2.1 Specimen collection and DNA sequencing

A total of 20 specimens of C. ‘sedna’ were collected by hand on SCUBA at depths of 5–15 m at multiple sites in the Caribbean Sea and eastern Pacific Ocean, including near the type locality in the Sea of Cortez (Figure 1, Table 1). Two additional specimens were provided by Florida Natural History Museum and Museum of Comparative Zoology and sequence data for three individuals were sourced from Genbank (Table 1). All specimens were preserved in 96% ethanol in preparation for molecular analyses. Genomic DNA was extracted from a sample of the foot tissue using a DNeasy blood & tissue kit (Qiagen) following manufacturer's guidelines. PCR amplicons were purified using either Exo-SapIT or gel purification. Two mitochondrial genes, cytochrome c oxidase subunit I (COI) and 16S ribosomal RNA (16S), and one nuclear gene, adenine nucleotide translocase (ANT), were targeted for sequencing. For PCR and sequencing, Folmer et al. (1994) primers were used for COI, Palumbi et al. (1991) primers were used for 16S, and Audzijonyte and Vrijenhoek (2010) primers were used for ANT. Amplicons were bidirectionally sequenced on an Applied Biosystems 3730 capillary sequencer. Forward and reverse sequences were edited and assembled in Sequencher (Gene Codes Corporation) and Geneious Prime 2021 (http://www.geneious.com/) and aligned with MAFFT (Katoh et al., 2002). A second alignment was trimmed to equal length in Geneious Prime 2021 for haplotype network construction.

2.2 Phylogenetic analyses and species delimitation

Both individual gene and concatenated gene datasets were used for constructing maximum likelihood (ML) trees in RAxML (Stamatakis, 2006) implemented in the raxmlGUI v2.0 (Edler et al., 2020). Trees were built with a GTR + G model and 1000 bootstrap replicates with Doriprismatica atromarginata as an outgroup for the concatenated dataset, COI and 16S, and with Tyrinna evelinae as an outgroup for ANT. Total branch length (TBL) of each gene tree was summed and used to derive a mutation rate ratio for mitochondrial (μm) and nuclear loci (μn) following Duda (2021). For μm, an average TBL of both genes was used. Both the Assemble Species by Automatic Partition (ASAP) algorithm (Puillandre et al., 2021) and the Bayesian bPTPmodel (Zhang et al., 2013) were used to partition the COI dataset into unique genetic clusters using default parameters. A minimum spanning haplotype network was generated with trimmed COI data using the haploNet function in the pegas package in R (Paradis, 2010).

2.3 Estimating divergence rates

Pairwise distances were calculated for COI, 16S, and ANT in MEGAX (Kumar et al., 2018) using a maximum composite likelihood model and pairwise deletion. To calculate a rate of divergence for this species, we used the minimum sequence divergence between geminate pairs and a closing date of the Isthmus of Panama of 2.8 mya (O'Dea et al., 2016). We chose to use the more conservative minimum sequence divergence rather than mean divergence to prevent the inclusion of accumulated intraspecific changes. Rates of divergence were calculated for all genes. The substitution rate is presented as half of the divergence rate. To determine whether the data were evolving in a clock-like fashion, a maximum likelihood clock test was conducted in MEGAX (Kumar et al., 2018) using a Tamura-Nei model and rate heterogeneity for complete COI, 16S and ANT sequences. Divergence rates from other geminate pairs of marine invertebrates were retrieved from the literature for comparative purposes, along with relevant life history information. Life history information, including lifespan, reproductive mode and larval dispersal, were obtained for the genus where possible but may represent broader taxonomic patterns in some instances.

3.1 Deep divergence within and among ocean basins

A total 22 COI, 22 16S and 19 ANT sequences were generated for C. ‘sedna’ in this study, with 25 total sequences of COI and 16S including those sourced from GenBank. Data for these specimens and GenBank accession numbers are provided in Table 1. The final COI alignment was 658 bp in length, the 16S alignment was 459 bp and the ANT alignment was 553 bp. A trimmed COI alignment of 590 bp was used for the haplotype network. Mean intraspecific divergence of C. ‘sedna’ was 8.7% (range 0.0%–15.7%) at COI, 3.9% (range 0.0%–8.1%) at 16S and 0.3% (range 0.0%–0.7%) at ANT (Tables S1–S3).

The COI + 16S + ANT ML phylogeny showed that western Atlantic and eastern Pacific C. ‘sedna’ formed separate clades, with more structure in the western Atlantic than in the eastern Pacific (Figure 2a). These same patterns were recovered in individual gene trees (Figure S1). Three western Atlantic clades were strongly supported in the combined phylogeny (BP > 90), although support at interior nodes was weak (Figure 2a). These results were reinforced by the species delimitation analyses that partitioned western Atlantic individuals into three clusters, corresponding to geographic region (Bahamas, Panama, and Florida) (Figure 2a). The minimum divergence between individuals in the western Atlantic and eastern Pacific was 13.6% at COI, 5.7% at 16S and 0.5% at ANT. Within the western Atlantic clade mean intraspecific divergence at COI was 4.0%, with a mean divergence of 5.4% between Florida and Bahamas, 6.5% between Bahamas and Panama, and 6.5% between Florida and Panama. The minimum sequence divergence at COI among these clades were 5.4%, 6.3%, and 6.2%, respectively. Within the eastern Pacific clade mean intraspecific divergence at COI was 0.45%.

The same patterns were also recovered in the haplotype network analysis, with different ocean basins showing significant differentiation, as well as regions within the western Atlantic showing deep haplotype divergence and no shared haplotypes (Figure 2b). In total, 8 and 9 unique haplotypes were identified from 12 and 13 western Atlantic and eastern Pacific individuals, respectively (Figure 2b). Within the Pacific, all haplotypes were closely related to the central haplotype in Mexico and the Sea of Cortez.

3.2 Rate variation

Given a minimum sequence divergence of 13.6% among C. ‘sedna’ individuals on either side of the Isthmus of Panama, and an estimated age of separation of 2.8 million years, COI is expected to diverge at a rate of 4.9% per million years. For 16S, with a minimum sequence divergence of 5.7%, this rate was 2.0% per million years. For ANT, with a minimum sequence divergence of 0.5%, this rate was 0.2% per million years. Divergence rates of 4.9%, 2.0%, and 0.2% correspond to substitution rates of 2.5% for COI, 1.0% for 16S, and 0.1% for ANT. COI and 16S rates are 25 and 10 times higher than the ANT rates, respectively. This variation in divergence rate also underpins differences in pairwise distance among loci (Figure S2). Total branch length of the COI, 16S, and ANT trees was 0.299, 0.136, and 0.006, respectively, with a μm/μn value of 39.5. The clock test was insignificant for COI, 16S, and ANT (p = .9, p = .7, p = .06), indicating that the null hypothesis of equal evolutionary rates throughout the tree could not be rejected (Table 2).

3.3 Systematics

Chromodorididae Bergh, 1891

Chromolaichma Bertsch, 1977

Type Casella sedna Ev. Marcus & Er. Marcus 1967

Chromolaichma hemera sp. nov. (Figures 1-3).

Zoobank registration. LSID: 42E0650E-C2F3-4A2E-A993-2CED24D17F82

Chromolaichma sedna—Matsuda & Gosliner 2018 (in part).

Dorisprismatica sedna—Johnson & Gosliner 2012 (in part), Goodheart et al. 2016.

Glossodoris sedna—Bertsch 1988, Lyons 1989, Humann & Deloach 2002, Debelius & Kuiter 2007 (in part), Behrens & Hermosillo 2005 (in part), Valdés et al. 2006.

3.3.6 Remarks

Chromolaichma was only recently resurrected by Matsuda and Gosliner (2018) to encompass three species (C. dalli, C. edmundsi, C. sedna). In previous family-level revisions, Rudman (1984) considered Chromolaichma as a synonym of Glossodoris, while Johnson and Gosliner (2012) placed it in ‘Doriprismatica’, later referred to as Doriprismatica.

Morphological conservation between C. hemera sp. nov. and C. sedna was noted by Bertsch (1988) who described them as nearly identical, noting slight differences in background color (not as white as C. sedna) and with possibly more crenulations in the mantle (Figure 4). Chromolaichma hemera sp. nov. also exhibits a wider yellow submarginal band than C. sedna (Figure 4). Bertsch noted that the reproductive system was identical to that illustrated in the original description of C. sedna (Marcus & Marcus, 1967) and that the radular morphology matched the holotype, and fell within the range for specimens described from the Gulf of California (Bertsch, 1978). Overall, he did not consider these slight morphological differences to warrant a new species, but the addition of molecular data that demonstrates deep genetic divergences does support the establishment of a new species for the western Atlantic and Caribbean specimens. Molecular data also demonstrates deep divergences among western Atlantic and Caribbean clades, which may represent additional cryptic species. A lack of geographically intermediate samples and additional morphological data prevents their description here, so for now, we refer to all lineages in the western Atlantic and Caribbean clades as C. hemera sp. nov. sensu lato. Upon interrogation of publicly available images, there are notable differences in rhinophore colouration between individuals in the lesser Antilles and the rest of the Caribbean, but additional work is needed to verify this preliminary finding. Finally, the name C. sedna is now restricted to specimens in the Gulf of California and the eastern Pacific.

4.1 A new species in the western Atlantic and deep divergences within populations

This study uncovered strong geographic structuring of C. ‘sedna’, with reciprocal clades in the western Atlantic and eastern Pacific, providing support for a previously unrecognized geminate species pair. The minimum sequence divergence among individuals on either side of the Isthmus (13.6% at COI) is much greater than intraspecific divergences reported for other nudibranchs (0%–6%) (e.g. Fritts-Penniman et al., 2020; Korshunova et al., 2018; Layton et al., 2018), and molluscs more generally (0%–5%) (e.g. Layton et al., 2014), and it aligns with divergences in other geminate species (e.g. 6.3%–26.5% in arcid bivalves, Marko & Moran, 2009; 8.9% in decapods, Pileggi et al., 2014). The monophyly of western Atlantic and eastern Pacific clades was strongly supported in the phylogenetic analyses and diagnostic morphological features set them apart. For instance, there are notable differences in the color of the marginal and submarginal bands and there are more crenulations in the mantle of the western Atlantic clade. Together, this evidence supports their status as separate species and a new name is established here for the western Atlantic clade (C. hemera sp. nov.). The discovery of a new species of nudibranch adds to a recent body of work employing molecular data to propel species discovery in this diverse group of invertebrates (e.g. Do et al., 2020; Epstein et al., 2019; Matsuda & Gosliner, 2018), but the nudibranch fauna of the western Atlantic has been less well studied than the Indo-Pacific.

Our study also uncovered significant population structure within western Atlantic C. hemera sp. nov., as evidenced by deep haplotypic divergence among populations, negating the hypothesis that this species was recently introduced to the area through ship transport (e.g. Goodheart et al., 2016; Humann & Deloach, 2002). Bertsch (1988) mistakenly reported the location for a specimen shared with him by Lyons when the first tropical western Atlantic specimens were reported (see Lyons, 1989 for correction). Thus, Bertsch incorrectly reported that all known specimens were from a restricted locality, initiating the idea that they may represent an introduction, although not explicitly stating that. The earliest known specimens in the western Atlantic were identified from the Bahamas in 1978 (Lyons, 1989), only 11 years after the description of the eastern Pacific species C. sedna. The deep molecular divergences among tropical western Atlantic populations do not support a recent introduction scenario and instead may represent a cryptic species complex (C. hemera sp. nov. sensu lato). Conversely, we uncovered a lack of structure within the eastern Pacific clade. Although surprising, this has also been shown in Tripneustes sea urchins (Lessios et al., 2003) and the pronghorn spiny lobster (Panulirus penicillatus) (Chow et al., 2011) and it may be due to differences in oceanographic currents, habitat complexity and glacial sea level changes in the region.

There are several possible explanations for the deep divergence among western Atlantic clades. First, these deep divergences might reflect missing taxa given the high extinction rate in the region following the closure of the Isthmus (Lessios, 2008; Marko et al., 2015), or they might reflect rapid diversification following the split (Thacker, 2017). The discovery of a distinct genetic clade at each region in the western Atlantic (e.g. Panama, Bahamas and Florida) is congruent with patterns in other marine invertebrates in the area. For instance, DeBiasse et al. (2016) uncovered deep divergences among populations of the brooding sponge, Callyspongia vaginalis, with isolation among Central American and Caribbean clades and a further phylogeographic break between populations in Florida and the Bahamas. Alternatively, deep sequence divergences may be the result of accelerated rates of evolution. This is a particularly plausible scenario for nudibranchs given that the COI divergence rate derived here (4.9% per myr) is significantly higher than rates reported in other (shelled) gastropods, although very similar to rates in other shell-less gastropods (e.g. Elysia, Ellingson & Krug, 2016).

4.2 Rate heterogeneity among geminate lineages

Rate heterogeneity has been invoked to explain variation in sequence divergence observed across marine geminate pairs (Marko, 2002), but this is the first study to summarize these rates. Here, we found significant rate heterogeneity among marine invertebrate genera that share a similar lifespan, reproductive mode and dispersal potential, suggesting that life history traits do not drive differences in mitochondrial rates. These findings align with recent work by Saclier et al. (2018), who found that while nuclear substitution rates were linked to generation time in isopods, mitochondrial substitution rates were independent of life history traits. Some of the seminal work linking mutation rates to life history traits was focused on only a few taxa and loci (e.g. Müller & Albach, 2010), thereby limiting our understanding of rate variation across the genomes of most species. In fact, we uncovered significant differences in nuclear and mitochondrial substitution and mutation rates, as evidenced by an elevated μm/μn ratio, similar to findings in other gastropods (Duda, 2021). Given the evidence for rate variation among both taxa and loci, future work will look to reassess substitution rates using transcriptome data from recent phylogenomic studies (Layton et al., 2020; Moles & Giribet, 2021). However, because COI, 16S, and ANT remain some of the most commonly used genes in species-level nudibranch investigations, the lineage-specific rates presented here will help improve understanding of divergence scenarios in this diverse group.