Historical faunal exchange between the Pontocaspian Basin and North America

Abstract Ecrobia is a genus of small brackish‐water mud snails with an amphi‐Atlantic distribution. Interestingly, the species occurring in the northwestern Atlantic, Ecrobia truncata, is more closely related to the Pontocaspian taxa, Ecrobia grimmi and Ecrobia maritima, than to the species occurring in the northeastern Atlantic and Mediterranean Sea. At least three colonization scenarios may account for this peculiar biogeographical pattern: (1) a recent human‐mediated dispersal, (2) a historical transatlantic interchange, and (3) a historical transpolar interchange. To test these three scenarios, we used five operational criteria—time of species divergence, first appearance in the fossil record, dispersal limitation as well as environmental filtering and biotic interactions along the potential migration routes. Specifically, we inferred a time‐calibrated molecular phylogeny for Ecrobia and reconstructed a paleogeographical map of the Arctic Ocean at 2.5 million years ago (Mya). Based on the five operational criteria, scenarios 1 and 2 can likely be rejected. In contrast, all criteria support scenario 3 (historical transpolar interchange). It is therefore suggested that a bird‐mediated and/or ocean current‐mediated faunal interchange via the Arctic Ocean occurred during the Late Pliocene or Early Pleistocene. This dispersal was likely facilitated by reduced distances between the Eurasian and North American/Greenland landmasses, marine introgressions, and/or a stepping‐stone system of brackish‐water habitats in northern Siberia, as well as a lack of competition along the migration route. As for the direction of dispersal, the scientific data presented are not conclusive. However, there is clearly more support for the scenario of dispersal from the Pontocaspian Basin to North America than vice versa. This is the first study providing evidence for a natural faunal exchange between the Pontocaspian Basin and North America via the Arctic Ocean.

As for the first scenario, there are numerous records of humanmediated dispersals of Pontocaspian taxa to North America and vice versa, starting as early as the thirteenth century (Petersen, Rasmussen, Heinemeier, & Rud, 1992). Since 1985, 70% of the species that have invaded North America are native to the Pontocaspian Basin (Ricciardi & MacIsaac, 2000). Thus, Ecrobia specimens could have been transported in either direction, for example, as larvae in ballast waters of ocean vessels.
Our working hypothesis follows scenario 3 and assumed a transpolar exchange of Ecrobia individuals, even though dispersal of brackish water taxa between North America and central Asia across the Arctic has not been reported before. To test this hypothesis in the context of the three scenarios proposed, we combined genetic and fossil data for the mud snail genus Ecrobia with ecological, paleogeographical, and biogeographical information. As test statistics, we used a set of five operational criteria (for details see Discussion), involving evolutionary (i.e., time frame of species divergence, age of fossil records) and community characteristics (i.e., dispersal limitations, environmental filtering, biotic interactions). We also composed a paleogeographical map of the Arctic Ocean at 2.5 million years ago (Mya) to infer potential changes in coastlines that may have aided dispersal between populations occurring in North America and the Pontocaspian Basin.
Our study might be of relevance for marine biologists and systematists interested in this ecologically and evolutionary important taxon, for biogeographers studying long-range dispersal processes of aquatic organisms in space and time, and for (paleo)geographers investigating the biological consequences of past environmental changes.

| Materials
This study includes 67 individuals, representing the distribution ranges of all accepted species of Ecrobia (Figure 1). They were collected between 1985 and 2017 from 55 localities (Table A1 in Appendix 1) as part of this or related studies. All specimens were F I G U R E 1 Photographs of ethanolpreserved specimens of Ecrobia spp.
hand-picked and preserved in 80% ethanol in the field. Surveys were conducted in concordance with CBD regulations ("Nagoya Protocol"). Voucher material and DNA samples were deposited in the University of Giessen Systematics and Biodiversity Collection (Diehl, Jauker, Albrecht, Wilke, & Wolters, 2018) in Germany.

| DNA isolation, amplification, and sequencing
DNA was isolated using the CTAB protocol described in Wilke, Davis, Qiu, and Spear (2006). Two mitochondrial markers, the cytochrome c oxidase subunit I (COI) and the large subunit rRNA (16S) genes, were amplified using the primers LCO1490 and HCO2198 (Folmer, Black, Hoeh, Lutz, & Vrijenhoek, 1994) and 16Sar-L and 16Sbr-H (Palumbi et al., 1991), respectively. PCR amplification was performed with an initial denaturation step at 95°C for 1 min, followed by 35 amplification cycles (denaturation at 95°C for 30 s, annealing at 52°C for 30 s, and elongation at 72°C for 30 s), and a final elongation step at 72°C for 3 min.
Bidirectional Sanger sequencing was either conducted on a Long Read IR2 4200 (LI-COR) sequencer or an ABI 3730 XL (Life Technologies).
The 16S rRNA sequences were aligned with AliView 1.23 (Larsson, 2014), using the secondary structure model for the family Hydrobiidae suggested by Wilke et al. (2013). The protein-coding COI sequences, which do not contain insertions and deletions in the Hydrobiidae, were unambiguously aligned using the software package BioEdit 7.2.5 (Hall, 1999). As the first base pairs behind the 5′ end of each primer were difficult to read, the fragments were trimmed, resulting in 638-and 505-bp long overlapping fragments for the COI and 16S genes, respectively.

| Molecular-clock analyses
Prior to the molecular-clock analyses, best-fit substitution models for the COI and 16S data sets were selected using jModelTest 2 (Darriba, Taboada, Doallo, & Posada, 2012;Guindon & Gascuel, 2003) based on the Akaike information criterion (AIC). The models suggested for the COI and 16S partitions were HKY+Γ and GTR+I+Γ, respectively. Molecular-clock analyses were then performed using *BEAST (Zhang, Ogilvie, Drummond, & Stadler, 2018) as implemented in BEAST 1.8.4 (Drummond, Suchard, Xie, & Rambaut, 2012). As outgroups, we used two other members of the subfamily Hydrobiinae, that is, Peringia ulvae and Salenthydrobia ferrerii ( Table   A1 in Appendix 1). Note that we did not test for substitutional saturation as both genes are not considered to be saturated within the Hydrobiidae (Wilke et al., 2001(Wilke et al., , 2013. For calibrating the molecular clock, we used two independent means. First, the beginning and end of the Messinian salinity crisis (MSC), which occurred ca. 5.96-5.33 Mya (see Krijgsman, Hilgen, Raffi, Sierro, & Wilson, 1999), were used as lower and upper bounds to time-calibrate the split between P. ulvae and S. ferrerii (see Wilke, Schultheiß, & Albrecht, 2009). In addition, the substitution rate of the COI partition was constrained using the marker-and trait-specific Protostomia clock rate of 1.24%- Two independent replicates for both strict-and relaxed-clock analyses were run for 40,000,000 generations each, sampling every 2,000th generation and using the birth-death model (Gernhard, 2008) as species tree prior. Convergence of parameters was ensured in Tracer 1.5 (Rambaut & Drummond, 2007), revealing ESS values for all parameters being >200. Replicates were combined using LogCombiner (BEAST package, 50% burn-in). The maximum clade credibility (MCC) tree was identified using TreeAnnotator (BEAST package, no additional burn-in). Strict-and relaxed-clock analyses were compared using the Bayes factor (BF) analysis as implemented in Tracer 1.5 by running 1,000 bootstrap replicates. The BF analysis slightly favored the relaxed-clock analysis over the strict-clock analysis (ln P relaxed: −3,794.0, ln P strict: −3,798.9, BF = 2.10). The posterior distributions of the strict-and relaxed-clock analyses were visualized in DensiTree 2.2.5 (Bouckaert, 2010), which is part of the BEAST 2.4.3 package (Bouckaert et al., 2014).

| Paleogeographical reconstruction
In order to assess potential historical dispersal events of brackish-water taxa, it is important to have a basic understanding of the paleogeography and paleocoastline locations. As preliminary analyses indicated that the split between the North American and Pontocaspian Ecrobia taxa may have occurred during the Late Pliocene or Early Pleistocene (Wilke, 2003), we created a paleogeographical map of the coastlines for this time period based on various sources. These included Knies et al. (2014) for the width between Greenland and Svalbard, Butt, Drange, Elverhøi, Ottera, and Solheim (2002) for the emerged Barents Sea region, Torsvik, Carlos, Mosar, Cocks, and Malme (2002) for the Late Pliocene North Atlantic coastline, and Vinogradov (1967)

| Phylogenetic inference and molecular-clock analyses
Tree likelihoods, Bayesian posterior probabilities (BPP), and estimated divergence times were very similar between the species trees inferred by *BEAST under the strict-and the relaxed-clock models.
The underlying gene tree for the favored relaxed-clock model based on all 67 Ecrobia specimens is provided as Figure A1 in Appendix 2.
The molecular-clock analyses indicated an age for the split be-

| Paleogeographical reconstruction
Northern Hemisphere glaciations played an important role in shaping the landscape at high northern latitudes during the last ~1 My. Erosion at the base of these ice sheets was significant and essentially stripped the continent from its sedimentary cover. As such, preexisting higher elevations were significantly lowered ( Figure 3). This is most notable in the Barents Sea shelf, which at present is submerged and has an average depth of 230 m. Redistributing the sediment volume at the bottom of the ocean in front of what would have been the Barents ice sheet back onto this shallow marine shelf indicates that prior to 1 Mya the Barents Shelf must still have been largely continental (Butt et al., 2002).

| D ISCUSS I ON
The main goal of this study was to test for three biogeographical scenarios that could potentially account for the peculiar sister-

| Support for biogeographical scenarios
Based on our operational criteria (Table 1), scenario 1 (human-mediated transport) can be dismissed. First, the split between the North American and Pontocaspian species (ca. 1.0-2.7 Mya, Figure 2) clearly predates the earliest human migrations ca. 25,000 years ago (Price, 2018). Second, Ecrobia spp. are known from the North American and Pontocaspian fossil records since the Pleistocene (Spencer & Campbell, 1987) and Middle Miocene (Büyükmeriç, Wesselingh, & Alçiçek, 2016), respectively. Therefore, the North America-Pontocaspian faunal interchange of Ecrobia specimens was very likely mediated by natural (non-human) mechanisms.
F I G U R E 2 Species trees based on relaxed-clock *BEAST analyses for Ecrobia spp. and two outgroup species. Left: Maximum clade credibility tree with posterior probabilities (black numbers), mean ages in Mya (gray numbers) and respective 95% HPD (blue bars). Right: DensiTree visualization of the *BEAST posterior distribution. The most popular tree is blue, the next most popular red, the third most popular green, and the rest is dark green Scenario 2 (historical transatlantic interchange) is supported by both the time frame of divergence between North American and Pontocaspian taxa as well as by the fossil records (see above).
Moreover, brackish-water habitats along the Atlantic coastlines could have served as stepping-stones for the active or passive dispersal of Ecrobia specimens, thus mitigating the effects of environmental filtering. However, the minimum straight-line ocean distance between the Pontocaspian area and North America during the early Pleistocene via the Atlantic Ocean (see Figure 3) was >7,500 km, thus constituting a considerable dispersal limitation. Moreover, while crossing the Mediterranean Sea/Atlantic Ocean, Pontocaspian species would have encountered considerable competition from congeners (criterion biotic interactions in Table 1). The latter criterion is of particular concern, as the genus Ecrobia constitutes a nonadaptive radiation. Within such radiations, biotic interactions will very likely result in competitive exclusion (Gittenberger, 1991;Wilke, Benke, Brändle, Albrecht, & Bichain, 2010). Due to these dispersal limitations and potential biotic interactions, a historical transatlantic interchange seems unlikely.

Scenario 3 (historical transpolar interchange) is supported by
both evolutionary criteria-time of species divergence and the earliest available fossil records (see above, Table 1). It is also supported by the three-community criteria-dispersal limitations, environmental filtering, and biotic interactions.  Figure 3). Such water bird flyways between the Caspian Sea and the Arctic Ocean still exist today (Boere & Stroud, 2006). Further dispersal along the coastlines of the Arctic Ocean could have been facilitated by bird-mediated transport (sensu Haase et al., 2010) and larval drift or rafting of adult individuals along the Beaufort Gyre, which has been persisting since the Middle Miocene (Matthiessen, Knies, Vogt, & Stein, 2009). In fact, Thiel and Haye (2006) suggested that Ecrobia specimens are able to disperse over several thousand kilometers by rafting. Overall, the dispersal distance in scenario 3 is at least 3,000 km shorter than in scenario 2.
Moreover, habitat suitability along the potential migration route from the Pontocaspian area to the Arctic Ocean (or vice versa) was F I G U R E 3 Paleogeographic reconstruction of North Atlantic and Pontocaspian coastlines during the Early Pleistocene, ca. 2.5 Mya (dark gray lines), superimposed on the present-day geography (light gray lines). The red arrows indicate potential pathways along which a Late Pliocene-Early Pleistocene marine connection between the Arctic Ocean and the Caspian Sea may have existed. Dots represent sampled localities for Ecrobia spp. according to Table A1  Early Pleistocene were higher than today and the sea-ice cover was periodically reduced (Matthiessen et al., 2009;Melles et al., 2012).
In addition, the Barents Sea shelf was still exposed (e.g., Butt et al., 2002), providing ample space for brackish-water habitats. This may have further mitigated potential effects of environmental filtering.
Finally, and most importantly, there is no evidence for the occurrence of other species of Ecrobia along the potential migration route for a transpolar interchange. As biotic interactions (i.e., species competition) play a major role in the distribution of members of nonadaptive radiations , the absence of this strong filter likely facilitated the spread of Ecrobia sp. across the Arctic Ocean.
Combining all pieces of evidence derived from our operational criteria, the most parsimonious explanation for the faunal interchange of North American and Pontocaspian mud snails would be a historical transpolar dispersal. Thus, our working hypothesis cannot be rejected. To our best knowledge, this is the first evidence for such an interchange dating back to the Late Pliocene/Early Pleistocene.

| Direction of faunal interchange
Whereas our study provides strong evidence for a historical interchange of North American and Pontocaspian mud snails, the direction of dispersal, that is, North America to the Pontocaspian Basin or vice versa, remains to be answered.
Considering natural dispersal events between the Eurasian part of the Arctic Ocean and the Pontocaspian Basin between 1.8 and 0.01 Mya, there are at least 24 Arctic invertebrate taxa that have entered the Pontocaspian Basin from the north with glacial meltwaters (Orlova, 2000). The well-studied Caspian seal, for example, is most likely the descendant of ancestors inhabiting the polar seas (Arnason et al., 2006;Davies, 1958;Fulton & Strobeck, 2010;McLaren, 1960;Palo & Väinölä, 2006). These findings would lend some support to the assumption that the Pontocaspian Basin was the sink and not the source for the dispersal of Ecrobia mud snails. Combining the pieces of evidence provided above, we think that dispersal of mud snails from the Pontocaspian Basin (or brackish-water systems in northern Siberia) to North America provides a straightforward and parsimonious explanation for the peculiar biogeographical patterns seen in Ecrobia spp.

| Limitations of our study
A potential limitation of our study is the lack of nuclear markers. We did amplify the internal transcribed spacer 2 (ITS2) for 46 individuals of Ecrobia spp. However, the resulting phylogenetic tree was unresolved due to incomplete lineage sorting. Moreover, when combining this nuclear dataset with our mitochondrial dataset, support values in the phylogenetic trees did not improve. However, the major conclusions of this paper are drawn from a combination of molecular, fossil, and biogeographical information, with the individual findings being broadly in accordance. Therefore, we believe that additional nuclear data would not have changed the overall outcome of this study.
Moreover, due to the unfavorable ratio of only five studied species but four distribution areas (i.e., Mediterranean Sea, Eastern Atlantic, Western Atlantic, and Pontocaspian), a meaningful ancestral area reconstruction for Ecrobia spp. is not possible, which could potentially have helped to unravel the direction of dispersal. However, the indirect evidence provided above partly compensates for the lack of such an analysis. In addition, an ancestral area reconstruction would not have had a direct effect on the outcome of our scenario testing, which favors scenario 3-a historical transpolar interchange of mud snails. Agency of Applied Ecology. We thank the editor and two anonymous referees for their valuable and constructive comments.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
JV and TW designed the study with contributions of CGCvB, BS, DD and CA. JV, CGCvB and BS conducted the analyses. JV and TW wrote the manuscript with contributions of CGCvB, BS, DD and CA.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in the analyses are available from GenBank (for GenBank accession numbers see Table A1 in Appendix 1).