S4.

High-conflict lineages exhibit a history of introgression

The shared genetic basis of a trait can be caused by independent mutations in the same or tightly linked gene(s), shared ancestral polymorphisms, or introgression. Given the substantial history of gene flow in this species complex^60,62–65, we next assessed patterns of introgression. Using four-population tests, we find extensive evidence of introgression across the complex, as the minimum value of D is significant in 83% of trios (mean D statistic=0.059; M. caespitosa as the outgroup), some of which is caused by ancient introgression among ancestors of extant lineages (Figure S5). Given this extensive introgression, disentangling its role in shaping the distribution of hybrid seed inviability is not straightforward, particularly in currently sympatric taxa.

However, there is a particularly strong signal of introgression between two high-conflict lineages— M. corallinus and Southern M. decorus. These lineages exhibit replicated patterns of hybrid seed inviability with other taxa in the complex (i.e. with both Northern M. guttatus and Northern M. decorus; Figure 2,3³⁵), our complementation test above supports a shared genetic basis between these two lineages for inviability with Northern M. guttatus, and they are currently allopatric, potentially allowing for more straightforward tests of introgression. We therefore focused on investigating the history of introgression between Southern M. decorus and M. corallinus.

We performed several tests using only Northern M. decorus, Southern M. decorus, and M. corallinus, with M. caespitosa as an outgroup. While we sequenced 17 Southern M. decorus individuals, we found a high degree of relatedness within populations. We therefore limited most genomic analyses to six unrelated individuals. In a PCA, Southern M. decorus is intermediate to Northern M. decorus and M. corallinus along PC1, which explains 26.67% of the genomic variance (Figure 4B). Multiple lines of evidence support a high fraction of M. corallinus ancestry segregating within Southern M. decorus at a genome-wide level. PCAdmix indicates that there are two ancestries among these three taxa, with Southern M. decorus being ~29% M. corallinus and 71% Northern M. decorus (Figure 4C). Additionally, we estimated a global admixture proportion of 18.2% M. corallinus with HyDe (Figure 4D). Lastly, D and f4 gave consistent estimates of admixture (0.204 and 0.255, respectively). In total, these patterns support a consistent and significant contribution of M. corallinus ancestry into Southern M. decorus.

To characterize local patterns of introgression in Southern M. decorus we identified windows exhibiting high f_dM, indicating putative introgression between Southern M. decorus and M. corallinus. Furthermore, given that the Northern and Southern lineages of M. decorus are closely related³⁵, we should expect relatively low values of F_ST between them, except at regions of the genome that have introgressed between M. corallinus and Southern M. decorus. Such regions should exhibit high F_ST between both Northern M. decorus and each of M. corallinus and Southern M. decorus, but low F_ST between M. corallinus and Southern M. decorus. Using these criteria, we find signals of introgression across the genome, typically in small blocks (Figure 5A). These results were consistent with topology weighting, wherein we found support for a phylogeny that grouped M. corallinus and Southern M. decorus at low levels across the genome, but little support for the phylogeny that grouped M. corallinus and Northern M. decorus, in line with a scenario of introgression rather than incomplete lineage sorting (Figure S5A). We used ancestryHMM to infer local ancestry within Southern M. decorus (Figure S5B,C,G–I), and to estimate the timing of admixture at 4,703 generations (Figure S5I). Ancestry from M. corallinus is correlated with f_dM (r²=0.511; p<2.2e-16; Figure S5D), but reveals high levels of individual ancestry heterozygosity, suggesting that M. corallinus alleles are segregating in Southern M. decorus populations.

Figure 5. The Southern M. decorus genome is a genomic mosaic of Northern M. decorus and M. corallinus.

A) Distribution of f_dM values measured in 250 SNP windows. Light green dots are in the top 1% of f_dM values. Dark green dots represent the windows that also overlap with the top 1% of windows with low F_ST between M. corallinus and Southern M. decorus and high F_ST between Northern and Southern M. decorus. B) The putative inversion on chromosome 13. The black line represents the smoothed average of F_ST between M. corallinus and Southern M. decorus minus F_ST between Northern and Southern M. decorus. Orange blocks indicate the location of genes with known roles in imprinting or endosperm expression: (1) Napin, (2) MSI homolog, (3) DOG1, and (4) FIE homolog. C) A PCA of SNPs within the inversion region places most Southern M. decorus samples intermediate to Northern M. decorus and M. corallinus, except for DL5, which falls closer to M. corallinus. D) Each unrelated Southern M. decorus sample is colored by the most likely ancestry at each position, determined by AncestryHMM. DL5 is largely homozygous for M. corallinus ancestry, while the remaining unrelated samples are largely heterozygous. See also Figure S5 and Table S2.

The distal end of chromosome 13 exhibits a striking signal of introgression (Figure 5B). This region harbors a chromosomal inversion known to be distinct between Northern M. decorus and M. guttatus⁶⁸ (Joint Genome Institute, pers. comm.). It has previously been implicated in hybrid seed inviability between M. guttatus and M. tilingii⁶⁹, and preliminary mapping data supports its role in hybrid seed inviability between Northern M. decorus and M. guttatus (Coughlan unpublished data). This inversion contains multiple genes that are potential targets of parental conflict (Table S2). Two of these genes are known to be expressed in the M. guttatus endosperm⁷⁰, although it is currently unknown whether either are imprinted in Mimulus. These genes are Mimulus homologs of Napin, a storage protein found in maturing seeds⁷¹, and DOG1, which plays an important role in seed dormancy⁷². There are also two genes that are imprinted in Arabidopsis, homologs of MSI and FIE. FIE is also a master regulator of imprinting in Arabidopsis endosperm^73–77.

We hypothesized that this inversion has introgressed from M. corallinus into Southern M. decorus and may play an important role in the nearly complete hybrid seed inviability between the two sister lineages of M. decorus. However, the role of this inversion is not straightforward and cannot explain all patterns of hybrid seed inviability across the M. guttatus species complex as M. corallinus and many other M. guttatus lineages share the same orientation, yet also exhibit hybrid seed inviability⁶⁸. Additionally, the inversion is heterozygous in 15/17 Southern M. decorus samples (5/6 unrelated samples; Figure 5C,D), with the remaining individuals carrying two copies of the M. corallinus orientation. This is perplexing, as segregation of the inversion in the offspring of Southern M. decorus should result in segregating hybrid seed inviability–a pattern that we do not see. However, these patterns may be explained if the genetic basis of hybrid seed inviability between the two lineages of M. decorus is sufficiently redundant (i.e. composed of several large affect alleles with redundant effects). Further genetic mapping will be required to explicitly link this chromosomal inversion with hybrid seed inviability between the two lineages of M. decorus.

Discussion

We show that hybrid seed inviability is widespread and has evolved rapidly and repeatedly in the Mimulus guttatus species complex. Hybrid seed inviability segregates among the widespread M. guttatus, as well as several other species in the group, yet patterns of hybrid seed inviability are more strongly associated with geography than phylogeny. Crossing patterns within this group also conform to predictions of parental conflict theory: crosses exhibit canonical parent-of-origin effects on growth, asymmetries in size are correlated with inviability, incompatibility is at least partially linked with the nutritive endosperm, and patterns of crossing are better predicted by inferred history of conflict than phylogenetic relatedness. Lastly, two instances of hybrid seed inviability appear to have a shared genetic basis, and these two taxa also exhibit a substantial history of introgression. Although more research is needed to explicitly identify genes involved in hybrid seed inviability and assess their histories of introgression, our results highlight the potential for intragenomic conflicts to drive the rapid evolution of incompatibility alleles, and for introgression to underlie the invasion of incompatibility alleles into new species. We posit that introgression of selfish alleles could then cause cascading reproductive isolation between populations of what was once the same species.

Parental conflict as a driver of hybrid seed inviability

Our data are consistent with the parental conflict theory, which posits an ongoing arms race between paternally-acting, resource-acquiring alleles and maternally-acting, resource-repressive alleles within populations. When populations with different histories of conflict hybridize, incompatibilities between these alleles are revealed. The strength of conflict within a population can be influenced by several demographic and life history factors, particularly mating system^37,42,78. However, all taxa in this study are likely outcrossing, based on their floral morphologies and levels of genetic diversity, yet exhibit signatures of different histories of parental conflict. Demographic differences that influence the number and/or relatedness of effective fathers in a population may also influence the strength of parental conflict (see^35,79). Additionally, it is also possible that ecological differences between the Sierras and other parts of the range mediate cost/benefit ratio of maternal investment, in turn influencing the maternal response to paternal excess (although we do not see consistent differences in final seed size among Sierran and non-Sierran species which may result from ecologically-mediated selection in seed investment; Figure 2D–F; Figure S1). In this particular group, significant work on the biogeographic and demographic histories will be required to better understand why histories of conflict appear to be so variable on such short timescales.

Patterns of hybrid seed inviability in the M. guttatus species complex conform to predictions of parental conflict. First, a hallmark phenotype of parental conflict is parent-of-origin growth defects in F₁ hybrids, which we observe in all crosses. Second, the extent of parent-of-origin growth effects are highly correlated with seed viability. Third, parental conflict theory predicts that incompatibilities should manifest in the endosperm. Therefore, it should be possible to “rescue” embryos by providing them with an external source of nutrition. Our embryo rescue results are less conclusive than the patterns in seed size, but they trend in the expected direction, and we were able to rescue hybrid embryos in crosses that typically do not germinate. Moreover, the procedure itself may have negatively influenced final germination rates, as exemplified in our control embryo rescues (which in some cases are significantly worse than the regular germination assay; Figure 2). Lastly, inferred histories of conflict are highly predictive of the outcome of future crosses. Despite being more distantly related, all Sierran lineages were entirely compatible with one another, as well as with high-conflict Southern M. decorus. By contrast, Sierran M. guttatus and M. corallinus were completely incompatible with the lower-conflict lineage, Northern M. decorus.

The introgressed chromosome 13 inversion also supports a role for parental conflict. Several genes located in or near our top f_dM peaks are expressed in the Mimulus endosperm⁸⁰ (e.g., DOG1 and Napin) or imprinted in Arabidopsis and contribute to parent-of-origin effects when disrupted in that species^73,75 (e.g., FIE and MSI). In addition to being imprinted themselves, both FIE and MSI function as part of the polycomb repressive complex, which establishes the histone modifications needed for imprinting and proper endosperm development^77,81. Moreover, this inversion has been implicated in hybrid seed inviability in at least two crosses (Coastal M. guttatus x M. tilingii⁶⁹ and Northern M. guttatus x Northern M. decorus; Coughlan unpublished data). While some questions remain (i.e. excess levels of heterozygosity within the inversion), we hypothesize that introgression of this inversion from M. corallinus into what is now Southern M. decorus led to cascading reproductive isolation with what is now Northern M. decorus. Explicit tests of its role in generating hybrid seed inviability between Southern M. decorus and M. guttatus will require additional genetic mapping.

Our work adds to a growing body of evidence for parental conflict across diverse taxa^{35,46,47,82–88}. Widespread, asymmetric hybrid seed inviability are not only consistent with other studies in diverse angiosperms, but also mirror the parent-of-origin phenotypes observed in mammals, where the placenta performs an analogous function to the endosperm^{10,12,26,35,37,40–43,46,48,89–94}. These parallels suggest that parental conflict could be a common driver of early-onset hybrid embryonic death, especially in cases where isolation is rapidly evolving and/or actively segregating, as is the case in the M. guttatus species complex.

A polymorphic incompatibility sorted by geography

One striking finding from our work is that hybrid seed inviability segregates within the widespread M. guttatus. Polymorphic incompatibilities are common in nature^6,95–98, and may represent incompatibilities en route to fixation (or loss⁹⁹). Yet, it is often unclear how such alleles would avoid removal from the population, especially when rare^100–102. If the alleles themselves are selfish, such as targets of parental conflict, the selective advantage conferred by these alleles may aid in overcoming their deleterious epistatic effects^55,56,103. Theory and empirical work has demonstrated the ability of selfish elements such as meiotic drivers to spread despite deleterious effects to the host^55,103–106, but an equivalent body of work is lacking for alleles evolving via parental conflict.

These patterns of hybrid seed inviability do not align with inferred phylogenetic relationships presented here or alternative topologies presented elsewhere¹⁰⁷. All Sierran M. guttatus have a similar crossing pattern and are compatible with one another, yet they are not a monophyletic group (Figure 1A). Although the M. guttatus species complex inhabits a broad range^58,59, the Sierras represent a unique geographic location where two diverged clades meet (i.e. Northern and Southern M. guttatus⁶⁰), and is home to several endemic species^{60,63,108–110}. Although more phylogeographic work is needed, one compelling hypothesis is that the Sierras may have contained unique refugia during the last glacial maximum, allowing for population divergence and the accumulation of incompatibility alleles. As ranges expanded, such incompatibility alleles may have spread to nearby populations in the Sierra foothills. In line with this hypothesis, we find excessive allele sharing between the Northern and Southern clades of M. guttatus in the Sierra mountains and foothills (Figure S5). Similar population histories within the Sierras are known in other species¹¹¹. Alternatively, incompatibility alleles may predate the split of the Northern and Southern clades, and may have independently fixed in the Sierra mountains and foothills. Given that the entire complex exhibits significant allele sharing^60,62–64, distinguishing patterns of incomplete lineage sorting from introgression will require a much deeper understanding of the relationship among populations in the Sierras and beyond (though see^60,112). Despite this, one clear incidence of introgression is that of M. corallinus into the now allopatric Southern M. decorus.

Introgression as a source of incompatibilities

Our work supports a shared genetic basis of hybrid seed inviability and a history of introgression between M. corallinus and Southern M. decorus. Given the low divergence between Southern M. decorus and its sister species, Northern M. decorus, we infer that this introgression may have led to cascading reproductive isolation between the two M. decorus lineages.

Many questions remain about this case of introgression. First, the current geographic distance between M. decorus and M. corallinus suggests that introgression is not ongoing, but it is unclear how and where this introgression originally occurred. The distribution of the M. guttatus species complex is likely a result of post-glaciation recolonization⁶⁰, which may have facilitated hybridization. Our estimate of the timing of admixture corresponds to a period of geological dynamism in the southern Cascades, wherein many of the Cascade Lakes were formed (4,703 generations; 4–14kya, assuming a 1–3 year generation time¹¹³). It is also possible that the source of introgression was a close relative of M. corallinus that was not sampled or is no longer extant. Indeed, regions that we identified as fixed for M. corallinus ancestry within Southern M. decorus are not identical (D_xy=0.017; F_ST=0.007). Still, even with the high diversity of this species complex⁶⁸, our data support M. corallinus as the most likely source of the introgression (Figure S5).

Introgression may be an underappreciated contributor to reproductive isolation. Although often viewed as homogenizing¹¹⁴, introgression can generate novel biodiversity^{51,52,115,116}. In between these extremes, introgression may move incompatibility alleles between populations^54,117. In the case of Northern and Southern M. decorus presented in this study, introgressed alleles from M. corallinus may have provided an advantage in parental competition to isolated populations of M. decorus, leading to the origin of the Southern lineage. We have previously shown that seeds developing alongside hybrids from low-conflict fathers are larger than seeds growing beside full siblings¹¹⁸, suggesting that at the level of individual alleles, selfish solicitation of maternal resources confers a direct benefit to the individual seed (and a direct cost to sibling seeds). However, how small increases in seed size translate into fitness in the wild remains unknown. Additional theory will be needed to understand the conditions under which such introgression is likely to be advantageous²⁷. Few incidences of introgressed incompatibilities have been reported^54,119, but given the ubiquity of introgression across eukaryotes^49,50, they may be more prevalent than previously thought.

Resource Availability

Lead Contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dr. Megan Frayer (megan.frayer@yale.edu).

Materials availability

Seeds from field collections and experimental crosses are available by request to the lead contact.

Data Availability

• Sequence data have been deposited at the NCBI Sequence Read Archive as BioProject PRJNA1333461 and are publicly available as of the date of publication.

• Data from crosses have been deposited at Dryad (https://doi.org/10.5061/dryad.xsj3tx9tr) and are publicly available as of the date of publication.

• All original code has been deposited at Zenodo (https://doi.org/10.5281/zenodo.18556668) and is publicly available as of the date of publication.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

Experimental Model and Study Participant Details

The Mimulus guttatus species complex is a group of very closely related wildflowers native to the western North America^120–122. Like many species complexes, the taxonomy of this rapidly radiating group is fairly complex. At current, the M. guttatus species complex contains several named species, many of which are fairly geographically restricted (i.e. M. corallinus, M. glaucescens, M. laciniatus, M. decorus), as well as the widespread, and ecologically and genetically diverse M. guttatus. Importantly, M. guttatus is paraphyletic and comprises two major clades (Northern and Southern), and all other species in the complex are nested within these clades (i.e. Figure 1A), potentially serving as a model for budding speciation^63,108,123. In addition to members of the M. guttatus species complex, we use the closely related species Mimulus caespitosa as an outgroup for many of our genomic analyses.

Plant Rearing

For all experiments, seeds were cold stratified at 4°C for 5 days before being transferred to long days in the University of North Carolina greenhouses (16hrs light, 22°C). Follow up crosses, including the complementation test and plants reared for embryo rescues, were completed under the same conditions at the Yale University greenhouses. For all crosses, the buds were emasculated 1–2 days before opening to prevent autogamous self-fertilization, and pollen from a freshly opened flower of the sire was applied to a receptive stigma of the dam.

Method Details

Crossing surveys

To assess hybrid seed inviability within the widespread M. guttatus, we grew 3–5 individuals from 1 inbred line/maternal family for each of 23 populations of M. guttatus, which span a large part of the species’ geographic range (Figure 1B; Table S1). Upon flowering, we crossed all accessions to each other in a nearly full diallel design, totalling 426 unique cross directions (out of a possible 506), with a mean of 4 replicate crosses per combination (range: 1–12, totalling 1,742 crosses). Based on the results of these crosses, we performed follow-up crossing surveys that incorporated other species within the complex: Mimulus decorus, M. corallinus, and M. glaucescens. Previously, we found that M. decorus was two genetically unique lineages, Northern and Southern, that vary in their ability to cross with M. guttatus, putatively via differences in the strength of parental conflict within each lineage³⁵. Therefore, we crossed 12 accessions of M. guttatus (which spanned the observed variation in crossing pattern in the initial survey) to 4 populations of Southern M. decorus (high-conflict lineage; mean of 4 replicate crosses/direction, 730 crosses total), and 3 populations of Northern M. decorus (low-conflict lineage; mean of 3 replicate crosses/direction, 256 crosses total). For M. glaucescens, we crossed individuals from 2 populations to 21 M. guttatus populations used in the original survey, as well as 2 populations of Northern M. decorus and 4 populations of Southern M. decorus (mean of 4 replicate crosses, 458 crosses total). For M. corallinus, we crossed 3 accessions of M. corallinus to one accession of each Northern M. guttatus, Sierran M. guttatus, Northern M. decorus, and Southern M. decorus (3 replicate crosses/direction, 95 crosses total).

Estimation of seed viability, germination, and size

We estimated seed viability in two ways. First, we counted all seeds formed per fruit and characterized viability based on morphology (e.g., shriveled/flat corresponding to inviable, or round corresponding to viable). To confirm germination proclivity, we performed a germination assay where 10 seeds from each unique cross combination were plated on 0.6% agar, cold stratified for 1 week, then placed in a growth chamber under long-day conditions (18hrs light, 21/18°C day/night) at Yale University. We seeded 3 replicated plates per unique cross. Plates were monitored weekly for 4 weeks for the emergence of a radicle, which was deemed a germinant. These two measurements of viability are highly correlated (r² = 0.81, DF = 532, p < 0.0001), but offer different insights; while one characterizes abnormal phenotypes, the other directly assesses death. To determine seed size, we photographed seeds on white paper with a Nikon D3500 DSLR digital camera and a size standard. We measured the width of 25 seeds for each fruit manually using ImageJ¹²⁴.

Embryo Rescue

We further wanted to determine if hybrid seed inviability was caused by defects in the endosperm rather than the embryo itself. To test this, we performed an embryo rescue experiment where early developing seeds were grown on Murashige and Skoog media supplemented with 4% sucrose to functionally compensate for the endosperm by providing external nutrition needed for growth. Under long days at the Yale University greenhouses, we grew four individuals from a single inbred line representing each of the four species used in this assay: Northern M. guttatus (IM62), Sierran M. guttatus (CCC9), M. corallinus (SILF), and Southern M. decorus (OD11). Given that all incidences of hybrid seed inviability reported here involve Northern M. guttatus, we crossed Northern M. guttatus reciprocally with the other three species and included intraspecific crosses as controls. For each cross direction, we performed three replicate crosses and dissected the ovaries eight days after pollination (DAP). Ten seeds per cross were placed on Petri dishes with media, sealed with 3M micropore tape, then incubated in growth chambers under long-day conditions (18 hrs light, 22/20°C day/night). Seeds were scored for germination weekly for four weeks.

Complementation test

In this study, we report multiple incidences of hybrid seed inviability and shared patterns of incompatibility across phylogenetically and geographically distinct taxa. Given that we find the repeated occurrence of high-conflict lineages within the M. guttatus complex, we next sought to understand if the genetic basis of the incompatibility was shared among incidences. To do so, we performed a complementation test between two geographically distinct taxa that showed similar patterns of hybrid seed inviability when crossed with Northern M. guttatus- Southern M. decorus and M. corallinus. We first crossed these two species reciprocally, then self-fertilized F₁s from each cross direction to create reciprocal F₂ populations (Figure 4). We predict that if the genetic basis of hybrid seed inviability with Northern M. guttatus was largely shared between M. corallinus and Southern M. decorus, all F₂s should produce inviable seeds when acting as the sire in crosses with Northern M. guttatus, just as we see for both M. corallinus and Southern M. decorus. By contrast, if the genetic basis was independent, then F₂s should segregate for their crossability with Northern M. guttatus, resulting in some fraction of F₂s that produced viable hybrid seeds. To test these predictions, we planted five replicates of each parent, along with 15 F₁s from each cross direction, and roughly 100 F₂s from each cross direction. By including both reciprocal directions of F₁s (with M. corallinus acting as both the dam and sire) and F₂s (with M. corallinus acting as both the grand-dam and grand-sire), we aim to control for any differences in parent-of-origin effects within this cross. After omitting individuals that did not flower, we were left with three Southern M. decorus, four M. corallinus, 15 Southern M. decorus x M. corallinus F₁s, 12 M. corallinus x Southern M. decorus F₁s, and 170 F₂s (83 with Southern M. decorus as the grand-dam and 87 with M. corallinus as the grand-dam). We crossed each experimental plant with an inbred line of Northern M. guttatus (IM62), with IM62 serving as the dam and each experimental plant as the sire for all crosses. We performed an average of three replicate crosses per plant (range: 1–10), totaling 761 crosses. Based on our previous results, hybrid seed inviability should manifest as large, flat, disc-like seeds. We assessed the percentage of viable seed per fruit.

Genomic Analyses

In order to contextualize the patterns observed above with the underlying genetic relationships between lineages, we used whole genome sequences from the lines used in our experiments, combined with numerous new and publicly available sequences. We sequenced 99 new accessions and combined this with 73 publicly available accessions (Table S1). For new sequences, we extracted DNA from bud and leaf tissue using the ThermoFisher GeneJET Plant Genomic DNA kit. Illumina libraries were prepped and sequenced using NovaSeq6000 by the Yale Center for Genome Analysis. For all accessions, we cleaned reads using fastp v0.23.2 (using option “-q 20”; Chen et al. 2018) and aligned to the IM62v3 reference genome⁶⁸ using the BWA mem algorithm v0.7.17¹²⁶. We removed duplicates using Picard v2.25.6 (options “VALIDATION_STRINGENCY=LENIENT”; http://broadinstitute.github.io/picard). We called variants using bcftools mpileup and call (V1.16^127,128), and filtered for quality and depth (-e “FMT/DP<4 | FMT/GQ<20 | QUAL<40”) using bcftools v1.16.

Quantification and Statistical Analysis

Estimation of seed viability, germination, and size

To assess cross compatibility and parent-of-origin differences in seed size, we ran a series of linear mixed models using the lme4 package in R¹²⁹, where either the proportion of viable seeds or seed width was the response variable, cross type was a fixed effect, and the identity of the maternal and paternal populations were each random effects. In both sets of analyses, cross type contained four levels (within-lineage for each parent and each reciprocal F₁). For each model, we then assessed differences among cross types with a type III Anova using the car package and identified significant comparisons using contrasts from the emmeans package in R^130,131.

Additionally, to assess whether patterns of hybrid seed inviability found in the Sierras were best described by the geographic and/or phylogenetic distance of the parental populations, we constructed a linear model with the average proportion of viable seeds across all replicates and cross directions as the response variable and the pairwise phylogenetic distance (calculated using the R package ape¹³²) and geographic distance (calculated as the Haversine distance) between accessions/sampling localities, respectively as fixed effects. We then assessed the significance of fixed effects with a type III Anova using the car package, as above. For these analyses, we limited the crossing data to any crosses involving Sierran M. guttatus and M. glaucescens, as we explicitly wanted to describe patterns of crossability that involved this Sierran lineage.

Lastly, to assess whether asymmetries in hybrid seed size were correlated with viability, we first calculated the relative difference between reciprocal F₁s in both seed size and viability as ((Species 1 × Species 2) - (Species 2 × Species 1)) / ((Species 1 × Species 2) + (Species 2 × Species 1)), where the values of these reciprocal crosses is the average seed size, proportion viable seed per cross, or average germination rate³⁵. Using this equation, values close to 0 denote F₁s with highly symmetric values of seed size or viability between crosses.

Embryo Rescue

For each cross direction, we tested for significant difference between the germination assay and 8 DAP embryo rescue using unpaired non-parametric Wilcoxon rank-sum test in R.

Complementation Test

We performed a simplified power analysis to assess under what genetic conditions we would be able to detect segregating phenotypes among our F₂s. For all models, we assumed x unlinked loci with equivalent and additive effect sizes, and assessed the expected fraction of F₂ individuals that would exhibit complete inviability with M. guttatus. Here, x varied from 1 to 10 loci, and we simulated four levels of effect size: each locus having a 100% lethal effect, 50% lethality, 25% lethality, or 10% lethality. We then performed a Fisher’s Exact test between our observed counts of individuals that produced no viable seeds (versus those that produced at least some viable seeds) with what would be expected under each set of genetic conditions. We find that, except under the most extreme parameters (i.e. 6 or more loci, each of which exhibits 100% lethality or 9 or more loci each of which exhibits 50% lethality; Figure S4), we have the power to identify a lack of segregation among F₂s. We note that these are quite conservative estimates, as other studies have found that hybrid seed inviability typically involves a smaller number of paternally acting alleles (i.e. 3–6), each of which reduces viability by 20–50%^37,41,69 (Coughlan, unpublished).

Genomic Analyses

We estimated relatedness among samples using king (v 2.3.2¹³³). Next, we constructed a phylogeny for all lineages using IQTree2¹³⁴. We removed individuals with less than 3X coverage, as well as those suspected to be early generation hybrids, leaving 158 sequences. We used only biallelic sites and sites with a minimum minor allele frequency of 0.05. We selected the best substitution model, TVM+F+ASC+R6, using ModelFinder¹³⁵, and then we used that model to infer a consensus tree using 1000 bootstraps (Figure 1A).

In order to broadly assess diversity, divergence, and introgression among these lineages, we calculated D and f4 statistics using Dsuite⁶⁷. We split samples into their respective lineages by combining field identification with the inferred relationships from the phylogeny, as well as PCAdmix run using all samples. We ran all combinations of M. guttatus species complex lineages with Dtrios using a more distantly related and allopatric species–M. caespitosa–as an outgroup. We compared the arrangement of the trios that minimized the value of D to the arrangement predicted by the phylogeny generated above. We also ran Dtrios constrained by the phylogeny, and used these results to obtain f-branch values⁶⁷.

We further compared the genomes of the two high-conflict lineages for which we had performed the complementation test (M. corallinus and Southern M. decorus), as well as the low-conflict sister species of Southern M. decorus: Northern M. decorus. We used PCANGSD¹³⁶ to perform PCA and PCAdmix for global ancestry grouping. Based on these initial findings, we found a striking pattern of ancestry sharing between the two high-conflict lineages. To investigate this shared ancestry further, we constructed explicit tests of introgression using Dsuite and calculated admixture proportion for southern M. decorus using HyDe⁶⁶, using standard settings with M. caespitosa as an outgroup.

In order to investigate patterns of introgression along the genome, we used Dinvestigate to calculate f_dM in sliding windows of 250 SNPs with a 50 SNP step. We selected fourfold degenerate sites using degenotate¹³⁷, and used these to calculate F_ST with Pixy¹³⁸ using the same sliding windows in which f_dM was calculated. We further investigated regions of the genome that appear introgressed by estimating topology weighting along the genome using TWISST¹³⁹. TWISST input files were made using the genomics_general scripts available from Simon Martin (https://github.com/simonhmartin/genomics_general), as recommended by the TWISST manual, again using M. caespitosa as an outgroup.

To understand the individual-level patterns of introgression, we inferred local ancestry within our Southern M. decorus genomes using ancestryHMM¹⁴⁰. We used all M. corallinus and Northern M. decorus individuals to calculate population allele frequencies for the parental populations using custom scripts. To assign genetic distances between loci, we used a linkage map generated from crosses between Northern M. decorus and M. guttatus (Schwarz, unpublished); for sites in between SNPs on the linkage map, we assumed a linear increase in genetic position with physical position. We used a starting ancestry proportion of 20% M. corallinus ancestry, and allowed the timing of that ancestry to be inferred by the program. We compared three different sets of SNPs: 1) all sites that had variation in the parental populations, 2) fixed differences between the parents, and 3) sites with an allele frequency difference between the parents greater than 50%. We only included SNPs that had no missing data in the parental samples (Figure S5G–I). We focus on the results from using all variable sites as these analyses yielded the most consistent results with other genomic analyses.

Finally, our M. corallinus samples displayed significant population structure, with four individuals forming a distinct group in the phylogeny and PCA. Therefore, we repeated all of our analyses with and without those individuals. There was not a substantial difference in the results, so we present only the results using all M. corallinus individuals here.

Additional Resources

There are no additional resources to report for this manuscript.

Supplementary Material

1

Table S1. Accessions used for crossings and genomic analyses, Related to STAR Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Sucrose	Fisher	Catalog number BP220-1
Murashige and Skoog media	MP	Catalog number 2623220
Powder Micropropagation type II	Caisson	Catalog number A037-1KG
Bacto Agar	BD Biosciences	Catalog number 214010
Critical commercial assays
GeneJET Plant Genomic DNA kit	ThermoFisher	Catalog number K0791
Deposited data
Phenotypic data from crossing survey	This study	Dryad DOI: 10.5061/dryad.xsj3tx9tr
Resequenced genomes	This study; Coughlan et al.35, Sandstedt et al.44, Brandvain et al.62, Ivey et al.63, Garner et al.69, Coughlan et al.110, Mower et al.141, Hellsten et al.142, Puzey et al.143, Colicchio, Hamm et al.144	BioProject PRJNA1333461; See Supplemental Table 1 for all accession numbers
Experimental models: Organisms/strains
Mimulus seeds	Collected by J. M. C.	N/A
Software and algorithms
fastp	Chen et al.125	v0.23.2; https://github.com/OpenGene/fastp
BWA mem algorithm	Li and Durbin126	v0.7.17; https://github.com/lh3/bwa
Picard	http://broadinstitute.github.io/picard	v2.25.6; https://broadinstitute.github.io/picard/index.html
bcftools	https://samtools.github.io/bcftools/howtos/install.html	v1.16; https://github.com/samtools/bcftools
R: lme4	Bates et al.129	N/A
R: emmeans	Fox and Weisberg130	N/A
R. car	Lenth131
R: ape	Paradis et al.132	N/A
king	Manichaikul et al.133	v2.3.2; https://www.kingrelatedness.com/
IQTree2	Minh et al.134	v2.4.0; https://iqtree.github.io/
Dsuite	Malinsky et al.67	https://github.com/millanek/Dsuite
PCANGSD	Meisner and Albrechtsen136	https://github.com/Rosemeis/pcangsd
HyDe	Blischak et al.66	v1.0.2; https://hybridizationdetection.readthedocs.io/
degenotate	Mirchandani et al.137	https://github.com/harvardinformatics/degenotate
Pixy	Korunes and Samuk138	v1.2.7; https://pixy.readthedocs.io/en/latest/
TWISST and genomics scripts	Martin and Van Belleghem139	https://github.com/simonhmartin/twisst; https://github.com/simonhmartin/genomics_general
ancestryHMM	Corbett-Detig and Nielsen140	https://github.com/russcd/Ancestry_HMM

Highlights.

• Hybrid seed inviability is widespread in the Mimulus guttatus species complex

• Patterns of hybrid seed inviability are better explained by geography than phylogeny

• Hybrid seed inviability may evolve selfishly via parental conflict

• Introgression may have facilitated the movement of such selfish alleles among species