Research Article |
Corresponding author: T. Keith Philips ( keith.philips@wku.edu ) Academic editor: Oana Teodora Moldovan
© 2020 Olivia F. Boyd, T. Keith Philips, Jarrett R. Johnson, Jedidiah J. Nixon.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Boyd OF, Philips TK, Johnson JR, Nixon JJ (2020) Geographically structured genetic diversity in the cave beetle Darlingtonea kentuckensis Valentine, 1952 (Coleoptera, Carabidae, Trechini, Trechini). Subterranean Biology 34: 1-23. https://doi.org/10.3897/subtbiol.34.46348
|
Cave beetles of the eastern USA are one of many poorly studied groups of insects and nearly all previous work delimiting species is based solely on morphology. This study assesses genetic diversity in the monotypic cave carabid beetle genus Darlingtonea
mitochondrial DNA, Mississippian Plateau, Pennyroyal, phylogeography, Southern Appalachians, troglobites, troglobionts
Variation within a species is usually not random, but structured in some way and typically forms a metapopulation with various levels of deviation from panmixia (
Patterns of gene flow among caves in karst areas vary mostly in accordance with the geographical distribution of subterranean limestone (e.g.,
The cave-rich limestone of the MP is bisected by the Cumberland Saddle, a low point in the Cincinnati Arch formation, which separates the MP into two regions: the MP-I to the west and the MP-II to the east (Fig.
Isolating barriers between cave systems restrict gene flow and promote divergence among populations of cave organisms, effectively dividing parts of cave systems into subterranean islands (
Darlingtonea Valentine, 1952 is a monotypic genus of cave carabid beetle found in a narrow distributional band from north-central Tennessee (known from a single cave near the Kentucky border) extending northeastward into east-central Kentucky (mainly the northern part of “MP-II” in Fig.
Regarding the origin and diversity of North American cave trechines, most authors have favored some version of a “Pleistocene-effect” model (
Other authors have found isolation and divergence in allopatry to be an unsatisfactory model for cave colonization in other taxa, which may be better viewed as a parapatric ecological transition or ‘adaptive shift’ occurring in the presence of gene flow via diversifying selection (
It is currently unclear what factors have led to the evolution of any morphological or genetic diversity within Darlingtonea kentuckensis. Darlingtonea kentuckensis has a broader than average distribution compared to most terrestrial Eastern North American troglobionts based on our review (Philips et al. unpublished). Both
The exceptional species diversity in North American cave trechines (
If important barriers to dispersal for cave trechines in the MP-II region exist, hierarchical tests of population genetic structure should reveal a general pattern of low diversity within and high diversity among clusters of genetically similar populations. Specific geographic barriers between these genetic clusters that may be responsible for population structure can then be hypothesized and should make geographic sense without being purely attributable to the influence of isolation and genetic divergence by distance. Patterns may also reveal the presence of cryptic species or subspecies.
The Kentucky and upper Cumberland rivers represent the two primary watersheds in the MP-II. Further, the divide between the watersheds of the Kentucky and Rockcastle rivers in northern Jackson County (
Caves also fall into smaller, “minor” watersheds (Table
List of Darlingtonea populations included in a study of mitochondrial haplotypes, including population (taxon) reference codes, locality information, collection dates, sample size, faunal region, local watershed and GenBank accession codes. Faunal region 1.
Taxon Code | Cave | County | Collection Date | N | Faunal Region | Local Watershed/Code (River Drainage) | GenBank accession number |
---|---|---|---|---|---|---|---|
BLO | Blowing | Wayne | 1-Mar-2014 | 3 | 4 | Otter Creek/OT (CR) | MN880837, MN880838, MN880839 |
CLF | Clifford Pearson | Estill | 14-Aug-2014 | 2 | 1 | Station Camp Creek/SC (KR) | MN880814, MN880815 |
CLI | Climax | Rockcastle | 31-Jul-2014 | 4 | 2 | Roundstone Creek/RO (RR) | MN880810, MN880811, MN880812, MN880813 |
FLE | Fletcher Spring | Rockcastle | 15-Mar-2014 | 3 | 2 | Skegg Creek/SK (RR) | MN880827, MN880828, MN880829 |
GSP | Great Saltpeter | Rockcastle | 15-Aug-2014 | 4 | 2 | Roundstone Creek/RO (RR) | MN880817, MN880818, MN880819, MN880820 |
HIC | Hicksey | Jackson | 14-Aug-2014 | 4 | 1 | Station Camp Creek/SC (KR) | MN880806, MN880807, MN880808, MN880809 |
HIS | Hisel | Jackson | 1-Aug-2014 | 1 | 1 | Station Camp Creek/SC (KR) | MN880805 |
HRT | Hurt | Wayne | 12-Jul-2014 | 4 | 4 | Beaver Creek/BE (CR) | MN880846, MN880847, MN880848, MN880849 |
JES | Jesse | Wayne | 28-Sep-2013 | 4 | 4 | Otter Creek/OT (CR) | MN880836, MN880840, MN880844, MN880845 |
JGR | John Griffin | Jackson | 31-Jul-2014 | 4 | 2 | Horse Lick Creek/HL (RR) | MN880801, MN880802, MN880803, MN880804 |
KOG | Koger | Wayne | 28-Sep-2013 | 1 | 4 | Beaver Creek/BE (CR) | MN880850 |
LAI | Lainhart #1 | Jackson | 1-Aug-2014 | 4 | 1 | Station Camp Creek/SC (KR) | MN880798, MN880799, MN880800, MN880816 |
LAK | Lakes | Jackson | 31-Jul-2014 | 3 | 2 | Horse Lick Creek/HL (RR) | MN880792, MN880796, MN880797 |
MOR | Morning Hole | Jackson | 14-Aug-2014 | 2 | 1 | Station Camp Creek/SC (KR) | MN880794, MN880795 |
MUL | Mullins Spring | Rockcastle | 15-Mar-2014 | 2 | 2 | Roundstone Creek/RO (RR) | MN880821, MN880822 |
PHC | Pine Hill | Rockcastle | 15-Mar-2014 | 3 | 2 | Roundstone Creek/RO (RR) | MN880830, MN880831 |
PIN | Piney Grove | Pulaski | 20-Oct-2013 | 3 | 3 | Pitman Creek/PI (CR) | MN880855, MN880856, MN880857 |
POU | Pourover | Pulaski | 20-Oct-2013 | 4 | 3 | Buck Creek/BU (CR) | MN880858, MN880859, MN880860, MN880861 |
RCH | Richardson’s | Pulaski | 20-Oct-2013 | 4 | 3 | Pitman Creek/PI (CR) | MN880866, MN880867, MN880868, MN880869 |
ROA | Roadside | Pulaski | 4-Jul-2012 | 1 | 3 | Pitman Creek/PI (CR) | MN880862 |
SAV | Savage (Copperas Saltpeter) | Clinton | 28-Sep-2013 | 2 | 4 | Spring Creek/SP (CR) | MN880834, MN880835 |
SOR | Sinks of Roundstone | Rockcastle | 15-Aug-2014 | 2 | 2 | Roundstone Creek/RO (RR) | MN880832, MN880833 |
SRI | Sinks and Rises | Jackson | 1-Aug-2014 | 3 | 2 | Horse Lick Creek/HL (RR) | MN880790, MN880791, MN880793 |
STA | Stab | Pulaski | 4-Jul-2012 | 4 | 3 | Buck Creek/BU (CR) | MN880851, MN880852, MN880853, MN880854 |
STL | Steele Hollow | McCreary | 12-Jul-2014 | 3 | 4 | Little South Fork/LS (CR) | MN880841, MN880842, MN880843 |
TEA | Teamers | Rockcastle | 15-Aug-2014 | 4 | 2 | Roundstone Creek/RO (RR) | MN880823, MN880824, MN880825, MN880826 |
WIND | Wind | Pulaski | 4-Jul-2012 | 4 | 4 | Pitman Creek/PI (CR) | MN880863, MN880864, MN880865, MN880870 |
Collecting localities (Figs
Beetle specimens were collected by hand into 95% ethanol and placed at -20 °C for short-term storage within 48 hours of collection. Ethanol was changed after processing (individuals from each locality were sorted by genus and inventoried) and whole specimens from each location were stored together in 95% EtOH at -80 °C. Table
Depending on the number of specimens available, up to four Darlingtonea individuals per cave were sequenced (for a total of 81 specimens) to capture a sample of within-population mitochondrial cytochrome oxidase subunit I (COI) haplotype diversity (Table
An ~850 bp COI target region was amplified from genomic DNA using the primer pair “Pat” and “Jerry” (
Sequences were aligned using CLUSTALW (
Partial COI sequences were collapsed into haplotypes using the online tool FaBox (
Arlequin 3.5 (
FST estimates the degree of differentiation among subpopulations within the total population. The closer FST is to 1, the greater the extent of allelic fixation or identity within populations (
Distance matrices and network connections among COI haplotypes were also calculated in Arlequin. Fixation indices (
An unrooted split network based on a NeighborNet algorithm was generated in SplitsTree (
Network connections among haplotypes were gathered directly from Arlequin output data, and a minimum spanning network of COI haplotypes was constructed using the program HapStar (
Due to the nearly identical external morphology in adults, male genitalia was also examined in a specimen from each cave sampled to see if any differences could be found and if so, to see if there was any correlation between groups discovered via the genetic analysis.
Successful PCR amplification was found to be less reliable for older samples (some as old as five years), despite storage at -80 °C in 95% or stronger ethanol. Despite careful optimization of thermal cycling conditions, agarose gel purification of PCR products was found to considerably improve sequence read quality and was performed for most samples included in the final data set.
The distribution of cave collection sites and proportions of haplotypes from 27 populations are shown in Figs
The analysis of molecular variance (AMOVA), from which F-statistics (FST, FCT, and FSC) were calculated to describe nucleotide sequence diversity at hierarchical levels, within and among groups from each hypothesis of structure are summarized in Table
AMOVA for the a posteriori structure hypothesis III, based on five distinct genetic clusters from a neighbor-joining network of COI sequences produced the greatest difference between FCT and FSC among all three analyses. In other words, when nucleotide diversity is partitioned among hierarchical levels, variance in nucleotide diversity is maximized among groups and minimized within groups. The northernmost 15 sampled populations make up three genetic clusters within an approximately ten-kilometer physical radius of one another. In this arrangement, no haplotypes are shared between the three groups, and the clusters contradict both a priori hypotheses about the locations of important major and minor water barriers to gene flow, especially in the northern part of the MP-II. Mantel tests of group submatrices found population pairwise FST to be independent of geographic distance within each cluster. Among all 15 of these populations, only a maximum of 14% of the observed variation can be explained by geographic distance.
Examination of male genitalia generally showed only slight differences among cave localities examined (Fig.
Distribution of cave collection sites and proportions of haplotypes from 27 populations of Darlingtonea kentuckensis in eastern Kentucky, USA. Circle area corresponds to number of individuals sampled per locality. Different colors indicate different haplotypes; similarity in hue qualitatively indicates sequence similarity. KR: Kentucky River; RR: Rockcastle River; CR: Cumberland River; MVF: Mount Vernon Fault; DD = drainage divide between Kentucky and Rockcastle rivers.
AMOVA statistics, fixation indices, and results of hypothesis tests for structure hypotheses I (four faunal regions), II (ten watersheds), and III (five genetic clusters).
Source of variation | Degrees of freedom | Sum of squares | Variance components | Percentage of variation |
---|---|---|---|---|
AMOVA I | ||||
Among groups | 3 | 149.425 | 2.19461 (Va) | 56.30 |
Among populations within groups | 23 | 108.884 | 1.44888 (Vb) | 37.17 |
Within populations | 58 | 14.750 | 0.25431 (Vc) | 6.52 |
Total | 84 | 273.059 | 3.89780 | 100 |
Fixation Indices: I | ||||
FSC | 0.85069 | Vb and FSC : P(random > observed) = 0.00000*** | ||
FST | 0.93476 | Vc and FST : P(random < observed) = 0.00000*** | ||
FCT | 0.56304 | Va and FCT : P(random > observed) = 0.00000*** | ||
AMOVA II | ||||
Among groups | 9 | 196.197 | 2.16311 (Va) | 60.85 |
Among populations within groups | 17 | 62.112 | 1.13762 (Vb) | 32.00 |
Within populations | 58 | 14.750 | 0.25431 (Vc) | 7.15 |
Total | 84 | 273.059 | 3.55503 | 100 |
Fixation Indices: II | ||||
FSC | 0.81730 | Vb and FSC : P(random > observed) = 0.00000*** | ||
FST | 0.92846 | Vc and FST : P(random < observed) = 0.00000*** | ||
FCT | 0.60846 | Va and FCT : P(random > observed) = 0.00000*** | ||
AMOVA III | ||||
Among groups | 4 | 221.073 | 3.27840 (Va) | 81.93 |
Among populations within groups | 22 | 37.236 | 0.46852 (Vb) | 11.71 |
Within populations | 58 | 14.750 | 0.25431 (Vc) | 6.36 |
Total | 84 | 273.059 | 4.00124 | 100 |
Fixation Indices: III | ||||
FSC | 0.64818 | Vb and FSC : P(random > observed) = 0.00000*** | ||
FST | 0.93644 | Vc and FST : P(random < observed) = 0.00000*** | ||
FCT | 0.81935 | Va and FCT : P(random > observed) = 0.00000*** |
Frequencies of COI haplotypes and their proportions, color coded for each hypothesis of structure; circle area corresponds to number of individuals assigned to each group. Overlain transparent dots show collecting localities. A Four faunal regions of hypothesis I (fifth region unsampled in this study: see discussion and
A–C Minimum spanning networks of COI haplotypes, color-coded for each hypothesis of structure. A Four faunal regions of hypothesis I B ten watersheds of hypothesis II C five genetic clusters of hypothesis III D A split network of 85 COI sequences revealing the five genetically distinct clusters of hypothesis III.
Representative male genitalia from 17 of the sampled caves: 1 Wells Cave; 2 Pine Hill Cave; 3–5 Wind Cave; 6 Richardson’s Cave; 7, 8 Lainhart #1 Cave; 9, 10 and 15, 16 Pourover Cave; 11, 12 John Griffin Cave; 13 Climax Cave; 14 Hicksey Cave; 17, 18 Stab Cave; 19, 20 Piney Grove Cave; 21, 22 Dykes Bridge Cave; 23 Great Saltpeter Cave; 24 Teamers Cave; 25 Mullins Spring Cave; 26 Jesse Cave; 27 Steel Hollow Cave. Note that Wells and Dykes Bridge Caves were not included in the genetic study.
FST measures allelic identity within populations, or among-population variation. Across partitioning schemes, FST values close to one indicate that individuals within populations are more similar to each other than to individuals in other populations, corroborating the idea that in general, cave populations in this study are isolated from one another. Structure hypotheses I and II were developed based on a priori information about the locations of cave collection sites relative to (I) two hypothesized major geographic barriers to gene flow or (II) ten watersheds of higher-order streams. Results of AMOVA for evaluating structure hypotheses I and II indicated that for both hypotheses, the majority of total variation (56–61%) is accounted for by variation among the groups defined under each hypothesis. These results support both structure hypotheses I and II over a null hypothesis of panmixia. Due to the similarity of results for both structure hypotheses I and II and because they are not mutually exclusive, neither can be concluded to better represent geographic structure of genetic diversity among the populations sampled. Hence both the major rivers and even some of the smaller watersheds may be geographic barriers to gene flow. High estimates of FSC relative to FCT (Table
Structure hypothesis III was developed based on the five genetic clusters resulting from a split network. The boundaries for the five population clusters in this hypothesis were determined solely by clustering based on genetic distances among sequences, independently of any a priori geographic information. AMOVA statistics for structure hypothesis III (Table
If geographic distance is strongly positively correlated with genetic distance, gaps in sampling (rather than specific geographic features acting as barriers to gene flow) could be responsible for at least some of the observed clustering of populations. Results of partial Mantel tests (Table
Results of Mantel tests (10000 permutations) of association between geographic distance and population pairwise FST within and among groups from hypotheses I and III, containing the same 15 northern MP-II populations partitioned in different ways.
Hypothesis (group #) | Populations included | % variation explained by geographic distance | Pobs>sim(α=0.05) |
---|---|---|---|
III (1) | CLF, HIC, LAI, MOR, HIS, SRI, JGR, LAK, CLI | <1 | 0.468 |
III (2) | FLE, PHC, SOR | <1 | 0.6637 |
III (5) | TEA, MUL, GSP | <1 | 0.673 |
I (1) | CLF, HIC, LAI, MOR, HIS | 7 | 0.761 |
I (2) | SRI, JGR, LAK, CLI, MUL, GSP, TEA, SOR, PHC, FLE | 19 | 0.0035 |
all northern MP-II | CLF, HIC, LAI, MOR, HIS, SRI, JGR, LAK, CLI, FLE, PHC, SOR, TEA, MUL, GSP | 14 | 0.0033 |
all 27 populations | 18 | 0.0001 |
The Mount Vernon fault (Fig.
The sampling scheme of our study makes it difficult to extricate signal due to population structure from that due to IBD for the two genetic clusters on either side of the Cumberland River, which are strongly clustered spatially (Fig.
Overall, the limits of neither major nor minor watersheds alone adequately model the observed distribution of genetic diversity across sampled populations of D. kentuckensis. Geographic distance and landscape features, both stratigraphic and fluvial, appear to have each contributed to this distribution. Determination of the boundaries of cryptic species or subspecies, inference of their pattern of relatedness, and identification of predictive characteristics of isolating barriers will require further sampling of additional populations and more complete and/or additional molecular loci.
Based on CO1 data alone, there is a wide range of divergence values between taxa that can be defined as separate species on their own evolutionary trajectory from other lineages (
This research was supported by funds from TKP and JRJ, in addition to internal funding from Western Kentucky University through a Graduate Student Research Grant and a Research and Creative Activities Program (RCAP) grant. Naomi Rowland provided valuable advice and assistance with molecular work. We thank Dr. Matthew Niemiller for his general guidance and valuable comments on the manuscript, Dr. Karen Ober for her contribution of specimens, data, and advice, and Elise Valkanas for her phylogenetic studies on eastern North American cave beetles (supported by an NSF- REU grant) which inspired and contributed to this project. We thank Den and Sheila Roenfeldt for supporting a travel fellowship which allowed OFB to present this research and gather feedback from colleagues at the International Society for Subterranean Biology’s 2016 conference. Our appreciation to Arnaud Faille and one anonymous reviewer whose insightful comments improved the manuscript. We thank the many landowners who granted permission for us to sample caves on private property, as well as Jim Currens, Julian Lewis, Ben Miller, Jason Polk, Lee Florea, Kurt Helf, The Rockcastle Karst Conservancy, the Green River Grotto, John Andersland, Rob Neidlinger, and the Kentucky Speleological Society.