A total of 22 A. brunnea, 14 A. vanderspoeli (based on the descriptions of van der Spoel 1976) and 55 A. turriculata were examined in this study. All specimens have biogeographical data and have been analysed in at least one additional way (molecular, morphological or both, Table 1, Figure 1). Specimens were collected during several research cruises (Table 1) using various plankton nets. Collection methods have been described for all research cruises (Schmidt 1932; Hirai et al. 2015; Burridge et al. 2017a; Wall-Palmer et al. 2018b; Wall-Palmer et al. in preparation). Specimens from the DANA cruise of 1921 (two A. brunnea, two A. vanderspoeli, and two A. turriculata) were kindly made available by the Natural History Museum of Denmark (NHMD). All biogeographical data have been visualised using the software QGIS (QGIS Development Team 2016).

Figure 1. 

Collection locations for A. brunnea group specimens analysed in this study. Members of this species group are known to inhabit all oceans from 40N to 30S (Wall-Palmer et al. 2018b).

Specimens were identified based on the species keys of Richter and Seapy (1999), Seapy et al. (2003) and van der Spoel (1976), and imaged using a Zeiss automated z-stage light microscope at Naturalis Biodiversity Center, Leiden. Specimen shells were destroyed during the process of DNA extraction, so archived images provide the opportunity to verify in cases of uncertainty between morphologically identified specimens and molecular results. All images are available through the Barcode of Life Data System (BOLD; http://www.boldsystems.org) and are linked to physical DNA extracts at Naturalis Biodiversity Center (Table 1).

Due to the small size of atlantid shells it is difficult to orientate them in a reproducible way for reliable morphometric measurements. Therefore, micro-computed tomography (micro-CT) was used to generate 3D-models for a subset of specimens in order to create 2D-slices through the larval shell for reproducible morphological measurements. A total of five A. brunnea, seven A. vanderspoeli, and 19 A. turriculata were imaged using a SkyScan 1172 high resolution micro-CT scanner. In total 958 images were collected per specimen using the following parameters: no filter, medium resolution camera, an image pixel size of 1.31–2.23 µm, a source voltage of 57 kV, a source current of 177 µA and an exposure time 600–800 ms. The rotation pitch was 0.2°, with averaging frames of 4, and random movement of 10. The software Avizo 9.0 was used to generate a 3D-surface of each shell, and to make 2D-slices perpendicularly through the earliest part of the suture of the protoconch (Figure 2, protoconch pointing upwards) by setting landmarks on the larval shell. Morphological analysis of the 2D-slices was carried out using the software IMAGEJ (Schneider et al. 2012). The apical angle (Figure 2A), maximum larval shell height (Figure 2B), maximum larval shell width measured perpendicular to the spire axis (Figure 2C) and adult shell diameter excluding the keel (Figure 2D) were measured from the 2D-slice three times and the means were calculated. Additionally, the number of whorls in the larval shell and the total number of whorls in the shell were counted using the method of Seapy (1990). For statistical analyses of the morphological data, the software PAST v3.12 was used (Hammer et al. 2001). Normal distributions were tested using the Shapiro-Wilk test (Shapiro 1965). To determine the most important variables, a Principal Component Analysis (PCA) was carried out. T-tests were used to detect significant differences between the clades with regards to those informative variables.

Figure 2. 

Examples of shell parameters measured from 2D slices of micro-CT scans. Measurements include A apical angle B larval shell height C maximum larval shell width, and D maximum adult shell diameter. The position of the slice through the 3D model to create a 2D image is shown in E relative to the suture.

Scanning Electron Microscopy (SEM) images were also made on a subset of specimens to investigate the surface shell morphology and ornamentation for each clade. Two specimens of A. vanderspoeli and three specimens of A. turriculata were scanned using a JEOL JSM-7600F Field Emission SEM. For A. brunnea, images of two specimens from Wall-Palmer et al. (2018b) were used.

DNA extraction was carried out on whole specimens using the NucleoMag 96 Tissue kit on a Thermo Scientific KingFisher Flex magnetic particle processor with a final elution volume of 75 µl. A ~920 bp fragment of the nuclear 18S rRNA gene was amplified using primers 18S-KP-F (Burridge et al. 2017b) and 1800R (Vonnemann et al. 2005). A ~868 bp fragment of the nuclear 28S rRNA gene was amplified using primers C1-F (Dayrat et al. 2001) and D3-R (Vonnemann et al. 2005). Additionally, a ~658 bp fragment of the mitochondrial cytochrome c oxidase subunit 1 gene (mtCO1) (Hebert et al. 2003) was amplified using primers jgLCO1490 and jgHCO2198 (Geller et al. 2013). All primers were tailed with M13F and M13R for Sanger sequencing (Messing 1983). PCR reactions contained 17.3 µl ultrapure water (milliQ), 2.5 µl Qiagen 10× PCR buffer CL, 0.5 µl 25 mM MgCl₂, 1 µl 10 mg/ml Life BSA, 1.0 µl 10 pMol/µl of each primer, 0.5 µl 2.5 mM dNTPs, 0.25 µl 5U/µl Qiagen Taq and 1 µl of DNA extract (diluted either 5 or 10 times). For all three genes, PCR was carried out using a BIO-RAD Thermal Cycler and starting with an initial denaturation step of 3 min at 96 °C, followed by 40 cycles of denaturation of 15 s at 96 °C, annealing of 30 s at 50 °C, extension of 40 s at 72 °C and a final extension step of 5 min at 72 °C. PCR products were sequenced by BaseClear (Leiden). All sequences are publicly available through BOLD (see Table 1 for accession numbers). Molecular results have been previously published for some of the specimens (Wall-Palmer et al. 2018b, in preparation).

Sequences (forward and reverse strands) were assembled, aligned and verified in Geneious R8 and multiple sequence alignment was performed using MEGA-X (Kumar et al. 2016a). The resulting length of sequence alignments for CO1, 28S and 18S were 658 bp, 868 bp and 920 bp respectively. Phylogenetic relationships were estimated based on single-gene Maximum Likelihood analyses performed in RAxML 8.2.9 (Stamatakis 2014) using the General Time Reversible (GTRCAT) model followed by bootstrap analyses of 1000 replicates (CO1) or 1500 replicates (18S, 28S). Phylogenetic analysis of the three genes combined (2447 bp, GTRCAT, partitioned, 3000 replicates) included) included a subset of 14 specimens (plus three specimens of A. helicinoidea), with two specimens per clade per region (in regions where clades are present). All three genes were available for all of the specimens (Table 1). CO1 Genetic distances between and within clades/species were calculated using a Kimura-2-parameter (KP2) substitution model with uniform rates in MEGA-X (Kumar et al. 2016b). Three specimens of Atlanta helicinoidea were used as an outgroup in all phylogenetic analyses.

Phylogenetic analyses of CO1 (Figure 3) and combined genes (Figure 4) recover these three species with moderate to high bootstrap support. The single gene phylogenies of 18S and 28S resolve deeper relationships, but do not resolve the more recent divergences at the species level within the A. brunnea species group (Suppl. material 1: Figures S1–S2). All of the single and combined gene phylogenetic analyses support the A. brunnea species group as being monophyletic with 100 % node support. In the combined gene phylogeny (Figure 4), A. vanderspoeli forms a monophyletic group with 97% node support, supporting inference that this is a valid species. Atlanta brunnea also forms a well-supported (80 % for the combined genes) monophyletic group, but this is further split into two clades. Atlanta turriculata forms a monophyletic group (Indian and Pacific Oceans), but this is only moderately supported (68% node support).

Figure 3. 

A Distribution maps showing the collection locations for each clade identified in B. The collection location of specimens of A. turriculata forma B identified by van der Spoel (1976) from offshore of Ternate Island is marked with a white triangle B maximum likelihood phylogeny based on the mitochondrial cytochrome c oxidase subunit 1 gene, with strong bootstrap support for four clades within the A. brunnea group. Atlanta vanderspoeli is supported as a valid species, and A. brunnea is formed of two geographically isolated clades. Bootstrap support (%) for nodes is displayed and branch lengths are proportional to the amount of inferred change, as indicated by the scale bar (mean number of nucleotide substitutions per site).

Figure 4. 

Maximum likelihood phylogeny of the A. brunnea species group based on analysis of the combined genes CO1, 28S and 18S with a total alignment of 2447 bp. All four clades within the A. brunnea species group are monophyletic with strong bootstrap support. Bootstrap support (%) for nodes is displayed and branch lengths are proportional to the amount of inferred change, as indicated by the scale bar (mean number of nucleotide substitutions per site).

Genetic distances for CO1 support the position of A. vanderspoeli as a valid new species. Atlanta brunnea and A. vanderspoeli are separated by a CO1 genetic distance of 13–16% (Table 2). These genetic distances are comparable to the CO1 genetic distances between the previously described species within this species group, A. brunnea and A. turriculata (18–20 %). CO1 genetic distances between A. brunnea, A. vanderspoeli, A. turriculata and the outgroup species A. helicinoidea are 22–25%, 23–24% and 21–22%, respectively (Table 2).

Within A. brunnea, the molecular data identify a further separation into two well supported geographic clades, one containing the Indian and Pacific Ocean population and one containing the Atlantic Ocean population (Figure 4, node support of 75% and 99% respectively for the combined genes). There is a CO1 genetic distance of 6–7% between specimens from the Atlantic Ocean and specimens from the Indian and Pacific oceans. This genetic distance is less than distances between other species within the A. brunnea group, and therefore, one of these two populations is likely to be an incipient species. However, the type locality of A. brunnea is not known for certain, Gray (1850) having written only ‘A. brunnea, Eydoux’ to accompany his drawing of this species. Here we accept the lectotype locality because we assume that Gray illustrated the same specimens that Souleyet (1852) named A. fusca (originally given the vernacular name “Atlanta brune”), as they both originated from specimens collected by Eydoux on the “Bonite” expedition and the illustrations are almost identical (Suppl. material 1: Figure S3). The type locality for A. fusca is the Indian Ocean and so we assume here that the type locality of A. brunnea is also the Indian Ocean. Therefore, the newly detected Atlantic Ocean population should be considered as the incipient species.

Principal Component Analysis (PCA) identifies the apical angle of the protoconch and the number of whorls in the larval shell as the most informative morphological characters to distinguish between A. brunnea, A. vanderspoeli, and A. turriculata (Figure 5). The apical angle of A. brunnea is the widest (58.7°–71.0°), followed by A. vanderspoeli (35.9°–45.8°) and A. turriculata (17.4°–32.2°, Suppl. material 2: Table S1, Figure 6). The difference in apical angle between A. brunnea and A. vanderspoeli, and between A. turriculata and A. vanderspoeli is significant (t-test, p=<0.001 for both; Figure 6). Differences in the number of whorls in the larval shell are also identified by the data. Atlanta brunnea has 3.75–4.25 whorls, A. vanderspoeli has 3.25–4.00 whorls and A. turriculata has 4.00–4.50 whorls in the larval shell (Suppl. material 2: Table S1). The height to width ratio of the larval shell, maximum number of whorls (in the whole shell) and adult shell diameter overlap for the three clades and cannot be used as reliable identifying features (Suppl. material 2: Table S1). The low number of specimens does not allow an assessment of morphological differences between the Atlantic Ocean and the Pacific and Indian Ocean specimens of A. brunnea.

Figure 5. 

Principal Component Analysis (PCA) performed on apical angle, height: width ratio and the number of whorls in the larval shell. Species identity confirmed for most specimens (N = 25) using DNA barcoding (CO1). Two specimens of each species (total N = 6) derive from the DANA collection, and could not be DNA barcoded (formalin-fixed). Morphometric data are reported in Suppl. material 2:Table S1.

Figure 6. 

Shell apical angles of A. brunnea (Atlantic, Pacific and Indian oceans), A. vanderspoeli (Pacific Ocean) and A. turriculata (Pacific and Indian oceans) are significantly different and do not overlap.

Ornamentation is limited in its use as an identifying feature for this species group, and the fine micro-ornamentation is only really visible with the use of SEM. All three species have a prominent ridge, or carina that runs slightly above mid-whorl height of the larval shell. The larval shell of A. brunnea is easily identified by the heavy zig-zag, and in places, net-like reticulate spiral lines covering the surface (Figure 7B). However, the ornamentation of A. vanderspoeli is more difficult to tell apart from A. turriculata. In both, the larval shell is more sparsely ornamented than in A. brunnea. In A. vanderspoeli, interrupted spiral lines and small projections roughly arranged in lines are found above the carina of each whorl, and zig-zag ornamentation is found below the carina of each whorl (Figure 7E, Figure 8). In A. turriculata, there is generally little or no ornamentation above the carina of each whorl, but in common with the other two species, zig-zag ornamentation is found below it (Figure 7H).

Figure 7. 

Scanning Electron Microscopy (SEM) and stacked light microscopy images of representative specimens of A. brunnea: A, B DANA_3929VIII C SN105_08, A. vanderspoeli: D–E DANA_3558VII (Holotype, NHMD-232132) F KH1110_15 and A. turriculata: G, H DANA_3929VIII I SN105_19. Apical angle is the most useful morphological feature for distinguishing between the species (B–C, E–F, H–I). The shell of A. brunnea is always brown (C); however, the colour of A. vanderspoeli and A. turriculata shell and soft tissues (F, I) can vary and these are not reliable features for identification.

Figure 8. 

SEM images of the A. vanderspoeli holotype from station DANA_3558VII (NHMD-232132) A apical view of the entire shell showing the rapid inflation of the shell B magnified view of the micro-ornamentation C magnified apical view showing the extent of the carina.

Data gathered in this study expand the known distribution of the A. brunnea group (Wall-Palmer et al. 2016, 2018b). Atlanta brunnea was found to have the broadest distribution, in the Atlantic Ocean (north and south subtropical gyres), in the Indian Ocean and in the Pacific Ocean (Figure 3). Atlanta brunnea was not found in the central southern Pacific Ocean, although this is based upon limited sampling coverage in this region. Conversely, A. vanderspoeli was only found in the equatorial and southern Pacific Ocean (0.80N–29.95S, 127.28E–100.00W). The two species only overlap in the south west and the east Pacific Ocean (Figure 3). The molecular results presented here show that the Atlantic Ocean population of A. brunnea is a geographically divergent species. Atlanta turriculata was found throughout the Pacific and Indian oceans, overlapping with the distribution of A. vanderspoeli in the equatorial and South Pacific Ocean. In agreement with previous studies, A. turriculata was not found in the Atlantic Ocean (Figure 3).

Atlanta turriculata was described by d’Orbigny (1836) from the South Pacific Ocean. This is the region in which the distributions of A. turriculata and A. vanderspoeli overlap and so it is necessary to check that d’Orbigny described and illustrated specimens of A. turriculata and not the morphologically similar A. vanderspoeli. Unfortunately, type material of A. turriculata could not be found at either the Natural History Museum, London (NHMUK), or the Muséum National d’Histoire Naturelle, Paris (MNHN), where it should be present (syntype material located at NHMUK was discovered to be mislabeled specimens of Oxygyrus). In addition, the illustrations of A. turriculata made by d’Orbigny are not sufficiently detailed to determine which species he actually drew. Although the more widely geographically distributed, narrow-turreted spired species is now well known as A. turriculata, we find it necessary to solve this potential confusion by depositing a neotype specimen of (what is now widely known as) A. turriculata (see Systematics section for details, Figure 9).

Figure 9. 

Atlanta turriculata neotype RMNH.MOL.342213, housed at the Naturalis Biodiversity Center.

Additional diversity within the A. brunnea species group has long been recognised, but, until now, this has not been thoroughly investigated. Van der Spoel (1972) first noticed A. vanderspoeli when he remarked on an atlantid form similar to A. turriculata, calling it A. turriculata forma B. Initially recognised from the soft tissues, van der Spoel described A. turriculata forma B as generally having a lack of tubercules on the operculum, which were present in A. turriculata. Later, van der Spoel (1976: 146–147) described the shell as having ‘whorls in the spire [that] increase more rapidly in width so that the spire in relation to the body whorl is larger than in [A. turriculata]’. Van der Spoel deposited specimens of this forma, which are now in the collection of the Naturalis Biodiversity Center, Leiden (Figure 10). More recently, Wall-Palmer et al. (2018b) detected an additional clade within the A. brunnea group using a CO1 phylogeny of a global collection of specimens. This clade was more closely related to A. brunnea, and was referred to as A. brunnea form B. Here we considered three species concepts to determine whether this new clade, referred to in this study as A. vanderspoeli, was a valid species; the morphological species concept, the biological isolation (biogeographical) species concept and the phylogenetic species concept (De Queiroz 2007). Under each of these concepts, our results verify that A. vanderspoeli is different from the two most closely related species, A. brunnea and A. turriculata. Although there were only a relatively small number of specimens available for morphological analysis, A. vanderspoeli shells were found to be morphologically distinct, having an apical angle that does not overlap with its two closest relatives. The combined three gene phylogeny supports A. vanderspoeli to be genetically distinct from A. brunnea and A. turriculata, with genetic distances comparable to other species within the group. Atlanta vanderspoeli also appears to have a relatively restricted distribution, only being found (thus far) in the equatorial and South Pacific Ocean. The pteropod species Cuvierina tsudai Burridge, Janssen & Peijnenburg, 2016 has a very similar geographical distribution (Burridge et al. 2015, 2016). The known distribution of A. vanderspoeli only overlaps with A. brunnea and A. turriculata at the very edges of their populations. Relatively poor spatial resolution of molecular data in the Pacific Ocean means it is possible that the distribution of A. vanderspoeli is also wider (e.g. in the north Pacific). The species is therefore not strongly geographically isolated from its two most closely related species. However, these results are compatible with the phylogenetic species concept, because even though the species overlap geographically and likely do meet in nature, there must be no interbreeding as evidenced by the combined gene phylogeny (and 28S, Suppl. material 1: Figure S1).

Figure 10. 

Five specimens of Atlanta turriculata forma B identified by van der Spoel and held in the collection at the Naturalis Biodiversity Center. These specimens are now designated as paratypes of Atlanta vanderspoeli RMNH.MOL.342212.

Results of the phylogenetic analysis also highlight a probable incipient species. The morphospecies A. brunnea was found to have two genetically different populations, one in the Indian and Pacific Oceans, and one in the Atlantic Ocean. The genetic distance between the two populations is relatively small, but suggests that the populations of A. brunnea in the Atlantic Ocean must be isolated and becoming a new species. This result agrees with Wall-Palmer et al. (2018b) who suggested the existence of an additional genetic lineage but due to insufficient data this could not be confirmed as a distinct species. Richter (1961) also noted differences in the Atlantic population of A. fusca (= A. brunnea), describing them as having smaller, more pigmented shells when compared to specimens in the Indian Ocean. Differences in the operculum and radula were also reported by Richter (1961). We did not detect any morphological differences between the two populations of A. brunnea, however, only shell ornamentation and the measurements of the larval shell have been analysed here.