MOTU Analysis

e relationship between numbers of MOTUs and the sequence divergence cut-off used to define them is shown for both investigated taxa in Figures 7.5 and 7.6. Both markers show a stable MOTU count over a large range of cut-off values suggesting the presence of a barcoding gap. e phyllosoma sequences were assigned to 3 MOTUs (1-9% divergence cut-off). Leptocephalus sequences were assigned to 5 MOTUs (1-2% divergence cut-off).

7.3.5 Community Composition

An original aim of this study was to assess the composition of larval assemblages across the survey area. is was not possible due to the limited number of identifiable specimens.

Figure 7.6:COI MOTU results. A stable region is obtained for 5 MOTUs, corresponding to 1-2% sequence divergence.

7.4 Discussion

7.4.1 Improving PCR success

e study highlights the difficulties of DNA barcoding for larval specimens. In particular, PCR amplification success was low despite comprehensive troubleshooting. Primers were selected according to their high specificity to target species, and PCR protocols proved to be very effective when tested on adult muscle tissue. However, unlike adult tissue, phyllo- soma and leptocephalus tissue is weakly developed, transparent and thin, resulting in low DNA yields. So low, in fact that the DNA concentration in many samples was near the de- tection limit of the NanoDrop spectrometer. More starting material, longer lysis times and lower elution volumes were chosen in an attempt to concentrate DNA during extraction, but without success. Concentrating DNA extracts furthermore has the drawback of con- centrating PCR inhibitors, such as mucopolysaccharides which make up a large proportion of leptocephalus tissue (Pfeiler, 1991).

e long storage of the leptocephali under semi-dry conditions and the resulting tissue degradation may have had impacts on DNA quality, although extracted concentrations of DNA were too low to assess fragment length distributions or similar parameters.

Low PCR success rates for marine larvae are not confined to this study. In an effort to identify over 2000 marine invertebrate larvae using COI, 16S or 18S rRNA markers, Heimeier et al. (2010) reported amplification rates of ca. 30% and similar DNA contami- nation issues.

7.4.2 Specimen identification

Morphological analysis successfully distinguished Palinurids from Scyllarids for 24 out of 25 samples. Molecular analysis proved to be useful at identifying phyllosomata further, as out of 11 sequences, 10 led to a genus level identification.

e 9 phyllosomata most closely matched toJ. lalandiiin GenBank (99% sequence sim-

ilarity) could not be identified beyondJasusfrom the 16S phylogeny. It is possible to spec-

ulate on their identity however. A total of 8 phyllosomata clustered with J. lalandiiand

J. edwardsii. ere are 7 known Jasusspecies found globally (Holthuis, 1991). Jasus la- landiiis found in Indian Ocean, and is typically located off the coast of Southern Africa

(Groeneveld et al., 2007). Jasus edwardsiiis found throughout coastal waters of southern Australia and New Zealand (Booth and Ovenden, 2000). It is therefore likely that these 8 phyllosomata areJ. lalandii, which have been swept off the shelf region by the Agulhas re-

turn current. e remainingJasusphyllosoma is placed on a well-supported sister branch

alongsideJ. tristanii, even though the closest GenBank match isJ. lalandii. Jasus tristanii

was thought to be endemic to the islands of the Tristan da Cunha group in the southern Atlantic, but a recent molecular analysis suggests that it is in fact conspecific toJ. paulensis, a species found around the St. Paul and Amsterdam islands in the Southern Indian Ocean and on seamounts to the northeast of these islands (Groeneveld et al., 2012). e identity of this larva is therefore likely to beJ. paulensis.

e phyllosoma identified asPanulirusis perhaps aP. homarussubspecies. GenBank

lists a 97% match and the phylogeny clearly places it in a well-supported sister branch to

P. h. homarus, which suggests that it is different fromP. h. homarus, based on . ere are

3 describedP. homarus subspecies, of which 2 are found in the Indian Ocean (Holthuis,

1991), however 16S sequences are only available forP. h. homarus. A comparison withP.

h. rubellus, which is found off the coasts of Madagascar and Southern Africa (Holthuis, 1991), may provide more clues as to the identity of this phyllosoma.

Little can be said about the phyllosoma identified as a scyllarid. e closest GenBank

match came at only 91% sequence similarity asScyllarus pygmaeus, a species found in the

Mediterranean and South Atlantic (Holthuis, 1991). ere are far fewer published 16S sequences for this family than for palinurids, which is likely explained by the fact that many more palinurid species support large fisheries worldwide, and hence have been more intensely studied (Spanier and Lavalli, 2006).

e 16S gene was originally chosen because of the availability of Achelata 16S sequences in GenBank, and because 16S was much more readily amplified than COI. It is clear, how- ever, that 16S does not adequately resolve the relationships betweenJasusspecies.

Leptocephali

Few leptocephali could be morphologically identified to family level. Molecular analyses proved to be of limited use, as out of 17 COI sequences, just 2 were identified as matching to

Gorgasia(Congridae), and one assigned to species level asSerrivomer beanii(Serrivomeri- dae). e 3 leptocephali assigned to a genus or species in BOLD were identified using the species level identification database and hence are likely to be correctly identified, bear-

ing in mind the limitations of barcoding discussed previously. Gorgasia, or garden eels, is a family of reef-associated eels which counts 19 known species (Michael, 1998; Fricke, 1999), of which several are found in the Indian Ocean, although their distribution across

the oceans is poorly studied. S. beanii is a deep-sea species found in the Southwest In-

dian Ocean, with locations around Cape and Natal off the coast of South Africa, and near Reunion (Smith and Castle, 1986; Fricke, 1999)

Near matches of ca. 89% to 97% sequence similarity were obtained in the all barcode records database for the remaining 14 sequences, and they all listed species belonging to the Congridae family. ese results have to be interpreted with caution. One might con- clude that although a species level match cannot be made, the sequences displaying higher similarity may belong to the Congridae family. Only 27% of the 800+ known anguilliform species have a COI barcode (www.fishbol.org ; accessed August 2013), with a majority of BOLD barcodes covering species from the Congridae family. erefore, a specimen with a near match to a species from the Congridae family may actually belong to an entirely different family that is not yet represented in the database.

With the exception ofS. beanii, which has a COI barcode available in both GenBank

and BOLD, phylogenetic analyses were constrained by the fact that none of the barcodes for species found to be near matches in BOLD are in the public domain, and hence could not be included in the analysis. Attempts were made to find species from the same genus or family, but once again this was not always possible. Sister clades A and B point to the presence of two unidentified anguilliform species among the collected larvae. Morpholog- ical analysis suggests that some species within both clade A and B may be Congridae, but these identifications were not assigned with a high degree of confidence.

Sister clades C and D cluster withA. shiroanago, a species from the Northwest Pacific

(Masuda and Muzik, 1984) and the onlyAriosomaspecies with a COI barcode in GenBank

(but not in BOLD). However, all members of clade C were identified in BOLD asGorgasia

sp., for which there was no publicly available COI sequence. ere are 26Ariosomaspecies, with at least 3 found in the western Indian Ocean, but little is known about their biogeog- raphy and dispersal pathways across the study area (Castle, 1968). ese results suggest

that clades C and D belong to the Congridae family, with clade C representing aGorgasia

species and clade D a closely related second species, possibly anAriosomaspecies. More-

over, this tree and other molecular phylogenies (Inoue et al., 2010) recover the Congridae family as polyphyletic, which does not lend a high degree of confidence in the identification of clade D (Meyer and Paulay, 2005).

Overall, the COI gene seems reasonably adequate for delimiting species within the tree, as illustrated by the Anguillidae. Currently, however, far more sequences would be needed for species level matches. A thoroughly sampled taxon is particularly important when families are recovered as polyphyletic, as species are likely to be misidentified (Meyer and Paulay, 2005). In addition, the current morphological categorisation of Anguilliforms is disputed and in need of a review (Inoue et al., 2010; López et al., 2007).

7.4.3 MOTU concordance

Table 7.7:MOTU concordance

Taxon Identification Results MOTU Results

Phyllosomata 3 — 4 clades 3 MOTUs

Leptocephali 5 — 6 clades 5 MOTUs

e number of MOTUs recovered were three and five for phyllosomata and leptocephali, respectively. is corresponds well to the results of the phylogenetic analyses. e three

crustacean MOTU’s correspond to theJasuscomplex, thePalinurussp. specimen and the

scyllarid specimen. For the eels the five MOTUs correspond to clades A-D in the phylogeny

and theS. beaniispecimen.

7.4.4 Larval distributions across the survey area