Ecrobia grimmi in brackish Lake Sawa, Iraq: indirect evidence for long-distance dispersal of hydrobiid gastropods (Caenogastropoda: Rissooidea) by birds

Mudsnail species of the genera Hydrobia and Ecrobia ( junior synonym Ventrosia) belong to morphostatic or nonadaptive radiations. Although these genera are anatomically well defined, their species can only be confidently identified using molecular data. During the last decade, the status, distribution and evolutionary history of these coastal, brackish water taxa have been clarified in large part. In general, congeneric (sub)species have nonoverlapping ranges replacing each other ecologically on large geographical scales (Wilke & Davis, 2000; Wilke, Rolan & Davis, 2000; Wilke & Pfenninger, 2002; Wilke, Pfenninger & Davis, 2002; Wilke, 2003; Kevrekidis, Wilke & Mogias, 2005). However, the discovery of isolated inland populations of these snails, such as an Ecrobia in Lake Sawa in Iraq, does not fit into this picture. Lake Sawa, which is located in the southwestern part of Iraq, is a mixomesohaline water body. It has a maximum depth of 5.5 m and no surficial influx or outflow. The lake is fed by groundwaters of the Euphrates and Dammam aquifers and its water chemistry, which is unique among Iraqi lakes, suggests that it is probably of relatively recent karstic origin and not a relic of a mid-Holocene marine incursion (Jamil, 1977; Naqash, Banat & Al-Shamee, 1977; Plaziat & Younis, 2005; Hassan, Al-Saadi & Alkam, 2006). The distances between Lake Sawa and the nearest populations of brackish-coastal hydrobiids are over 800 km to the Caspian Sea and over 900 km to the Mediterranean (Fig. 1). The purpose of this note is to identify the hydrobiid snail that lives in Lake Sawa and to infer its origin based on molecular analyses. We also discuss, within this context, the importance of birds as vectors for the dispersal of coastal hydrobiid gastropods, which has been contested (e.g. Barnes, 1988). Snails in Lake Sawa were collected by hand or washed from stones and fixed in ca 85% ethanol. Dissections indicated that they belonged to the genus Ecrobia. In order to identify the species we sequenced a 638-bp fragment of the cytochrome c oxidase subunit I gene (COI) of three specimens and compared it with that of congeners whose sequences were extracted from GenBank (for E. ventrosa we selected five individuals spread across its entire geographic range) as well as to unpublished sequences from our ongoing phylogenetic studies. Hydrobia acuta acuta, Peringia ulvae and Salenthydrobia ferreri served as outgroups (Table 1). DNA preparation and PCR followed standard procedures (Wilke et al., 2006; Haase, 2008). For sequencing, PCR products were sent to AGOWA (www.agowa .de). Base composition was homogeneous, a prerequisite for model-based phylogenetic analyses. Phylogenetic analyses were conducted based on the optimality criteria maximum parsimony (MP) and maximum likelihood (ML) as well as by neighbour-joining (NJ) and Bayesian inference (BI). Optimal substitution models (HKY þ G) were inferred using jModeltest v. 0.1 (Posada, 2008) and MrModeltest v. 2.3 (Nylander, 2004), respectively. Trees were reconstructed in PAUP* v. 4b10 (Swofford, 2003) and MrBayes v. 3.1.2 (Ronquist & Huelsenbeck, 2003). For MP and ML, starting trees were obtained via stepwise addition. The addition sequence was random and random trees were used as starting points. Tree search went through 10,000 replicates, each limited to 10 rearrangements, which were done by TBR-branch swapping. Bootstrapping was performed with 10,000 (MP, NJ) and 1,000 (ML) replicates, respectively. For BI, four Metropolis-coupled Markov chains each searching the tree space in two parallel runs were sampled every 100th generation of a total of 2,000,000. The burn-in was conservatively set to 500,000 generations. At this point, the standard deviation of split frequencies had already dropped below 0.01, suggesting that both searches were already sampling around the same optimum. Convergence of parameters was monitored using Tracer v. 1.4 (A. Rambaut & A. Drummond, http://tree.bio .ed.ac.uk/software/tracer/) as well as the criteria implemented in MrBayes. The three specimens from Lake Sawa had identical sequences. They differed from the Caspian sequence of E. grimmi in only three positions and therefore are most likely conspecific. However, assigning a specific name can only be done tentatively in this case owing to the large number of nominal hydrobiid taxa that were described from the Caspian and Aral Seas by Starobogatov and co-workers (see Filipov & Riedel, 2009, and references cited therein). Filipov & Riedel (2009) suggested that only a single species is involved, with the oldest available name for this snail being Hydrobia grimmi Clessin & Dybowski, 1888. Hence, we tentatively assign the hydrobiid snail that lives in the Caspian Sea and Lake Sawa to this species, pending detailed taxonomic study of the hydrobiid taxa of the Ponto-Caspian region. The tree reconstructions are summarized in Figure 2 based on the ML topology, which had the highest resolution. The closest relative of E. grimmi was, as expected based on distributional evidence, E. maritima, which lives in the Black Sea and adjacent Mediterranean (Kevrekidis et al., 2005). Surprisingly, the sister species of this clade was E. truncata, a Western Atlantic species. Ecrobia ventrosa, which is widely distributed in the Mediterranean and northeast Atlantic (Wilke & Davis, 2000), was sister to the above species. An unnamed species from the Tunisian island of Djerba branched off at the base of the tree; this will be dealt with in a comprehensive phylogeographic analysis of the entire genus that is in progress. The support values of the four tree-reconstruction methods were very heterogeneous (Fig. 2). Posterior probabilities usually are expected to be higher than bootstrap values. However, this was only the case in two of the seven nodes supporting the species and their relationships. Two nodes were Correspondence: M. Haase; e-mail: martin.haase@uni-greifswald.de