Focal Populations
The goal of this study was to compare the P matrices of allopatric
generalist pupfish populations to sympatric populations of pupfish
containing both generalist and specialist species, and to estimate the
P-matrices of F2 hybrid pupfish to make inferences about the underlying
mechanisms of craniofacial traits on SSI. To that end, we measured and
compared craniofacial traits of fish from: five Caribbean populations of
pupfish that contained only generalist species (Lake Cunningham (New
Providence Island, Bahamas), Flamingo Pond (Acklins, Bahamas), Lake
George (Rum Cay Island, Bahamas), Etang Saumatre (Dominican Republic),
and Laguna Oviedo (Dominican Republic)); six SSI populations that do not
contain the full radiation of pupfishes, and instead contain only or
mostly generalist species (Wild Dilly Pond, Reckley Field Pond, Pain
Pond, Moon Rock Pond, Six Pack Pond, and Mermaid Pond); and three SSI
populations of pupfish that contained the full radiation of pupfishes in
sympatry (Crescent Pond, Little Lake, and Oyster Pond), although only
generalist and scale-eater specimens were available for measurement from
Oyster Pond (Table 1; Figure 1). A portion of the specimens from the
above ponds were initially collected, measured, and analyzed in a
previously published article (Martin, 2016), and complete collection
details can be found there. Additional justification and information
about the use and categorization of these populations can be found in
appendix A. For simplicity, we collectively refer to the five Caribbean
populations of pupfish as the ‘Caribbean’ group/population (N = 61), the
six populations that do not contain the full pupfish radiation as the
‘generalist-only’ group/population (N=85), and the three populations
containing all three pupfish species in sympatry as the ‘radiating’
group/population (N=42).
To make inferences about the potential mechanisms underlying
craniofacial traits on SSI, we measured traits of F2 scale-eater X
snail-eater hybrid offspring and estimated phenotypic variation and
covariation of these traits. We produced F2 hybrids by first crossing a
single male snail-eater with a single female scale-eater to produce F1
offspring. We repeated this process independently for the Crescent Pond
and Little Lake populations. At least four F1 offspring from each
population were then allowed to interbreed to produce F2 hybrids. In
total we produced and measured 301 F2 hybrids from Crescent Pond and 194
hybrids from Little Lake (Table 1). These measurements were used for
separate QTL mapping study of these crosses (St. John et al. 2021).
When comparing F2 hybrid phenotypes and P matrices we assume a simple
additive model of inheritance (Falconer 1996; Roff 1997). The
assumptions of this model include: 1) that there are two alleles per
locus, 2) that mendelian laws of segregation are adhered to, 3) that
loci across the genome are in linkage equilibrium, and 4) that there are
only additive genetic effects (i.e., no dominance or epistatic effects).
With these assumptions we expect that the distribution of phenotypic
traits in the F2 generation should follow a 1:2:1 ratio, where
intermediate phenotypes are most common and phenotypes aligning with
either parental phenotype are less common and equally distributed, and
we expect covariation between traits to correspond to recombination
events, which for the sake of simplicity we assume to uniformly
distributed across the genome. To investigate deviations from these
assumptions we estimate expectations of additivity by calculating the
average of parental traits from the F0 generation and calculate
expectations of variation within traits using the parental trait values
and hybrid population sample sizes (Appendix E: Tables E1, E2, & E3).