Systematics and historical biogeography of the Aphanius dispar species group (Teleostei: Aphaniidae) and description of a new species from Southern Iran

1.1 Aphanius dispar (Rüppell, 1828)

Aphanius dispar as currently understood (“A. dispar” in the following) is found over an area that extends from the easternmost Mediterranean Sea, the Red Sea, and the Persian Gulf to the Indian Ocean and the southern coastline of Pakistan (Wildekamp, 1993) (Figure 1a,b). Its principal habitats are coastal lagoons, but it is capable of thriving in inland waters to which it may have migrated from the coast in the past (Krupp, 1983; Reichenbacher, Feulner, & Schulz-Mirbach, 2009). A notable feature of “A. dispar” is its astonishing ecological tolerance. It can survive in all types of endorheic and exorheic river systems, including very small creeks and large streams, hot sulfur-rich springs (in Iran), and the springs in the Dead Sea Valley (Krupp, 1983; Teimori, Schulz-Mirbach, et al., 2012; Wildekamp, 1993).

Individuals of “A. dispar” may reach 50 mm (very rarely up to 70 mm) in total length and are characterized by a moderately slender body with a prominent pigmentation pattern consisting of blotches, flank bars, or a combination of the two. The name of the species refers to the clear sexual dimorphism with respect to pigmentation (only the males have two black vertical stripes on the caudal fin) and anal and dorsal fin lengths (which are larger in males) (Krupp, 1983; Rüppell, 1828; Villwock, Scholl, & Krupp, 1983; Wildekamp, 1993).

Formerly, two subspecies were accepted, that is, A. dispar dispar (Rüppell, 1828) and A. dispar richardsoni (Boulenger, 1907). One other subspecies has also been described as A. dispar stoliczkanus (Day, 1872), but this has been considered as a synonym for A. dispar dispar in more recent work (e.g., Wildekamp, 1993). Data on the phenotypic variation between isolated populations of “A. dispar” have been provided by several studies (Keivany & Ghorbani, 2012; Krupp, 1983; Masoudi et al., 2016; Teimori, Schulz-Mirbach, et al., 2012; Villwock et al., 1983; Wildekamp, 1993), and also variation in otolith morphology within “A. dispar” was described (Reichenbacher, Feulner, et al., 2009; Reichenbacher, Kamrani, Esmaeili, & Teimori, 2009; Reichenbacher, Sienknecht, Küchenhoff, & Fenske, 2007; Teimori, Schulz-Mirbach, et al., 2012; Teimori, et al., 2012). Furthermore, the impact of environmental parameters on the growth, feeding habits, and reproduction of “A. dispar” has been studied (Frenkel & Goren, 1997, 2000; Homski, Goren, & Gasith, 1994; Lotan, 1969).

It is obvious from the aforementioned studies that the wide geographical distribution of “A. dispar” and its presence in what are now isolated habitats have resulted in considerable intraspecific diversity with regard to fin sizes, pigmentation, morphometric and meristic characteristics, and otolith morphology. However, the only studies that have yet considered “A. dispar” based on molecular genetic analyses are those of Hrbek and Meyer (2003), Chiozzi et al. (2017), and Freyhof et al. (2017). Hrbek and Meyer (2003) showed that “A. dispar” splits into two major clades (a Persian Gulf clade and a Red Sea clade) and does not represent a monophyletic species because it also includes A. ginaonis (Holly, 1929) (see Figure 1c). In light of the work of Hrbek and Meyer (2003), it seemed likely that the broad range of intraspecific variation in “A. dispar” reflects diversification due to geographical isolation and that the taxon encompasses more than one species. A similar conclusion was presented by Chiozzi et al. (2017), who demonstrated that Aphanius cf. dispar in the Danakil Depression in northeastern Africa reveals high genetic diversity and is clearly separated from “A. dispar sensu stricto.” Freyhof et al. (2017) showed the validity of A. richardsoni from the Dead Sea drainages, and of A. stoliczkanus from the coastal areas of the Persian Gulf, the northern Arabian Sea east to Gujarat in India, the Gulf of Oman, and some endorheic basins in Iran and Pakistan. These authors also described A. kruppi Freyhof, Weissenbacher and Geiger, 2017 as new species within the A. stoliczkanus group from Oman.

In this study, we (i) clarify the phylogenetic relationships between the Persian Gulf, Red Sea, and Dead Sea populations of “A. dispar”, (ii) define the species subsumed within the taxon and describe a new species from southern Iran, (iii) present a time-calibrated phylogeny using fossils as calibration points, and (iv) discuss the biogeographical history.

2.1 Ethics statement

This study was carried out in strict accordance with applicable national and international guidelines. The study was approved by the Ethics Committee of the Biology Department of the Shiraz University (SU-909830).

2.2 Sampling

The dataset comprises 286 individuals (+ four tissue samples) of “A. dispar” from 21 sites in the Persian Gulf and Gulf of Oman, the Red Sea, and the Dead Sea (Table 1; Figure 1). These individuals were used for molecular genetic and morphological analyses (see Table 2 for details). In addition, we collected 10 specimens of A. ginaonis (Holly, 1929) (site 2) and eight individuals of A. furcatus Teimori et al., 2014 (sites 1, 3, 4, 8) from the Hormuzgan basin.

The sites in the Persian Gulf area comprised (i) river drainages in SW Iran (Helleh and Karoun-Tigris drainage), S Iran (Hormuzgan basin), and SE Iran (Makran-Jazmourian-Mashkid drainages), and (ii) coastal sites in the Gulf of Oman, the United Arab Emirates, and Bahrain (see Figure 1a,b). The samples from the Red Sea and the Dead Sea, respectively, come from a stream near Djibouti (drainage of Lake Assal) and a spring close to Ain Abata in Jordan. Details of the sampling sites and numbers of studied specimens (including comparative materials) are given in Table 1. No material was available from A. kruppi and A. sirhani, and our differential diagnosis concerning these species is based on Freyhof et al. (2017).

2.3 Molecular genetic analyses

All samples of “A. dispar,” A. ginaonis and A. furcatus shown in Table 1 were used for the molecular genetic studies. Sequences from GenBank were added for the outgroup taxa, that is, further species of Aphanius from Iran, Turkey, and the Mediterranean Sea, and some other cyprinodontid taxa (see Table 2).

DNA was extracted from fragments (on the order of a few mm³) of muscle tissue from the right side of the caudal peduncle, using a commercial DNA extraction kit (DNeasy Tissue Kit, Qiagen) and following the manufacturer's protocol. Three mitochondrial genes (cytochrome b (cyt b), 12S rRNA, and tRNA-Val) were amplified by polymerase chain reaction (PCR) using the specific primers shown in Table 3. The amplification reaction was performed as follows: initial denaturation at 94°C (2 min), 35 cycles of denaturation at 94°C for 45 s, annealing at 58°C for 60 s and extension at 72°C for 90 s, and a final extension phase at 72°C for 10 min. The PCR mixtures were prepared with 2.5 μl of 10× Standard Taq Reaction Buffer (New England BioLabs), 0.5 μl of 10 mM dNTPs, 0.5 μl of each primer (0.2 RM final concentration), 0.2 μlTaq DNA polymerase and 1–2 μl DNA (<1,000 ng), and brought to a final volume of 25 μl with ddH₂O. The products were visualized on a 1% agarose gel and then purified by precipitation with PEG solution (PolyEthylene Glycol, modified from A. Rosenthal, Coutelle, & Craxton (1993), which contains 10.0 g PEG, 7.3 g NaCl, and 45 ml ddH₂O. Forward and reverse strands were sequenced with the PCR primers and BigDye 3.1 chemistry following the manufacturer's protocol (Applied Biosystems) on an ABI3730 automated sequencer in the Genomic Sequencing Unit at LMU Munich. Sequences were deposited in GenBank (Table 2).

After primer trimming in Geneious 10.1.3, the new data sets consisted of 1,134 alignable characters for cytb, and a total of 518 characters for 12S rRNA and tRNA-Val. To provide a data matrix for analyses, sequences from all studied genes were trimmed to the size of the smallest fragment to minimize the amount of missing data; the final data set (cyt b + 12S rRNA and tRNA-Val) consisted of 1,064 bp.

Sequences of the cytochrome b gene were aligned using MUSCLE 3.6 (Edgar, 2004), as implemented in Geneious 10.1.3 (Biomatters Ltd, Auckland, New Zealand). Alignments of 12S rRNA and tRNA-Val were constructed manually based on secondary structural models. All regions whose alignment was ambiguous were excluded from the phylogenetic analyses.

The models best fitting the concatenated data set were obtained with JModelTest (Darriba, Taboada, Doallo, & Posada, 2012), using the Akaike information criterion (AIC) (Akaike, 1973) and Bayesian information criterion (BIC). JModelTest output based on both criteria suggested the GTR + G + I model as the best fit for the current data (Table S1). Accordingly, Bayesian inference was accomplished by MrBayes v. 3.1.2 program (Ronquist & Huelsenbeck, 2003) using the GTR + G + I model.

MrBayes carries out Metropolis-coupled Markov chain Monte Carlo sampling. Bayesian analyses employed four Markov chains, with random starting trees, default priors, and chains with 1 × 10⁶ generations length. Analyses were terminated once the average standard deviation of split frequencies for the simultaneous analyses fell below 0.01. Maximum-likelihood reconstruction searches of the concatenated dataset were conducted with RAxML v. 7.2.5 (Stamatakis, 2006) using a GTR + G + I model of nucleotide substitution, randomized MP starting trees, the fast hill-climbing algorithm, GTR rates, with CAT approximation of rate heterogeneity and fast bootstrap (2,000 bootstrap replicates).

2.4 Morphological analyses

As mentioned in the introduction, “A. dispar” exhibits a strong sex dimorphism, which is visible in the pigmentation pattern and coloration of the males and females, respectively, but can also concern morphometric and meristic characters (see Teimori, Schulz-Mirbach, et al., 2012). As a result, morphometric analyses and meristic counts were carried out separately on all studied male and female specimens (see Supporting Information, Tables S2 and S3). Eighteen linear distances were measured to the nearest 0.5 mm using a Vernier calliper. Measurements were standardized based on standard length (SL), pre-anal distance, head width, head length (HL), eye diameter, and minimum body depth, and 17 morphometric variables were calculated. Meristic counts included the numbers of the dorsal, anal, pectoral, and pelvic fin rays and gill rakers under a stereomicroscope. Vertebra counts include the last vertebra (= uroterminal centrum) and were based on X-ray photographs of specimens of all studied species, with the exception of A. ginaonis. All data are presented in Tables S2 and S3 of the Supporting Information.

Morphometric variables and meristic data were analyzed using PASW 21.00 (SPSS Inc, 2015) and PAST (Paleontological Statistics, version 2.17c; Hammer, Harper, & Ryan, 2001). Most of the morphometric variables were normally distributed (Shapiro–Wilk, p > .05). The exceptions were “length of pelvic fin/HL,” “length of pelvic fin/SL,” and “length of dorsal fin/SL’.” Principal component analysis (PCA) and canonical discriminant analysis (CDA) were used to display the differences between the groups in multivariate space (see Section 2).

2.5 Divergence time estimates

The extinct species †A. princeps Gaudant & Rovira-Sendrós, 1998; from the late Early Miocene (16–17 Mya) of Spain is the oldest fossil that can be securely assigned to Aphanius because it displays the characteristic dentary and tricuspid jaw teeth of the genus (see Gaudant & Rovira-Sendrós, 1998). We used this fossil to define a minimum age for Aphanius. Other fossil species of Aphanius cannot be used in our analysis because their phylogenetic position within the tree is not known due to the lack of a character matrix for Aphanius that is based on morphological characters (and thus applicable to a fossil). For the same reason, the only fossil species from Iran, that is, A. persicus from the Late Miocene (7–9 Mya) of northwestern Iran (see Gaudant, 2011; Reichenbacher et al., 2011) cannot yet be used as additional calibration point. We have set the maximum age for Aphanius at 34 Ma using the oldest known fossil of the Old World killifishes, that is †Prolebias stenoura Sauvage, 1874 (see Gaudant, 2012).

Divergence time was estimated in BEAST v1.7 (Drummond, Suchard, Xie, & Rambaut, 2012) using the uncorrelated log-normal relaxed clock model (UCLN) (Drummond, Ho, Phillips, & Rambaut, 2006) and the birth–death model. One major advantage of BEAST is its ability to estimate the topology and divergence time simultaneously (Drummond & Rambaut, 2007). BEAUti (Bayesian Evolutionary Analysis Utility version) v1.6.1 (Drummond et al., 2012) was utilized to generate initial xml files for BEAST. A starting chronogram that satisfied all monophyly and initial divergence was generated under penalized likelihood in r8s v1.71 (Sanderson, 2002) using the RAxML tree. The substitution model was GTR + G + I with four rate classes. Five replicates of the Markov chain Monte Carlo (MCMC) analyses were run for 100 million generations. The maximum credibility tree, with means and 95% highest posterior density of divergence times, was identified with TreeAnnotator v1.6.1 (Rambaut & Drummond, 2010).

2.6 Nomenclature

The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence, the new names contained herein are available under that code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved, and the associated information viewed through any standard web browser by appending the LSID to the prefix “http://zoobank.org/”. The LSID for this publication is urn:lsid:zoobank.org:pub: DA995390-A7F3-4455-A9B5-E1F2436FAB3C. The electronic edition of this work was published in a journal with an ISSN and has been archived and is available from the following digital repositories: PubMed Central, LOCKSS.

3.1 Molecular genetic analysis

Maximum likelihood and Bayesian estimation methods resulted in similar topologies of relationships among the studied groups and revealed a clear differentiation among the specimens previously identified as “Aphanius dispar” (Figures 2, S1 and S2). Four distinct clades can be recognized, which reflect the main areas of distribution of the taxon: (i) the Red Sea, (ii) the Dead Sea, (iii) Hormuzgan (S Iran), and (iv) the Persian Gulf & Gulf of Oman (Figure 2). The Hormuzgan clade is sister to the Persian Gulf & Gulf of Oman clade, and these two represent the youngest of the four. They, in turn, display a sister group relationship to the Dead Sea clade, and all three are sister to the Red Sea clade, which is the oldest one. Furthermore, two subclades can be distinguished within the Hormuzgan clade, one is represented by A. ginaonis, while the second includes all other specimens from the Hormuzgan Basin studied here (Figure 2).

Thus, the phylogenetic relationships revealed by the analyses of the three mitochondrial genes imply that the designation “A. dispar” actually covers four distinct species. The species from the Hormuzgan basin is introduced here as A. hormuzensis sp. nov. The remaining three correspond to the former subspecies and are designated here as A. dispar (Rüppell, 1828), A. richardsoni (Boulenger, 1907), and A. stoliczkanus (Day, 1872).

3.2 Systematics

Subdivision Teleostei sensu Arratia (1999)

Order Cyprinodontiformes Berg, 1940

Suborder Cyprinodontoidei Parenti, 1981

Family Aphaniidae Hoedeman, 1949

Genus Aphanius Nardo, 1827 (type species A. fasciatus Valenciennes, in Humboldt and Valenciennes, 1821)

Aphanius hormuzensis sp. nov. urn:lsid:zoobank.org:act:DA995390-A7F3-4455-A9B5-E1F2436FAB3C (Figure 3c)

Holotype Male (ZM-FISBUK 157) (Figure 3a).

Paratypes 11 males (22.1–33.9 mm SL), five females (ZM-FISBUK 164–179).

Referred specimens See Table 2.

Type locality Iran, Hormuzgan Province (S-Iran), 330 m, Mehran River, Gotab village, 15 km south of Bastak (E 54°,15′,46.1″ N 27°,08′,39.8″), 29 August 2015, Azad Teimori.

Further localities See Table 2.

Material used in the molecular genetic analysis See Table 2.

Etymology The name refers to the basin name “Hormuzgan” in S Iran.

Diagnosis This species is well distinguished by combination of the fish coloration (in males) and a unique combination of otolith characters. The males have mostly 13–17 vertical light flank bars and two crescent-shaped dark brown stripes on the caudal fin. The vertical flank bars are usually straight (Figure 3a), but can also be somewhat scattered (Figure 3c) or be replaced by vertical series of white spots (Figure 3d). Otoliths are rounded trapezoid to rounded triangular with a rounded or relatively straight dorsal rim and no dorsal tip; the rostrum and antirostrum are relatively short and terminally rounded, the rostrum is slightly longer than the antirostrum (see Figure 5). In addition, the new species has a slightly higher total number of vertebrae (27–28) than A. dispar (24–25) and A. richardsoni (26) and also slightly more caudal vertebrae (16) than A. dispar (13) and A. richardsoni (15) (see Tables S2 and S3; Figure 4).

3.3 Analysis of phenotypic variation in the Aphanius dispar species group

As is usually the case in Aphanius, the morphometric and meristic characters of the species of Aphanius studied here show considerable overlap (Tables S2 and S3). As a result, it is difficult to separate them on the basis of a single morphometric or meristic character. This is consistent with the outcome of earlier work, which in most cases has found broad overlap in morphometric and meristic characters between closely related species of Aphanius (e.g., Coad, 2009; Esmaeili, Teimori, Gholami, et al., 2014; Hrbek et al., 2006; Villwock et al., 1983). Nevertheless, compared to the other studied species, A. ginaonis tends to have a longer head (relative to the SL), while A. dispar usually has the highest number of dorsal fin rays (8–11) and lowest numbers of total (24–25) and caudal (13) vertebrae (see Tables S2 and S3). Furthermore, species separation is possible in multivariate space. The canonical discriminant analysis based on the morphometric variables revealed high overall classification success both for males (95.0%) and females (95.90%) (Table 4; Figure 6). In contrast, principal component analysis (PCA) of the morphometric and meristic characters failed to separate species, neither among the males nor between the females (plot not shown).

3.4 Analysis of otolith variation in the Aphanius dispar species group

Otoliths of Aphanius and other genera among the Cyprinodontiformes are well known for their species-specific characteristics (e.g. Gholami et al., 2014; Reichenbacher & Reichard, 2014; Reichenbacher et al., 2007). Here, we refer to the otolith characters that were introduced in Reichenbacher et al. (2007) and have proven to be useful for species separation in subsequent work (Gholami, Esmaeili, & Reichenbacher, 2015; Reichenbacher, Kamrani, et al., 2009; Teimori, Esmaeili, et al., 2012; Teimori et al., 2014).

The general otolith morphology of the A. dispar species group is characterized by rounded, triangular, or trapezoid otolith shapes, a well-developed rostrum, a prominent antirostrum, and an almost straight sulcus that is bent toward the ventral rim terminally (see Figures 5b,c and 7a,b for otolith terminology). Aphanius dispar (in its new definition) is unusual insofar as its sulcus is very flat and the posterior part of the cauda is not clearly delimited, as in the other three species.

Apart from A. stoliczkanus, the otoliths of which are quite variable, each species can be identified by its characteristic otolith contour. In A. dispar, the shape is triangular to triangular trapezoid and a dorsal tip is present in some specimens (Figure 5k–n); in A. richardsoni, the contour is triangular (Figure 5j); in A. ginaonis, the shape is ovate to elongate and a dorsal tip is lacking (Figure 5g–i); and in A. hormuzensis sp. nov., the contour is rounded trapezoid or rounded triangular and no dorsal tip is present (Figure 5a–f).

Furthermore, the species are well separated from each other based on the dimensions and proportions of rostrum, antirostrum, and excisura. In both A. dispar and A. richardsoni, the relative height of the rostrum is greater than in the other two species, while the rostrum is relatively longer in A. dispar compared to A. richardsoni. In the two species from the Hormuzgan Basin, A. ginaonis, and A. hormuzensis sp. nov., otolith morphology appears to reflect their sister group relationship, because a relatively long and flat dorsal rim (without a dorsal tip) and a deeply incised sulcus are present in both species. However, the A. ginaonis otolith is clearly differentiated from that of A. hormuzensis sp. nov. by its greater relative rostrum length and excisura depth (see Figure 5g–i versus a–f)).

The otoliths of A. stoliczkanus from the population east of the Hormuz Strait possess a consistent overall morphology (Figure 7i–p): The otolith shape is rounded triangular with a prominent dorsal tip, the sulcus is not as flat as in A. dispar or as deep as in the two species from the Hormuzgan Clade, and the ventral and posterior rims are crenulated. As for the population west of the Hormuz Strait, only the specimens from the coastal sites of the Persian Gulf (Bahrain, Khor Hulaylah) show an overall otolith morphology similar to that seen in the population east of the Hormuz Strait, but their otolith rims are not crenulated (Figure 7h). In contrast, trapezoid to triangular shapes characterize the otoliths of A. stoliczkanus from the inland sites of Iran and Iraq, with the populations displaying distinctive variations in rostrum, antirostrum, and excisura sizes and proportions (see Figure 7a–g).

3.5 Revision of the Aphanius dispar species group

Aphanius stoliczkanus west of the Hormuz Strait

Aphanius stoliczkanus west of the Hormuz Strait is distributed in brackish rivers, freshwater springs, hot sulfur springs, and coastal habitats around the Persian Gulf and in inland sites of Iran and Mesopotamia (see Teimori, Schulz-Mirbach, et al., 2012, as A. dispar), but is not found in the Hormuzgan Basin. In the Karoun Basin in SW Iran, A. mesopotamicus Coad, 2009 is found sympatrically with A. stoliczkanus (Masoudi et al., 2016; as A. dispar group). Based on the genetic, morphometric, meristic, and otolith data presented here, it appears that the population of A. stoliczkanus west of the Hormuz Strait is highly variable (Figure 7a-h, Tables S2 and S3). This is additionally confirmed by a distinctive variation in the contour of the caudal fin and also in the pigmentation. Generally, the caudal fin is slightly widened distally and its end is truncated or slightly indented. However, in the specimens from coastal sites of the Persian Gulf (i.e., Khor Hulaylah (in UAE) and Bahrain), the caudal fin is narrow and tapers toward the tip, while in the Tigris Basin specimens (SW Iran), it has a broad, fan-like shape (Figure 2). The dark brown bars on the caudal fin can be slightly or strongly bent. In their study, Freyhof et al. (2017) have also considered the populations from the United Arab Emirates (UAE) and Bahrain as A. stoliczkanus (Figure 7).

4.1 Historical biogeography

Our new phylogenetic data reveal a clear pattern of distribution of the species previously identified as “A. dispar” (Figure 8). Aphanius dispar (in its new definition) occurs along the coasts of the Red Sea. The Dead Sea Basin is the home of A. richardsoni. Mesopotamia and the region fringing the Persian Gulf west and east of the Strait of Hormuz—with the exception of S Iran (Hormuzgan Basin)—are populated by A. stoliczkanus. The Hormuzgan Basin in S Iran is so far unique in that two closely related species of Aphanius, that is A. hormuzensis sp. nov. and A. ginaonis, live in this small basin.

The biogeographical history of A. dispar and its descendants has been controversially discussed in previous works based on morphological comparisons of the populations from the different regions. Steinitz (1954) argued that a Pleistocene transgression of the Mediterranean Sea into the Jordan River valley had led to the immigration of the ancestor of A. richardsoni and that the today's Dead Sea Valley represents the refuge of its ancient Pleistocene distribution. This idea was taken up again by Por (1975) and Kornfield & Nevo (1976). Akşiray (1955) also argued that A. richardsoni originated during the Pleistocene, but, unlike the other authors cited, he proposed that A. richardsoni could have reached the Dead Sea Valley via dispersal from the Red Sea, by crossing the watershed between the two basins. Thus, all authors agree in proposing a Pleistocene age for the split of A. richardsoni, but disagree from where the ancestor of A. richardsoni has reached the Dead Sea Valley, that is, from the Mediterranean Sea or from the Red Sea. However, the time-calibrated tree of the A. dispar species group (Figure 9) suggests that the split between A. dispar and its sister group (including A. richardsoni) occurred at some time between the late Miocene and the Pliocene (9–4 Mya) and is thus older than that assumed in the above-mentioned studies.

Krupp (1983) was the first to propose a model of past migration routes for A. dispar (in the old definition) that took account of its distribution in Mesopotamia and the Persian Gulf. According to him, A. dispar started to migrate during the early Pliocene (c. 5 Mya) from the Mesopotamian Basin to the Persian Gulf, the Arabian Sea, and further to the SE (Indian Ocean) from where it reached the coastal habitats of the Gulf of Aden and the reefs of the Red Sea. Like Akşiray (1955), Krupp (1983) thought that Pleistocene changes in the drainage systems might have enabled the Red Sea populations of A. dispar to invade the Dead Sea Valley. There, its populations became isolated by the middle or late Pleistocene (75,000–12,000 years B.P.), which eventually led to the evolution of A. richardsoni. However, in the case of this scenario, A. richardsoni should represent the youngest split in the phylogenetic tree, which is evidently not the case (Figure 2). Moreover, the chronogram (Figure 9) suggests that A. richardsoni originated prior to the Pleistocene, as indicated above.

The question remains whether A. richardsoni colonized the Dead Sea Valley from the north, via a transgression of the Mediterranean Sea into the Jordan River System, or by a transgression from the South, that is from the Red Sea. Kornfield and Nevo (1976) provided comparative electrophoretic evidence indicating that A. dispar from the easternmost Mediterranean Sea, A. dispar from the Red Sea and A. richardsoni from the Dead Sea are equally different from each other. This may indicate that a single event isolated the three populations from each other—and that the populations in the easternmost Mediterranean Sea might actually represent a further, as yet unrecognized new species. Such an isolating event could have happened in the aftermath of the early Pliocene transgression (c. 5 Mya) of the Mediterranean Sea up to the depression of the Dead Sea Valley (Lake Kinneret, Jordan Valley, Dead Sea) (Garfunkel, 1997; Horowitz, 2001; Rosenthal, Flexer, & Moller, 2006; Stein et al., 1997), which was connected to the Red Sea at that time (Ben-Avraham, Garfunkel, & Lazar, 2008). In the course of the Pliocene, the sea retreated and the shallow marine connection between the Red Sea, the Dead Sea Valley, and the Mediterranean Sea was closed (Ben-Avraham et al., 2008; Coleman, 1993; Garfunkel, 1997; Rosenthal et al., 2006). We assume that it was this Pliocene closure event that isolated the ancestral populations of A. dispar in the Red Sea, Dead Sea, and Mediterranean Sea from each other (Figure 8). This idea is compatible with the time-calibrated tree (Figure 9), which suggests a 95% credibility interval extending from the late Miocene (9 Mya) to the middle Pliocene (4 Mya) for the split between A. dispar and its relatives.

The subsequent split that separated A. richardsoni from the common ancestor of the species of Aphanius that make up Clades III and IV (Figures 2 and 9) probably happened within a relatively short time (see Figure 9). This implies that the common ancestor reached the shores of the Persian Gulf and the Gulf of Oman via the brackish water bodies of the Dead Sea Valley in the course of the middle Pliocene (Figure 8). This migration model is in principle consistent with Krupp's (Krupp, 1983) hypothesis that the A. dispar populations in the Persian Gulf originated in Mesopotamia, but it differs with respect to the time at which this happened (middle Pliocene versus Pleistocene).

Furthermore, the chronogram (Figure 9) reveals that some time must have elapsed between the arrival of the ancestor of Clades III and IV in the Persian Gulf and Gulf of Oman regions and the split between these clades, resulting in A. hormuzensis sp. nov. and A. stoliczkanus. It appears that this split can be linked to vicariance events during the Pleistocene, when tectonic uplift and compression led to the emergence of isolated drainage systems in SE Iran (Makran and Mashkid), S Iran (Hormuzgan), and SW Iran (Helleh) (e.g., Hatzfeld et al., 2010; Kompani-Zare & Moore, 2001; Regard et al., 2004). Clade III (Hormuzgan clade, A. ginaonis, and A. hormuzensis sp. nov.) diverged perhaps as a result of the development of unusual habitats such as hot sulfur springs in this part of Iran (Kosari Torbehbar & Haghnazar Liseroudi, 2015). A second example of a recently diverged endemic species from the Hormuzgan Basin (unrelated to the descendants of A. dispar) is A. darabensis (see Esmaeili, Teimori, Gholami, et al., 2014).

The two populations that can be recognized within A. stoliczkanus are probably the result of a vicariance event during the global cooling and sea-level fluctuations that occurred in the Pleistocene. The population west of the Hormuz Strait may have found a refuge in the Mesopotamian rivers (Tigris, Euphrates) and the rivers that rise in the Zagros Mountains. The population east of the Hormuz Strait would not have been affected by the fall in global sea level because this region was deeper and remained inundated (Lambeck, 1996; Teller, Glennie, Lancaster, & Singhvi, 2000). It appears that the Holocene sea-level rise at 11,000 BP and re-flooding of the Persian Gulf did not lead to mixing between the populations of A. stoliczkanus west and east of the Hormuz Strait. Most probably, the Strait of Hormuz acted as an ecological barrier to gene flow, since the waters of the Persian Gulf have higher salinities and exhibit more extreme shifts in water temperature than the Gulf of Oman (Hoolihan, Premanandh, D'Aloia-Palmieri, & Benzie, 2004).

The young (Pleistocene) age of the splits between and within Clades III and IV also explains why morphometric and meristic characters are of little use in separating the individual species. The same observation has previously been made with respect to the inland species of Aphanius from the Iranian plateau and Anatolia, which are all very similar in phenetic characters, but clearly different based on molecular genetic analyses (Esmaeili, Teimori, Gholami, et al., 2014; Esmaeili, Teimori, Sayyadzadeh et al., 2014; Gholami et al., 2014; Teimori, Esmaeili, et al., 2012).