At the confluence of vicariance and dispersal: Phylogeography of cavernicolous springtails (Collembola: Arrhopalitidae, Tomoceridae) codistributed across a geologically complex karst landscape in Illinois and Missouri

Abstract The processes of vicariance and dispersal are central to our understanding of diversification, yet determining the factors that influence these processes remains a significant challenge in evolutionary biology. Caves offer ideal systems for examining the mechanisms underlying isolation, divergence, and speciation. Intrinsic ecological differences among cavernicolous organisms, such as the degree of cave dependence, are thought to be major factors influencing patterns of genetic isolation in caves. Using a comparative phylogeographic approach, we employed mitochondrial and nuclear markers to assess the evolutionary history of two ecologically distinct groups of terrestrial cave‐dwelling springtails (Collembola) in the genera Pygmarrhopalites (Arrhopalitidae) and Pogonognathellus (Tomoceridae) that are codistributed in caves throughout the Salem Plateau—a once continuous karst region, now bisected by the Mississippi River Valley in Illinois and Missouri. Contrasting phylogeographic patterns recovered for troglobiotic Pygmarrhopalites sp. and eutroglophilic Pogonognathellus sp. suggests that obligate associations with cave habitats can restrict dispersal across major geographic barriers such as rivers and valleys, but may also facilitate subterranean dispersal between neighboring cave systems. Pygmarrhopalites sp. populations spanning the Mississippi River Valley were estimated to have diverged 2.9–4.8 Ma, which we attribute to vicariance resulting from climatic and geological processes involved in Mississippi River Valley formation beginning during the late Pliocene/early Pleistocene. Lastly, we conclude that the detection of many deeply divergent, morphologically cryptic, and microendemic lineages highlights our poor understanding of microarthropod diversity in caves and exposes potential conservation concerns.

Genetic isolation is a primary driver of molecular divergence and ultimately speciation, but determining the factors that promote or constrain genetic diversity remains a significant challenge in evolutionary biology. Patterns of diversity in caves are often attributed to vicariance or dispersal, but the relative influence these processes have on the evolution and contemporary distributions of cave fauna has been widely debated (see Culver, Pipan, & Schneider, 2009;Porter, 2007). However, it is generally accepted that patterns of diversity in caves are likely shaped by a complex interaction of intrinsic factors (e.g., species-specific differences in ecology, life history, or biology) that can influence dispersal capacity and extrinsic factors (e.g., geographic barriers or climate change) that can enhance or limit dispersal opportunity Porter, 2007).
Phylogeography, the study of processes that influence the contemporary geographic distributions of species' populations by utilizing genetic data can provide insights into the relative influences of evolutionary factors driving patterns of genetic isolation and divergence in biological communities (Avise, 2000;Avise et al., 1987). For instance, phylogeographic congruence among codistributed species can implicate vicariance caused by "hard" geographic barriers or environmental changes affecting entire communities (Lapointe & Rissler, 2005), whereas conflicting phylogeographic patterns may be attributable to intrinsic differences that can affect species dispersal capacity across "soft" potential genetic barriers (e.g., Goldberg & Trewick, 2011;Hodges, Rowell, & Keogh, 2007;Hurtado, Lee, & Mateos, 2013). With cave organisms, the majority of research studies have been limited to single species (e.g., Dörge, Zaenker, Klussmann-Kolb, & Weigand, 2014;Faille et al., 2015) or cryptic species complexes with allopatric distributions (e.g., Gómez et al., 2016;Rastorgueff, Chevaldonné, Arslan, Verna, & Lejeusne, 2014). The arthropod class Collembola (springtails) offers a nearly unparalleled opportunity for elucidating the interplay of factors that affect speciation and molecular diversification in subterranean ecosystems. These small, wingless, insect-like arthropods are among the most abundant, diverse, and well-adapted organisms in caves (Christiansen, 1965;Thibaud & Deharveng, 1994), and are considered important subterranean examples of adaptive radiations (Christiansen & Culver, 1969) and parallel speciation (Christiansen, 1961(Christiansen, , 1965Christiansen & Culver, 1968). Their small size (body length often less than 1 mm), low vagility, and close associations with cave habitats facilitate their isolation, resulting in a high degree of endemism (Niemiller & Zigler, 2013) and cryptic species (Juan & Emerson, 2010). For example, the springtail genus Pseudosinella alone contains more than 100 species found in caves worldwide, many of which are known only from a single cave system (Hopkin, 1997). Most importantly, cave-dwelling springtails have varying levels of ecological specificity to, and dependence upon, cave habitats.
Although surface species are commonly found in caves as accidentals (i.e., they may fall or get washed into caves, but cannot maintain populations in caves), the majority of collembolans occurring in caves can maintain permanent subterranean populations and are either classified as troglobionts (i.e., obligate cave-dwellers that are never encountered on the surface and often have conspicuous troglomorphic adaptations associated with cave habitats) or eutroglophiles (i.e., facultative cave-dwellers that also occur in surface habitat and usually lack apparent troglomorphy) (see Sket, 2008 for current ecological classifications of subterranean animals). Because troglobiotic and eutroglophilic springtails can be codistributed (Katz et al., 2016;Soto-Adames & Taylor, 2013), extrinsic evolutionary processes are likely exerting similar selective pressures upon them. Therefore, opposing patterns of genetic structure among these species distributed across the same geographic area can reflect intrinsic factors, such as differences in the degree of ecological association with cave habitats (cave dependence) that can affect a species' capacity to disperse across geographic barriers (Pérez-Moreno et al., 2017;Weckstein et al., 2016). Disparate geographic distributions among closely related surface springtails provide some indirect evidence that varying dispersal capacity may be associated with differences in species-specific traits (Costa et al., 2013;Katz, Giordano, & Soto-Adames, 2015), and Christiansen and Culver's (1987) biogeographic study of cave springtails revealed that more pronounced troglomorphy can be correlated with smaller geographic ranges.
Long-term local persistence and small geographic ranges are typical for troglobionts, and by definition, these species cannot maintain surface populations to facilitate dispersal between discontinuous subterranean habitats. Therefore, patterns of genetic differentiation in troglobionts are likely driven primarily by isolation due to physical barriers and reflect vicariance. On the contrary, we expect isolation by distance (IBD) to be the primary driver of genetic variation in eutroglophiles owing to their propensity to disperse across surface habitats.
To test these predictions, we incorporate a suite of molecularbased approaches to (a) delimit cryptic species in the focal complexes, (b) detect molecular signatures of isolation to identify potential genetic barriers, and (c) estimate evolutionary relationships and divergence times to elucidate the roles of vicariance and dispersal in shaping patterns of cave-dwelling springtail diversity throughout the Salem Plateau-a major cave-bearing karst region of the Ozark Plateau that spans the Mississippi River Valley in Illinois and Missouri. Recent molecular-based biogeographic investigations of Ozark cave biodiversity have been useful for addressing evolutionary hypotheses for salamanders (Phillips, Fenolio, Emel, & Bonett, 2017) and fish broadly distributed across the Mississippi River Valley (Niemiller et al., 2012). However, the phylogeography of cave invertebrates has yet to be evaluated for the Salem Plateau. Fine-scale phylogeographic patterns of cave springtails distributed across the Mississippi River may be used to investigate the impact of intrinsic and extrinsic factors (e.g., the degree of cave dependence and geographic barriers) on the evolution of cave organisms, broaden our limited understanding of subterranean microarthropod diversity, and assess biogeographic interpretations that may help clarify the complex, yet poorly understood, geological history of the Salem Plateau.

| Study system, focal taxa, and field collections
The complex geological landscape of the Salem Plateau (Figure 1) provides the ecological context for testing biogeographic hypotheses of vicariance and dispersal. This once continuous karst region, now bisected by the Mississippi River Valley, is located south of St.
Louis and covers just eight counties but contains thousands of sinkholes and includes the largest cave systems in Illinois and Missouri F I G U R E 1 Salem Plateau cavebearing karst spanning the Mississippi River border of Illinois and Missouri (gray) (adapted from Panno et al. (1997Panno et al. ( , 1999 (Panno, Weibel, & Li, 1997 Notes. a Refers to high-density sinkhole areas in the Salem Plateau karst study area defined for Illinois (Panno et al., 1997(Panno et al., , 1999Venarsky et al., 2009) and Missouri (Burr et al., 2001;Panno et al., 1999) (see Figure 1). abundance on organic debris and rock surfaces in cave entrances and twilight zones, and less frequently and in smaller numbers in cave dark zones. preparation, but the fragile cuticles were easily damaged when handled and small individuals were nearly invisible making them difficult to recover. Therefore, the heads of specimens, which include important diagnostic morphology (e.g., the arrangement and morphology of setae), were dissected and stored separately prior to DNA extraction as back up vouchers for those cases where the now-translucent bodies were not recovered.
COI and 16S are particularly useful for evaluating population-level variation as they exhibit high levels of genetic variation and have been used extensively for species-and population-level phylogenetic research in springtails (Hogg & Hebert, 2004). Collembola are generally characterized by extremely high levels of molecular diversity (Katz et al., 2015); therefore, more slowly evolving loci, 28S and histone-3, were included to provide stronger phylogenetic signal among more distantly related taxa. Histone-3 and 28S D1-3 were excluded for Pogonognathellus due to inconsistent amplification.
See Supporting information Appendix S1 for list of all taxa included in this study, including sample information and all sequences with corresponding GenBank (Benson et al., 2013) accession numbers.
See Supporting information Appendix S2 for PCR and sequencing primers, including a description of the PCR protocol and sequence alignment methods used in this study. The outgroup taxa listed in Supporting information Appendix S3 were chosen based on their affinities with the target taxa and availability of sequences in GenBank.

| Detecting and delimiting cryptic diversity
The presence of cryptic diversity was detected by incorporating a number of different tests. First, we calculated uncorrected pairwise COI distance frequencies for all sampled specimens with PAUP* 4.0a build 159 (Swofford, 2002) and plotted distance frequency histograms to detect the presence of interspecific variation within each targeted morphospecies. A gap between the greatest putative intraspecific and smallest putative interspecific pairwise distances can be interpreted as the boundary between species-and populationlevel variation (Meier, Zhang, & Ali, 2008).
To determine how interspecific variation was geographically distributed, we performed a hierarchical analysis of molecular variance (AMOVA) for COI, 16S, and 28S using all taxa sampled for each target morphospecies using Arlequin v. 3.5.2.2 (Excoffier & Lischer, 2010). Haplotypes were grouped within samples, among samples in caves, and among caves with 50,000 permutations performed to assess significance. The presence of strong genetic structuring within samples or among samples in caves can be an indicator of cryptic diversity because sexual isolation is typically required to maintain high levels of genetic variation occurring in sympatry.
We also delimited putative species boundaries using a General Mixed Yule Coalescent (GMYC) analysis (Pons et al., 2006). This method uses ultrametric gene trees to identify the interface between population-and species-level branching patterns and demarcates genetically cohesive clades as independent evolutionary units known as operational taxonomic units (OTUs). The GMYC analysis was performed on COI gene trees using the single threshold delimitation method implemented in the splits package (Ezard, Fujisawa, & Barraclough, 2009)

| Tests for genetic structure
The relative role of cave dependence and its influence on springtail dispersal capacity remain unclear, in part, because the identities of genetic barriers are not known for cave-dwelling springtails. To identify barriers to Pygmarrhopalites and Pogonognathellus dispersal, we evaluated and compared levels of genetic structure across cave boundaries and the Mississippi River Valley. In addition, we also included sinkhole area boundaries in the genetic structure analyses.
The most sampled OTUs for each target morphospecies, identified by the GMYC analysis, were chosen as focal OTUs for population analyses to avoid attributing deeply divergent and structured lineages to population-level variation, rather than to species-level variation (Fouquet et al., 2007). Hierarchical AMOVAs were performed independently with Arlequin for COI and 16S for both focal OTUs by grouping haplotypes within samples, among samples within barriers, and among samples across barriers. Significance was assessed with 50,000 permutations.
Patterns of population structure resulting from dispersal and genetic drift, rather than of vicariance across geographic barriers, are common in animals with low mobility and can usually be attributed to a model of IBD (Costa et al., 2013;Timmermans et al., 2005). To determine whether geographic distance is significantly correlated with genetic distance, we performed a Mantel test (Mantel, 1967;Sokal, 1979; but see Diniz-Filho et al., 2013;Legendre, Fortin, & Borcard, 2015) for each locus. We also evaluated the significance of genetic structure across barriers while controlling for geographic distance using a partial Mantel test (Smouse, Long, & Sokal, 1986), which allows for the comparison of two variables (i.e., pairwise genetic distances and position relative to geographic barrier) while controlling a third (i.e., geographic distances Templeton-Crandall-Sing (TCS) haplotype networks (Clement, Snell, & Walker, 2002) for COI and 16S were estimated with PopART (Leigh & Bryant, 2015) to visualize and compare phylogeographic structure across genetic barriers for Pygmarrhopalites and Pogonognathellus focal OTUs.

| Phylogenetic inference, divergence time estimation, and topology tests
To further investigate the interplay of vicariance and dispersal capacity on cave springtail diversity, we conducted a Bayesian phylogenetic analysis using BEAST 2 to infer evolutionary relationships and to estimate divergence times for all sampled lineages of Pygmarrhopalites and Pogonognathellus. Two independent datasets were analyzed and compared: the Pygmarrhopalites dataset (COI, 16S, 28S D1-3, 28S D7-10, histone-3; 3,358 total bp) and the Pogonognathellus dataset (COI, 16S, 28S D7-10; 2,059 total bp).
External rates were used for molecular clock calibrations rather than fossil information because springtails lack an adequate fossil record and phylogenetic framework for calibrating molecular clocks.
Katz (2018)  Following guidelines proposed by Kass and Raftery (1995), a twice logarithm BF difference (2 × log e BF) of higher than 6 was considered strong evidence against the null hypothesis.

| Evidence for cryptic diversity
Uncorrected pairwise COI distance frequency histograms revealed extraordinarily high genetic distances among sampled specimens within each morphospecies: up to 35% for Pygmarrhopalites and 18% for Pogonognathellus ( Figure 3). COI distances above 8%-15% in springtails are typically recognized as interspecific when used in combination with independent evidence (Katz et al., 2015). Moreover, COI distances form bimodal distributions for both morphospecies, each separated by a 10% gap (Figure 3) which can be interpreted as a boundary between intra-and interspecific genetic variation (Meier et al., 2008), providing preliminary support for the presence of cryptic diversity within both target morphospecies.
The results of the initial AMOVA that incorporated all sampled taxa identified high levels of genetic structure within caves and within samples, supporting the presence of sympatric cryptic species (Table 2): Between 40% and 60% of genetic variation in COI, 16S, and 28S was structured among samples within the same cave for both genera. Genetic variation in COI, 16S, and 28S (24%, 21%, and 29%, respectively) was also structured within samples for Pygmarrhopalites, but this pattern was not recovered for Pogonognathellus (COI, 2%; 16S, 0%; 28S, 41%).
The GMYC analyses revealed 14 putative species: 10 Pygmarrhopalites OTUs (A1-10) and four Pogonognathellus OTUs (T1-4) ( Figure 6).  Figure 4). Pygmarrhopalites A10 and Pogonognathellus troglomorphy (e.g., elongated antennae and thread-like unguiculus) and was most similar to Pygmarrhopalites pavo (Christiansen & Bellinger, 1996), a troglobiont reported from caves in Virginia (Christiansen & Bellinger, 1996), West Virginia (Fong, Culver, Hobbs, & Pipan, 2007), Tennessee (Lewis, 2005), and Missouri (Zeppelini, Taylor, & Slay, 2009). We believe the unique differences in morphology, in combination with molecular evidence, support the recognition of all Pygmarrhopalites OTUs as distinct and potentially new species. Because some cryptic lineages may be of higher conservation concern, it is imperative to identify and describe these lineages for potential management initiatives (Delić, Trontelj, Rendoš, & Fišer, 2017;Niemiller, Graening, et al., 2013). However, we chose to refrain from giving OTUs formal species names at this time because a comprehensive taxonomic review is required to describe new species and to clarify the status of existing species-a task beyond the scope of this study.

| Phylogeny, divergence times, and topology tests
The rate-calibrated phylogenetic analysis based on the multilocus dataset produced trees with high support for all OTUs identified by the GMYC analysis, and molecular divergence time estimates revealed that all OTU diversification predated the Pliocene (Figure 6). The multilocus phylogeny also shows that two additional OTUs (Pygmarrhopalites A3 and A4) contain both Illinois and Missouri lineages, but did not form monophyletic groups by region relative to the Mississippi River. All other OTUs were short-range endemics, from a single cave (A1, A2, A6, A8, A9, T1-3) or from neighboring cave systems within the same sinkhole area (A5, A7) ( Figure 6).

TA B L E 5
Mantel test results (a, COI; b, 16S) to identify isolation-by-distance (IBD) patterns and correlations between genetic distance and geographic barriers after controlling for geographic distance in Pygmarrhopalites A10 and Pogonognathellus T4 The topology tests (
Mantel tests confirmed geographic distance to be a significant driver of genetic isolation for both taxa (Table 5), suggesting springtails are weak dispersers regardless of ecological classification. After controlling for geographic distance using partial Mantel tests, we still recovered significant positive correlations between genetic distance and sinkhole areas and between genetic distance and position relative to the Mississippi River for Pygmarrhopalites, but not for Pogonognathellus (Table 5). The haplotype networks ( Figure 5), phylogenetic trees (Figures 6 and 7), and topology tests (Table 6) also corroborate these findings providing similar patterns of ge- Relative to the Mississippi River Valley and sinkhole area boundaries, we observed a very different pattern when genetic variation was partitioned among caves: Cave boundaries were identified as significant genetic barriers for Pogonognathellus only, whereas patterns of genetic structure among caves identified by the AMOVA for Pygmarrhopalites A10 (Table 4a) were not supported after accounting for geographic distance (Table 5). In this case, patterns of genetic structure among caves are driven by IBD for Pygmarrhopalites A10 (not Pogonognathellus T4) suggesting that troglobiotic Pygmarrhopalites are capable of dispersing between caves. Although this finding appears to contradict the hypothesis that troglobiotic species are less capable of dispersal across geographic barriers, it can still be explained by differences in cave habitat preferences.
Aquatic interstitial subterranean connections joining neighboring cave systems may enable subterranean dispersal during flooding events for Pygmarrhopalites A10. This is supported by shared 16S haplotypes between neighboring cave systems (PAC, HSC, and STC) ( Figure 5b). Groundwater connections (e.g., alluvial aquifers, epikarst systems) have been implicated as "interstitial highways" that can provide subsurface dispersal pathways for a wide range of subterranean arthropods (e.g., Lefébure et al., 2006;Ward & Palmer, 1994), but Collembola are not normally considered members of the interstitial groundwater community as they cannot complete life cycles while submerged (Deharveng, D'Haese, & Bedos, 2008).
However, growing evidence suggests that they are not only present in these habitats, but can occur in abundance and comprise diverse communities (Bretschko & Christian, 1989;Deharveng et al., 2008; Cave-to-cave subterranean dispersal is unlikely or infrequent for Pogonognathellus because species in this genus do not occur in interstitial habitats and prefer floor or wall surfaces near cave entrances rather than dark zone habitats. This is supported by strong genetic structuring among caves for Pogonognathellus T4 indicating that cave-to-cave dispersal is extremely rare for this species despite having naturally occurring surface populations that could presumably F I G U R E 6 Time-calibrated trees for (a) Pygmarrhopalites and (b) Pogonognathellus inferred by Bayesian phylogenetic analysis. Clade posterior probabilities are indicated at each node. Divergence times are represented by blue bars at each node with their length corresponding to the 95% HPD of node ages. OTUs identified by the GMYC analysis are indicated to the right of each clade (A1-A10 and T1-T4). Focal OTUs chosen for population structure analyses (A10 and T4) are highlighted in gray boxes (see Figure 7 for close-up of A10). Single-site endemic OTUs are labeled in red. Taxon labels correspond to cave name abbreviation, sample #, state, specimen # (see Table 1 for cave abbreviations and Appendix 1 for sample information). Scale bars represent substitutions/site/Ma To assess the effect of cave dependence on patterns of molecular variation, we were required to make informed assumptions about species ecology, including the classification of Pygmarrhopalites A10 as a troglobiont. For many small cave-dwelling animals, such as springtails, it is often impossible to ascertain with certainty that a species only occurs in caves (Christiansen, 1962); a species reported only from caves could also be a common soil species, having yet to be reported from surface habitats; the distinction between cavernicolous habitats and other subsurface microhabitats may be weak or nonexistent for small animals; and troglobionts often lack obvious troglomorphy. Despite these concerns, we are confident that the combination of troglomorphy, close morphological affinities to known troglobiotic species, and their exclusive occurrence in dark or deep twilight cave zones (Supporting information Appendix S1) provides sufficient evidence that Pygmarrhopalites A10 is a troglobiont.
The degree of cave dependence is certainly a major factor influ-  (Christiansen & Bellinger, 1996), and males have also been reported for P. pavo, a species that is morphologically similar to Pygmarrhopalites A10 (Christiansen & Bellinger, 1996).

| Biogeography: evidence for vicariance across the Mississippi River Valley
The climatic and geological changes during the Pleistocene and their impacts on the distribution and diversity of North American cave F I G U R E 7 Close-up of clade Pygmarrhopalites A10 from Figure 6a illustrating timing information from estimates of molecular divergence and geological evidence supporting vicariance across the Mississippi River Valley: (a) 2.90-4.76 Ma (95% HDP) (blue bar) divergence time between Missouri and Illinois lineages (separated by gray dashed line), posterior probabilities at each node lower than 1 are not displayed; (b) late Pliocene/ early Pleistocene timing (dashed arrow) of initial Mississippi River entrenchment (Cupples & Van Arsdale, 2014) and increased river discharge (Cox et al., 2014); (c) 3.25 ± 0.26 Ma (green column) timing of initial Green River karst incision and excavation (Granger et al., 2001); (d) 2.41 ± 0.14 Ma (orange column) timing of first glacial melt (Balco et al., 2005  fauna have been well documented (Porter, 2007). For example, the modern course of the Ohio River, formed by changing climate during the Pleistocene, bisects a major cave-bearing karst region along the Indiana-Kentucky border. Niemiller, McCandless, et al. (2013) demonstrated that this river is a major biogeographic barrier, facilitating the divergence and subsequent isolation and speciation of troglobiotic cavefish populations. Like the Ohio River, the Mississippi River has also been implicated as a "hard" geographic barrier to dispersal for many surface species (e.g., Soltis, Morris, McLachlan, Manos, & Soltis, 2006), but its influence on the evolutionary history of cavedwelling organisms has yet to be evaluated, in part, because the geological history of the Mississippi River and its influence on regional cave-bearing karst remain poorly understood.
Molecular divergence times of Pygmarrhopalites A10 populations spanning the Mississippi (Figures 6 and 7), patterns of genetic structure (Tables 4c, 5; Figure 5), and topology tests ( shortly after, this process took place for the Mississippi River. The corroboration of timing information derived from both biological and geological data (Figure 7) supports the hypothesis that climatic and geological events beginning in the late Pliocene initiated and maintained genetic isolation between troglobiotic springtail populations in Illinois and Missouri, but the exact mode of gene flow across the preglacial Mississippi River and tributaries, prior to their genetic isolation, is not known. It is plausible that sections of karst were periodically isolated and rejoined by shifting meanders and periods of low flow, later removed by Plio-Pleistocene entrenchment and excavation, providing intermittent subterranean passage for cave organisms until the late Pliocene or early Pleistocene.
The lack of genetic structure across the Mississippi River (Tables 4c and 5; Figure 5) and nonmonophyly ( Over-reliance of mtDNA can produce misleading phylogenetic and biogeographic conclusions due to introgression, hybridization, paternal inheritance, and incomplete lineage sorting (Funk & Omland, 2003), but none of these processes have been reported for Collembola, except hybridization (Deharveng, Bedos, & Gisclard, 1998;Skarzynski, 2004). The development of more sensitive, nondestructive DNA extraction and genomic sequencing methods will certainly help alleviate these issues, improve the precision and accuracy of divergence time analyses, and bring springtail genetics into the big data era.

| Cryptic diversity, short-range endemism, and implications for conservation
Recent discoveries of cryptic species have challenged our current understanding of biological diversity (Fišer, Robinson, & Malard, 2018), and this paradigm shift is particularly evident in subterranean habitats where ideal conditions have fostered widespread cryptic speciation, including examples of recent divergence in cavefish (Niemiller, McCandless, et al., 2013), morphological stasis in amphipods (Trontelj et al., 2009), and morphological convergence in springtails (Christiansen, 1961). Therefore, it was important in this study to detect the presence of cryptic diversity and delimit OTUs prior to phylogeographic comparisons, to avoid interpreting interspecific variation as population-level genetic structure. Large gaps in genetic distance frequencies ( Figure 3) and the presence of strong interspecific genetic structure within caves (Table 2)  The detection of short-range endemics, genetic isolation, and apparent cryptic diversity has major conservation implications.
Reduced dispersal capacity observed for Pygmarrhopalites can increase their susceptibility to human disturbances such as land use practices, climate change, pollution, and invasive species-all of which pose major threats to fragile cave ecosystems (Culver & Pipan, 2009a;Taylor & Niemiller, 2016). In fact, growing concerns of karst groundwater contamination (Panno, Krapac, Weibel, & Bade, 1996) prompted Pygmarrhopalites madonnensis (Zeppelini & Christiansen, 2003), a troglobiotic springtail known from a single cave in Monroe Co., Illinois, to be listed as state endangered (Mankowski, 2010). This is concerning considering that our data indicate that single-site endemics are not only extremely common but may also comprise a large majority of troglobiotic springtail diversity throughout this region. Lastly, unrecognized cryptic species complexes with allopatric ranges, presumed to be a single widely distributed species, may lead to misguided biodiversity conservation and management decisions.

| CON CLUS IONS
Salem Plateau caves and their springtail inhabitants provide a model system for comparative phylogeographic studies addressing important questions in evolution and subterranean biogeography. We characterized and compared patterns of molecular diversity between species in the genera Pygmarrhopalites and Pogonognathellus, which led to three important findings. First, conflicting phylogeographic patterns between troglobiotic and eutroglophilic species distributed across the same geographic barriers suggests that different degrees of cave dependence can have major impacts on the dispersal capacity and genetic connectivity of cave organisms. Second, estimates of genetic structure and molecular divergence indicate that climatic and geological processes during the late Pliocene/early Pleistocene were major factors driving isolation between populations of troglobiotic cave organisms in Salem Plateau karst spanning the Mississippi River in Illinois and Missouri. Lastly, the large number of deeply divergent lineages and high rates of short-range endemism detected in this study exposes a major knowledge gap in our understanding of cave microarthropod diversity and highlights potential conservation concerns under growing threats to cave biodiversity. Additional phylogeographic research and the development of genomic datasets for cave springtails will further contribute to our understanding of how and why organisms occupy, persist in, and adapt to cave environments-information critical for the development and implementation of conservation strategies needed to manage and protect cave biodiversity (Porter, 2007). Institute. We would also like to thank the faculty, staff, and graduate students of the Department of Entomology at the University of Illinois at Urbana-Champaign for their support.

AUTH O R CO NTR I B UTI O N S
A.D.K. contributed to research design, collected and analyzed data, and wrote the manuscript. S.J.T. conceived of the project, contributed to research design, provided access to cave sites, and assisted in data collection and manuscript writing. M.A.D. assisted in writing the manuscript and provided substantial molecular laboratory resources that contributed to data collection.

DATA ACCE SS I B I LIT Y
All DNA sequence data from this study have been submitted to GenBank and are available under accession numbers MH269419-MH269696 and listed in Supporting information Appendix S1.