Habitat characteristics are partially determined by abiotic conditions and variation in these conditions are typically distributed across gradients or clines (Nowicki et al., 2017). Dispersal across gradients or clines by organisms into novel habitats presumably submits organisms to novel selection pressures (Geber and Eckhart, 2005). Directional adaptive divergence on the fringe of populations is one way that organisms can colonize novel habitats (Waterhouse et al., 2018). This perspective assumes that populations distribute themselves randomly and biogeographic distributions are mediated by deleterious genetic effects causing barriers to dispersal (Lange and Marshall, 2016). For motile organisms, behavior can play a role in maintaining biogeographic distributions through active avoidance of less hospitable habitats (Resetarits, 2005; Edelaar et al., 2008; Fišer et al., 2016). However, it is not fully understood how such behaviors evolve. Behaviors also allow organisms to respond to changes in conditions more rapidly than intragenerational acclimatization or intergenerational adaptation (Little and Seebacher, 2017). For example, reptiles maintain preferred body temperatures by behaviorally altering the thermal environment they experience, which is a much more rapid response than evolving new thermal tolerance or producing new enzyme conformations.

Shared phenotypes can occur across many organisms of variable relatedness (e.g., numerous lineages sunbathe, and sunbathing is not limited to just reptiles). A commonly-highlighted example of this is the convergent “subterranean-adapted” stygobiont phenotype [e.g., elongated appendages for sensing the environment in the absence of light, reduced or absent pigmentation to conserve energy expenditures, lower metabolic rates, and vestigial or absent eyes (Fong et al., 1995; Aden, 2005; Culver et al., 2010)]. These adaptations are observed across distantly related cave-dwelling animal lineages, and the stark phenotypic contrasts between subterranean- and related surface-adapted species (Pipan and Culver, 2012) suggests that the surface-subterranean habitat ecotone can be used to study the importance of adaptive phenotypes in maintaining population level biogeographic distributions. Selection in a habitat devoid of light is thought to be responsible for the apparent convergent morphological traits of many subterranean fauna across diverse lineages (Jeffery, 2005; Culver et al., 2010; Pipan and Culver, 2012; Carlini and Fong, 2017). Although these traits are considered adaptations to dark conditions, the same traits are presumably maladapted to conditions commonly found in epigean habitats. Lower fitness of subterranean fauna in epigean habitats is likely due to a lack of sight, pigment, and camouflage, increasing vulnerability to predation (Tobler, 2009; Horstkotte et al., 2010; Klaus and Plath, 2011) and mutagenic radiation (Protas and Patel, 2008; Bilandžija et al., 2017). Likewise, epigean species are considered maladapted to subterranean environments due to the metabolic cost of maintaining eyes in the absence of light (Moran et al., 2015), higher metabolic rates and an inability to deal with prolonged periods of starvation due to scarcity of food in resource-limited subterranean environments (Aspiras et al., 2015; Nair et al., 2020), disruption of circadian rhythms (Espinasa et al., 2016), and difficulty finding mates (Riesch et al., 2011). Thus, fitness costs associated with surface- and subsurface-adapted taxa dispersing out of their respective habitats likely mediates contemporary biogeographic distributions to some extent. However, in the case of subterranean-epigean ecotones (e.g., spring orifices, seeps), it is not known if biogeographic distributions are behaviorally mediated or if they are only maintained by suffering deleterious consequences.

The evolutionary consequences of an organism venturing out of available optimal habitats are variable and depend on phenotype (Langecker, 2000). If light is a factor driving biogeographic distributions of organisms at surface-subterranean ecotones, then it should be expected that the ability to detect and respond to light has been selected for in eyeless subterranean species that periodically encounter light (such as those occupying shallow spring or seep habitats). Fišer et al. (2016) demonstrated that epigean and subterranean species living along surface-subterranean ecotones will use light as a habitat choice mechanism, allowing surface and subterranean species to remain within their appropriate niches. In contrast, an ability to detect light should not have been selected for in subterranean species which rarely if ever encounter light (Langecker, 2000). Therefore, we hypothesized that patterns in light response would reflect habitat distributions more so than an inferred history of common ancestry, reflecting selection as a major factor determining phenotype even in the presence of standing genetic variation. We predicted that subterranean species which encounter light regularly will exhibit photophobic responses to light, thereby allowing them to avoid epigean habitats. We also predicted that species which occupy deeper aquifer habitats would exhibit little response to the presence of light (i.e., photoneutrality), because these species do not encounter light at all and are predicted to be naïve to it. In respect to surface species that encounter light daily, we predicted that exposure to light will not be stressful (i.e., photoneutrality), but they may behaviorally respond in a variety of ways (i.e., photophobia, photoneutrality, or photophilia).

We tested these predictions in a series of experiments on several closely related amphipods of three genera in the family Crangonyctidae: Crangonyx, Stygobromus, and Sicifera. This family of amphipods was selected because some species have eyes and live in epigean environments while others are eyeless, and the eyeless species occupy different microhabitats with differing light availability (Gibson et al., 2008). Also, eyeless stygobiontic amphipods in this family (i.e., Stygobromus allegheniensis) (Espinasa et al., 2016) and all the species in the stygobiontic genus Niphargus (Simčič and Brancelj, 2007; Borowsky, 2011; Fišer et al., 2016) have the capacity to detect light despite vestigial or no eyes. To assess photophilia, photoneutrality, and photophobia, we performed experiments that quantified behavioral avoidance of direct light, and physiological stress response due to exposure to light. Trait data were used in phylogenetic comparative methods to assess the evolution of these traits in response to selection imposed by different microhabitats.

This study characterized phenotype using live organisms (captured from the wild) that can be reared and reproduced in captivity, thus allowing responses to be studied in a controlled environment. Phenotypic attributes other than morphology can be the result of complex evolutionary processes that allow adaptation to novel environments without appreciable changes in morphology. These attributes are not quantifiable using preserved archived specimens and require the study of live organisms. Studying ecological adaptation of live organisms can be confounded by altered organismal behavior or physiology due to stress associated with captivity, and ecological interactions that are difficult or impossible to control in field studies.

Four of the six crangonyctids were behaviorally photophobic (eyed species: Crangonyx sp., Sicifera sp., and eyeless: Stygobromus pecki and S. russelli) while the remaining two eyeless species (S. flagellatus and S. bifurcatus) were photoneutral. This result indicates that at least some Stygobromus, and perhaps other species of eyeless Crangonyctidae, have the capacity to detect light although there is currently no obvious anatomical or molecular evidence for this. This finding is consistent with recent studies on subterranean amphipods in Europe (Fišer et al., 2016), subterranean beetles in Australia (Langille et al., 2018), and subterranean crayfish in the USA (Li and Cooper, 2002), suggesting photoreceptors could be widespread in subterranean eyeless arthropods. Although our sample size was relatively small (n = 5–10 individuals per species), the relatively low CVs suggest that these estimates adequately capture variation for these traits within these species.

All Stygobromus species are eyeless, and Stygobromus was recovered in our analysis to be a monophyletic taxon, thus the eyeless phenotype appears to have evolved once in a common ancestor of all extant Stygobromus. Eyelessness in stygobiontic species is hypothesized to be due to strong selection against visual organs in lightless environments because eyes are energetically wasteful in subterranean environments where sight is not possible and food is scarce (Green and Romero, 1997; Moran et al., 2015). However, along with the ANSs, the paraphyletic distribution of photic behavior and respiratory phenotypic trait means across the phylogeny suggests that photoneutrality vs. photophobia evolved independently in various species.

Unfortunately, because we were only able to obtain a maximum of two species per habitat category, we were not able to quantitatively test the relationship between habitat association and phenotype. This limitation is unavoidable, as the species sampled are rare, often at low densities, and in the case of some habitats, the species sampled are the only locally available crangonyctids adapted to their respective habitats. To overcome this, we also performed an ANS for habitat association which found that epigean vs. subterranean habitat associations are monophyletic among these species, though subterranean phenotypes are known to have evolved independently multiple times in Crangonyctidae (Holsinger, 1967; Cannizzaro et al., 2021; Marin and Palatov, 2021).

Though the phylogenetic distribution of trait means were found to be paraphyletic with little to no phylogenetic signal, there was phylogenetic signal for the CVs of phenotypes. At least in the case of Stygobromus, this might be due to neutral processes under relaxed selection because photophobia is not relevant in deep phreatic waters (i.e., these species never encounter light and therefore any phenotypes relating to light responses cannot be advantageous and are therefore either neutral and evolving due to genetic drift, or negative if they are energy expensive).

The habitat model was found to be the strongest for two of the four evolutionary model comparisons, suggesting that habitat largely determines phototaxis in this group of species ( Table 3 ). We hypothesize that this result could be because behavioral photophobia is maintained by selection in interstitial and spring-associated species to protect members of these species from entering inhospitable environments (i.e., an epigean environment with visual predators). Even though the eyeless phenotype may be highly advantageous in subterranean habitats, it is presumed to be highly disadvantageous in surface habitats because sightlessness in an epigean environment is likely an evolutionary dead end due to predation and other nonlethal risks including scarcity of suitable mates (Tobler, 2009; Horstkotte et al., 2010; Klaus and Plath, 2011). Therefore, selection on near-surface sightless species to remain in a subterranean environment likely outweighs any fitness disadvantages associated with maintaining light sensitive photoreceptors.

Presently, it is unclear if convergent phenotypes in Stygobromus are the result of selection operating on novel or standing genetic variation. However, research in other stygobiontic organisms has suggested that convergent characteristics in related organisms can be attributed to genetic changes that are unique to each instance of adaptation (Protas and Patel, 2008), suggesting that adaptation of similar phenotypes can be due to unique lineage-specific genomic changes rather than reuse of standing genetic variation.

A possible explanation for the inverse relationship between the CVs of the behavioral and respiratory phenotypes ( Figure 5 ) is balancing of the trait phenotype with the level of familiarity of the stimulus. In other words, an environmental stimulus to which a population is naïve to is expected to be stressful, whereas a familiar environmental stimulus is expected to be less stressful, and the population’s phenotype is expected to be tuned to stimulus through selection. In this study, both species collected from deep aquifer environments (S. flagellatus and S. bifurcatus) showed the highest CV for respiratory response while the epigean species showed the lowest; the opposite was found for behavior. This suggests that the respiratory trait contains neutral variation in the species that never encounter light while the trait is closer to fixed in the species that encounter light regularly. Considering that deep aquifer species essentially never encounter light, the broader variation in the respiratory response of deep aquifer species suggests that neutral variation exists for how stressful light is perceived to be by these naïve species. Likewise, it is not surprising that they would also have little variation in a photic behavioral response (i.e., organisms don’t adapt to stimuli they have never experienced before). In contrast, we see that the epigean species show low variation in stress response (i.e., light is a daily factor for these species, and they all find it non-stressful) but high variation in behavioral response. This would suggest that avoiding light bears relatively neutral consequences for these species, which makes sense: if light is deleterious to a species that encounters light every day, it is unlikely for that species to persist. Following this logic, we see that the ecotone species show moderate variation for both phenotypes. This is not surprising as these species are likely under directional selection with respect to how they deal with light, and therefore have ‘pruned down’ neutral variation. In summary, if an organism was completely naïve to an environmental stimulus, the stimulus could be quite stressful; however, a stimulus that is common to the evolution of the species may elicit an adaptive behavioral response without causing much stress.

The phylogenetic signal for the phenotypic CVs suggests that the relationship between the mean and the amount of variation for a trait in a population may be phylogenetically constrained. This could be due to certain lineages having higher standing genetic variation for the measured phenotypes, or at least some other mechanism of maintaining higher population-level phenotypic variation, which may affect the evolvability of a lineage. In essence, a population with an exceptionally low CV for a trait could be said to be fixed for that trait because there is almost no intrapopulation variation for selection to operate on; whereas a population with a relatively high CV could be expected to have individuals within the population affected quite differently by the same selection pressures due to considerable differences in their phenotype. Therefore, populations with higher trait CVs could be said to have greater evolutionary potential as a population because individuals are expected to have variable and unequal fitness and therefore, as a population, have the possibility of tolerating greater variation in selection.

These findings support the hypothesis that, under certain circumstances, motile organisms can evolve behaviors in response to environmental indicators to avoid inhospitable habitats and deleterious effects associated with those habitats. Thus, it is likely that behavior mediates nonrandom dispersal and habitat occupancy, functioning to maintain biogeographic distributions of populations (Binckley and Resetarits, 2005; Resetarits, 2005; Edelaar et al., 2008; Fišer et al., 2016; Kralj-Fišer et al., 2020). If standing genetic variation can affect the evolution of behavior, and behavior can affect population structure and thus gene flow, then it is possible that behaviors and genetic structure form a positive feedback loop affecting speciation.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material .

The manuscript presents research on animals that do not require ethical approval for their study.

Conceptualization: MW, PN, JG, BS, and WN. Methodology: MW, PN, JG, DR, BS, and WN. Software: MW. Validation: MW, PN, JG, DR, BS, and WN. Formal analysis: MW. Investigation: MW, PN, and DR. Resources: JG, DR, BS, and WN. Data curation: MW, PN, and DR. Writing—original draft preparation: MW. Writing—review and editing: MW, PN, JG, DR, BS, and WN. Visualization: MW. Supervision: JG, BS, and WN. Project administration: MW, BS, and WN. Funding acquisition: BS and WN. All authors contributed to the article and approved the submitted version.