Systematics and diversification of Anindobothrium Marques, Brooks & Lasso, 2001 (Eucestoda: Rhinebothriidea)

Tapeworms of the genus Anindobothrium Marques, Brooks & Lasso, 2001 are found in both marine and Neotropical freshwater stingrays of the family Potamotrygonidae. The patterns of host association within the genus support the most recent hypothesis about the history of diversification of potamotrygonids, which suggests that the ancestor of freshwater lineages of the Potamotrygonidae colonized South American river systems through marine incursion events. Despite the relevance of the genus Anindobothrium to understand the history of colonization and diversification of potamotrygonids, no additional efforts were done to better investigate the phylogenetic relationship of this taxon with other lineages of cestodes since its erection. This study is a result of recent collecting efforts to sample members of the genus in marine and freshwater potamotrygonids that enabled the most extensive documentation of the fauna of Anindobothrium parasitizing species of Styracura de Carvalho, Loboda & da Silva, Potamotrygon schroederi Fernández-Yépez, P. orbignyi (Castelnau) and P. yepezi Castex & Castello from six different countries, representing the eastern Pacific Ocean, Caribbean Sea, and river basins in South America (Rio Negro, Orinoco, and Maracaibo). The newly collected material provided additional specimens for morphological studies and molecular samples for subsequent phylogenetic analyses that allowed us to address the phylogenetic position of Anindobothrium and provide molecular and morphological evidence to recognize two additional species for the genus. The taxonomic actions that followed our analyses included the proposition of a new family, Anindobothriidae fam. n., to accommodate the genus Anindobothrium in the order Rhinebothriidea Healy, Caira, Jensen, Webster & Littlewood, 2009 and the description of two new species—one from the eastern Pacific Ocean, A. carrioni sp. n., and the other from the Caribbean Sea, A. inexpectatum sp. n. In addition, we also present a redescription of the type species of the genus, A. anacolum (Brooks, 1977) Marques, Brooks & Lasso, 2001, and of A. lisae Marques, Brooks & Lasso, 2001. Finally, we discuss the paleogeographical events mostly linked with the diversification of the genus and the protocols adopted to uncover cryptic diversity in Anindobothrium.


Phylogenetic analyses
Two major analytical protocols were applied according to the main goals of this study. The first set of analyses addressed the phylogenetic position of Anindobothrium within major lineages of cestodes closely related. The second analysis used molecular data as a tool for species discovery and delineation within Anindobothrium.
Phylogenetic position of Anindobothrium. Nucleotide sequences of 18S rDNA and 28S rDNA (D1-D3 regions) were first submitted to phylogenetic analysis by direct optimization (DO; [35]) using POY (version 5.1.1; [36]) under parsimony as optimality criteria. Initial tree searches included 10 iterations of two independent searches for 1 h 30 min using the command search [i.e., search(max_time:0:01:30]) assuming equal weights for all character transformations. This search was conducted in a 10 X 2.83 GHz Intel 1 Core™2 Quad Processor Q9550 computer cluster. A DO sensitivity search [37] was performed using nine alignment parameter sets in which gap extension costs varied from one to eight and transformation costs (transversions and transitions) from one to four with an opening gap cost twice that of gap extension cost rendering the following alignment cost ratios for opening gaps, extension gaps, transversions, and transitions, respectively: 0:1:1:1, 2:1:1:1, 2:1:1:2, 2:1:2:1, 2:2:1:1, 2:2:1:2, 2:2:2:1, 2:4:1:1, 2:4:1:2, and 2:4:2:1. For each parameter set, tree space was explored by two independent searches for 2 h [i.e., search(max_time:0:02:00]) in the same computer cluster environment as the previous analysis but using five nodes. After compiling candidate trees by DO, we submitted unique topologies to tree refinement by tree-fusing algorithm [38] and re-diagnosis by iterative pass alignment (DO/IP+Fuse; [39]). Our final analytical step under this optimality criterion was to verify the results obtained with DO/IP+Fuse by performing a phylogenetic analysis of the implied alignment (sensu Wheeler [40]) generated by the previous step in TNT [41] using its New Technology searches [41,42] with the following parameters: rep 100, ratchet 50, fuse 20, hold 10. We evaluated nodal support by using Goodman-Bremer values (GBS, [43][44][45]; see [46]). To obtain this metric, we considered the shortest tree found by TNT based on the implied alignment above and executed a modified version of the script BREMER.RUN distributed with TNT. This script considered 1,000 replicates with 10 repetitions of ratchet and drift [41,42] in constrained searches and the remaining default parameters. Finally, putative transformations for selected branches were compiled using the consensus tree obtained by TNT with YBYRÁ [47].
We also analyzed the previous datasets using maximum likelihood (ML) to identify nodes sensitive to a different optimality criterion. We started by submitting the implied alignment generated by DO/IP+Fuse to model selection in jModeltest (version 2.1.6; [48,49]) considering 88 candidate models ranked by AICc scores. Following, tree searches were performed using the parallel implementation of GARLI (version 2.0; [50]) applying 1,000 independent search replicates and remaining default parameters of GARLI configuration file. This search was conducted by implementing 20 searches replicates in 50 X 2.83 GHz Intel 1 Core™2 Quad Processor Q9550 computer cluster. Log-likelihood difference support (LLD; sensu Lee et al. [51]; see [52,53]) was calculated for selected nodes using constrained negative searches in GARLI under the same configuration settings as the initial tree search.
Species discovery within Anindobothrium. Our first approach was to perform a simultaneous phylogenetic analyses of nuclear regions 18S rDNA, 28S rDNA (D1-D3), Cal, and ITS-1, and the mitochondrial region of COI for all representatives of Anindobothrium from three major bodies of water: eastern Pacific Ocean, Caribbean Sea, and Neotropical freshwater systems. We submitted the nucleotide sequences to phylogenetic inference by direct optimization (DO, [35]) using POY (version 5.1.1; [36]) under parsimony as the optimality criterion. Tree search was performed by three independent searches for 30 min using the command search (i.e., search(max_time:0:00:30)) assuming equal weights for character transformations in the same computer cluster environment mentioned above using 10 nodes. All trees compiled by DO were re-diagnosed by iterative pass algorithm (IP, [39]) and the implied alignment (sensu Wheeler, [40]) was submitted to TNT (Golloboff et al. [41]) to verify the results using xmu algorithm with 1,000 replicates and holding at the most 10 trees per replicate. A similar analysis was conducted partitioning the dataset into nuclear and mitochondrial regions. The first partition was analyzed in POY as described above and COI was only analyzed in TNT with the same settings as before. Selected clades were diagnosed using YBYRÁ [47] within each partition.
We also performed a phylogenetic analysis using ML as optimality criterion based on the implied alignments resulted from previous analyses. Model selection for the concatenated dataset and for each nuclear region was performed in jModeltest considering 88 candidate models ranked by AICc scores. Under this optimality criterion, phylogenetic analyses were performed for each gene region separately, as well as for two concatenated datasets: one considering only nuclear genes and the other including all regions. We analyzed each concatenated dataset using two different partition models. One considered a single substitution model for all regions and the other considered individual substitution models. Independently of dataset or partition model, tree searches were performed using parallel implementation of GARLI applying a total of 1,000 independent search replicates in 10 X 2.83 GHz Intel 1 Core™2 Quad Processor Q9550 computer cluster. The best partition model was selected based on AICc information criterion.
Congruence between phylogenetic patterns and morphological data was observed by compiling morphometric and meristic data for 137 marine specimens of Anindobothrium. The dataset included the newly collected material as well as the type series of A. anacolum. A total of 29 measurements were selected, most of which are traditionally used in the taxonomy of the group and missing entries were filled with the mean within clade recovered by phylogenetic inference. All statistical analyses were performed in R as follows. The first step was to identify and exclude all highly correlated measurements (i.e., r > 0. 70). Then, a Principal Component Analysis (PCA) was performed to identify whether or not there were any morphological patterns in our data congruent with phylogenetic patterns. Putative groups suggested by phylogenetic analyses and PCA were tested by Linear Discriminant Analysis (LDA; [54,55]). The error rate of the discriminant function was evaluated by 1,000 iterations of 10-fold cross validation procedure. This dataset and R scripts are available in the repository Dryad under doi: 10.5061/dryad.gr0sb.

Phylogenetic analyses
Phylogenetic position of Anindobothrium. To address the phylogenetic position of Anindobothrium within the selected lineages of cestodes, 18S and 28S nucleotide data were generated for 27 terminals (Table 1). These terminals included two haplotypes of Caulobothrium sp. found in Potamotrygon sp. from the Delta of Orinoco, two members of Rhinebothroides sp. collected from Potamotrygon wallacei de Carvalho, Rosa & de Araujo from the Rio Negro, three specimens of A. lisae found in freshwater potamotrygonids from the Rio Negro and Orinoco river basins, and 20 other haplotypes of Anindobothrium, most of which collected from species of Styracura. In addition to these haplotypes, the dataset included one individual of Anthocephalum hobergi (Zamparo, Brooks & Barriga, 1999) Marques & Caira, 2016 from Urobatis tumbesensis (Chirichigno & Mc Eachran) off the coast of Ecuador, which was used as outgroup taxa. To this dataset, we added 67 selected terminals from Healy et al. [2], Caira et al. [7], Ruhnke et al. [8], and Marques and Caira [10] (Table 2). These additional sequence data included 16 out-group terminals representing members of the Litobothriidea (three), Cathetocephalidea (two), Lecanicephalidea (four), Onchoproteocephalidea (one), Phyllobothriidea (one), and "Tetraphyllidea" (five); and 51 rhinebothriideans, which included members of all families according to Ruhnke et al. [8] and Marques and Caira [10]. The complete dataset considered 93 terminals. Sequences of 18S ranged from 1,350 to 1,412 unaligned base pairs (bp)-MAFFT alignment (MAFFTaln) resulted in sequences of 1,464 bp-, and sequences of 28S ranged from 803 to 873 unaligned base pairs-MAFFTaln of 1,025 bp.
Tree search using direct optimization under equal weights for all transformations was based on 1,279 builds followed by TBR (Tree-bisectioning and redraft, [56]), 22,032 cycles of tree fusing [38], and 622 iterations of ratchet [42]. Sensitivity search included 186 builds followed by TRB, 2,437 cycles of tree fusing, and 88 iterations of ratchet. Combined, the tree search under direct optimization found 72 unique topologies ranging from 6,407 to 6,663 steps in length. The re-diagnosis of these 72 unique topologies under iterative pass algorithm followed by tree fusing rendered a single topology with 6,350 transformations. The implied alignment submitted to TNT, with 1,698 bp for 18S and 1,299 bp for 28S, resulted in two topologies with 6,346 steps in length. These two trees differed on internal arrangements of apical terminals with near-zero branch lengths. Fig 1A displays the summary results of this analysis, including GBS support for selected nodes. A topology with all terminals is provided in supplementary results (S1 Fig).
Implied alignment resulted from the analyses above was submitted to ML phylogenetic inference assuming GTR+Γ+I as the substitution model. This analysis resulted in a topology with -lnL 28523.9385, which summary is presented in Fig 1B along with values of Likelihood Length Difference as a measure of support for selected clades. The detailed sister-group relationships hypothesized by ML analysis is presented in supplementary results (S2 Fig).
Both optimality criteria supported the monophyly of families but suggested different sets of sister-group relationships (Fig 1A and 1B). These differences involve clades with relatively lower support. According to the parsimony analysis, Anindobothrium is sister to the Anthocephaliidae, whereas ML topology suggested it is sister to the clade Anthocephaliidae +Escherbothriidae. Although the internal branches supporting sister-group relationships among families and genera, such as Anindobothrium and the "New genus 11" (Rhinebothriinae n. sp.1 of Healy et al. [2]), have relatively lower support, all families have a relative high support (Fig 1A and  1B). Also, branches leading to clades for recognized families, representatives of Anindobothrium nested, and for the "New genus 11" possess sets of molecular synapomorphies that could be used as putative diagnostic nucleotides for each of them (Fig 2). For instance, Anindobothrium is supported by 30 transformations from the 18S region, three of which are observable (non-gap), unique, unambiguous synapomorphies, and 44 from 28S region, four of which are observable (non-gap), unique, unambiguous synapomorphies. The amount of transformations inferred for Anindobothrium (74) is smaller than what was recovered for the Echeneibothriidae de Beauchamp, 1905 (86) but larger when compared to all other families (Fig 2). Implied alignment data files and consensus tree used to provide diagnosis for each clade of the Rhinebothriidea are available in the repository Dryad under doi:10.5061/dryad.gr0sb.
Species discovery within Anindobothrium. Molecular data: Our phylogenetic analyses utilized 24 terminals, 23 haplotypes of Anindobothrium and one specimen of Anthocephalum hobergi used to root the tree (Table 1). For this dataset, unaligned sequences of 18S ranged  The biogeographical representation of this dataset included three individuals identified as A. lisae (one) ex Potamotrygon schroederi from the Rio Negro and (two) ex P. orbignyi from the Rio Negro and Mid-Orinoco river basins, three terminals identified as A. anacolum from Trinidad & Tobago parasitizing Styracura schmardae and one from Lake Maracaibo infecting P. yepezi. We also included 15 members of Anindobothrium, among which 10 individuals were from Belize and one from the Caribbean coast of Panama recovered from S. schmardae, and four worms parasitizing S. pacifica from the eastern Pacific coast of Panama (Table 1).
Tree search using direct optimization under equal weights for all transformations was based on 126 builds followed by TRB, 13,465 cycles of tree fusing, and 79 iterations of ratchet, found 144 trees, 1,249 steps long. The re-diagnosis of all trees using iterative pass, utilizing 1,386 bp for 18S, 1,148 bp for 28S, 393 bp for Calmodulin, 549 bp for COI, and 803 bp for ITS, found all of them to have 1,245 steps. The phylogenetic analysis of the implied alignment resulted in 11 trees equally parsimonious with 1,245 steps. The consensus tree of these 11 topologies (Fig 3A) suggested the monophyly of A. lisae, the haplotypes of Anindobothrium from eastern Pacific, and a clade comprised by members of the genus from Belize and Panama (Caribbean). However, A. anacolum resulted as paraphyletic. The same phylogenetic pattern was observed when the partition for COI was analyzed separately in TNT. This analysis found 12 topologies at a cost of 371 steps, for which the consensus tree is presented in Fig 3B. Our analysis based on direct optimization of nuclear genes included 184 builds, 22,826 cycles with tree fusing, and 156 iterations of ratchet and found 215 trees with 868 steps. The re-diagnosis of these trees by iterative pass found all of them to have 863 steps and the implied alignment analyzed in TNT rendered two topologies with same cost. Contrary to COI, the nuclear genes recognized four monophyletic groups within Anindobothrium. Among those not recognized by COI, A. anacolum resulted as a monophyletic group (Fig 3C). Implied alignment data files and consensus tree used to provide diagnosis for each clade of Anindobothrium are available in the repository Dryad under doi:10.5061/dryad.gr0sb.
The ML analysis rendered similar results. Model selection suggested that the best fitting model for the concatenated dataset was TVM+Γ+I, whereas for 18S, 28S, Cal, COI, and ITS-1 it was TPM3uf+Γ, TIM2+Γ, TIM1+Γ, TPM1uf+I, and TPM1uf+Γ, respectively. For the simultaneous analysis of all data we utilized two partition models. One analysis utilized the model TVM+Γ+I for all concatenated partitions and the other assumed distinct substitution models for each gene regions separately. The AICc favored the partition model in which different substitution models were assigned to each region (Table 3). This analysis recovered the same phylogenetic pattern as obtained by parsimony analysis (Fig 3A). The phylogenetic analysis of COI using ML corresponded to most of the nodes recovered by parsimony analysis (see Fig 3B). For nuclear genes, the partition model also favored different substitution models for each partition ( Table 3). The ML analysis of nuclear regions displayed the same topology as obtained for the parsimony analysis (see Fig 3B). However, ML analysis of individual partitions did not recover some nodes.
A comparison of KP2 distances among representatives of the marine clades showed that COI was the most divergent region used in this study (Table 4). On average, sequences of COI from haplotypes of Anindobothrium collected in the eastern Pacific Ocean differed in 13.8% from representatives of the genus collected in the Caribbean Sea. Between the clades in the Caribbean, A. anacolum differed from A. inexpectatum sp. n. in 10.5% (see Table 4). Calmodulin and ITS were together the second most divergent regions. In general, for both regions, specimens from the eastern Pacific differed in 2.7% from those collected in the Caribbean and the differences between A. anacolum and A. inexpectatum sp. n. were 1.0% and 1.3%, respectively. Finally, 18S and 28S regions showed little variation within marine haplotypes of Anindobothrium (Table 4).
Haplotype networks of the mitochondrial region COI and the nuclear markers 18S, 28S, Calmodulin, and ITS revealed some interesting patterns (Fig 4). The most obvious is the observation that there is no shared haplotypes among the biogeographical areas sampled (i.e., freshwater rivers of South America, eastern Pacific Ocean, and the Caribbean coasts of Central and South America). As expected, the haplotypes of A. lisae are the most divergent among Anindobothrium. For genes with low substitution rates, such as 18S and 28S, putative new species were segregated from others by few mutational steps. For instance, A. carrioni sp. n. and A. inexpectatum sp. n. were segregated from A. anacolum by 3 and 2 mutational steps, respectively, according to 18S sequences (Fig 4). A similar pattern is observed for 28S in which A. anacolum is separated from these two new species by a single mutational step. However, these newly recognized lineages are well segregate from others by faster evolving genes (i.e., Calmodulin, COI, and ITS; see Fig 4).
Morphological data: The morphological dataset utilized 29 morphometric variables for 137 specimens. We focused on the marine representatives of Anindobothrium considering the following representation: 29 specimens of A. anacolum from the type locality-including five individuals from the type series (USNPC 73969, holotype and HMWL 20265a-d, paratypes), 24 individuals attributed to A. anacolum from Trinidad & Tobago, 41 worms from the Caribbean representing the clade formed by specimens collected in Styracura schmardae from Belize # Taxa, number of taxons analyzed; # Branches, number of branches; # EPSM, number of free (estimated) parameters in substitution models; K, total number of free parameter, which includes topology, branch lengths, and free parameters in the substitution model(s); # Char, number of characters utilized in ML analyses (unique patterns); lnL, negative Log-Likelihood scores; AIC, Akaike Information Criterion score of partition models; AICc, Corrected Akaike Information Criterion score of partition models.
https://doi.org/10.1371/journal.pone.0184632.t003 (31) and Panama (10), and 32 specimens of Anindobothrium collected from S. pacifica from the eastern Pacific coast of Panama. Eight of the initial 29 morphometric variables were found to be highly correlated (r > 0.70, see supplementary S1 Table) and were excluded from further analyses. The PCA utilizing 21 morphometric variables suggested that PCA1 explains 23% whereas PCA2 explains 15% of total variance ( Fig 5A). The centroid around the means (95% confidence) suggested that individuals assigned to A. anacolum clustered together with most of the specimens from its type series and worms collected in the type locality as well as from Trinidad & Tobago. The results of this analysis revealed a great overlap between members of this genus collected in the eastern Pacific Ocean and from the Caribbean coast of Panama/Belize. These two populations overlapped at the edge of the centroid of A. anacolum. The loadings of PCA1 indicate that total length, bothridial length and scolex width have greater influence in that component, whereas for PCA2, most of the variance is due to the length of poral testes and cirrus sac dimensions (S2 Table).
The LDA was performed to evaluate whether the recognized marine lineages based on molecular data could be discriminated by morphological data. Our analysis was able to discriminate representatives of marine clades ( Fig 5B). The proportion of traces (i.e., the percentage separation achieved by each discriminant function) was 59% for LDA1 and 41% for LDA2. Loadings for each discriminant function suggested that terminal mature proglottid ratio and number of testes are the most important measurements for LDA1, whereas the number of   mature proglottids and vitelline follicles length were most important measurements for LDA2 (S3 Table). The cross validation procedure indicated a discriminant function error rate of 3%. In summary, our results indicated that the phylogenetic position of Anindobothrium was sensitive to optimality criteria, especially in nodes that displayed relative lower support. However, we found that the amount of molecular divergence in the branch supporting the monophyly of Anindobothrium was as great as, if not greater than, those found in branches of presently recognized families within the Rhinebothriidea. For Anindobothrium, we were able to recognize four independent lineages, two of them already described, A. anacolum and A. lisae, whereas the other two require formal description. The recognition of these lineages was not only based on molecular data but also supported by the LDA analysis. Based on these results the proposed taxonomic actions are as follows.

Nomenclatural acts
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. 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:0CC9ACE3-9691-4043-8695-B70B6C4A2342. 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 and LOCKSS.

Remarks:
The phylogenetic analysis of molecular data provided unequivocal evidence that Anindobothrium is a member of the Rhinebothriidea. This results corroborate Ruhnke's [3] observation that the presence of stalked bothridia found in members of Anindobothrium would support the assignment of this genus to this order (see also Healy et al. [2] and Ruhnke et al. [9]).
The sub-ordinal classification of the Rhinebothriidea, as we know to date, was proposed by Ruhnke et al. [8] which recognized four families: Anthocephaliidae Ruhnke, Caira & Cox, 2015, Echeneibothriidae, Escherbothriidae Ruhnke, Caira & Cox, 2015, and Rhinebothriidae Euzet, 1953. Despite the great taxonomic representation of their dataset, no specimens of Anindobothrium were included in their analysis. Our results revealed that the position of Anindobothrium is sensitive to optimality criterion since the MP topology suggested that this genus is sister to Anthocephaliidae, whereas the ML topology suggested that Anindobothrium is sister to the clade Anthocephaliidae+Escherbothriidae. Therefore, based on the phylogenetic patterns recovered, we would not be able to assign Anindobothrium to any family of the Rhinebothriidea. In addition, the amount of molecular divergence of the branch leading to haplotypes of Anindobothrium is comparable to those supporting families and the morphology of the genus that does not conform with the diagnoses of the families provided by Ruhnke et al. [8]. Hence these observations justify the erection of a new family to accommodate Anindobothrium.
The family Anindobothriidae fam. n. can be distinguished from the Echeneibothriidae by the absence of a myzorhynchus in the adult stage and by the position of genital pore (anterior vs. mid-posterior). It differs from the Rhinebothriidae by possessing a clear anteroposterior orientation of the bothridia characterized by a conspicuous apical sucker and by the partial or total interruption of the vitelline follicles by the ovary. It closely resembles the Anthocephaliidae and the Escherbothriidae but can be easily distinguished by the possession of post-vaginal testes. Below we provide a revised key of the Rhinebothriidea to accommodate the Anindobothriidae fam. n.
Key to families of the Rhinebothriidea (modified from Ruhnke et al. Scolex with four stalked bothridia; myzorhynchus absent. Bothridia typically longer than wide, with or without longitudinal septa, with apical sucker and marginal loculi, with or without two rows of facial loculi. Mature proglottids longer than wide. Testes numerous, arranged in two irregular columns; post-poral field present. Vas deferens extending anteriorly from mid-proglottid to enter to cirrus sac at anterior margin, more porally than anti-porally; external seminal vesicle absent. Genital pores marginal, irregularly alternating from 15 to 44% from anterior end of proglottid; genital atrium shallow. Cirrus sac in anterior 1 4 of proglottid, thinwalled, tilted posteriorly, containing eversible coiled cirrus armed with spinitriches. Vagina extending from ootype along midline of proglottid to anterior margin of cirrus sac and laterally, becoming sinuous, to open into genital atrium anterior to cirrus sac; vaginal sphincter present; seminal receptacle absent. Ovary H-shaped in frontal view, tetralobed in cross section; ovarian margins lobulate. Vitellarium follicular, in two lateral bands; bands extending length of proglottid, interrupted by terminal genitalia, partial or total interruption by ovary. Uterus median, ventral, sacciform, with poorly differentiated lateral diverticula or lacking diverticula, extending from ovarian isthmus to anterior margin of proglottid. Excretory vessels four in number, arranged in dorsal and ventral pairs at lateral margins of proglottid. Parasites of the Potamotrygonidae (Myliobatiformes).
Remarks: The amended diagnosis made several contributions to that provided by Marques et. al. [4] and included modifications to accommodate the new findings on the morphology of this taxon. For instance, Marques et al. [4] described Anindobothrium as possessing bilobed bothridia, while in fact its members have elongated ones. Moreover, the absence of longitudinal septa became contradictory with the examination of additional specimens of A. anacolum (see below) and the presence of marginal loculi is found in all species we now recognize within the genus. Also, the vitelline follicles may be partial interrupted by the ovary. Since Anindobothrium is the only member of the Anindobothriidae fam. n., the genus can be differentiated from all other genera within the Rhinebothriidea by the same characters that differentiate the families of this order (see the key to the families above).
Remarks. The acquisition of additional specimens from the type locality and other localities in the Caribbean Sea and adjacent waters (i.e., Lake Maracaibo) allowed us to better understand the distribution, host association, and morphological variability of Anindobothrium anacolum. In addition, we provided for the first time the description of the microtriches morphology and a molecular diagnosis for this species.
Anindobothrium anacolum seems to have a restricted distribution in the Southern Caribbean Sea. Among all localities sampled in the Caribbean Sea, which included the coasts of Belize, Panama, Colombia and Trinidad & Tobago, we were only able to collect this species off the coasts of the last two countries. Furthermore, we also collected few specimens from Lake Maracaibo in Maracaibo infecting Potamotrygon yepezi, a freshwater potamotrygonid.
The presence of A. anacolum in Potamotrygon yepezi was unexpected. This species was only known to parasitize S. schmardae and most members of the Rhinebothriidea seems to exhibit oioxenous specificity for their hosts (sensu Euzet & Combes [57]; Ruhnke et al. [8]). Be that as it may, this could represent an accidental infection that needs further investigation.
The examination of the type series and additional material revealed that the original description provided by Brooks [58] as well as the diagnosis of this species amended by Marques et al. [4] did not provide a fair account on the bothridial morphology of this taxon. Both previous studies provided a description of the bothridia based on the holotype (USNPC 73969), which was poorly prepared and did not permit the verification of the presence of a longitudinal septum and the marginal loculi on the bothridia. Moreover, Marques et al. [4] illustrated a poorly defined anterior marginal loculus ( figure 1 in [4]), which we considered as an apical sucker. Therefore, in the present redescription, bothridial structures were observed and described in newly collected specimens.
The additional material provided a better understanding of the morphological variability of this species. Their examination allowed us to increase the range of some structures, which revealed to be more variable than previously reported. Both previous studies found that the total length was 6.8-15.4 but we found it to be 3.1-15.1. Brooks [58] observed 23-24 loculi, while Marques et al. [4] found between 42 and 44. The present study revealed a number of loculi of 39-55, which closely corresponds to the range reported by Marques et al. [4]. This difference could be attributed to the presence of longitudinal septa, not seen by Brooks in 1977. For genital pore position, both studies reported it to be 19-29% from anterior end, whereas, we found 15-32%.  008 1032 1,036 1,042 1,052 1,059 1,065 1,066 https://doi.org/10.1371/journal.pone.0184632.t005

Systematics and diversification of Anindobothrium
Anindobothrium anacolum can be easily differentiated from the only species previously assigned to the genus, A. lisae, by the morphology of the bothridia. Anindobothrium anacolum possesses marginal and facial loculi due to the presence of a longitudinal septum and transversal septa on bothridia, whereas A. lisae has only marginal loculi. As will be shown below, all marine species recognized for Anindobothrium possess bothridial architecture similar to A. anacolum, hence can be likewise differentiated from A. lisae. Despite the differences in the bothridial morphology, the proglottids of both species are similar by possessing numerous testes arranged in two irregular columns with post-poral field present, cirrus sac at the anterior 1 4 of proglottid, ovary H-shaped in frontal view, follicular vitelaria arranged in two lateral bands extending the length of proglottid, which is interrupted by terminal genitalia and partial or total interruption by the ovary. Finally, A. anacolum can be distinguished from all species of the genus by a set of 7 molecular synapomorphies (Fig 3C; Tables 5 Systematics and diversification of Anindobothrium Resdescription. [Based on the type series comprised of holotype (CHIOC 34375) and three paratypes (HWML 16379a and INPA 400a,b), and 34 additional mature specimens, which included 29 whole mounts, three worms observed with SEM, and two serially-sectioned]: Worms acraspedote, apolytic, 2.5-11.7 mm (25) long, composed of 7-24 (26) proglottids ( Fig 6B). Scolex with greatest width 386-1,197 (32), composed of four, stalked bothridia (Figs 6B and 10A). Bothridia orbicular-elliptoid shaped, 225-643 (17) long by 297-838 (12) wide, with 40-58 (10) marginal loculi and an apical sucker, 20-79 (8) long by 30-79 (9) wide (Fig 10B and 10C). Transverse and longitudinal septa absent. Short cephalic peduncle. Proximal surface of apical sucker covered by acicular filitriches (Fig 10D); medial portion of proximal bothridial surface covered by gladiate spinitriches (Fig 10E). Distal surfaces of bothridia covered by acicular filitriches (Fig 10C, 10F and 10G), by papilliform ( Fig 10C) and capilliform filitriches (Fig 10F). Cephalic peduncle covered by capilliform filitriches (Fig 10H).
Immature proglottids wider than long. Mature proglottids 772-2,395 (28) long by 257-643 (28) wide, 1-4 (26) in number. Vas deferens not sperm-filled in terminal proglottids. Gravid   insertion of 113 base pairs between positions 1,094 and 1,106 with 7 synapomorphies-C/ 1,095, G/1,097, G/1,100, A/1,102, G/1,103, A/1,104, and G/1,105 (see Table 8 Remarks. The redescription of Anindobothrium lisae added information on the morphology and patterns of distribution of microtriches and cross sections of the ovary for this taxon. Also, we have a better understanding of the distribution, host association and morphological variability of this species. Finally, we provided a molecular diagnosis for A. lisae for the first time. Anindobothrium lisae was only known parasitizing Potamotrygon orbignyi from the Rio Negro [4]. However, the examination of newly collected material from Venezuela revealed that this species also infects P. schroederi from mid-Orinoco. Rio Negro and Orinoco are known to share freshwater fauna [59], including these two species of hosts. Interestingly, A. lisae has never been found in P. schroederi from the Rio Negro, despite the examination of 31 specimens from there, as well as one from Ilha do Catalão, in the confluence of the Rio Negro and the Rio Solimões. In addition, this cestode has not been observed in P. orbignyi after examining 12 specimens from mid-Orinoco (Marques, unpubl. data). We found no molecular evidence that can be used to distinguish those two populations of A. lisae since the only haplotype from Orinoco included in our study nested between two haplotypes from the Rio Negro (Fig 3A-3C).
The redescription of A. lisae provided additional information on morphological variability as compared to what was reported by Marques et al. [4]. Comparing the ranges provided in the original description with those reported here revealed a larger spectrum of variation for structures such as bothridial width (322-775 vs. 297-838, respectively), number of poral anterior (5-13 vs. 3-15, respectively), and anti-poral testes (21-38 vs. 15-43 respectively). Also, the redescription provided a better understanding of the bothridial morphology of this species, for which the original account was based on the interpretation and illustration of an immature specimen (see figure 2B in [4]). Furthermore, the examination of additional material showed that this species can have the vitelline follicles partially interrupted by the ovary, which was not observed in the original description.
The early divergence of this lineage of Anindobothrium is evident in its morphology and nucleotide sequences. As pointed out above (see Remarks for A.anacolum), the morphology of the bothridia in A. lisae, which lacks facial loculi and longitudinal septa, is different from all marine species of the genus. As a species, A. lisae can be diagnosed by the largest sets of molecular synapomorphies, which comprise a total of 67 sites, and which include two unique insertions in 28S sequences (see Tables 5-9). For diagnostic purposes, however, this is the most conservative approach since there are other positions that could be used in addition to the molecular synapomorphies listed above. For instance, positions 300/T and 357/T for COI (Table 9), 1,032/deletion for 18S (Table 5), 1,042/C and 1,101/C for 28S (Table 8), and 229/ deletion and 277/deletion for ITS (     (Fig 12B). Short cephalic peduncle. Proximal surface of apical sucker covered by acicular filitriches (Fig 12D) and medial portion of proximal bothridial surface covered by gladiate spinitriches (Fig 12E). Distal surfaces of bothridia covered by gladiate spinitriches and acicular filitriches (Fig 12C, 12F and 12G). Cephalic peduncle covered by capilliform filitriches (Fig 12H).
Immature proglottids wider than long. Mature proglottids 765-2,058 (39) long by 201-391 (39) wide, 2-6 (39) in number. Some terminal proglottids with sperm-filled vas deferens (Fig 11B). Gravid proglottids not observed. Testes round to oval, 22-57 (38) long by 19-44 (38) wide; 23-44 (37) 84-183 (39) wide, containing eversible coiled cirrus armed with spinitriches (Fig 11B and 11C). Genital atrium present. Genital pores 17-27% (39) of proglottid length from anterior end, irregularly alternating. The vagina runs anterior to the cirrus sac and then turns posteriorly towards the ootype. Ovary near posterior end of proglottid, bilobed in dorso-ventral view, tetra-lobed in cross-section (Figs 11B and 9F), symmetrical, 228-651 (39) long by 97-235 (39) wide at isthmus; antero-ventral lobes converging anteriorly to midline of proglottid, but not fusing; ovarian margins lobulate. Vitelline follicles extending throughout length of proglottid, 11-44 (39) long by 11-26 (39) wide; partial or total interruption by the ovary. Eggs not observed. Remarks. Anindobothrium inexpectatum sp. n. was found in specimens of S. schmardae collected off the coast of Belize and northeastern coast of Panama in the Caribbean Sea. This https://doi.org/10.1371/journal.pone.0184632.t008 species appears to have a disjunctive distribution with respect to A. anacolum, despite sharing the same host and close distributional ranges. This somewhat odd host association and distributional pattern would suggest the hypothesis that this new species should be considered a population of A. anacolum. We think we have enough evidence to support an alternative interpretation by considering those specimens recovered off the coast of Belize and northeastern coast of Panama in the Caribbean Sea as a new taxon: Anindobothrium inexpectatum sp. n. This alternative hypothesis is supported by molecular, morphological data and patterns of endemism in the Caribbean Sea, which will be addressed in the Discussion. We found 6 unique unambiguous synapomorphies for COI and 11 for the nuclear genes 18S and ITS that can be used to characterize A. inexpectatum sp. n. (Fig 3B and 3C; Tables 5, 7 and 9). Anindobothrium anacolum, on the other hand, can be diagnosed based on 7 unique unambiguous synapomorphies (see above). Therefore, in total, there are 23 sites that can be used to distinguish the two species (Tables 5-7 and 9). In addition to the sets of molecular synapomorphies that diagnose this species, we also found that specimens of A. inexpectatum sp. n. display non negligible amounts of molecular divergence-as inferred by K2P-distances (Table 4)-when compared to haplotypes of A. anacolum (e.g., 10.5% for COI, 1.3% for ITS, and 1.0% for Calmodulin). COI divergence is of particular interest, since empirical data for more than 13,000 congeneric pairs among 11 phyla of Metazoa revealed that most pairs (98%) have over 2% sequence divergence in this mitochondrial region [60]. Therefore, given the amount of sequence divergence and that A. inexpectatum sp. n. and A. anacolum can be diagnosed unambiguously by sets of character states from nucleotide data, we considered A. inexpectatum sp. n. as a new species of the genus.
Our hypothesis is also supported by morphological data. As the case for other marine species of the genus, A. inexpectatum sp. n. can be differentiated from the freshwater species A. lisae by the morphology of the bothridia (see Remarks for A. anacolum above). Among marine species, we were unable to find any discrete morphological attribute or morphometric discontinuity that could be used to distinguish A. inexpectatum sp. n. from A. anacolum. The multivariate statistics analyses of morphometric data showed, however, that A. inexpectatum sp. n. not only differs from the marine lineages of the genus but also it is most similar to a lineage of the genus found in the eastern Pacific coast of Panama (Fig 5A and 5B). The PCA analysis (Fig 5A), for instance, showed that the cluster of specimens of A. inexpectatum sp. n. overlap more with those from the eastern Pacific congener than to those of A. anacolum. The linear discriminant analysis (LDA; Fig 5B) is congruent with the phylogenetic pattern observed in the sense that this analytical tool was able to discriminate A. inexpectatum sp. n. from other marine lineages of the genus with an error rate of 3%. Therefore, the absence of a discrete morphological attribute or any morphometric discontinuity should not pose any problem to identify A. inexpectatum sp. n. on the bases of morphological data, since discriminant function analysis would serve this purpose.
Anindobothrium carrioni sp. n. urn:lsid:zoobank.org:act:0A955A70-E88D-4EFC-98AB-13851FB6EB4B (Figs 6D, 9G and 9H and 13 and 14)  G  T  A  T  A  T  A  T  T  A  T  T  T  A   A. lisae [MZUSP 7784]  G  G  T  A  T  A  T  A  T  T  A  T  T  T  A   Anth. hobergi [MZUSP 7756]  T  G  T  A  A  T  T  G  T  T  G  C  A  A  T https://doi.org/10.1371/journal.pone.0184632.t009 Etymology: This species is named in honor of Señor Agustín Carrión, a gifted and witty fisherman who guided us to find Styracura pacifica in the Gulf of Montijo during our collecting trip to the eastern Pacific coast of Panama.
The molecular evidence for the recognition of Anindobothrium carrioni sp. n. as a segregated lineage within the genus is overwhelming. Members of this species have a set of 25 unique unambiguous synapomorphies-9 from COI and 16 from nuclear regions (Fig 3B and  3C; Tables 5, 6, 7 and 9). Sequences of Anindobothrium carrioni sp. n. are the most divergent ones among marine species of the genus. In average, sequences of COI, Calmodulin and ITSthe most variable regions included in this study-differed from A. anacolum and A. inexpectatum from Caribbean in 13.8%, 2.7% and 2.7%, respectively (Table 4). This constitutes strong molecular support to recognize Anindobothrium carrioni sp. n. as a new species.
Despite the molecular divergence and sets of diagnostic character states, all marine species of the genus have a very similar morphology, especially with regards to the bothridial architecture that share the presence of longitudinal septa and facial loculi. The bothridial morphology can be used to distinguish the marine lineages from the only species known to be restricted to potamotrygonids in freshwater systems of South America, A. lisae, as discussed earlier. However, there is no discrete morphological character or discontinuous morphometric variable that could be used to distinguish A. carrioni sp. n. from A. anacolum and A. inexpectatum sp. n. But, based on the PCA analysis A. carrioni sp. n. seems to be phenetically closer to A. inexpectatum sp. n., since the area circumscribed by the 95% confidence interval around the centroids for each species overlap to a great extent (Fig 5A). Despite the overlap observed in the PCA plot, all marine species are discriminated in the LDA (Fig 5B). Therefore, as for A. inexpectatum sp. n., A. carrioni sp. n. could only be recognized morphologically by discriminant function analysis.

Phylogeny of the Rhinebothriidea and the position of Anindobothrium
The order Rhinebothriidea was erected by Healy et al. [2] as the result of a phylogenetic analysis based on molecular data for a selected group of the polyphyletic Phyllobothriidae, then an family of Tetraphyllidea. Healy et al. [2] circumscribed the taxonomic representation of their study upon Euzet's [63,64] concept of the Rhinebothriinae Euzet, 1953, which was proposed to accommodate phyllobothriids that lacked a myzorhynchus in adult forms and which had subdivided and unarmed bothridia. Healy et al. [2] found molecular support for a number of phyllobothiids-especially members of Echeneibothriinae and Rhinebothriinae-that were also characterized by possessing stalked bothridia. Accordingly, the original concept of the order Rhinebothriidea included members of Anthocephalum Linton, 1890, Echeneibothrium van Beneden, 1850, Rhabdotobothrium Euzet, 1953, Rhinebothroides Mayes, Brooks & Thorson, 1981, Rhinebothrium Linton, 1890, Rhodobothrium Linton, 1889, Scalithrium Ball, Neifar & Euzet, 2003, Spongiobothrium Linton, 1889, and the undescribed "New genera 1-4" (sensu Healy et al. [2]). In addition, based on the putative morphological synapomorphy of the order, Healy et al. (2009) suggested that some of the genera in the subfamily Phyllobothriinae such as Anthobothrium van Beneden, 1850 and Carpobothrium Shipley & Hornell, 1906 could ultimately be found to be members of the order. They also included as potential members of the new order Anindobothrium and Pararhinebothroides Zamparo, Brooks & Barriga, 1999 as they were also described as possessing stalked bothridia.
The internal relationships and composition of the Rhinebothriidea was revisited recently in two studies. Ruhnke et al. [8] expanded the taxon sampling of Healy et al. [2] by adding putative members of the order, in addition to most species of Anthocephalum recognized to date. Based on their phylogenetic hypothesis, they recognized four families within the Rhinebothriidea: Rhinebothriidae, Echeneibothriidae, Anthocephaliidae and Escherbothriidae, from which the latter two families were newly erected. Marques and Caira [10] corroborated Ruhnke et al.'s [8] suspicion that Pararhinebothroides not only was a member of the Anthocephaliidae, but in fact a member of Anthocephalum. In both studies, however, no members of Anindobothrium were included.
Our phylogenetic analysis provided unambiguous evidence that Anindobothrium is a member of the Rhinebothriidea (Fig 1A and 1B). However, the phylogenetic position of this genus seems to be unstable, as are most internal nodes within the order. A comparison between the phylogenetic hypotheses proposed by Ruhnke et al. [8] and Marques and Caira [10] is an example of this instability (Fig 15A and 15B). Although both studies suggested that the Anthocephaliidae and the Escherbothriidae are sister taxa, they proposed different phylogenetic arrangements for the remaining taxa. Comparing the present results to previous studies, we observed that each study provided different sets of sister-group relationships for the families (see Figs 1 and 15). All studies had different taxon sampling schemes and some phylogenies were based on different optimality criteria. However, we believe that most of the discrepancies among these studies are related to nodes that have relatively low support due to the apparent limited resolution power of 18S and 28S rDNA. Hence, the inclusion of additional markers in future studies could provide a more stable resolution for the inter-relationships among families of the Rhinebothriidea.
Although sister-group relationships within the Rhinebothriidea are unstable, all familiesincluding the Anindobothriidae-are well supported by the data regardless of optimality criteria (Fig 1A and 1B). This support stems from the large sets of molecular synapomorphies for each clade, albeit not all of them are unique and unambiguous (Fig 2). In addition, all members of this monotypic family possess genital pores at the anterior 1 4 of the proglottids and have testes anterior and posterior to the cirrus sac. Therefore, Anindobothriidae can be circumscribed by both molecular and morphological characters.

Species diversity within Anindobothrium
The diversification of Anindobothrium can be correlated to all major paleogeographic and biogeographical events that shaped the biotas of the Netropical freshwater systems, the tropical eastern Pacific and the tropical western Atlantic Oceans. These events comprise the colonization of the fluvial systems of South America by marine lineages, including an ancestor potamotrygonid lineage, during the Paleogene Period-between the early Miocene and mid-Eocene (i.e., 22.5-46 Mya) [11,12,14,15]-and the isolation of the transisthmian marine fauna during the Late Pliocene-3.2-2.7 Mya [61,62,65].
The phylogeny of Anindobothrium is congruent with the history of colonization of freshwater stingrays. The current hypothesis of the origin of these freshwater stingrays suggests that the ancestor of this lineage colonized the rivers of South America after the marine incursions in the northern region of that continent during the Miocene [11,14,15,17]. As postulated initially, freshwater potamotrygonids formed a clade sister to amphi-American species of Himantura Müller & Henle, which were recently transferred to a new genus, Styracura, within the Potamotrygonidae by de Carvalho et al. [18]. According to the phylogeny of the host and its biogeographical implications, the divergence of the freshwater lineage took place prior to the diversification of Styracura spp., possibly as a result of the closure of the Isthmus of Panama (see below). The phylogeny of Anindobothrium mirrors the phylogeny of the host, suggesting an event of codivergence (sensu Page & Charleston [19]) at the split between marine and freshwater lineages of Anindobothrium (Fig 16).
It is puzzling, however, that so far, no other freshwater species of Anindobothrium have been found and A. lisae is the only representative of the genus in freshwater potamotrygonids. Other lineages of cestodes, which had the same historical fate as Anindobothrium, diversified after the colonization of the river systems of South America. There are 35 species of freshwater stingrays distributed throughout all major river basins of South America [67,68]. These hosts house at least 6 valid species of Acanthobothrium, 8 of Potamotrygonocestus Brooks & Thorson, 1976, 5 of Rhinebothroides, 7 of Rhinebothrium, and 2 of Paroncomegas Campbell, Marques & Ivanov, 1999. In addition, for all of these genera there are many species waiting formal description (Marques, unpub. data) as a result of an extensive collecting effort in the past years that led to the examination of *1,300 host specimens of freshwater potamotrygonids. Yet, A. lisae is the only species in the genus, which compared to many other cestodes found in freshwater potamotrygonids has a very restricted distributional range: upper-middle Orinoco and Rio Negro-two river basins that are known to share freshwater lineages since they are connected by the Casiquiare river [59]. The reasons behind the lack of diversification of Anindobothrium in freshwater systems, as compared to other lineages of cestodes in the same habitat or its marine congeners, still need to be elucidated. One possibility might be the life cycle requirements for members of this genus, which could differ from that of other genera found in freshwater potamotrygonids. However, very little is known about the life cycles of these cestodes.
The paleogeographic history of the tropical eastern Pacific and tropical western Atlantic Oceans seems to have imposed the pattern of diversification we observed in marine lineages of Anindobothrium and their hosts. After the marine incursions into South America in the Miocene-credited to have allowed the colonization of South American rivers by many marine lineages (see Lovejoy et al. [17] and references therein)-the region that now comprises the tropical eastern Pacific and tropical western Atlantic Oceans underwent drastic changes due to the uptlift of the Isthmus of Panama that now separates these bodies of water. The seaway that connected these two regions started narrowing and shallowing by the middle Miocene 69,70]) causing the interruption of gene flow between shallow marine animal populations in the Late Pliocene *3.2 Mya [62]. This major paleogeographic event is known to have driven global oceanic reorganization and major biotic change on land and at sea [62], as it not only allowed the exchange of fauna between North and South America (see review by Leigh et al. [65]) but also imposed independent evolutionary trajectories of marine organisms now segregated to two oceans affected by changes in current patterns, salinity, temperature, and primary productivity [61]. As a result, the uplift of the Isthmus of Panama affected the taxonomic composition of its adjacent waters [65].
The pattern of diversification of marine lineages of Anindobothrium and their hosts is congruent with this paleogeographical event. Species of Styracura have transisthmian distribution in which S. pacifica is known for the tropical eastern Pacific Ocean and S. schmardae is found along the Caribbean coast of Central America and northern coast of South America. The distribution of Styracura spp. resembles the pattern documented for many other species pairs of marine lineages with transisthmian distribution (see [61,62] and references therein) for which as early as 1908 Jordan [71] coined the term "geminate species". The split between Anindobothrium carrioni and the clade formed by the Caribbean species of this genus mirrors the divergence between their host (S. pacifica and S. schmardae) suggesting codivergence between host and parasite lineages (Fig 16). In addition, the molecular divergence between transisthmian lineages of Anindobothrium is within the range of what is documented for many other groups of Metazoa for which species pairs are believed to have originated by the uplift of the Isthmus of Panama [61].
Lessios [61] compared the K2P nucleotide sequence distances for 38 genomic regions involving 115 pairs of geminate clades of echinoids (9), crustaceans (38), fishes (42), and molluscs (26). He concluded, among other things, that 34 clades had diverged at the time of the Isthmus of Panama completion, when gene flow was interrupted between the isolated areas (i.e., 3.2-2.7 Mya [61,62,65]). For the most prevalent genomic region used in his study, Lessios [61] suggested that ranges of K2P distances for COI of 8.7-13.5% for echinoids, 4.1-8.7% for crustaceans, 3.2-5.5% for fishes, and 7.4-9.2% for molluscs were an indication that those clades diverged at the time of the Isthmus completion. Although we acknowledge that genetic divergence is not produced exclusively by vicariant events, since other factors can affect it as well (e.g., different effective population sizes, differential mutation rates, or distinct modes and intensities of selection), it is interesting to notice that the sister clade of marine lineages of Anindobothrium (i.e., A. carrioni + A. anacolum/A. inexpectatum) presented a K2P distance of 4.9-16.2%, averaging 13.8% for COI. These values may suggest that these lineages were segregated prior to the final stages of the Isthmus completion, as thought to be the case for most of the geminate pairs included in Lessios's [61] study.
The final event of diversification in marine lineages of Anindobothrium involved the split between A. anacolum and A. inexpectatum hosted by S. schmardae from the Caribbean Sea. This putative event of associate-lineage duplication (sensu Page & Charleston [19]) was surprising, albeit it is congruent with the patterns of endemism reported for the Caribbean [66,72].
Although the tropical eastern Pacific Ocean and the Caribbean Sea are sister areas with a common faunal heritage, the uplift of the Isthmus of Panama had great ecological impact on both now separated bodies of water. This is most evident in the Caribbean, which, compared to the ancestral area, became a mainly oligotrophic and rich in coral reefs albeit in some parts it retained eutrophic environments prevalent before the closure of the Isthmus of Panama [61,65,66]. The heterogeneity of the Caribbean has driven areas of endemism that have been recognized as early as 1950's [73] (see also Robertson & Cramer [66] and references therein). Robertson & Cramer [66] is the most recent and comprehensive study on patterns of distribution of shorefishes of the Caribbean. They analyzed *800,000 species site records which included 1,559 species of elasmobranchs and bony fishes from the Caribbean reported for the upper 100 m of the water column of continental and insular shelves. They found evidence to recognize three biogeographical provinces: the Northern Province-the Gulf of Mexico and southeastern USA; the Central Province-the West Indies, Bermuda and Central America; and the Southern Province-Northern South America (see Fig 16). Two of these provinces are occupied by different species of Anindobothrium.
The distribution of the sister species A. anacolum and A. inexpectatum are congruent with the biogeographical provinces recognized by Robertson & Cramer [66] (Fig 16). Anindobothrium anacolum seems to be restricted to the Southern Province, which included the entire continental shelf of northern South America from Colombia to northern Guyana [66]. This biogeographical province is quite different from the Central Province from which A. inexpectatum is reported. The continental shelf of northern South America is characterized by having a tropical and eutrophic environment due to high nutrient inputs from coastal wind-driven upwelling systems and outflows from large rivers that drain from South America. This is quite different from the Central Province characterized by eutrophic environments in which primary production occurs on the sea floor [65]. To date, only a single species of Styracura is known from the Caribbean Sea and, as some species included in Robertson & Cramer's study, S. schmardae seems to be distributed throughout the Central and Southern Provinces. However, the pattern of host association for the Caribbean species of Anindobothrium calls for a close examination of the population structure of S. schmardae, since there is a possibility that this host group is, in fact, a species complex.
The apparent recent diversification of marine lineages of Anindobothrium may respond to the cohesive morphological attributes of these species. To the best of our knowledge there is no discrete morphological attribute that could be used to identify any of the three species. The same applies for the recognition of morphometric discontinuities, which have been the prevailing criterion for species recognition in related groups of cestodes found in elasmobranchs and other vertebrates. The only morphological signal we were able to recover to diagnose these species was using discriminant function analysis (Fig 5). Linear discriminant functions have already been used successfully by others to recognize parasite lineages [74,75] and other groups of Metazoa (e.g., [76,77]). Our study suggests that the inclusion of discriminant function analysis into the toolbox of cestode systematics might be fruitful, especially in the absence of molecular data, which for Anindobothrium provided discrete nucleotide data to diagnose all the species we now recognize for the genus. We are not advocating that species should be erected on the basis of molecular or discriminant analyses alone-although some have taken this path [78,79]. The approach adopted in the present study should be an example of integrative taxonomy (sensu Padial et al. [80,81]), a concept based on integration by congruence of evidence generated from the analysis and evaluation of data from different sources (e.g., molecular, morphological, biogeographical, among others). This approach, as was undertaken in the present study, has the potential to support robust hypotheses for species.

Integrative taxonomy and cryptic species in cestodes
The morphological homogeneity of these three marine species of Anindobothrium and their distribution in two species of batoid fishes draws attention to a component on the diversity of cestodes that might have been neglected in the past, that is cryptic species-those assumed to be only recognized by the examination of molecular data. Reports of cryptic species have been published for cestodes before (see reviews in [82,83]), but to the best of our knowledge, no accounts of cryptic species exists for cestodes infecting elasmobranchs. Even so, compared to other groups of Metazoa, there is a small number of publications addressing cryptic speciation in cestodes. Pérez-Ponce de León & Nadler [82] reported 15 studies referring to cryptic species for this group from 1999 to 2009. Recently, this list was revised by Pérez-Ponce de León & Poulin [83] who found 14 studies from 1978 to June 2016 for cryptic species in cestodes. Independent of the status of A. inexpectatum as a truly cryptic species or a pseudocryptic one-those not initially recognized as phenotypically distinct due inadequate analysis of morphological data [84,85]-, this species would not have been recognized by traditional approaches applied by taxonomists in the group. The taxonomy of cestodes, in particular those lineages parasitic in elasmobranchs, has historically been based on morphological discontinuities and on the assumption that cestode fauna of a species of elasmobranch does not vary substantially across its distribution [86]. Since cryptic lineages have been shown to be common throughout the metazoan taxa (see [87][88][89][90][91], and references therein), we should expected to find them also among cestode species.
Many studies have found cryptic species by evaluating the molecular diversity of wide spread taxa composed by populations that are morphologically similar-if not indistinguishable (e.g., [79,[92][93][94][95][96][97][98]). These studies illustrate the importance of recognize cryptic speciation as part of the process that produces biodiversity since the failure to detect cryptic species can result in underestimation of biodiversity [91]. The importance of recognizing this component of our biota relies on the fact that most questions in evolutionary biology (e.g., speciation), ecology (e.g., ecosystem development), conservation biology (e.g., conservation priorities) or biogeography (e.g., diversification processes) depend, to a great extent, on the accurate recognition of the lineages that comprise that biodiversity [80]. For host-parasite systems, the circumscription of host and parasite lineages shape our understanding of host specificity, influence our efforts to control parasitic diseases and may determine our ability to provide robust hypotheses for historical association events [83,99].
As pointed out by Goldstein & DeSalle [100], the documentation of cryptic species demands creative approaches. The inclusion of scanning electron microscopy is a common practice in the taxonomy of elasmobranch cestodes to document tegumental structures (i.e., microtriches patterns) and it has been useful in taxonomic decisions (see Caira [86] and references therein). Also, the concern for selecting characters not subject to fixation artifacts and the search for those thought to display interspecific variation are credited to have improved the taxonomy of cestodes over the years [86]. Molecular data have contributed to the taxonomy of many groups of cestodes, however timidly so. A non-exhaustive survey of cestode species described in the past 10 years revealed that more that 300 species were described during this period (source: Global Cestode Database [101] and ISI Web of Knowledge; December 2016), from which 47 species were described in studies that included molecular data (e.g., [7,8,). In addition to the list provided by Pérez-Ponce de León [83] accounting for studies revealing cryptic species in cestodes we found 8 more publications [125][126][127][128][129][130][131][132] (Source: ISI Web of Knowledge, March 2017). Among all theses studies, only one described a species after recognizing it as cryptic, which turned out also to be distinct morphologically [131] (see also Nakao et al. [128]). Marques et al. [133], for instance, explicitly recognized that their study provided molecular data to circumscribe two different species of Didymobothrium Nybelin, 1922, but refrained to make any nomenclatural changes as they were unable to provide morphological diagnoses.
Molecular data has greatly accelerated the identification of cryptic species (see Padial et al. [81], and references therein), but the sole identification of cryptic species will not solve the taxonomic challenges we face today. Formally described species are not only the basis for biological classification but also the only means by which we can effectively communicate and link information to existing knowledge [134]. Be that as it may, molecular data are rarely included in formal descriptions [134,135] and there is still the prevailing notion that a morphological diagnosis is required to describe species, although no current codes for biological nomenclature specify the class of characters upon which descriptions ought to be based [136]. Thus, the trend observed in cestode systematics mirrors the practice observed in other groups.
Especially in cases of cryptic species, we should acknowledge that molecular data could be valuable as diagnostic characters in the absence of morphological discontinuities and there is no epistemological justification to exclude them. The goal of modern taxonomy is to provide sets of characters upon which we can erect species hypotheses. Systematists should be under no obligation to stick to any class of data (i.e., phenotypic and/or genotypic) in their taxonomic practice. Rather, we should embrace the virtues of "integrative taxonomy", which seems to be the most profitable path not only to incorporate all information available at the time into species description but also to provide a more robust empirical foundation for systematics [100]. Systematics, as a science, should proceed via free exploration of ideas, and not by peer pressure to give preference to any class of characters as the primary data source for species descriptions nor by restricting the methods available to investigators [137]. We predict that as we closely look at cestode fauna across biogeographical areas and apply a variety of tools available for species discovery, we will encounter closely related species morphologically very similar but yet displaying discrete sets of molecular characters, which may also be congruent with other attributes (e.g., host association, paleogeographical history, among others). Our only concern should be to provide testable hypotheses about the structure of biodiversity [138], which can be corroborated or refuted in light of new empirical data.  Venezuela. We are indebted to Dr. Angel Javier Vega and M.Sc. Leysi del Carmem from Centro Regional Universitario de Veráguas-Universidad de Panamá (Santiago de Veráguas, Veráguas, Panama) for facilitating our field trip in Panama and for the assistance during the field trip on that country. We thank Dr. Anna J. Phillips from the National Museum of Natural History (Smithsonian Institution, Washington, D.C., U.S.A.) and Dr. Gabor Racz from the Harold W. Manter Laboratory (University of Nebraska, Lincoln, Nebraska, U.S.A.) for the loan of the type material of Anindobothrium anacolum, and also Dr. Marcelo Knoff from the Coleção Helmintológica do Instituto Oswaldo Cruz (Rio de Janeiro, RJ, Brazil) for allowing us to examine the type material of A. lisae. We also would like to thank B.Sc. Beatriz V. Freire for her help in gathering the molecular data. Finally, we appreciated comments and criticisms provided by Dr. Bjoern C. Schaeffner, Florian Reyda, Kirsten Jensen and Veronica M. Bueno on earlier versions of this manuscript. This work was supported in part with funds from NSF PB&I grants DEB 0818696 and DEB 0818823; grants from the State University of New York College at Oneonta Research Foundation; and by FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo) grant # 2014/10220-0. Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.