Predicting the occurrence of host-associated differentiation in parasitic arthropods: a quantitative literature review.
Parasite populations associated with different host species can encounter a variety of isolating reproductive barriers, leading to each population independently accumulating genome-wide genetic differences due to their host associations. This phenomenon is called host-associated differentiation (HAD) and has been proposed as an indicator of early diversification among parasitic arthropods. Although many parasite-host case study systems have been tested for the genetic signature of HAD (e.g., FST≥0.15 between sympatric, host-associated populations in the absence of allopatry), it is unknown which isolating reproductive barriers best predict the general occurrence of HAD. HAD development has been attributed to biological and ecological factors that either directly generate reproductive isolation between parasites living on different hosts, such as 'immigrant inviability' (i.e., lower fitness of immigrants in non-native environments), or that promote the accumulation of host-specific genetic adaptations, such as the gallmaking feeding mode. In fact, some of these factors are shared across multiple case studies, suggesting that the occurrence of HAD is generalizable and can be predicted based on the incidence of significant biological and ecological factors. By means of a discriminant function analysis (DFA), this research assessed 108 arthropod parasite-host case studies for ecological and biological factors significantly correlated with the occurrence of HAD and whether these factors could be used to distinguish the presence of HAD from its absence. The DFA demonstrated that case studies that developed HAD could be distinguished from case studies that did not develop HAD. The results of the DFA were corroborated by a 'non-iterative partial least squares' (NIPALS) discriminant model and a nominal logistic regression. Case studies with HAD could be robustly separated from case studies without HAD based on the incidence of these predictive factors: immigrant inviability, gallmaking, endophagy, recent range invasions of either hosts or parasites, differential host phenology, and differential parasite morphology. These results were used in an infinite random forest analysis to generate a hierarchy of conditional probabilities that separated HAD presence from absence. The results provide researchers with a tool for reliably predicting which untested parasite-host system would likely develop HAD. Immigrant inviability, gallmaking, and their combination were strongly correlated with the presence of HAD, which indicated parasite-host systems with these traits were highly likely to develop HAD. Contrary to expectation, endophagous feeding was negatively correlated with HAD presence, which indicated phytophagous endophagous feeders (excepting gallmakers) were highly unlikely to develop HAD. Furthermore, parasitoids were shown to be just as likely to develop HAD as not. Unfortunately, potentially significant predictive factors (e.g., allochrony) were excluded from analysis because too few case studies have been specifically tested for these factors. Furthermore, this analysis was biased by the lack of 'negative' publication results and the overrepresentation of research laboratories that primarily study HAD. Future research should accumulate novel HAD case studies that specifically test for allochrony, differential microbial associations, and morphological differentiation.