Morphology and Species Composition of Southern Adriatic Sea Leptocephali Evaluated Using DNA Barcoding

Leptocephali are the characteristic larvae of the superorder Elopomorpha that are difficult to identify at the species level. In this study, we used DNA barcoding (i.e. short genetic sequences of DNA used as unique species tags) coupled with classical taxonomic methods to identify leptocephali in the southern Adriatic Sea. This information will provide an assessment of the biodiversity of the eel larvae in this region. A total of 2,785 leptocephali were collected, and using external morphology were assigned to seven morphotypes: Ariosoma balearicum, Conger conger, Gnathophis mystax, Facciolella sp., Nettastoma melanurum, Dalophis imberbis and Chlopsis bicolor. Collectively, these seven morphotypes are considered to be a good proxy for the Anguilliformes community (the main order of the Elopomorpha) in the southern Adriatic Sea (to date, seven families and sixteen species have been recorded in this region). Interestingly, the higher number of G. mystax larvae collected suggests an increased abundance of this genus. To validate the morphological identifications, we sequenced 61 leptocephali (at a 655 bp fragment from the cytochrome oxidase subunit 1 mitochondrial region) and developed barcode vouchers for the seven morphotypes. Using genetic information from reference databases, we validated three of these morphotypes. Where reference sequences were unavailable, we generated barcodes for both adult and juvenile forms to provide additional genetic information. Using this integrated approach allowed us to characterize a new species of Facciolella in the Adriatic Sea for the first time. Moreover, we also revealed a lack of differentiation, at the species level, between G. mistax and G. bathytopos, a western Atlantic Ocean species. Our morphological and barcode data have been published in the Barcoding of the Adriatic Leptocephali database. This work represents the first contribution to a wider project that aims to create a barcode database to support the assessment of leptocephali diversity in the Mediterranean Sea.


Introduction
To overcome these difficulties and improve overall species identification, DNA-based methods have become routinely used to support and/or integrate with classical taxonomic approaches (e.g., [24,25,26]). To date, several molecular genetic methods have been indeed successfully used to identify eels at the different life cycle stages [7,[27][28][29][30]. DNA barcoding, a method that uses a standardized, universal DNA region as a unique species tag has been particularly effective. This technique allows for rapid and accurate species identification [31,32]. It has also proven efficient in solving taxonomic ambiguities, revealing cryptic diversities [33,34] and identifying early life stages [35][36][37][38]. Despite its benefits, the use of DNA barcoding for identifying leptocephali is still in its infancy [30,39]. However, its application could address several outstanding issues, providing a tool that can: i) match leptocephali with their respective adult forms (where adults sequences are available), ii) evaluate whether similar leptocephali morphotypes represent single molecular operational taxonomic units [40] and iii) offer an alternative identification approach when specimens are severely damaged, rendering morphological identification impossible.
In this study, we aimed to improve overall identification efficiency and develop the first records of anguilliform larval assemblage in the southern Adriatic Sea by using an integrated identification approach that combined traditional taxonomic methods with the use of species specific DNA barcode vouchers.

Sample collection
A total of 2,785 leptocephali, were collected from 35 oblique hauls during Deep Water Cruises carried out in the southern Adriatic Sea in August 2010, 2011 and 2012 by the vessel m/n Andrea (Fig 1). These hauls were performed at each station using a midwater trawl with an approximate mouth opening of 100 m 2 and a cod-end mesh size of 20 mm. The nets were lowered to depths between 0 and 900 m, with a different maximum depth reached at each station. Specimens were fixed on board using 70% ethanol, with a subsample fixed in a 95% solution for the genetic analyses. As leptocephali are classified as zooplankton, neither special permits nor ethics approval were required for their collection.

Morphological analyses
Total length (to the nearest 0.1 mm) was recorded for all leptocephali. To reduce the shrinkage caused by the fixation and preservation method, and to better observe the morphological features, the larvae were restored with filtered seawater prior to the measurements being taken. Larval morphology was examined using a Zeiss Stemi 2000 C stereomicroscope and compared with available dichotomous keys and descriptions for leptocephali, primarily Grassi [19] and Smith [41]. The key morphological characteristics used to identify eel larvae are body shape, the organization of the hypurals in the caudal skeleton, the length and shape of the gut (i.e., simple or with swelling or thickness), pigmentation patterns and the total number of myomeres (TMs). In specimens that were hard to identify, additional morphometric and meristic elements (e.g., head length, predorsal and preanal length, predorsal and preanal myomeres) were also considered. Specimen images were captured using a digital camera (Canon Power-ShotG5) mounted on the Zeiss Stemi 2000 C stereomicroscope or using a photo scanner (Epson Perfection V600 Photo). The pictures were calibrated (with an accuracy of 0.005 mm) and handled using ImageJ software (National Institute of Health, USA, version 1.49J R). For each leptocephalus species identified, we wrote a short descriptions that focused on the key diagnostic morphological features. These descriptions were supplemented with pictures of representative individuals.

Molecular genetic analysis
For each morphotype, one to 32 representative individuals were selected for molecular genetic analyses. This selection included also all the individuals that had been difficult to identify using the morphological characteristics. Approximately 25 mg of muscle tissue was dissected from each sample and kept in a 95% ethanol solution at -20˚C until it was used. After a 30 min incubation period at 37˚C to remove residual ethanol, the total genome was extracted using a NucleoSpin1 Tissue Kit (MACHEREY-NAGEL GmbH & Co), following the manufacturer's instructions. A fragment of approximately 655 bp from the 5 0 region of the cytochrome oxidase subunit I (COI) gene was amplified using various combinations of the fish-specific primers described in [42]. Polymerase chain reactions (PCRs) were carried out in a total volume of 25 μL. This contained 3 μL of leptocephali DNA as a template, 0.2 μM of each primer, 1.25 U of TaKaRa Ex Taq1 DNA Polymerase (Takara), 1X TaKaRa Ex Taq1 Buffer and 200 μM of each deoxynucleotide solution. The thermal cycling conditions were: 1) an initial denaturation at 95˚C for 2 min, then 2) 35 cycles of amplification (30 s at 94˚C, 30s at 54˚C and 1 min at 72˚C), followed by 3) a final extension at 72˚C for 10 min. Negative controls were included in each amplification reaction. All PCR reactions were performed in a TProfessional basic thermocycler (Biometra) and PCR products of the expected length size (i.e., 655 bp) were checked using 1.5% agarose gel electrophoresis after staining with GelRed™ (BIOTIUM). Products were purified with an ExoSAP-IT1 kit. The products were then labeled using the BigDye1 Terminator v.1.1 Cycle Sequencing Kit (Applied Biosystems, Inc.) and sequenced bidirectionally using an ABI PRISM1 3100-Avant Genetic Analyzer (Applied Biosystems). A total of 61 individual sequences, belonging to six morphotypes, were generated (Table B in S1 File). In addition, the COI sequences of a juvenile D. imberbis and an adult F. oxyrhyncha were obtained to further validate the DNA barcodes. This step was undertaken because no reference sequences were available from the public databases for these species (Table B in S1 File).
All the 63 sequences produced (61 leptocephali, one juvenile and one adult) were analysed by using, BioEdit Sequence alignment Editor [43] and MEGA 6 software [44] and were matched against the BOLD (Barcode Of Life Data system, Version 3 http://www.barcodinglife. org) and GenBank (http://www.ncbi.nlm.nih.gov/genbank/) databases to confirm their morphological identifications. We assigned each specimen to their taxonomic rank, according to their similarity values (SV): species (SV ! 98%), genus (92% ! SV < 98%) and family (! 85% SV < 92%). These newly generated sequences were merged and aligned with the published sequences we obtained from the databases. These results were used to build three neighbourjoining (NJ) trees [45], one for each family identified, using the uncorrected p-distances and 1,000 bootstrap replications generated by MEGA 6 [44]. This provided a graphical representation of the patterns of divergence between the species we analysed. A COI sequence of Albula vulpes (Elopomorpha: Albuliformes) was used as an outgroup in each NJ tree. Where species assignment was uncertain, we also computed the between-and within-genetic p-distances using MEGA 6. Finally, we performed character based species classification analyses (BLOG software; default settings; [46]) to corroborate our SV and NJ tree results.
The representative sequences and electropherograms of all the species we identified in this study, as well as information on the primers we used, were uploaded to the new barcoding database 'Barcoding of the Adriatic Leptocephali' (BAL) which can be accessed through the Barcode of Life Data System (BOLD, http://www.barcodinglife.org).

Results and Discussion
Using available dichotomous keys, a total of 2,785 specimens were identified and assigned to four families and seven species of Anguilliformes (Table 1). The majority of individuals were in the larval stage (N = 2,780), with a small number in the metamorphosing stage (N = 5). The results presented below focus on individuals in the larval stage. Seven eel families are known to occur in the Adriatic Sea [3]; consequently, we felt that our study specimens were representative of the area's eel community. Furthermore, each family was well represented, with all species belonging to the Congridae, Nettastomatidae and Chlopsidae families considered (Table A in S1 File). Gnathophis spp. were the most commonly identified genus, a finding that likely reflects their higher abundance in the Adriatic Sea. Note, however, that we did not have any information about their localised abundance at the sampling sites. In total, we barcoded 61 leptocephali that we also identified using morphological taxonomy. This step provided ex-novo species-specific specimen vouchers that we uploaded to the new BAL barcoding database. These sequences were, on average, 603 bp in length (Table B in S1 File). None of the amplified sequences showed insertions or deletions or stop codons, indicating that they were functional mitochondrial COI products. Meanwhile, the similarity analysis allowed us to validate, to the species level, three morphotypes (A. balearicum, C. conger and N. melanurum) with SV ! 98% (Table 1). For the remaining morphotypes, the limited number of DNA barcodes from voucher specimens only allowed us to confirm identity to the family level (i.e., Chlopsidae for Chlopsis bicolor; SV = 85.9% and Ophichthidae for D. imberbis; SV = 87%, Table 1) or the genus level (i.e., Facciolella sp.; SV = 93%). To address these uncertainties, we compared our leptocephali barcodes with homologous sequences obtained from an adult F. oxyrhyncha and a juvenile D. imberbis (we were unable to obtain either for C. bicolor). This enabled us to validate D. imberbis to the species-level (SV ! 99%; Table 1), although we could not verify the Facciolella sp. due to the low SVs returned against F. oxyrhyncha (SV = 93%; Table 1). With G. mistax, both the similarity search and BLOG analyses matched our specimens to the barcode voucher for Gnathophis bathytopos (SV = 99%; Table 1), revealing a taxonomic incongruence. As the N. melanurum sequences had not been released on the BOLD database at the point of analysis, the BLOG [46] software was only capable of confirming the identification of A. balearicum and C. conger (Table 1). Therefore, these specimens were not included in any further analysis. The NJ trees, computed using the p-distances instead of the Kimura-2 parameters [47], as suggested by Collins and Cruickshank [48], provided a useful description of the data (S1-S3 Figs).
The following section presents a brief overview of the morphological and genetic results obtained in this study (summarized in Table 1). Table 1. Summary of the morphological and molecular genetic identification of the southern Adriatic leptocephali assemblage. The morphometric and meristic counts measurements were taken from a subsample of each species: total length (TL), total number of myomeres (TM) and not assigned (NA). The leptocephali of A. balearicum (Fig 2) are one of the most collected and studied anguilliform larvae in the Atlantic and Pacific Oceans. Consequently, several morphological characterizations are available (e.g., [16,[49][50][51][52]). However, A. balearicum is rarely documented in the Mediterranean Sea and the only morphological descriptions from this region are those of Ophisoma balearicum larvae (later A. balearicum). These were collected from the Straits of Messina and reported by Grassi [19]. Recently, Bojanić et al. [53] recognized some leptocephali accidentally caught in the middle of the Adriatic Sea as A. balearicum, based on their general morphology, pigmentation and morphometric characteristics. Among our specimens, A. balearicum was clearly recognizable (Fig 2; Table 1 and S1 File) as their morphology matched the descriptions (e.g., [16,19,50,51  performed using the Bold System also appropriately identified these sequences as A. balearicum (SV>98%). The BLOG analysis confirmed these results. Finally, in the Congridae family NJ tree, our A. balearicum sequences were all grouped together with the published sequences (from the public database) in a unique cluster that was clearly differentiated from Ariosoma meeki (S1 Fig). Conger conger (Linnaeus,1758; Congridae)

Species
The C. conger (Fig 3) leptocephali we examined displayed the typical taxonomic larval characteristics, as identified by Grassi ([19], S1 File). Our TM range (TM = 148-153; Table 1) aligns with the ranges reported by Grassi ([19], TM = 148-155) and Aboussan ( [56], TM = 148-153) for the larval stage of this species in the Mediterranean Sea. In contrast, the values previously identified in the Atlantic Ocean are higher (TM = 154-163; [57][58][59][60][61]). This discrepancy may be partially related to faulty counting techniques [58,62]. It is also possible that a distinction between high-and low-count forms, similar to that seen in A. balearicum [63], may exist. However, TM counts are not a compelling method to distinguish between different Conger species, i.e., the European C. conger and the American C. triporiceps share many morphological features and their TM ranges overlap. In this situation, lateral pigmentation can assume a diagnostic value since it seems to be absent in C. triporiceps leptocephali [51,59], whereas a series of stellate chromatophores are present along the lateral line of C. conger [19]. However, this characteristic requires careful interpretation as pigmentation patterns can undergo ontogenetic changes or may not be visible in poorly preserved specimens. Thus, genetic investigations have been suggested [58] and applied [29] as useful tools to support the unequivocal, systematic identification of this larvae. The results of our barcode analyses validated our morphological identification of the C. conger specimens, with all 11 barcodes achieving a SV ! 99%. The BLOG analysis corroborated these results and in the NJ tree, our sequences were clustered together with the public sequences in a clearly separate group, away from other Conger species (S1 Fig). Gnathophis mystax (Delaroche, 1809; Congridae) The larval morphology of G. mystax (Fig 4) is similar to C. conger; however, several peculiar characteristics allow it to be easily differentiated (Table 1, S1 File). In our specimens, we found that the general morphology was comparable to Grassi's [19] description of Congromuraena mystax (later renamed G. mystax). One of its most distinctive morphological features is the shape of its last hypural, which shows a pronounced dorsal hump. This characteristic can be used to distinguish this species from similar Conger larvae. In addition, Gnathophis leptocephali have longer, more acute snouts [41] and a dense, double series of punctuate melanophores along the top, as compared to the Conger larvae. In contrast, Gnathophis spp. show highly similar morphologies [51,64] and their TM counts cannot be considered as a differentiating characteristic. The TM range reported by Grassi [19] for G. mystax leptocephali in the Mediterranean Sea is 133-139, but given the ranges reported by other authors (e.g., TM = 127-135 [57] and TM = 132-147 [54]), we think it should be wider (i.e., TM = 127-147). Our TM range (TM = 130-136) fell within this wider range. This widened range also overlaps with the TM ranges observed for G. bathytopos (TM = 126-141; [64]) and G. capensis (TM = 132-140; [65]) in the Atlantic Ocean. Furthermore, these species exhibit almost identical morphologies to G. mystax [51,65]. In view of these similarities, our molecular genetic analyses could provide an efficient tool to differentiate between these three species. Unfortunately, we discovered that there were no reference sequences available for any of the 32 G. mystax barcode sequences we obtained. Moreover, the blast match unexpectedly assigned these barcodes as G. bathytopos (SV ! 98%). The BLOG analysis results supported this preliminary result, also classifying our sequences as G. bathytopos, while the NJ tree clustered G. mystax with G. bathytopos. However, the NJ tree did differentiate this mixed group from all the other Gnathophis clusters (S1 Fig). The genetic distance between the two species was the same magnitude as the within species distance (p-distances: 0.009 vs 0.004; Table C in S1 File). While our results all suggest that our specimens were G. bathytopos, our knowledge of eel distributions suggested they were wrong. G. mystax is found in both Mediterranean Sea and the north-eastern (southern Portugal to Morocco; [66]) and western Atlantic Ocean [67] regions, whereas G. bathytopos is restricted to the western Atlantic Ocean. However, these conflicting facts lead us to draw an intriguing hypothesis: while the two species share every morphological feature, their numbers of myomeres are far from conclusive and the barcoding analyses essentially only detect one clade. Therefore, G. mystax and G. bathytopos could possibly be the same species, described in two distinct geographical regions and the morphological differences (detected mainly in the vertebrae counts between the adult forms of these two species; [68]) could merely represent a phenotypic polymorphism. Another explanation may be a lack of resolution in the barcode sequences at the species level. These hypotheses need to be considered further using nuclear sequences and further investigations on the adult forms. Overall, we found that the barcode analysis is an efficient method for differentiating G. mystax from G. capensis and G. nystromi (S1 Fig and Table C in S1 File).

Chlopsis bicolor (Rafinesque, 1810; Chlopsidae)
We identified three individuals as Chlopsis bicolor. These specimens clearly displayed the morphological characteristics of the larval form of this species, as described in the literature ( [19,69]; Fig 5; S1 File). The pre-metamorphic stage was also confirmed by the presence of pectoral fins. In C. bicolor leptocephali, these fins gradually shrink and disappear during the metamorphosis from larva to juvenile [70]. The COI sequence obtained from this leptocephalus, blasted against the BOLD database, showed the highest similarity percentage match with C. bicollaris (SV = 85%). Given this low value, we were only able to identify this specimen to the family level (Chlopsidae). As reference sequences for C. bicollaris were not publicly available and we hadn't collected any adults or juveniles to analyze, we were unable to progress this investigation any further.

Facciolella sp. (Whitley, 1938; Nettastomatidae)
Approximately 30 specimens in our collection exhibited the general morphological traits reported for the Facciolella larvae (Table 1; Fig 6; S1 File). For these specimens, we identified an average TM value of 244 (TM range = 240-250; Table 1). Globally, this genus is represented by six valid species, with F. oxyrhyncha the only species reported in the Mediterranean Sea [71]. It was first described from its larval stage as Leptocephalus oxyrhynchus by Bellotti [72] but one of the most extensive descriptions was given by Grassi [19]. This description was based on specimens that were collected in the Straits of Messina and erroneously described as Saurenchelys cancrivora. A further specimen has been recorded by Stramigioli et al. [73] in the southern Adriatic Sea. However, in spite of their similarities with F. oxyrhyncha [19], our putative Facciolella sp. specimens actually returned higher morphological similarity with Saurenchelys halimyon, another species described by van Utrecht [74]. However, there were some morpho-meristic incongruences with this species, namely its higher TM count (273). Van Utrecht [74] noted sufficient differences between his specimens and the closely related S. cancrivora (now F. oxyrhyncha), as described in Grassi [19], that they could be considered a new species of nettastomid eel. However, S. halimyon has never been formally accepted. In this regard, molecular genetic data could provide useful information. Indeed, the eight barcode sequences we analyzed only showed SVs of 92-93% with F. oxyrhyncha, F. gilberti and the COI sequence we obtained from the adult F. oxyrhyncha (EMBL accession number LT158010, http://www.ebi.ac.uk/). The BLOG analysis [46] was unable to correctly assign these specimens, while the NJ tree confirmed the similarity search results, clustering all our samples together in a well-defined molecular operational taxonomic unit. It assigned our adult F. oxyrhyncha specimen to the F. oxyrhyncha cluster (S2 Fig). The pairwise genetic distances between Facciolella sp. and F. oxyrhyncha and F. gilberti (0.070-0.059; Table C in S1 File) corroborated this result; they were considerably higher than the mean within-species distance of Facciolella sp. (0.000; Table C in S1 File). While these analyses need further validation, they do suggest the presence of a previously undescribed species in the southern Adriatic and Mediterranean Seas.
Nettastoma melanurum (Rafinesque, 1810; Nettastomatidae) The morphological characteristics of these leptocephali were clearly visible, enabling us to easily identify them (Fig 7; Table 1, S1 File). However, performing the meristic counts was difficult, especially in the caudal region where the myomeres were very close to each other and hard to distinguish. Nevertheless, the TM estimate (TM:~200, Table 1) and the results of our broader morphological assessment corresponded to those described in the literature [16,19]. The results of the barcode analysis also confirmed this morphological assessment. Finally, our BOLD analysis matched our barcode sequence to N. melanurum voucher sequences very well (SV = 99%). However, the BLOG analysis was unable to rank our query sequence, as these voucher sequences were not publicly accessible at the time of our analysis and hence, we were unable to use them. As a consequence the NJ analysis clustered our query sequence with the congeneric species Nettastoma parviceps (S2 Fig). Dalophis imberbis (Delaroche, 1809; Ophichthidae) The armless snake eel D. imberbis belongs to the family Ophichthidae (subfamily Ophichthinae) and has been recorded in both the eastern Atlantic Ocean and Mediterranean Sea [71]. Blache [16] described several ophichthid larvae from the Gulf of Guinea, including some Dalophis sp. However, none of these specimens showed all the morphological characteristics of D. imberbis. Currently, the most detailed description of these leptocephali and their developmental stages at the species level can be found in [19]. Our specimens displayed complete morphological agreement with those described by Grassi [19] for Sphagebranchus imberbis (later D. imberbis, Fig 8; Table 1 and S1 File). Although ophichthid larvae are characterised by typical gut swellings and/or thickenings [41,56,64], this characteristic is not described for D. imberbis larvae [19]. However, gut swellings tend to disappear in the Dalophis larvae [16] during their development and were not visible in our specimens. The only thickening we noted in our specimens was at the gallbladder (Fig 8).
The similarity search analysis (performed on 5 barcode sequences) returned the largest SVs (87%) with representative species of Ophichthidae (Ophichthus zophochir) and Congridae (Uroconger lepturus, Oxyconger leptognathus). Voucher sequences for D. imberbis or other congeneric species were absent from the public databases. Accordingly, the BLOG analysis was unable to classify our query sequences. However, they almost completely matched (SV > 99%) a sequence we obtained from a juvenile D. imberbis (EMBL accession number LT158011, http://www.ebi.ac.uk/) which allowed us to confirm our initial identification. The NJ tree grouped all the D. imberbis sequences into a well-defined cluster within Ophictidae (S3 Fig). Thus, the results of the barcode sequence analyses validated our morphological identification. In addition, our D. imberbis juvenile specimen represents the first voucher sequence to be deposited in a public database.

Conclusions
In taxonomy, identifying species using morphological characteristics is the classical approach. While they will continue to play an essential role in both identification and in developing a thorough understanding of ontogenetic changes, several issues do hamper their efficacy. The identification of anguilliform larvae, or leptocephali, is such an example where these issues can become significant. As leptocephali differ significantly from their adult forms, identifying them can be challenging, especially when considering multiple species from multiple regions. Morphological guides are available (e.g., [16,17]). However, for the larvae living in the Mediterranean, these works provide sufficient information to identify specimens to their family or genus levels, while their use in making species level identifications can be very limited. In this respect the most comprehensive descriptions are set out in Grassi [19]. To overcome morphological identification limits, molecular genetic tools, used in combination with more traditional taxonomic methods, can deliver more reliable results [75,76]. Of these genetic tools, DNA barcoding is increasingly being used to identify larval fish species [35,36,77,78], although it has rarely been used in the identification of leptocephali [30,38]. While the current lack of DNA barcodes for adult eels in public databases could limit the use of this technique for these species, nevertheless, molecular genetic data are crucial to unravel possible misidentifications (e.g., our Gnathophis species and [38]) or help distinguish well defined molecular operational taxonomic units (e.g., our Facciolella sp. specimens). It may also aide morphotype evaluation; for example, in determining whether previously identified morphotypes (using morphological tools) actually represent true species or perhaps hide cryptic species. This study demonstrates the suitability of DNA barcoding for the identification of leptocephali species across several anguilliform families and genera in the Mediterranean Sea. Future studies should seek to address the ambiguities we found in this study and continue the development of the public databases, through the contribution of more voucher specimens.
In this work, we provide the first assessment of eel diversity and abundance in the southern Adriatic Sea. We also described and photographed the morphological characteristics (validated by DNA barcoding) of G. mystax, Facciolella sp. and D. imberbis species. This contribution is significant as previously the only descriptions of these species appeared in the Italian monograph of Grassi [19], a reference which is only accessible to a restricted number of Italian research centres.