Integrative Taxonomic Study of the Purse Crab Genus Persephona Leach, 1817 (Brachyura: Leucosiidae): Combining Morphology and Molecular Data

Marine crabs of the genus Persephona Leach, 1817 are restricted to American waters of the western Atlantic and eastern Pacific Oceans. Subfamilial assignment of this taxon has varied between authors and its species composition remain in question. We conducted a comparative study based on morphology and molecular phylogenetics for all ten recognized species of Persephona, along with Iliacantha hancocki. We tested whether Persephona finneganae, P. lichtensteinii, and P. crinita represent a single species as suggested by some authors; whether specimens identified as P. punctata, P. mediterranea, and P. aquilonaris warrant treatment as separate species; and whether I. hancocki should be regarded as a junior synonym of P. subovata. Diagnostic morphological characters (of the carapace, chelipeds, and third maxillipeds) were used along with gonopod (male first pleopod 1) features and live coloration. The 16S rRNA and the Cytochrome Oxidase I (COI) (DNA barcoding) mitochondrial genes were used as molecular markers. Both morphological and molecular analyses revealed that putative specimens of P. crinita from Brazil and those assigned to P. finneganae were no different from specimens presently assignable to P. lichtensteinii. P. finneganae is regarded as a junior synonym of P. lichtensteinii, and we apply P. crinita only to specimens we examined from the Gulf of Mexico. Specimens from Brazil previously reported as P. crinita are herewith concluded to represent P. lichtensteinii. Additionally, P. townsendi is a junior synonym of P. orbicularis, Iliacantha hancocki is concluded to be a junior synonym of P. subovata, while P. aquilonaris and P. mediterranea are found to represent separate species. On the basis of our revisions, eight species of Persephona are considered valid, and the reported distribution for P. crinita is restricted.


Introduction
Pacific: Mexico, Costa Rica, and Panama (Fig 1). Specimens representing these regions, along with two other species of Leucosiidae to serve as outgroups, were used in an accompanying molecular analysis (Table 1). Two other members of the subfamily Ebaliinae were selected as outgroups based on their phylogenetic proximity to the genus and availability of data.
Intended for the molecular analyses, some fresh samples from the São Paulo state were collected complied with current applicable state and federal laws of Brazil (DIFAP/IBAMA/126/ 05; permanent license for collection of Zoological Material number 11777-MMA/IBAMA/SIS-BIO). All other specimens were obtained from museums and zoological collections as follow.  Collecting sites for specimens analyzed. An apostrophe the ['] to right of number indicates that both morphological and molecular analyses were carried out on material from that site. Otherwise, only morphological analyses were conducted on that material. The dotted lines correspond to limits between zoogeographic provinces proposed by Briggs and Bowen [5]. Western Atlantic: CrP = Carolinian Province; CbP = Caribbean Province; BrP = Brazilian Province; ArP = Argentinian Province. Eastern Pacific: CaP = Californian Province; CoP = Cortezian Province; PaP = Panamanian Province; PeP = Peru-Chilean Province.  The indication of biogeographic provinces followed the classification of Briggs and Bowne [5]. Western Atlantic coast: Carolinian (Gulf of Mexico-from Cape Romano, Florida to Cape Rojo, Veracruz and Atlantic-from Cape Hatteras, North Carolina to Cape Canaveral, Florida), Caribbean (from Bermuda, Cape Canaveral, Florida, to the Amazon River), Brazilian (from Amazon River south to Santa Catarina) and Argentinian (from Santa Catarina to the Valdez Peninsula, Chubut). Eastern Pacific coast: Californian (from Los Angeles, California to Magdalena Bay, Baja California Sur), Cortezian (all the Gulf of California), Panamanian (from Magdalena Bay south to Gulf of Guayaquil), Galapagos (Galapagos Archipelago), Peru-Chilean (from the Gulf of Guayaquil to Taitao, Aysén) and Juan Fernández (Juan Fernández Islands).

Molecular analyses
Molecular analyses were based on two mitochondrial genes, a fragment DNA of the 16S rRNA (16S) and the barcode region of the Cytochrome C Oxidase subunit I (COI). Both genes are widely used in phylogenetic studies of many invertebrates and decapod crustaceans [22][23][24][25]. More specifically for decapod crustaceans, both genes have been used in recent barcoding projects [26][27][28] and U.S. National Science Foundation Decapod Tree of Life phylogenetics studies [29][30][31] to delimit species boundaries and to clarify the evolutionary relationships among decapod crustaceans.
DNA extraction, amplification, and sequencing protocols followed Schubart et al. [32] with modifications according to Mantelatto et al. [33] and Robles et al. [34]. Total genomic DNA was extracted from muscle tissue of the chelipeds. The tissue was incubated for 48h in 600μL of lysis buffer at 55°C, with 200μL of proteinase K (PK); protein was separated by addition of 200μL of 7.5M ammonium acetate prior to centrifugation. DNA was precipitated by addition of 600μL of isopropanol cooled to the supernatant and then centrifuged; the resultant pellet was washed with ethanol 70%, centrifuged, dried and resuspended in 10-20μL TE buffer.
PCR products were purified using a SureClean Plus kit (following the vendor's protocols) and sequenced using the ABI Big Dye 1 Terminator Mix in an ABI Prism 3100 Genetic Analyzer 1 following Applied Biosystems protocols. The sequences obtained were confirmed by sequencing both strands; consensus sequences were obtained using BioEdit version 7.0.5 [37] from the two complementary sequences. Consensus sequences of the 16S and COI genes were aligned using ClustalW [38] as implemented in BioEdit [37], with default parameters. The COI sequences were checked for the presence of stop codons. All sequences were submitted to GenBank.
Phylogenetic analyses. The concatenated analyses were conducted based on a total of 1091bp (606 for the COI and 485 for the 16S genes, excluding the primer regions). Alignment of both gene sequences was unambiguous and the ILD test showed no significant incongruence. After confirming that the two genes have the same evolutionary history, the best-fitting model for sequence evolution of the combined COI and 16S was determined by JModelTest 2.1.4 [39], selected by the AIC (Akaike information criterion) method. This information criterion indicated the TPM1uf+I+G as the best-fit model of DNA sequence evolution, accounting for invariable positions and unequal rates of substitutions under a gamma distribution, with the nucleotide frequencies: A = 0.3626, C = 0.1266, G = 0.1618, and T = 0.3490; rates of nucleotide substitution A-C = 1.0000, A-G = 94.1738, A-T = 7.0694, C-G = 7.0694, C-T = 94.1738, G-T = 1.0000; proportion of invariable sites I = 0.6530; and gamma shape = 0.8590.
The Bayesian analysis (BAY) was performed with MrBayes 3.2.2 [40] with the parameters obtained from JModeltest. The search was run with four chains for 20,000,000 generations with trees being sampled every 10,000 generations.
Trace plots were visually inspected to assess convergence, mixing, and stationarity in Tracer v1.4. [41]. Once the split frequency in each analysis was 1% (reached well before 2 million generations = 200 trees), we found the maximum clade credibility tree (MCCT) using TreeAnnotator v1.5.4 [42] from the remaining 1800 saved trees. We obtained a 50% majority rule consensus tree using the same 1800 trees. Bayesian posterior probability [43]; values > 70% were shown on the resulting MCCT.
Maximum Likelihood analysis (ML) was performed with RAxML 7.0.4 [44], as implemented in CIPRES (Cyberinfrastructure for Phylogenetic Research). The model of evolution was the GTR+G+I, which is the default model for RAxML. The internal consistency of the branches was evaluated by the bootstrap method [45], and we selected the option to automatically determine the number of bootstraps to be run in RAxML. A total of 150 bootstrap pseudoreplicates were run, and confidence values > 50% were shown on the resulting trees.
Genetic distance analyses. Genetic distance analyses were applied in previous systematic studies of crustaceans and other animals [46][47][48][49][50]. Genetic distance calculations were performed using MEGA 5 [51], and two distance matrices were calculated using uncorrected distances (p-distance), based on COI and 16S sequences. We did not calculate a distance matrix using a model of evolution since it has been shown that using p-distance avoid over-parametrizing and there is no need to use complex distances measures when studying closely related sequences [52,53]. To help assess intraspecific and interspecific genetic distances, two frequency histograms were constructed with pairs of COI and 16S sequences.

Morphological analyses
Analyses were conducted to evaluate morphological characters used most commonly for identifying members of Persephona in classical and current systematic literature (primarily features of carapace shape including the front, carapace armor or ornamentation, the third maxilliped, and the chelipeds). Additionally, the morphology of the first male gonopod (Go1) was comparatively reviewed [12,54,55,56].
Scanning electron microscopy. For morphological study, Go1 were dissected from voucher specimens preserved in 70% alcohol and thereafter fixed in 3% glutaraldehyde, after which they were dehydrated in a graded ethanol series of 30%, 50%, 70%, 80%, 90%, 100% (30min changes each), dried in a critical point dryer with liquid CO 2 in the EMS 850 (Electron Microscopy Sciences 1 ), and sputter-coated with gold in Denton Vacuum Desk II coater. Examinations and micrograph exposures were conducted with a JSM 5410 (JEOL 1 ) scanning electron microscope (SEM).

Molecular analyses
Phylogenetic analyses. The ILD test showed no significant incongruence with the P = 0.874. Thus, we can consider the 16S and COI genes to have the same evolutionary history [57] and we could use the concatenated dataset for our phylogenetic analysis. Visual inspection of the two tree topologies obtained from the Bayesian and Maximum Likelihood analyses, based on all specimens analyzed, showed the topologies to be identical (Fig 2). All congeners of Persephona, along with Iliacantha hancocki, are joined in a highly supported clade in both the ML and BAY analyses. All congeners of Persephona along with Iliacantha hancocki are joined in a highly supported clade in both the ML and BAY analyses.
Analyzed specimens of "P. mediterranea" (here treated as P. aquilonaris) from Gulf of Mexico comprise a sister clade to P. mediterranea from Brazil. Together they form a highly supported sister group to the clade composed of P. punctata and P. edwardsii, with the latter clade exhibiting high support values in ML and BAY analysis.
Persephona subovata and Iliacantha hancocki cannot be distinguished from each other. The two specimens form a clade that is without support grouped with the clade encompassing P. aquilonaris, P. mediterranea, P. punctata, and P. edwardsii in ML and BAY analyses.
The species P. orbicularis and "P. townsendi" form a clade topologically placed as a sister to that containing P. aquilonaris, P. mediterranea, P. punctata, P. edwardsii, and P. subovata, with support in both the ML and BAY analyses. The topology also depicts P. crinita from Gulf of Mexico as the sister group of a clade composed by P. lichtensteinii, "P. finneganae", and  Table 1. specimens of "P. crinita" from South America (treated here as P. lichtensteinii). Within this latter clade, Caribbean specimens form a clade that is sister to all P. lichtensteinii from Brazil with high support in ML analysis.
When genetic distances of specimens assigned to "P. crinita" and "P. finneganae" from the Brazilian Province were compared to those for specimens of P. lichtensteinii from the same province, values ranged from 0 to 0.003 (Table 2); these distance values fell within the range of intraspecific variation observed for other Persephona spp. (Fig 3). In contrast, when these same specimens of "P. crinita", "P. finneganae" and P. lichtensteinii from the Brazilian Province were compared to specimens of P. crinita from the Gulf of Mexico (Carolinian Province), genetic distance values ranged from 0.102 to 0.104 (Fig 3; Table 2); these values fell within the range for interspecific distance.  Table 1. The genetic distance value between P. orbicularis and "P. townsendi" (both specimens from the Panamanian Province) was 0.017, falling within the gap between intraspecific and interspecific genetic distance values ( Fig 3A). The same was observed for specimens of P. mediterranea compared from three zoogeographic provinces. When two specimens from the Gulf of Mexico (one of "P. mediterranea" from the Carolinian Province and another from the Caribbean Province; see Fig 1) were compared to those from Brazil (two specimens from the Brazilian Province), values ranged from 0.031 to 0.033 (Fig 3), respectively. It is also noteworthy that when comparing the genetic distance between the two specimens of P. mediterranea from within the Brazilian zoogeographic province the value was 0.000, and when comparing the specimens of "P. mediterranea" from the Gulf of Mexico, the value ranged from 0.005 to 0.008 (Table 2).
Genetic distance values for compared specimens of P. lichtensteinii from two different zoogeographic provinces (Caribbean and Brazilian) fell in the interval between intraspecific and interspecific values at 0.026 (Fig 3). When all specimens of P. lichtensteinii from within a single province were compared, the value for genetic distance was 0.000 (Brazilian Province) and ranged from 0.000 to 0.003 (Caribbean Province) ( Table 2; Fig 3).
Compared between two specimens of P. punctata from two different zoogeographic provinces (Caribbean and Brazilian), genetic distance values also fell between intraspecific and interspecific values (0.036; Fig 3). The observed genetic distance value between specimens of P. subovata and "I. hancocki" (both specimens from the Panamanian Province) was 0.003 (Table 2), within the values typically observed when comparing specimens within a single species of Persephona.
Genetic pairwise distance values based on the 16S gene, showed a clear separation among morphologically well-defined species (depicted as a wide gap; Fig 4A). The values observed between compared individuals of P. crinita from the Gulf of Mexico (Carolinian Province) and between compared individuals of P. mediterranea from Brazil (Brazilian Province) ranged from 0.000 to 0.010 (mean ± SD = 0.050 ±0.007). Values among compared specimens of different species (e.g. P. crinita from the Gulf of Mexico and P. lichtensteinii; P. crinita from the Gulf of Mexico and "P. townsendi") ranged from 0.028 to 0.055 (mean ± SD = 0.042 ± 0.019) (Fig 4).
Pairwise distances within species of Persephona, based on the 16S gene, were obtained only for specimens of P. crinita, P. lichtensteinii, and P. mediterranea. Genetic distances between the specimens of "P. mediterranea" from the Gulf of Mexico (Carolinian Province) and one from Brazil (Brazilian Province) showed a genetic divergence of 0.010 (Fig 4, Table 2). These values fell within the gap that separates the intraspecific and the interspecific groups ( Fig 4A).
In comparison of P. lichtensteinii and "P. finneganae" from Colombia and Venezuela, respectively, with P. lichtensteinii from Honduras (three specimens from the Caribbean Province) the value was 0.005 (Table 2). When comparing this same specimen from Honduras (Caribbean Province) to "P. crinita" from Brazil (Brazilian Province), the value was 0.007 ( Fig  4; Table 2). However, even being in different provinces, comparing P. lichtensteinii and "P. finneganae" from Colombia and Venezuela, respectively, with P. lichtensteinii from Brazil the obtained value was 0.002. Thus genetic divergence values among specimens from the same zoogeographic province (Caribbean) were intermediate to values observed when comparing specimens from two different zoogeographic areas (Caribbean and Brazilian Provinces). These values also fell within the gap between what might be defined as an intraspecific rather than interspecific grouping (Fig 4A).

Morphological analyses
The morphological characters included number and size of spines or granules, proportion of width and length of the front, proportion of width and length of the chelipeds articles and coloration of carapace, all proving to be informative for identification of adults of Persephona ( Fig 5). In addition, the carapace coloration pattern of two species, P. aquilonaris and P. mediterranea is informative even in juveniles (Fig 6).  Scanning electron microscopy. The morphology of the Go1 is species-specific (Figs 7A-7G; 8A-8C and 8D-8F), but with no clear evidence of importance for higher phylogenetic groupings. The principal differences were found in shape of the apex, presence or absence of subapical setae, and presence or absence of lobule on the mesiodistal margin (Fig 8). Furthermore, Go1 of Persephona is a stable character for males of different sizes and apparently does not change once the male minimum size has been reached.

Discussion
Persephona crinita Rathbun, 1931, P. finneganae Rathbun 1933, and P. lichtensteinii Leach, 1817 Our phylogenetic analyses showed clear separation between two well-supported clades of specimens identified as P. crinita sensu lato. One clade corresponds to specimens from the Gulf of Mexico (Carolinian province) and a second included specimens of "P. finneganae", P. lichtensteinii, and those identified as "P. crinita" (Brazilian Province).
Persephona crinita is a species frequently reported in biodiversity studies of the Brazilian crustacean fauna [2,8,58,59,60]. However, specimens of "P. crinita" from Brazil usually are identified on the basis of a taxonomic key by Melo ([2]: 151), modified from the original one by Rathbun [61], and a drawing labelled as "P. crinita" that in fact depicts a color variant of P. lichtensteinii with poorly developed lateral spines. When analyzing specimens from the Gulf of Mexico, including the paratype of P. crinita (USNM 64254), it is evident that this species does not possess lateral spines but instead seven tubercles along the lateral margin, which are clearly depicted by Rathbun [61]. On the other hand, specimens from Brazil possess either seven spines, tubercles, or teeth that are of differing shapes and sizes, but these are noticeably more developed than are those of P. crinita. However, these marginal spines actually vary greatly in size, possibly due to wear resulting from the sand burrowing habits of these animals. To augment these sometimes variable spines as diagnostic characters, the morphology of the adult Go1 is clearly different in these two species. The opening in P. crinita (Fig 7A) is narrow and directed towards the mesiodorsal face while in P. lichtensteinii ( Fig 7C) the opening is directed towards the mesioventral face.
In addition to her noting the possession of seven marginal spines, tubercles, or excrescences as already mentioned, Rathbun [61], [3] used the presence of tubercles between the sub-hepatic and the lateral tubercles to differentiate P. crinita ( Fig 5A) from P. lichtensteinii (Fig 5B), a character also used to separate P. finneganae from P. lichtensteinii. However, this character is highly variable and can be evident or obscure, including on the type material [6,7]. The posterior elongate spine on the carapace was also used by Rathbun [3,62], to differentiate P. finneganae from P. lichtensteinii. It can be observed on individuals with or without tubercles, being present between the lateral and subhepatic spines, though it is not in itself a diagnostic character for P. finneganae. Another variable character utilized by Rathbun [62] to describe P. finneganae was the intestinal region of the carapace being partially defined by shallow furrows. This subjective character is present to varied degrees in all species of Persephona and is not of much value for diagnoses. Our examination of the carapace morphology and the Go1 of specimens assigned to P. finneganae, including the paratype (MZUSP 934) and the holotype (USNM 67989), support previous suggestions that P. finneganae is a junior synonym of P. lichtensteinii [6,7]. Furthermore, our genetic distance analysis placed divergence values between these two species and materials from Brazil previously assigned to "P. crinita" within the range of intraspecific variation (even sharing some haplotypes), supporting the hypothesis that the three "species" are morphological variants of P. lichtensteinii. Thus, we contend that specimens currently assigned to P. finneganae and "P. crinita" from Brazil should now be referred to as P. lichtensteinii. Based upon our phylogenetic analyses, some conclusions on recent evolution of the group can be drawn, especially since both ML and BAY analysis also recovered the same topology. Persephona crinita (from the Carolinian Province) was recovered as a sister group of P. lichtensteinii (from the Caribbean and Brazilian Provinces) in all analyses and always with strong support. These species are in turn joined with strong support to the clade encompassing all other species of Persephona, including those from the Pacific. Internal support for this overall clade is strong, even while segregating Atlantic and Pacific species into separate subclades, suggesting that they diversified following closure of the Panamanian Isthmus. Dynamics contributing to diversification in the western Atlantic after emergence of the Isthmus may have included redirection of the North Equatorial Current in the northern hemisphere, this propelling the Gulf Stream through the Yucatan Strait, with this current becoming more intense than  before the closure of the Isthmus [63]. In this scenario [63,64] it is not difficult to envision barriers to dispersal, regional selective pressures, or isolation and sorting events that could result in divergence of a western Atlantic species pair from what was possibly a common evolutionary stock.
Persephona aquilonaris  and P. mediterranea (Herbst, 1794) Specimens a priori identified as "Persephona mediterranea" from the Gulf of Mexico (Carolinian and Caribbean Provinces) were clustered in a well-supported clade together with specimens from Brazil (Brazilian Province) identified as P. mediterranea. However, it is clear from our phylogenetic analysis that there is some degree of genetic structure, which was also evident in the genetic divergence analysis. The genetic divergence distance based on COI was around 3%, which might or might not indicate variation expected at the population level [65]. Yet, in the analysis of 16S sequence data, the values for genetic distance were around 1%, which is above those observed for species separations among some other decapod crustaceans (i.e. species of mud crabs of the family Panopeidae; Schubart et al., [65]). Thus, we conclude that the analyzed specimens from the Gulf of Mexico (Carolinian and Caribbean Provinces) indeed represent a different species from those in South American waters. Persephona aquilonaris was described by Rathbun [62] for specimens from the northern hemisphere. However, the name was mostly abandoned following the work of Guinot-Dumortier [12]. When comparing specimens from Brazil to those from Louisiana, Texas, and eastern Florida (St. Augustine), including the type of P. aquilonaris (USMN 62057), we observed a few consistent morphological differences. The most definitive structural difference was found in morphology of the Go1, which exhibits a prominent lobe in Persephona mediterranea but not in P. aquilonaris (Fig 8A-8F). Gonopod morphology aside, their live coloration (which persists for short periods even after preservation) is also consistently different from other taxa. Although the color of the carapace for many specimens can be similar, the distribution pattern of the darker spots is usually distinct, especially in mature individuals. In adults of both species, the dark reddish to yellow-orange patterning of the carapace is somewhat bilaterally symmetrical, being subdivided by a mid-dorsal longitudinal track of light background color (Fig 6). However, in mature specimens of P. aquilonaris, the dorsal dark pattern consists of anastomosed small dark reddish brown spots, each usually with one to several light granules at its center, giving the spots a broadly ocellated appearance. The spots themselves tend to be loosely or continuously collected into longitudinally oblique tracts, separated by narrow lighter veins of background color, with the veins and lines of intersecting spots converging somewhat posteriorly and the pattern broken into isolated small spots posterolaterally (Fig 6A-6C; see also Williams [66]: 150, fig. 127, as P. punctata aquilonaris; Williams [4]: 288, fig. 223 as P. mediterranea). By contrast, in P. mediterranea, the darker components of the dorsal pattern usually include 7-14 large reddish brown to light orange or tan spots, each usually margined by a narrow darker line encircling multiple granules, giving large spots a narrowly ocellated appearance. These robust spots of somewhat varied darkness are separated by broad lighter veins of lighter background color, with the veins and rows of the large spots not obviously converging posteriorly nor broken into a pattern of isolated small spots posterolaterally (Fig 6D-6F; see also Herbst [67]: 150, tab. 37, fig. 2 as Cancer mediterraneus; Guinot-Dumortier [12]: 432, fig. 9 as P. aquilonaris; Melo [2]:153 unnumbered figure). In both species, subadult and immature specimens are less conspicuously patterned and very small juveniles may almost totally lack most dark pigmentation of the carapace.
On the basis of our molecular phylogenetic analysis, genetic distance analysis, morphological comparisons of the Go1, and comparisons of adult color patterns, we conclude that P.
aquilonaris and P. mediterranea should be treated as separate species. In general, P. aquilonaris applies to most populations distributed from New Jersey to Texas, with P. mediterranea applying to primarily those populations known to be distributed from the Antilles and Caribbean through southern Brazil, albeit with a slight overlap of the two species in tropical waters of the eastern Gulf of Mexico. While we separate P. aquilonaris and P. mediterranea, our phylogenetic analyses grouped them as sister species in a clade with high support values in ML and BAY analyses. We regard them as northern and southern counterparts that likely diverged from common stock over a long history of glacial advance and retreat following closure of the Panamanian Isthmus [63].
Persephona punctata (Linnaeus, 1758) and P. edwardsii Bell, 1855 It is clear in our analyses that the Atlantic P. punctata deserves species rank, independent of its former subspecies, P. aquilonaris, but less expected that it resolves as a sister species of the Pacific congener, P. edwardsii, in both analyses with high support for ML and BAY analyses. The interspecific genetic distance was about 17% for COI and around 3% for 16S. Furthermore, the carapace armature and Go1 morphology (Fig 7B and 7E) are considerably different. Nonetheless, these two species are the members of the genus in which the carpus and merus of the cheliped share the character of bearing tufts of setae on the inner margins. This appears to represent yet another example among marine decapod crustaceans of a species pair diverging from a common ancestor after gene flow was interrupted by closing of the Isthmus of Panama (~3.5 mybp [68]). A growing number of such amphi-American or trans-isthmian species pairs are now underpinned by molecular phylogenetic analyses that provide a measure of genetic divergence since separation [35,68,69,70]. It is noteworthy that lacking modern genetic tools, Rathbun [3] had already regarded P. subovata rather than P. edwardsii as the Pacific analog of the Atlantic P. punctata.
Persephona orbicularis Bell, 1855 and P. townsendi (Rathbun, 1893) Persephona orbicularis was recovered in the same clade as "P. townsendi" with high support in ML and BAY analyses. Furthermore, the genetic divergence value in COI was 1.7%, which is between values corresponding to intraspecific and interspecific genetic distance. The specimens are also very similar in their morphological characteristics. Rathbun [3] distinguished these species based on the shape and size of the spines in the sub-hepatic margin, which were reported to be long and acute in P. townsendi but shorter and dentiform in P. orbicularis. However, as previously noted, such characters are unreliable because these crabs burrowing habits may cause wear [71], [72]. It has been reported also that these crabs show sexual dimorphism in that spines are more developed in males of P. orbicularis than in females (Boone [73], pl. 11, figs. A, B). We were also able to confirm that these spines are more developed on the type male of P. townsendi (USNM 17382) than in the female type of P. orbicularis (USNM 17382). Those variable characters were, however, used by Rathbun [3] to separate these two species. In addition to our finding little genetic difference between these putative species, we found their Go1 structures to be remarkably similar (Fig 7D and 7G), without features to separate them. Thus, based on the molecular and morphological evidence, and their overlapping distributions, the two species are synonymized. As P. orbicularis is the older name available, P. townsendi becomes a junior subjective synonym of P. orbicularis.
Persephona subovata (Rathbun, 1893) and Iliacantha hancocki  In our analysis, the separation of P. subovata and "I. hancocki" cannot be supported. The genetic divergence value obtained with COI was only 0.3%, clearly within the range of intraspecific genetic variation, placing "I. hancocki" and equivalent to other existing species in the genus Persephona.
Previous study of the morphological similarities between these species revealed no significant differences upon which to base separations and encouraged further morphological studies with the inclusion of type materials [19]. From our studies of specimens representing "I. hancocki" and P. subovata from Costa Rica and Panama, along with type materials of both species (I. hancocki, USNM 69260; Myra subovata, USNM 17385), we could find no consistently definitive morphological characters to use for distinction. The characters used by Rathbun [74] to establish I. hancocki are all deemed to be variations broadly representative of P subovata. For example, the presence of a definite line of granules on the lateral margins, described to occur in P. subovata but not I. hancocki, is in fact present in the holotype of the later. Also the holotypes of I. hancocki as well as P. subovata exhibit a series of granules on the subhepatic region instead of a "rudimentary tooth" as described by Rathbun [74] exclusively for I. hancocki. Rathbun [3] noted that the chelipeds of I. hancocki are long, but much less slender than the others allied species. The inconsistency of the morphological characters separating these two species was also noticed by Hendrickx [19] who found differences only in "the size of the granules on the periphery of the carapace and the relative size of the posterior teeth". These two characters also appear to be variable and dependent on the maturity of the specimens, based on different specimens analyzed in the present study. Effects of ontogeny on the distance between the posterior teeth was also found to be variable in other species of Persephona (i.e. P. lichtensteinii) and is thus generally of questionable value. Thus, based on the molecular and morphological evidence, the two species must be synonymized. As P. subovata is the older name available, I. hancocki becomes a junior subjective synonym of P. subovata.
The above synonymization of I. hancocki does not affect the taxonomic status of the genus Iliacantha. There are five more species within Iliacantha for which we do not have genetic data to assess the validities of the two genera. However, we are able to separate both genera based on morphology of the cheliped alone.
The original characters described by Stimpson [75] to separate Iliacantha are mostly useless, i.e. the three spines at the posterior extremity of the carapace, used to diagnose Iliacantha, can be also found on Persephona; the extremities of the pterygostomian channels projecting considerably beyond the orbits and the fused abdominal segments 3-5 and 4-6, for males and females respectively, are also present on all members of Persephona [76]. The characters of the cheliped that can separate the two genera are as follow: Iliacantha has elongated chelipeds, slender fingers with similar thickness throughout its length. On the other hand, Persephona has the chelipeds rather massive, with large fingers whose thickness decreases from the base to the apex. Additionally, specimens of Iliacantha have a peculiar orientation of the hands, which are twisted in the way that the fingers open in a vertical, instead of a horizontal plane. In what refer to those characters, Iliacantha hancocki resembles more members of Persephona than those of Iliacantha.

Conclusions
Based on molecular and morphological analyses, we propose substantial modifications to the current taxonomy of the genus Persephona. We restrict the distribution of P. crinita to the Gulf of Mexico; we synonymize P. finneganae under P. lichtensteinii; we confirm that P. mediterranea and P. aquilonaris are both valid species, the first occurring in part of southern Florida, the Caribbean and South America, the second restricted to the Gulf of Mexico and eastern North America. We also show that P. townsendi is a junior synonymy of P. orbicularis, and that Iliacantha hancocki is a junior synonym of P. subovata. Following this revision, there are eight valid species of Persephona.  [84]. -Rouse, 1970: 24, fig. 65 [85]. -Felder, 1973: 4, 40, 42, pl. 5, fig. 4  Revised diagnosis.-Carapace with three marginal spines on posterior margin; adult background carapace color light beige to pale yellow, typically marked by strong dark reddish brown patterning of anastomosed spots distributed symmetrically to each side of light unspotted median line, dark spots separated by narrow lighter veins of background color, veins and lines of intersecting spots converging somewhat posteriorly, pattern broken into isolated small spots posterolaterally; front prominent, width usually two times length. Cheliped propodus usually more than three times longer than wide, carpus and distal part of merus, without tufts of setae on inner margins. Go1 distal opening large, directed towards mesioventral face; apex simple, without prominent lobe on mesiodistal margin, without subapical setae (Fig 8A-8C).
Distribution.-Western Atlantic: Gulf of Mexico.
Remarks.-The type of Persephona crinita was requested from the USNM where it is now kept (USNM63739). Unfortunately, the type remains on loan to another colleague and we were unable to examine it. However, using the original description of Rathbun [61], the figures of Correa-Sandoval, 1991: 12 [107]. -Hendrickx, 1993: 8 [108]. -Hendrickx, 1995: 129 [109]. Revised diagnosis.-Carapace with three marginal spines or tubercles, three long spines on posterior margin; adult carapace background color pale rose pink to bluish off-white, sometimes with poorly defined slightly darker rose area centered on and dominating most of dorsal surface, without symmetrical dark patterning of lines or spots; front slightly produced, width usually three times length. Cheliped propodus length usually more than three times width, carpus and distal part of merus with tufts of setae on inner margins. Go1 distal opening narrow, directed forward; apex simple, without prominent lobe on mesiodistal margin, with a row of setae from the base to apex (Fig 7B).
Remarks.-The type locality of P. edwardsii was considered to be the Galapagos Islands [18]. However, Garth [112] clarified that P. edwardsii was collected by Hugh Cuming, and only later said to have originated from the Galapagos Islands. Since this and other of Bell's reported collections could not subsequently be found in the Galapagos Islands, Garth removed P. edwardsii from the list of known Galapagos brachyuran crabs. Thus, the type locality of P. edwardsii remains in question.
Persephona Revised diagnosis.-Carapace with seven marginal spines or tubercles, three long spines or tubercles on posterior margin, two spines on lateral margin, two spines or tubercles on subhepatic margin; adult background carapace color pale rose pink to bluish off-white, normally uniform darker rose area dominating most of dorsal surface, often with diffuse withe color at the spines, without symmetrical dark patterning of lines or spots; front not prominent, width usually more than three times length. Cheliped propodus length usually exceeding 4 ½ times width, carpus and distal merus without tufts of setae on inner margins. Go1 distal opening narrow, directed towards ventral face; apex simple, without prominent lobe on mesiodistal margin, with subapical setae (Fig 7C).
Remarks: The holotype and paratype photographs of P. lichtensteinii (NHMUK White 1 97. b and NHMUK White 1 97.a, respectively) were analyzed, as well as the holotype and paratype of P. finneganae (USNM 67989 and MZUSP 934, respectively). Based on original description of Leach [76] we are sure about the identity of P. lichtensteinii and the synonym of P. finneganae to this.
Distribution.-Western Atlantic: Eastern Gulf of Mexico, Caribbean Sea, Antilles, Venezuela, Suriname, French Guiana, Brazil to Uruguay.
Remarks.-Bell ([18]: 292-293) analyzed the types specimens of Leach at the British Museum and concluded that P. lamarckii and P. latreillei were identical, considering the differences on the pterygostomian region as variation. He noted that the coloration of those specimens were not preserved, not being possible the identification of theses specimens as P. mediterranea (Cancer mediterraneus) by Herbst. He also commented that Browne does not mentioned coloration on his description to Cancer punctatus. Thus, Bell synonymizes both Leach species to Persephona guaia. Rathbun ([3]: 153) considered the three species as synonyms of the Persephona punctata punctata, and those animals with occurrence from West Indies to Brazil. The specimens from United States, from New Jersey to Texas, should be considered Persephona punctata aquilonaris, as stated on introduction. We analyses the photographs of the types of P. lamarckii (NHMUK White 1 96a), P. latreillei (NHMUK White 1 96d), P. guaia (OUMNH 13775), and the specimen of P. puntata aquilonaris (USNM 62057). Based on morphology, we could infer that P. lamarckii is clearly a junior synonym of P. punctata with acute margin on the pterygostomian region and front not prominent. While P. latreillei might be a junior synonym of P. mediterranea. The specimen has obtuse margin and prominent front, from West Indies. Although we do not have access to DNA, color pattern or gonopod morphology, we could infer based on Rathbun's and our pattern of distribution of the species P. mediterranea and P. aquilonaris, that the specimens from Caribbean to Brazil are indeed P. mediterranea and those from New Jersey to Gulf of Mexico are P. aquilonaris. According to DiMauro ( [106]: 173) the specimen P. guaia (OUMNH 13775) is registered as syntype of Bell, from West Indies. Based on the photography, we could identify the same pattern of coloration of P. mediterranea (Cancer mediterraneus) by Herbst, and the obtuse margin on pterygostomian region. Thus, we concluded that P. guaia by Bell is in fact a junior synonym of P. mediterranea. With the analysis of the specimen of P. puntata aquilonaris (USNM 62057) from St. Augustine, Florida we could identify obtuse margin and prominent front, but also the different pattern of coloration on carapace, as described in P. aquilonaris by Rathbun ([3]: 154). Thus, the holotype of P. punctata aquilonaris is in fact correctly as synonym of P. aquilonaris.
Persephona orbicularis Bell, 1855 Fig 7D and  Revised diagnosis.-Carapace with five marginal spines or tubercles, three long spines on posterior margin, two spines at subhepatic margin; adult background carapace color pale dark reddish to purplish, sometimes with defined slightly darker reddish patterning of veins distributed asymmetrically to each side of orange light spotted median line, area dominating most central part of the carapace; front not prominent, usually width more than three times length. Chelipeds propodus length usually more than four times width, carpus and distal part of merus without tufts of setae on inner margins. Go1 distal opening narrow, directed towards meisoventral face; apex simple, without prominent lobe on mesiodistal margin, with subapical setae (Fig 7D and 7G).
Remarks.-The type locality of P. orbicularis reported as Valparaiso, Chile by Bell [18]. The type specimen was brought by Mr. Miller, a surgeon in the British Royal Navy, and deposited into the Bell collection [18]. However, Boone [71] questioned the occurrence of P. orbicularis in Chile. Garth [17] also suggested that this species was never found in Chile. He argued that species of Persephona have a tropical distribution in the Americas tropics; thus P. orbicularis may have been collected farther north than Chile. Nonetheless, he suggested that the occurrence of P. orbicularis collected by Boone (Panama) it is within the range of distribution of P. townsendi, which could indicate that the last is synonymous of the first. Based on morphological and molecular analyses we conclude that the P. townsendi is a junior synonymy of P. orbicularis. The distribution of P. orbicularis goes from the Gulf of California to Peru. This represents another mistake in the type locality as it was observed for P. mediterranea (see remarks).
Revised diagnosis.-Carapace with five marginal spines or tubercles, 3 long spines on posterior margin, 2 tubercles at subhepatic margin; adult background carapace color pale purplish to bluish off-white, sometimes with defined slightly darker purplish patterning of anastomosed spots distributed symmetrically to each side of orange light spotted median line, dark spots separated by narrow lighter veins of background color, veins and lines of intersecting spots converging somewhat posteriorly, area dominating most of frontal dorsal surface, but diffused purple color, often with diffuse pale brown to orange near frontal margin and subhepatic margin; front not prominent, width usually four and half times length. Chelipeds with propodus usually more than 3 times and half times longer than wide, carpus and distal part of merus with tufts of setae on inner margins. Go1 with distal opening wide, directed towards ventral face; apex expanded laterally, without prominent lobe on mesiodistal margin, without subapical setae (Fig 7E).
Persephona subovata (Rathbun, 1894) Fig  Revised diagnosis.-Carapace with three marginal spines or tubercles, three long spines on posterior margin; adult background carapace color reddish or orange, with darker red area dominating most of central and frontal dorsal surface, without symmetrical dark patterning of lines or spots; front slightly produced, width usually two times length. Cheliped propodus length usually more than six times width, carpus and distal part of merus without tufts of setae on inner margins. Go1 distal opening narrow, directed towards meisodorsal face; apex simple, without prominent lobe on mesiodistal margin, with subapical setae (Fig 7F).
Palacios-Theil, M. Terossi and I. Miranda for help with molecular data, M. Massarani for the drawings, and R. Buranelli for some photos of Persephona. This is contribution number 179 from the UL Laboratory for Crustacean Research and 1020 from the Smithsonian Marine Station, Ft. Pierce, Florida.