Inferences about the underlying mechanisms of craniofacial traits from F2 hybrid P-matrices
In addition to making inferences about how variation in P-matrices may contribute to diversification in the wild, this study also sought to investigate the underlying mechanisms associated with these traits on SSI by investigating the similarities and differences in P-matrices and phenotype values of F2 hybrids from two separate crosses. We found that 6 of 18 traits met the null expectations of additivity in both populations (Figure 7), suggesting that the mechanisms underlying these traits generally fit the simple models often used in quantitative genetics which assume Mendelian segregation, independent assortment, and the presences of additive genetic effects (i.e., no dominance or epistatic effects; Falconer 1996; Roff 1997). Of the remaining 12 traits, we found that only orbit diameter failed to meet additive expectations in both crosses. Crescent Pond and Little Lake F2 hybrids both exhibited larger mean estimates of this trait than additive expectations, a pattern which could be produced through dominance or epistatic effects. Ten of the remaining traits met null expectations of additivity in Little Lake but not in Crescent Pond, and one trait, maxillary head protrusion, met expectations in Crescent Pond but not in Little Lake. This pattern could again be driven by dominance or epistatic effects which are likely different between ponds, however, it may also be driven by other factors such as new mutations in F2 hybrids, recessive alleles, lethality of specific allele combinations (i.e., Dobzhansky-Muller or constitutive incompatibilities), or gene drive (Rick and Smith 1953; Burkart-Waco et al. 2012; Chevin et al. 2014; Andersen and Rockman 2022). There is indeed some evidence for this in our current dataset. For example, many traits, especially in Little Lake, have much higher levels of variation in phenotype trait values and exhibit phenotypes that fall well outside those observed in the F0 parental types. Little Lake also shows higher levels of covariation between traits than we observed in Crescent Pond. The increased variation in trait values in both ponds may imply that new combinations of alleles or epistatic effects are responsible for this additional variation, while the high levels of covariation in Little Lake may suggest that these traits share a common mechanism that is not present in the Crescent Pond cross.
Increased variation and covariation within and between hybrid traits is commonly observed in nature, and there is even evidence that hybridization may drive speciation events in many systems including pupfish (Seehausen 2004; Bell and Travis 2005; Selz et al. 2014; Richards et al. 2019, 2021). On the other hand, F2 hybrids in Crescent Pond do not display phenotype value that meet or exceed the scale-eater parental values for traits such as palatine height, lower jaw length and dentigerous arm width (Figure 7), implying that dominance, segregation distortion, or lethality may be at play. While lethal combinations of alleles is possible it is more commonly expected to occur when species undergone divergence for longer periods of time than what is observed in the pupfish system (Coyne and Orr 1997). Therefore, future work needs to 1) identify the loci associated with these traits and 2) investigate whether these loci suffer from segregation distortion as a first step to untangling the underlying mechanisms of these traits.