Which trait relationships may promote or inhibit diversification on San Salvador Island
Within the above phylogenetic context, parsimony predicts that trait relationships in the SSI radiating group which diverge from both Caribbean and SSI generalist-only groups may be associated with diversification. Indeed, we find that dentigerous arm width, lower jaw length, palatine height, and maxilla length—traits that have high correlations with one another in our PCA— display either increasing or decreasing strength in their relationships across pupfish groups which supports this inference. To easily illustrate this pattern, we visualize the R2 values from linear models with the covariation values of dentigerous arm width, lower jaw length, palatine height, or maxilla length as the response variable and the covariation values of all other traits as the predictor variable and display the relationships which may promote or inhibit divergence on SSI (Appendix Figure D3). For example, dentigerous arm width, lower jaw length, palatine height, and maxilla length all have higher associations with suspensorium length and ectopterygoid in the Caribbean, weakened but intermediate relationships in the SSI generalist-only group, and the weakest relationships in the SSI radiating group. This pattern could suggest that increased independence between these traits occurred on initial colonization of SSI and further shifted during the radiating process. On the other hand, all four traits have a weak relationship with maxillary head protrusion in the Caribbean and stronger relationships on SSI.
The SSI generalist group also exhibits unique relationships between dentigerous arm width, lower jaw length, palatine height, and maxilla length and other traits that are not present in either the Caribbean or SSI radiating group. We see a stronger relationship with other jaw apparatus traits such as maxillary head height and the jaw closing and opening in-levers in SSI generalist groups, but also see stronger relationship with traits associated with general body shape and size such as pelvic girdle length, orbit diameter, head depth, and cranial height (Appendix Figure D3). We also see a stronger association between maxilla length, palatine height, and nasal tissue protrusion in the Caribbean and SSI radiation that is weakened in the SSI generalist-only group.
Dentigerous arm width, lower jaw length, palatine height, and maxilla length are traits with clear connections to feeding kinematics across many fish species (Hulsey and García De León 2005; Westneat 2005; Muñoz et al. 2018; DeLaurier 2019). Interpreting the patterns of associations with these traits can provide us with potential modules which may be important for feeding performance in SSI radiating groups and which explain the lack of diversification in SSI generalist-only populations. Four-bar linkages have been previously used to describe the sequence of steps involved in fish feeding kinematics (Grubich & Westneat, 2006; Westneat 1990, 2005; Wainwright et al., 2004; Hulsey & Wainwright, 2002), and our measurements of dentigerous arm width, lower jaw length, maxilla length, and palatine height correspond to portions of the oral four bar system. The adductor mandibulae complex also connects to the premaxilla, maxilla and the dentary via several buccopalatal ligaments aiding in the control of the jaw important for suction feeding (Hernandez et al. 2009; Datovo and Vari 2013). Similarly, the interoperculomandibular ligament is involved in opening the jaw during the first stage of feeding and rotation around the joint of the lower jaw and quadrate contributes to increasing gape opening during feeding (Martin et al., 2017; Westneat, 2005). The similar patterns of covariance between these traits along with their strong relationship to one another suggests they may form a module in pupfishes. While all these traits seem to act together across groups, the SSI generalist-only group is the only one that displays a strong relationship between this module and other traits such as head depth, maxillary head height, and jaw opening in-lever. This could mean that these additional relationships are constraining the jaw apparatus in SSI generalist-only groups and preventing them from diversifying into additional ecological niches.
Overall, we observed that the P matrix of generalist-only group on SSI was distinct, and generally more constrained, than those of the Caribbean and SSI radiating groups, a pattern that could be explained by several mechanisms: First this pattern may be due to founder effects (Barton & Charlesworth, 1984). For example, if the founding population of pupfish which first populated Wild Dilly Pond, Reckley Field Pond, Pain Pond, Moon Rock Pond, Six Pack Pond, and Mermaid Pond on SSI had more constrained P matrices than those that populated Crescent Pond and Little Lake, then available variation may have been insufficient to produce the scale-eating and/or snail-eating phenotypes. Alternatively, the generalist-only ponds may have once contained the full adaptive radiation, but the specialist species went extinct. We therefore may be observing differences in the P matrix attributes due to the Baldwin effect. The Baldwin effect predicts an initial burst of plasticity when an environment changes, as in when pupfish first colonized SSI, and subsequently predicts that plasticity will decrease as selection acts on the population to bring them to a stable fitness optimum (Baldwin, 1896, 1902; Crispo, 2007). For instance, if selection in Wild Dilly Pond, Reckley Field Pond, Pain Pond, Moon Rock Pond, Six Pack Pond, and Mermaid Pond populations led to divergence into generalist, snail-eater and scale-eater species and plasticity was then lost, an extinction event of the specialists would leave the generalist species alone in the pond with reduced plasticity. It should be noted that the Baldwin effect can occur without species divergence, and the loss of plasticity may occur following selection in the new environment. Further work needs to be done to understand which scenario is most likely for this system.