On the molecular and morphological evolution of continental and insular Cryptorchestia species, with an additional description of C.garbinii (Talitridae)

Abstract Semi-terrestrial talitrid amphipods of the genus Cryptorchestia (sensu Lowry and Fanini 2013) associated with freshwater-soaked leaf litter were known to occur in inland lakes of Turkey and at the shores of the Black Sea. Before 2013 they had been reported as Orchestiacavimana and later as Cryptorchestiacavimana. In our phylogenetic tree, inferred from a mitochondrial and nuclear gene dataset (cytochrome oxidase I (COI), and histone H3 (H3), respectively), we show that these Turkish populations belong to Cryptochestiagarbinii, a common and widespread continental species, which is closely related to C.cavimana (endemic to Cyprus) and C.ruffoi (endemic to Rhodes). For the Turkish and European populations of C.garbinii, we found low levels of both genetic differentiation and morphological variation, and an age-related size variability (increasing at each moult) of the small lobe in the male gnathopod I merus, the main taxonomically diagnostic character for Cryptorchestia. A mainland (C.garbinii) versus insular isolation and in situ speciation (C.cavimana, and C.ruffoi) in the two east Mediterranean islands of Cyprus and Rhodes is discussed in relation to terrestrial Cryptorchestia species endemic to North East Atlantic volcanic islands (Azores, Canary Islands, and Madeira). The incorporation of five Mediterranean and Atlantic Orchestia species in the Bayesian analysis of the two genes (COI, and H3) indicated that both genera Orchestia and Cryptorchestia are not monophyletic.


Introduction
The genus Cryptorchestia Lowry & Fanini, 2013 was erected to accommodate semi-terrestrial and terrestrial species, associated with freshwater, set apart from the marine, supralittoral beachhoppers in the genus Orchestia Leach, 1814. Up to this date, Cryptorchestia contains eleven species (Horton et al. 2018) and is mainly distributed over several North East Atlantic islands (Azores, Canary Islands, Madeira), Europe, and over the Eastern Mediterranean, including the islands of Rhodes and Cyprus .
Semi-terrestrial talitrid amphipods of the genus Cryptorchestia associated with freshwater-soaked leaf litter were known to occur at the shores of inland lakes of Turkey and also of the Black Sea. Before 2013 they had been reported as Orchestia cavimana Heller, 1865 and after that as Cryptorchestia cavimana (Heller, 1865), without considering that C. cavimana is endemic to Cyprus (Ruffo et al. 2014. Özbek et al. (2004) listed the Malacostraca species of the Iznik and Uluabat lakes and recorded O. cavimana (here identified as C. garbinii Ruffo, Tarocco and Latella, 2014) from those lakes. Later, Özbek (2011) summarized the distribution of Ponto-Caspian amphipod species in Turkey and mentioned the presence of O. cavimana (C. garbinii, present study) in Sapanca, Iznik, and Uluabat lakes in the Sakarya and Bursa provinces. A southern record of similar looking specimens was reported from Eğirdir Lake, which is located in the Lake District Region of Turkey (Özbek and Ustaoğlu 2005). Recently, Özbek et al. (2017) reported C. cavimana (probably C. garbinii, see below) from a creek near Silivri, a town bordering the northern coast of the Sea of Marmara.
As part of the studies on the taxonomy and distribution of the genus Cryptorchestia in countries bordering the Mediterranean Sea, talitrid specimens were collected at the shore in a river near Kiyiköy, and in Lake Iznik and Lake Sapanca. These localities are in the Marmara region of Northwest Turkey. Our phylogenetic analysis based on mitochondrial (mt) and nuclear gene sequences confirmed that these Turkish populations belong to C. garbinii. Here, we also used morphological comparative methods for the Turkish populations in relation to European conspecific populations, and provided an additional description of C. garbinii to report on intraspecific variation.
In order to investigate evolutionary patterns, our study compared the continental C. garbinii in relation to the insular C. cavimana (Cyprus), C. ruffoi (Rhodes), and Cryptorchestia species from NE Atlantic islands (Azores, Canary Islands, Madeira). Finally, our molecular framework, obtained by means of Bayesian analysis that included five Mediterranean/Atlantic Orchestia species, indicated that both the genera Orchestia and Cryptorchestia are not monophyletic.

Population sampling
The specimens analysed were collected from three localities in the Marmara region ( Fig. 1). A total of 23 specimens were examined for an additional description and measurements of Turkish C. garbinii populations. Most specimens were preserved in 70% ethanol, and some were mounted on glass slides in Faure's medium for morphological observation, others were immediately preserved in 96% ethanol in the field. The photo of a male (Fig. 2) was obtained with a stereo microscope Leica M 165c, mounted with a Leica DFC450 camera at the Museo Civico di Storia Naturale of Verona. The material is deposited in the Museo Civico di Storia Naturale of Verona (MSNVR), Verona, Italy, and the Naturalis Biodiversity Center (RMNH), Leiden, The Netherlands.

DNA isolation, PCR amplification, and DNA sequencing
Specimens of C. garbinii from Lake Iznik (Bursa Province), Lake Sapanca (Sakarya province), and a stream near Kiyiköy (Kirklareli province) ( Fig. 1) were collected in the end of August 2003 and stored in 96% ethanol. Genomic DNA was extracted from pereopods or whole organisms using QIAamp DNA Mini kit (QIAGEN). A PCR product of ca. 400 base pairs (bp) was amplified from the gene encoding the mt COI . The PCR-mediated reaction was performed using the primers BI-COI and SUBIR cited in Davolos and Maclean (2005). The PCR amplification conditions were cited in . The amplified fragments were checked by electrophoresis in 1% agarose gels and then used as templates for cycle sequencing reactions (BigDye chemistry) followed by DNA sequencing using BI-COI and SUBIR primers. In addition, a fragment of ca. 350 bp of the gene encoding the nuclear histone H3 (H3), was PCR amplified using the primers H3Of and H3Or cited in Davolos and Pietrangeli (2014). The PCR cycling parameters were cited in . The PCR products were verified and then sequenced using H3Of and H3Or primers, as above described.
The novel annotated sequences from the mt COI and the nuclear H3 genes from C. garbinii of this study have been submitted to the GenBank databases at the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov) ( Table 2). Other DNA sequences used for the Bayesian analysis of the present study were already deposited at NCBI by Davolos and reported in . In more details, the European populations of C. garbinii are represented by samples from France (Dijon), and Italy (Latium; Table 2). DNA sequences of C. cavimana (Davolos et al. 2018) and C. ruffoi Latella & Vonk, 2017 are cited in . We added DNA sequences from Cryptorchestia species of the northeast Atlantic volcanic islands examined in our previous study (C. canariensis (Dahl, 1950), C. gomeri (Stock, Figure 1. Occurrence of Cryptorchestia garbinii in the Marmara region of Turkey and the habitats (insets) where specimens were collected. The 15 mm male (inset) of C. garbinii was found in Lake Sapanca.
1989), C. guancha (Stock & Boxshall, 1989), and C. stocki (Ruffo, 1990) from the Canary Islands, C. monticola (Stock & Abreu, 1992) from Madeira, and C. chevreuxi (De Guerne, 1887) from Terceira, Azores . Moreover, additional unpublished sequence data of undescribed Cryptorchestia taxa from the islands of Graciosa and Flores, Azores, were obtained from the work of . We also included sequences from the mt COI and the nuclear H3 genes of five Orchestia species, two Mediterranean endemic species, O. montagui Audouin, 1826and O. stephenseni Cecchini, 1928, and the Mediterranean/Atlantic O. gammarellus (Pallas, 1976), O. mediterranea, Costa, 1853, and O. aestuarensis Wildish, 1987. These sequences were obtained from a former project of Davolos and De Matthaeis and are available from GenBank (see Table 2 for accession numbers). For the outgroup we used DNA sequences of Platorchestia platensis (Krøyer, 1845) from the Island of Capri, Italy (see Table 2).

Molecular phylogeny
A Bayesian phylogenetic inference analysis was carried out using the MrBayes v.3.2.6 software package (Ronquist et al. 2012). We concatenated the two genes into a single  Table 2. Platorchestia platensis was used in this study as an outgroup species. Marked in blue: Cryptorchestia species; marked in green: Orchestia species. Numbers at nodes correspond to Bayesian posterior probability (PP) support values; PP values greater than 0.5 are labelled. The Gen-Bank accession numbers of the DNA sequences from the COI and the histone H3 genes used in this study are reported in Table 2. alignment before Bayesian methods were conducted. The General Time Reversible plus gamma distributed rate variation among sites (G), and proportion of invariable sites (I) in the dataset, GTR+G+I substitution model, was selected for the two genes COI, and H3. Analyses were run for 10 million generations, with trees sampled every 1000 generations and a burn-in of 25%. Two independent runs with three hot chains and one cold chain were conducted, which were long enough to effectively sample each distribution and to reach convergence (e.g., the standard deviation of split frequencies was 0.004, the potential scale reduction factor (PSRF) was close to 1.0 for all parameters, and the effective sample size (ESS) was high for all the parameters). The first 25 % of trees, which represented the burn-in phase of the analysis, were discarded. The remaining trees were used for calculating posterior probabilities (PP) of recovered branches in the 50 % majority rule consensus tree. Subsequently, the tree was graphically visualized and exported by FigTree v1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/) for further editing with the program Inkscape (http://www.inkscape.org).

Molecular phylogenetic analysis
Mitochondrial and nuclear genes (COI, and H3, respectively) of the newly sequenced Turkish specimens of C. garbinii and other talitrids for a total of 27 taxa were used for phylogenetic analysis. Concatenation of DNA sequence data produced an alignment of 693 bp.
The results from the Bayesian inference phylogenetic analysis are presented in Fig.  2. In the COI-H3 phylogeny constructed with Mr Bayes, the phylogenetic clusters generally were well supported (PP above 0.8) and some clades were very well supported (PP > 0.95). A first clade grouped with full support (PP of 1) O. montagui and O. stephenseni, both endemic to Mediterranean marine supralittoral. Moreover, our phylogenetic analysis recovered a monophyletic clade (PP = 0.99) that was formed by the continental species C. garbinii, widely distributed from Turkey (present study) to central Europe, and two other freshwater Cryptorchestia species, C. cavimana, and C. ruffoi, which are endemic to the east Mediterranean islands of Cyprus and Rhodes, respectively. Within C. garbinii, apparently the European populations from France (Dijon) and Italy (Latium) are more related to the Turkish populations from Lake Iznik and Lake Sapanca, while the Kiyiköy population was genetically more related to the population of Lake Ohrid in Macedonia.
Finally, our phylogenetic analysis recovered a group of talitrid taxa from NE Atlantic area (PP of 0.82), which was divided into two subgroups. A monophyletic subgroup showed very well-supported relationships (PP of 1) for the terrestrial Cryptorchestia lineages of C. canariensis, and C. stocki endemic to Gran Canaria, Canary Islands, C. monticola endemic to Madeira, and undescribed Cryptorchestia taxa from the islands of Graciosa and Flores, Azores . The second monophyletic clade (PP of 0.83) was further divided into subclades. One subclade was formed by the terrestrial Cryptorchestia lineages of C. guancha endemic to Tenerife, Canary Islands, and C. chevreuxi from Terceira, Azores, while C. gomeri endemic to La Gomera, Canary Islands, and O. gammarellus (known to occur from fully marine to freshwater shores) as sister species. The other subclade (PP of 1) was formed by the two morphologically related but genetically different O. mediterranea and O. aestuarensis (Fig. 2), both species well known to occur even in low salinity conditions. Localities. The specimens of C. garbinii reported from Turkey were distributed in Lake Iznik (Bursa Province), Lake Sapanca (Sakarya province), and in a stream near Kiyiköy (Kirklareli province) (Fig. 1).
Diagnosis. Talitridae, with antenna I shorter than the combined peduncle segments of antenna II; accessory flagellum absent; mandible without palp, left mandible with 5-dentate lacinia mobilis; maxilla I with nine setae on inner lobe; maxilliped palp article IV reduced; gnathopod I male with a small, partially transparent lobe on merus; pereopod IV with bulged inner margin of dactylus; pereopods 3-7 cuspidactylate; telson variable in shape, longer than wide in holotype from Lake Garda, wider than long in the Turkish populations.
Additional description. Based on adult males with an average length of 14.58 mm (Table 1). Head. Antenna I (Fig. 3A) flagellum with five to seven articles. Antenna II (Fig. 3B) little shorter than half of the body length, peduncle article V longer than article IV, flagellum with 16-21 articles. Labrum (Fig. 3C) and labium (Fig. 3D) with very fine setules on anterior margin. Mandible (Fig. 3E, F) left with 5-dentate lacinia mobilis and 5-dentate pars incisiva, right mandible with multi-teethed lacinia mobilis. Maxilla I (Fig. 3G) with nine robust and crenelated setae on inner lobe. Maxilla II (Fig. 3H) with numerous apical setae, a long and finely pinnate seta on inner margin of inner lobe. Maxilliped (Figs 3I, 7C) basal lobe with three blunt teeth on anterior margin, axial margin lined with robust setae armed with setules; palp article IV reduced, but clearly visible.
Pleon. Epimeral plates (Fig. VIA, B), posterior margins weakly crenulate. Pleopods I-III (Fig. 6B) well developed, biramous; rami with slender setae; inner ramus slightly shorter than outer. Uropod I (Fig. 6C) with two rows of seven axial setae on peduncle; rami of equal length, with three to four apical setae (of which one on the inner ramus longer than others). Uropod II (Fig. 6D) peduncle with one row of four axial setae; rami of equal length, outer ramus with four apical setae of which two small, inner ramus with two apical setae. Uropod III (Figs 6E, 7E) peduncle with three to four distolateral setae, ramus with four apical setae of which one long, and two additional lateral   (Table 1). Antenna I short, 1.22 mm length in average, flagellum with four articles. Antenna  II long, 4.21 mm in average. Gnathopod I (Fig. 4C) subchelate; coxal plate lower margin with irregularly placed prominent setae; basis rectangular with several short setae; carpus with one robust and long seta between several smaller on lower margin; Figure 6. Cryptorchestia garbinii, male 12 mm, Lake Iznik. A epimeral plates I and II B pleopod II C uropod I D uropod II E uropod III F telson G female 13 mm, Lake Iznik, telson.
propodus with three bundled groups on the palmar margin; dactylus with unguis longer than palm. Gnathopod II (Fig. 4D), basis large, with strongly curved anterolateral margin; merus with lobe or patch with fine structures resembling a rasp surface; Figure 7. Cryptorchestia garbinii, male 16 mm, Lake Iznik. A gnathopod II, palmar margin and dactylus B gnathopod I C maxilliped D pereopod IV, propodus and dactylus E uropod III F telson. carpus and propodus expanded, with lateral lobes, propodus with rounded frontal margin and small claw.
Remarks. The specimens from the Turkish lakes Iznik and Sapanca show some slight morphological differences compared to the Lake Garda (type locality) specimens.
We interpret this as intraspecific variation. The lobe on the merus of the first gnathopod is smaller in the 12 mm male (Fig. 4B) than in the 16 mm male, especially when the twist in the carpus-propodus part relative to the merus forces the gnathopod to appear in an angle on a mounting slide. The feature is clearer in SEM photographs where it shows a rugose patch.
The specimens from Turkey were also compared to the geographically more adjacent freshwater species C. ruffoi from the island of Rhodes, Greece, to C. cavimana from Cyprus, and to C. kosswigi (Ruffo, 1949) from the Hatay province in southern Turkey. Differences are: a small lobe on merus of gnathopod I in male (large in C. ruffoi, not mentioned as present in C. cavimana by Heller (1865) in his German text but vaguely present in the figures); antenna I with 5-7 segmented flagellum (5-6 in C. garbinii of Lake Garda, four in C. ruffoi, six in C. cavimana, six in C. kosswigi); pereopods V-VII with wide bases and of short and stubby appearance (less wide in C. garbinii of Lake Garda, C. ruffoi and C. cavimana, broad in C. kosswigi but here the other segments are long and slender); gnathopod II female with very broad and rounded basis (less in C. ruffoi, not mentioned in C. cavimana, but prominently present in the C. cavimana that was partly redescribed from the British Isles (Lincoln 1979), and present to a lesser extent in C. kosswigi); pereopod IV with clear bulge on dactylus (not mentioned in C. garbinii of Garda lake, not present in C. ruffoi and not mentioned in C. cavimana, present in C. kosswigi). We consider these differences as interspecific variation of morphological features.

Discussion
Our molecular phylogeny based on mitochondrial and nuclear protein-coding genes helps to understand evolutionary relationships between continental (C. garbinii) and insular Cryptorchestia species from the East Mediterranean region and from the Canary Islands, Madeira, and Azores (North East Atlantic area) examined in this study and reported in Fig. 2 and Table 2.
The molecular data indicated a low level of genetic differentiation within the mainland Turkish and European populations of C. garbinii. It is noteworthy that DNA sequence data from Kiyiköy on the west side of the Bosphorus indicated that this Turkish coastal population of C. garbinii was highly related to that from Lake Ohrid, Macedonia, while the DNA sequences from the Turkish lake inhabitants on the east side from Iznik (Bursa Province) and Sapanca (Sakarya province) in the north-west pointed to a strong relationship with those from Europe, here represented by samples from France (Dijon), and Italy (Latium) (Fig. 2). The low intraspecific genetic diversification within C. garbinii and its wide geographic distribution (Figs 1-2) both lead to the supposition that dispersal events have occurred relatively recently. The evolution of C. garbinii could have been a pattern of recent east-to-west dispersion, with a more recent northward expansion. A more extensive sampling of other representative localities of C. garbinii is required to resolve the phylogeography of this species in greater detail.
Our molecular data further suppose an insular isolation and in situ speciation in two east Mediterranean islands, namely C. cavimana from the island of Cyprus, and C. ruffoi from the island of Rhodes in south-eastern Greece. Moreover, our study supports the monophyletic clade formed by C. garbinii, C. cavimana, and C. ruffoi (Fig. 2), all with a clear small lobe on the merus of gnathopod I.  added to this group C. kosswigi from the Turkish coast that clearly has a small lobe on the male gnathopod I merus (currently no DNA sequences available). These species represent Cryptorchestia sensu stricto and are generally found in Mediterranean riparian habitats. Therefore, the pattern of evolution within this semi-terrestrial talitrid lineage appears to have been a process with mainland and insular speciation, with the plausible hypothesis of a common ancestor that occurred in freshwater of the east Mediterranean region.
However, our molecular analysis indicates Cryptorchestia not monophyletic ( Figure 2) and helps to understand various ecological, morphological, and geographic distribution aspects. It is important to note that semi-terrestrial Cryptorchestia species from the east Mediterranean region live in riparian habitats, while the terrestrial species from the islands of the north-east Atlantic area live in humid, evergreen broadleaf laurel forest (laurisilva; see  and reference therein). Moreover, Cryptorchestia has been proposed on a single character (Lowry and Fanini 2013), the gnathopod I merus with a very small lobe on the posterior margin (in Orchestia there is a palmate lobe only on male carpus and propodus of gnathopod I), of which the presence is variable within the genus. By focusing on the Atlantic terrestrial Cryptorchestia species that live in laurisilva, they appear to be phylogenetically and morphologically distinct and can be divided into two groups (Fig. 2). The first group is formed by C. guancha, C. gomeri, and C. chevreuxi, which share superficially a similar morphology on the absence of the small lobe on the merus of gnathopod I, and they have O. gammarellus as sister species. A loss of the lobe in these Cryptorchestia species from Macaronesia cannot be ruled out, but it is important to note that they share a common ancestor with the clade formed by O. mediterranea and O. aestuarensis. Therefore, a plausible hypothesis is that their putative marine common ancestor probably had the sexually dimorphic pereopods I (without the small lobe on the merus), II, and VII. The second group is formed by north-east Atlantic volcanic island Cryptorchestia species, some of which (C. stocki, and undescribed Cryptorchestia species from the Azores ) have a small lobe on the merus of gnathopod I. Therefore, our study suggests that the morphological convergence of this character in unrelated Mediterranean and Atlantic Cryptorchestia species is likely (Fig. 2).
It seems reasonable to assume that the Atlantic volcanic archipelagos, with a sequence of island emergence and ageing over millions of years, might have been colonized by talitrid lineages independently and on different times . However, we are aware that the phylogeny of Cryptorchestia species is far away from being resolved. An important phylogenetic gap exists between the Macaronesian Cryptorchestia species and putative close continental relatives, probably from southwest Europe/northwest Africa, even if extinction of the ancestral lineages after the colonization of the Macaronesian volcanic islands cannot be ruled out. A Bayesian phy-logeny, using multiple calibrations for divergence time estimation that includes a larger dataset of talitrid genera, such as Palmorchestia and Macarorchestia, from the North East Atlantic area and the Mediterranean region is one of the major scopes of future work. However, we have described here, through a multi-locus molecular phylogeny approach, three main clades with species classified as Cryptorchestia. Our Bayesian phylogeny showed that the genera Cryptorchestia and Orchestia as described appear polyphyletic within the Talitridae and indicated that using the current Cryptorchestia taxonomy is useful for the Cryptorchestia s. s., but that a taxonomic revision within this genus is required. Clearly, our studies confirmed the importance of insular speciation in both the northeast Atlantic and east Mediterranean Cryptorchestia lineages.