3 | Results
We successfully amplified and sequenced 174 individuals of 67 distinct species. Phylogenetic analyses of the resulting COI sequences revealed high diversity and a number of cryptic species within all groups, but especially within Lobata, which had not been evident from previous 18S rDNA sequence analyses. COI and 18Ssequence fragments were deposited in GenBank with accession numbers (xxx-xxx) (Table S1). Taxon-optimized primer sequences are indicated in Figure 1.
Analysis of multiple isolates of Bolinopsis suggested that the genus is polyphyletic. In addition, we confirmed morphological differences of distinct genera with COI sequence fragments, such as Deiopea , Kiyohimea , and Eurhamphaea , that were unresolved with the 18S fragment, and found support for subtle morphological differences between undescribed species within the generaLampocteis and Bathocyroe .
3.1 | Primer Design and Combinations
Successful primer pairs varied by taxon (Table S1, Figure 1, and branches of Figure 2). Primer combinations amplified a range of fragment lengths (Figure 1), so we trimmed the ends of the alignments used to generate phylogenies to exclude missing data. A few sets of primers such as F259/1060R worked well for many species, and reverse primer R866 was successful for many species of cydippids (Figure 2, starred primers). However, amplification of many genera such as Euplokamis, Ocyropsis, Beroe, and Lampea required customized primers. Some primers designed for one species also worked for other closely related taxa, such as Bfor259F, which was specific to B. forskalii but also amplified most Beroids. For several species, multiple combinations of primer pairs successfully amplified the COI fragment (Figure 1, 2).
  3.2 | Phylogenetics and Species Delimitation
The COI alignment included 174 sequences representing 67 MOTUs (molecular operational taxonomic units). It was trimmed to a 765 base-pair (bp) fragment and we used a GTR+i+Γ model for phylogenetic analyses based on jModelTest results (Figure 2). Mitochondrial loci in ctenophores are generally rich in the bases adenine (A) and thymine (T) (Pett et al. 2011). Average base composition for all ctenophores sequenced were ~ 50% A, 21% T, 16% C, and 13% G, and the vast majority of changes were at the third codon position.
In order to examine saturation between and within taxa, we plotted the proportion of transitions and transversions versus GTR distance among all sequences (Figure 3a). For ctenophores sequenced, we found that the proportion of transitions was greater than that of transversions, which is in contrast to the general rule (for mammal mitochondrial DNA), where transversions are often two-fold more common than transitions (Xia et al. 1996). Transitions were saturated for species that were more than ~25% distinct and transversions were saturated for species that were more than ~30% distinct. We saw the same pattern for the other mitochondrial loci that we could align without the presence of stop codons including; COII, COIII, CytB, ND4 and ND5 (Figure S1). In these genes as well, transversions were more common than transitions and there were high levels of saturation.
The high levels of saturation between distantly related species (Figure 3a) were reflected in the poor resolution of deeper level relationships amongst taxa in the Bayesian and maximum likelihood phylogeny trees. Phylogenetic relationships for COI among all the ctenophores were therefore portrayed in an unrooted tree without support values (Figure 2). In order to illustrate levels of diversity within the phylum, we also plotted within and between pairwise GTR distances for all individuals sequenced (Figure 3b). In comparison to other phyla, distances between MOTUs were relatively large, ranging between from 0–43%, and showed four peaks. The first peak represented within-species variability, around 0-4% GTR, although we had few within-species samples for this estimate. The second small peak centered at 10% GTR distance, was represented by closely-related species complexes. The two largest peaks were represented by the majority of our sequencing efforts and occurred at ~17% (within group; pink, blue, or green) and 30% GTR distance among different orders (gray) of ctenophores (Figure 3b).
The 18S alignment of the Lobata included 82 sequences for 29 MOTUs. This was trimmed to 1780 bp and had a TrN+i+Γ model determined with jModelTest for phylogenetic analyses. The 18S fragment was very conserved and included only 47 parsimony-informative sites within all sequence data for the Lobata and phylogenetic analyses showed little differentiation within and even between many genera (Fig 4a). Posterior probabilities and bootstrap values were relatively poor and unresolved for most relationships among taxa due to low resolution among species and genera (Figure 4a). Distinguishing taxa using the COIfragment limited to the Lobata was more successful since the marker is more variable than 18S. The COI fragment had 199 parsimony informative sites between 27 MOTUs (Figure 4b). Although deeper level relationships amongst more distantly related taxa were still not well supported and some saturation was evident (Figure 3a), analyses revealed high support for several cryptic species complexes and new MOTUs. Conversely, other species were revealed to be truly cosmopolitan with worldwide or at least ocean-basin wide distributions (Figure 4b).