Genomic data reveal similar genetic differentiation between species with vastly different dispersal capabilities and life histories

Abstract

Little is known about the life histories, population connectivity, or dispersal mechanisms of shallow groundwater organisms. Here we used RAD-seq to analyze population structure in two aquifer species: Paraperla frontalis, a stonefly with groundwater larvae and aerial adults, and Stygobromus sp., a groundwater-obligate amphipod. We found similar levels of connectivity in each species between floodplains separated by ~70 river km in the Flathead River basin of NW Montana, USA. Given that Stygobromous lacks the aboveground life stage of P. frontalis, our findings suggest that aquifer-obligate species might have previously unrecognized dispersal capacity.

1. Introduction

In 1974, Stanford and Gaufin [1] reported stoneflies in an alluvial aquifer supplying domestic water to a Montana community. Researchers have since documented diverse communities of macroinvertebrates, meiofauna, and microbes in shallow aquifers worldwide, and explored the conservation and ecological importance of floodplain habitats [e.g., 2,3]. These communities include insects that spend time above ground (amphibionts), as well as taxa (e.g. crustaceans, oligochaetes, and mites) that never leave interstitial spaces in aquifers (stygobionts). These animals occur up to 10 meters beneath riverine floodplains and as many as two km from the main river channels [4]. Amphibiotic stoneflies spend 1-3 years maturing in the aquifer before emerging as adults with an aerial lifespan of only a few days [5]. The life cycles of many stygobionts, including Stygobromus, are largely unknown.

Shallow aquifers offer many challenges to resident organisms including geologically bounded isolation, no light, variable water flow, and reduced availability of carbon, other nutrients, and oxygen [6,7]. Recent stygobiont research has focused on the ecology of shallow groundwater environments, noting the variable influence of many factors on spatial distribution, including bedrock geology, soil permeability, water chemistry and quality, groundwater levels, adjacent surface flows, riparian vegetation, and climate [8–11]. A recent study documented the role of methane in subterranean food webs [7]. Previous genetic studies identified widespread, long-term barriers to dispersal by groundwater species, even within drainages, despite potentially linking floods [12–15].

The ability of groundwater organisms to actively disperse within and between adjacent watersheds remains unclear. Life history surely plays a significant role in population connectivity, with the retention of ephemeral, winged life stages by amphibionts (e.g., stoneflies) offering dispersal advantages over taxa that never leave the groundwater (e.g., amphipods). Current genomic tools facilitate the study of local adaptation to atypical environments [16] and can resolve fine-scale differentiation in aquatic insects [17–19]. A better understanding of dispersal in these systems would greatly benefit biological conceptualization of connectivity along the river corridor, a major theme in river ecology and management [e.g., 20].

Here we address these issues with RAD-seq datasets for two groundwater species from two well-characterized floodplains of the Flathead River in northwestern Montana. This study design is novel and powerful, using cutting-edge techniques to compare population connectivity in two co-occurring, but ecologically distinct, species.

Methods

Study sites

Our samples are from two Flathead River floodplains that are separated by ~70 km: the Nyack and the Kalispell (Fig. 1). Both locations are underlain by high-porosity alluvial aquifers that are entirely recharged by river water. The aquifers are likely not connected in the subsurface because the floodplains are bounded by bedrock knickpoints [21], and each is known to contain a diverse array of meiofauna and macroinvertebrates [7,22]. These floodplains have been the focus of long-term research [e.g., 23], and permanent wells allow sampling of groundwater habitats.

Figure 1.

Sampling sites, from region to floodplain to individual wells, for two species of groundwater invertebrates included in this study.

Taxa and sampling

We sampled two taxa that exemplify major life history strategies: Paraperla frontalis, a hyporheic stonefly with a winged adult stage, and Stygobromus, a blind, pigmentless, groundwater-obligate crustacean. We hypothesized that the stonefly would display higher gene flow and less genetic structure between floodplains than the groundwater amphipod.

We collected specimens from seven permanent wells in June of 2011 and 2012, though both species were not always sampled from the same wells (Fig. 1). Because of limited sampling, all Stygobromus from Kalispell were considered one population in analyses. We extracted DNA from 96 individuals of each species and confirmed species identifications by COI barcoding before proceeding with RAD-seq.

SNP calling and filtering

We prepared RAD-seq libraries for both species following standard protocols using the restriction enzyme SbfI and unique 6-bp barcodes [24]. We sequenced 192 samples on two lanes of an Illumina HiSeq 2500 with 100-bp, single-end chemistry. Raw sequences were quality filtered with 90% of the bases required to have a quality score ≥20. Reads for Stygobromus were further trimmed to 80 base pairs to maximize read quality. We used the process_radtags script in Stacks v.1.19 [25] to demultiplex reads by barcode, removing any with an uncalled base. We called SNPs in the program Stacks [25] with a read depth (-m) of 5 and a maximum of two mismatches (-n) per locus and between catalog loci (-N). Additional SNP calling details are provided in Supplementary Materials. When necessary, we used PGDSpider 2.1.1.1 [26] for data conversion. To investigate the influence of data scale (SNP number) and missing data on our results, we constructed four datasets for downstream analyses. All filtering steps applied to both species and all datasets received the baseline filtering described above. (1) We removed all individuals with >50% missing data, and all loci genotyped in < 60% of individuals. We also removed all loci that were not present in at least 50% of individuals in each population. (2) No additional filters. (3) We removed loci genotyped in <25% of individuals. (4) We removed loci genotyped in <75% of individuals. For datasets 2-4, we also removed any individuals with more than one standard deviation of missing data above the mean in the raw SNP dataset.

Pair-wise differentiation and population structure

We calculated pairwise F_ST values using GENEPOP [27,28] for all population pairs using dataset #1. Next, we explored population structure in two ways for four datasets. We used Structure v2.3.4 [29], a Bayesian iterative algorithm, and a discriminant analysis of principal components (DAPC) implemented in the R package adegenet [30]. For both methods, we tested K=1-7 for P. frontalis and K=1-6 for Stygobromus. Structure analyses were performed on dataset #1 and DAPC analyses were performed on datasets #2-4. Complete details of population structure analyses are provided in Supplementary Materials.

Results

Sequencing and genotyping

We observed lower coverage depth and fewer SNPs despite a much larger SNP catalog (e.g., 2-3x) for Stygobromus versus P. frontalis. This is likely because the genome of Stygobromus is much larger than P. frontalis [see 31]. For each dataset, we identified: (1) 806 SNPs for 90 P. frontalis and 314 SNPs for 50 Stygobromus, (2) 3,187 SNPs for P. frontalis and 543 SNPs for Stygobromus, (3) 1,175 SNPs for P. frontalis and 476 SNPs Stygobromus, and (4) 279 SNPs for P. frontalis and 167 SNPs for Stygobromus. For datasets 2-4, we included 84 and 46 individuals of P. frontalis and Stygobromus, respectively (Table 1).

Table 1.

Sample sizes of the amphipod Stygobromus sp. and the stonefly Paraperla frontalis, including total numbers of individuals with DNA initially extracted and those that met the quality requirements for the final RAD-seq data sets (#1 and #2-4).

Pair-wise differentiation and population structure

We calculated pairwise F_ST values within and between floodplains for both species using dataset 1 (Table 2). However, because we kept just 10 Stygobromus individuals from the two Kalispell wells, we did not calculate pairwise F_ST within that floodplain. We found genetic differentiation between floodplains to be low for both species (F_ST=0.004 for Paraperla and F_ST=0.000 for Stygobromus). These low F_ST were likewise reflected in low population pairwise F_ST values (Table 2).

Table 2.

Pairwise F_ST values for samples of Paraperla frontalis and Stygobromus sp. as calculated from data set #1 which included 806 SNPs for 90 P. frontalis individuals and 314 SNPs for 50 Stygobromus sp. individuals. Bold names indicate different floodplains.

For population structure analyses, we identified optimal values of K=2 for P. frontalis and K=5 for Stygobromus (Figs. S1–2). DAPC analyses revealed similar levels of population structure for both species. For P. frontalis, the optimal K ranged from K=2 (datasets 2-3) to K=3 (dataset 4; Fig. 2A). For Stygobromus, the best-fit DAPC K was either K=5 (dataset 2), K=4 (dataset 3), or K=3 (dataset 4; Fig. 2B), with all K’s highlighting no obvious patterns of geographic structuring.

Figure 2.

Comparisons of population structure for DAPC analyses of data sets #2-4 which include varying numbers of SNPs for A) Paraperla frontalis and B) Stygobromus sp. For each plot, the number of clusters supported (K), filters employed, number of SNPs, and percent of missing data are included. Filter abbreviations include: >75% miss. and >25% miss. = removal of any loci with greater than 75% or 25% missing data, respectively. Each vertical bar represents one individual, and best-fit K’s were identified by plotting the Bayesian Information Criterion (BIC) for A) K = 1-7 and B) K = 1-6. Lower BICs indicate higher model support.

Discussion

Shallow alluvial aquifers are refugia for diverse, functionally unique species that are inherently difficult to study. Thus, we are only beginning to understand such fundamental attributes as their dispersal ability, population structure, mating habits, and adaptability. Our study compared winged stoneflies with groundwater amphipods, species we expected to exhibit vast differences in their dispersal capacity. Using hundreds to thousands of loci, we have shown a lack of regional structure in both our study species, a surprising result given their different life histories, predicted dispersal differences, and the barriers (geologic and geographic) separating populations.

There are several possible explanations for this result. First, the deep bedrock, phreatic, water channels underlying this region may connect distant floodplains, thus facilitating dispersal of one or both of our taxa. However, entry of groundwater taxa into such channels may be limited because they lack the direct surface-to-groundwater connection present in the hyporheic zone. Furthermore, the environment of these channels would be challenging to shallow aquifer taxa, with longer residence times and potentially limiting temperatures, carbon availability, and oxygen levels. Second, it is possible that Stygobromus individuals irregularly enter the river current through upwelling and passively disperse beyond bedrock nickpoints to become established in downstream floodplains. Flood events may contribute to their movement as well, though they would still have to be brought out of the hyporheos in some way. Even very low dispersal rates could lead to genetic homogeneity [32].

In summary, we applied powerful genomic tools to a fundamental biological question in a long-studied system and found a novel, surprising result. Our results clearly indicate that differences in dispersal capacity cannot be definitively assumed from life history differences, even something as extreme as purely subterranean versus winged life stages. Ongoing work will extend these results with more robust taxonomic, genomic, and geographic sampling. In the future, we expect to better resolve the historical and ecological drivers of groundwater genomic and phylogenetic diversity.

Author contributions

SJ, JS, and GL designed the study. SJ, CN, JS, and GL did field and/or lab work. SJ, BKH, and SH performed analyses. All authors contributed to manuscript preparation.

Data accessibility

Data will be deposited in GenBank and other repositories.

Funding

Funding was provided by Bucknell University Biology, NSF, University of Montana.

Competing interests

We have no competing interests.

Ethical statement

Invertebrates were sampled humanely under permits from USGS, FWS, and the state of Montana.

Acknowledgements

We thank all field and lab technicians, as well as Mike Miller, Sean O’Rourke, Omar Ali, and Laney Hayssen for laboratory assistance.