The cushion–star Parvulastra exigua in South Africa: one species or more?

Abstract The cushion–star Parvulastra exigua (Lamarck, 1816) is a widely distributed member of the temperate intertidal fauna in the southern hemisphere. In South Africa, it occurs in sympatry with the endemic Parvulastra dyscrita (Clark, 1923), the two species being differentiated predominantly by gonopore placement. Several recent studies have suggested that there may be additional cryptic species within the Parvulastra exigua complex in South Africa, based variously on color morphology, genetic evidence and the differential placement of the gonopores. This paper attempts to resolve whether one or more species are represented within Parvulastra exigua. A total of 346 Parvulastra exigua and 8 Parvulastra dyscrita were collected from sites on the west and south–west coasts of South Africa; morphological, anatomical and genetic analyses were performed to determine whether cryptic species and/or Parvulastra exigua specimens with aboral gonopores were present. Results show that neither cryptic species nor Parvulastra exigua specimens with aboral gonopores occur at these sites. This study thus refutes previous claims of the existence of aboral gonopores in South African Parvulastra exigua, and suggests that a single species is represented. The distinction between Parvulastra exigua and Parvulastra dyscrita is also confirmed, and features separating these two species are clarified and documented.


Introduction
The dwarf cushion-star Parvulastra exigua (Lamarck, 1816) is a prominent and widespread member of the temperate intertidal fauna in the southern hemisphere (Hart et al. 2006), occurring along the entire southern coastline of Africa from Namibia to Mozambique, in southeastern Australia and on several oceanic islands (Clark and Downey 1992). In South Africa, P. exigua occurs in sympatry with another endemic cushionstar, Parvulastra dyscrita (Clark, 1923), which has a larger adult size, occurs in lower densities and is found predominantly subtidally along the south and east coast, between False Bay and East London (Branch et al. 2010). Parvulastra dyscrita has a complex and intertwined taxonomic history with P. exigua (Table 1), which in part was driven by their morphological and ecological similarities. However, P. exigua and P. dyscrita have now been confirmed as two separate species in an unpublished thesis by Dunbar (2006), based on molecular and morphological (external gonopore position) evidence.
While the larger P. dyscrita has a fairly consistent mottled coloration ( Fig. 1), P. exigua demonstrates a high degree of color variation, with two major color morphs in South Africa demonstrating an allopatric distribution ( Fig. 1; Branch et al. 2010). The distribution patterns of these color morphs were studied in detail by Dunbar (2006), who noted a strong spatial divergence separated by a narrow zone of color morph sympatry around Cape Point. On the west coast, P. exigua were of a uniform khaki green color (similar to Australian populations), while more brightly colored, variegated individuals dominated along the east and south coasts. Dunbar (2006) also found the color morphs to demonstrate some degree of ecological divergence. The mottled morph is found predominantly in the high tidal zone within protected, bare rock, algae-encrusted and under boulder habitats with few algal tufts, while the green and intermediate color morphs were found in the mid tidal zone, with the green color morph inhabiting under boulder and bare rock habitats and tidal pools with little/no algae canopy and coralline algae. Similarly, the intermediate color morph also predominantly inhabited bare rock tidal pools, with little coralline algae, but also avoided those with algal tufts and an algal canopy.
Contrary to her expectations, Dunbar (2006) found no evidence of genetic separation between the two major color morphs, with perhaps temperature and/or predation maintaining the observed color polymorphism. In addition, a highly divergent haplotype was identified in twelve Kommetjie specimens, indicating the presence of a reproductively isolated cryptic species within a very narrow geographic range (Dunbar 2006). These individuals were documented as members of the intermediate color morph and exhibited a unique reddish-orange coloration (Dunbar 2006). Apart from this relatively subjective difference in color morph, Dunbar (2006) noted that this cryptic species appears morphologically similar to P. exigua, especially with regards to the presence of oral gonopores, but was found to be more closely related to the outgroup taxa Parvulastra parvivipara (Keough and Dartnall 1978) and Parvulastra vivipara (Dartnall 1969). Dunbar (2006) went on to suggest that this Kommetjie lineage should be classed as a new species, but to the authors' knowledge, no such species description was ever prepared.
The major differentiating features between P. exigua and P. dyscrita are the position of the gonopore and reproductive mode. Parvulastra exigua is an ovipositor that spawns predominantly from August to October (Lawson-Kerr and Anderson 1978;Byrne 1992). The sticky egg masses are laid via oral gonopores (Lawson-Kerr and Anderson 1978) on the undersides of boulders and give rise to distinct lecithotrophic benthic larvae (Byrne and Anderson 1994). By contrast P. dyscrita releases eggs into the water column via aboral gonopores, where they hatch into planktonic larvae.
However, in 2006, Hart et al. externally examined various preserved specimens of P. exigua from museum collections from South Africa, southern Australia and several islands (St. Helena, Amsterdam, St. Paul and Kerguelen) for evidence that some P. exigua populations might include cryptic species with a different mode of reproduction. Overall, 33% (21% in South Africa) of the P. exigua specimens examined (excluding individuals in which gonopore position was uncertain) were reported to have aboral gonopores, with such individuals occurring predominantly in South Africa and the St. Paul, Amsterdam and St. Helena islands. Hart et al. (2006) also analyzed mitochondrial DNA sequence data from the study by , leading them to tentatively suggest that a cryptic species of P. exigua (or more) exist in South Africa, probably with aboral gonopores. Moreover, Dartnall and Byrne (unpublished observation cited in Colgan et al. 2005) proposed the presence of a cryptic species with aboral gonopores in South Africa and a few oceanic islands. Step in taxonomic history Performed by P. exigua first described as Asterias exigua.
Whitelegge (1889) Mortensen (1921) Asterina exigua moved into the new genus Patiriella (often ignored by later authors). Verrill (1913) A new species with aboral gonopores, Asterina dyscrita described; suggestion made that it may only be a variety of Asterina exigua. Clark (1923) Asterina dyscrita placed into synonymy with A. exigua. Mortensen (1933) Asterina (Patiriella) exigua reviewed; it was proposed that there was a second species within exigua with aboral gonopores. Dartnall (1971) It was suggested that the species with aboral gonopores was Asterina dyscrita and moved to the genus Patiriella due to morphological similarity with Patiriella exigua. Clark (1974) Patiriella exigua and Patiriella dyscrita moved to the new genus Parvulastra which is distinguished from Patiriella based on ray width, ray plate alignment and is supported by previous molecular studies conducted by .
O' Loughlin and  Parvulastra exigua and Parvulastra dyscrita were confirmed to be two separate species based on morphological (external gonopore position) and molecular evidence (mtDNA COI). Another species that only occurs in Kommetjie was recognized within P. exigua, but no species description was recorded. Dunbar (2006) The studies of both Dunbar (2006) and Hart et al. (2006) point to the possibility (and presence) of cryptic species within P. exigua, yet with the absence of specimens from both studies and the lack of resolution on gonad structure, this remains unresolved. Therefore, the aim of this study was to resample P. exigua populations on either side of the morphological overlap to confirm whether genetically distinct specimens and/or P. exigua specimens with aboral gonopores are in fact present. To do this, mtDNA COI gene as well as geometric morphometric approaches were utilised.

Specimen collection
Where possible, 90 starfish were collected during low spring tide from intertidal rocky shores at each of four main collecting sites ( Fig. 2; compiled using QGIS v.2.6.1): Kalk Griffiths; individuals not to scale, with approximate sizes given in Table 5.
Bay and Hermanus (south-west coast), and Kommetjie and Britannia Bay (west coast) which lie within two bioregions. At each site, 30 specimens were collected from three vertical intertidal zones; the high-, mid-and low shores. An exception was Britannia Bay, where few starfish could be found in the lowest zone. Twelve additional specimens were added to the analysis from other sites around the coast to either enhance the Parvulastra dyscrita sample size, or to include Parvulastra exigua of unusual appearance or from unusual habitats, such as intertidal sandbanks (Table 2).

Morphology and anatomy
The oral and aboral surface of each specimen was photographed after collection to document the color pattern of each starfish. After preservation in 70% ethanol, several qualitative and quantitative characteristics, based predominantly on previous taxonomic descriptions from both known species, were recorded, using a dissecting microscope and digital caliper respectively (Table 3). In addition, marginal plate spines and tube feet were examined, but these characteristics were excluded from further analyses, as no differences between specimens were noted. The abactinal and actinal surface coloration of each individual specimen was also eliminated while only R/r values were considered, due to the high variability observed and possibility of skewing the analyses respectively. Specimens were not retained after examination.

Statistical analyses
Overall, 354 specimens were included in the multivariate analyses, which were performed on unstandardised and untransformed characteristic data using PRIMER v.6.1.5 (Plymouth Routines in Multivariate Ecological Research; Clarke and Gorley 2006). A non-metric multidimensional scaling (MDS) ordination, based on a resemblance matrix generated from Bray-Curtis similarities, was used to visually assess specimen similarity. Additionally, this ordination was utilized to identify outlying individuals which were included in the genetic analyses. Six groups were defined ( Fig.  3; Table 4), each comprising seven individuals that could possibly represent cryptic species within P. exigua. A seventh group, consisting of seven P. dyscrita, was included for comparison. The one-way ANOSIM (analysis of similarity) routine was performed to determine whether possible specimen groupings are associated with any of the documented characteristics, with the significance of the statistical tests assigned at the 5% level. Thereafter, SIMPER (similarity percentage analysis) was used to determine the characteristics that contribute to at least 90% of the difference between divergent cluster groups.

DNA extraction, PCR and sequencing
Overall, 49 specimens were selected for genetic analyses ( Fig. 3; Table 4); 44 were successfully sequenced. Approximately 25 mg of the specimen arm was removed for analysis. DNA was extracted using the NucleoSpin Tissue Kit (Machery-Nagel), following the manufacturer's instructions. Table 3. Starfish characteristics examined per specimen and used in multivariate analyses.

Characteristic Technique
Quantitative R/r Expression of body proportion. R = greater radius measured along the ambulacral groove. r = smaller, interambulacral radius. Both measured along three non-deformed arms per specimen and averaged.
Peristomial membrane diameter Used as an expression of body size. Measured twice per specimen and averaged.
Madreporite diameter Measured twice per specimen and averaged.
Papulae diameter Five diameters measured per specimen and averaged.
Oral plate spines Number of oral plate spines.
Oral plate erect spinulation Number of erect spines per oral plate.
Oral marginal plate spinulation Number of oral marginal spines.
Qualitative Color morph Green or mottled.
Abactinal surface texture Either clusters of spinelets, or evenly granular surface texture.
Adradial actinal spinulation Absent or present. Where present, noted whether these spines occurred in more, or less than, three arms.
Furrow/Ambulacral spinulation Classified according to relative number of plates with one spine, as well as the presence or abundance of three spines per plate.

Actinal intermediate plate spinulation
Classified according to the relative number of plates with one or two spines, starting position of the plates with two spines and the presence of plates with three spines.

External visible gonopore position
Having either oral gonopores or none (aboral gonopores are difficult to observe).
Gonopore position by dissection Dissected to determine gonad placement and definitively document gonopore position. Table 4. Specimen groups selected for genetic analyses. Each group comprises seven individuals identified on an MDS.

Mottled
Individuals from south-west coast with variegated coloration.

Green
Individuals from west coast with green coloration.

Orange
Individuals from west coast with orange coloration.
4. Two oral plate spines Individuals with two oral plate spines, as opposed to the four observed in most specimens collected.

Langebaan
Individuals from a sandflat habitat with deep aboral 'dents' and peculiar abactinal surface spinulation. Some also appear non-pentagonal.

Parvulastra dyscrita
Outgroup included for comparison.  Table 4, and used in the genetic analyses. Letters indicate location of specimen collection as seen in Table 2. Photos by R.P. Payne; individuals not to scale, with approximate sizes given in Table 5.
A partial section of the mtDNA cytochrome oxidase I gene was amplified by PCR, using a combination of primers; the invertebrate primers LC01490 and HCO2198 (Folmer et al. 1994), as well as P. exigua specific primers: Pexig_F1 (5'-CTTTCC-CACGAATGAACAAYATGAGC-3') and Pexig_R1 (5'-CCGAGGGCTCATAG-GAGAGGAGTGTC-3') (Mertens 2012). All amplifications were performed in 25 µL reactions, with the PCR protocol as follows: an initial denaturing step of 3 min at 94 °C, followed by 35-38 cycles of 94 °C for 30 s, an annealing temperature of 45 °C for 45 s, and 45 s at 72 °C, with a final extension of 10 min at 72 °C. The number of cycles (35-38) was dependent on the primers and DNA dilution used, which differed according to specimen. PCR products were visualized on a 1% agarose gel stained with ethidium bromide, and sequences were generated on an ABI-3100 automated sequencer, at the Stellenbosch University Central Analytical Facility.

Sequence analyses
Geneious v.6.1.6 was used to build an unrooted neighbor-joining tree with bootstrap support from consensus sequences that had a final length of 345 bp. Sequences were also analyzed using TCS v.1.21. (Clement et al. 2000) in order to generate a parsimony haplotype network. For the latter analysis, a 95% plausible connection limit was used. The parsimony haplotype network was visualized using Haploview (Barrett et al. 2005).

Results
Of the 354 specimens included in the analysis, eight were identified as Parvulastra dyscrita and the remainder as Parvulastra exigua. The two species are morphologically distinct (Fig. 1); Table 5 lists the main morphological differences between them. The collection location, as well as several characteristics including size, abactinal surface coloration, surface spinulation, surface texture and the presence or absence of adradial actinal spines can aid in the identification of specimens in the field.
All collected P. dyscrita specimens had muted shades of pink, white, brown and turquoise on the abactinal surface, with a bluish-yellow color orally (Fig. 1). The gonopore position was not externally visible in any of these specimens, but once dissected it was clear that all had aboral gonopores Representative individuals of P. exigua collected from the various study sites are depicted in Fig. 3, displaying a wide range of both shape and color. Specimens collected from Kommetjie and Britannia Bay (west coast sampling sites) exhibited a variety of uniform abactinal surface coloration, ranging from pale green, olive green, to orange, brown and blue, while those from Hermanus and Kalk Bay (south-west coast sampling sites) were all mottled morphs, of darker coloration. No green morphs were sampled from the south-west coast collection sites and neither were intermediate forms. Indi-  Clark 1923;Dartnall (1971); O' Loughlin and  Actinal intermediate plate spinulation Each plate with only one or two spines, with the latter occurring more frequently distally.
None. Difficult to detect. -

Gonopore position
Oral-two in each interradius, separated from oral plates by approximately three actinal plates.
Aboral. Dartnall (1971); Clark and Courtman-Stock (1976) O' Loughlin and  viduals collected from Langebaan Lagoon were dark green, with some exhibiting an orange shoulder. Actinal coloration was highly variable between and among sampling sites and the oral gonopore position could be seen in some preserved specimens, while no gonopores could be seen in others. Specimens collected from Langebaan Lagoon all demonstrated large aboral 'grooves' that could be confused for gonopores. However, when dissected, all P. exigua specimens displayed oral gonopores.

Inter-species differences
As can be identified in Fig. 4, two clear clusters, with an average dissimilarity of 35.96%, are evident in the MDS plot, representing P. dyscrita and P. exigua. This confirms that these represent distinct species, with gonopore position (ANOSIM, R = 0.981, p = 0.001), abactinal surface texture (ANOSIM, R = 0.981, p = 0.001), abactinal surface spinulation (ANOSIM, R = 0.757, p = 0.001) and oral plate spines (ANOSIM, R = 0.682, p = 0.001) playing a predominant role in this configuration. SIMPER results suggest that spinulation plays a major role in the delineation of these species, with erect oral plate, actinal intermediate plate, oral marginal plate and furrow/ambulacral spinulation contributing to 57.55% of the difference between the two species. An unrooted neighbor joining tree (data not shown) supports this distinction, with a clear separation of the sequences into two clades with 100% bootstrap support. This finding is supported by the haplotype network (Fig. 5), which failed to connect the two clades with more than 95% probability.

Intra-species variation
Within the P. exigua cluster (Fig. 4), the specimens differed significantly, with oral plate spines (two to four spines per oral plate or variable, ANOSIM, R = 0.677, p = 0.001) and abactinal surface spinulation (short columnar spines to similar spines but with slightly different shape, ANOSIM, R = 0.527, p = 0.001) playing a predominant role in this configuration. Overall there seems to be little morphological separation between P. exigua specimens. The phylogenetic tree (data not shown) and the haplotype network (Fig. 5A) show little genetic variation between P. exigua sampled at different locations, tidal heights or color morph. Two main haplotypes dominate Figure 4. Non-metric MDS ordination of all specimens analyzed. Two clear clusters displayed which represent the species Parvulastra dyscrita on the left, and the species Parvulastra exigua on the right. Circles indicate 75% similarity. The specimen causing a cluster overlap is of the species P. exigua, but has many of the morphological characteristics associated with P. dyscrita due to its large size and possibly collection location.
the haplotype network for P. exigua (Fig. 5A), both of which are well distributed across the sampling locations, with three unique haplotypes found along the west coast. The connections between the haplotypes indicate that they are genetically very similar (at the maximum five mutational steps distance). Overall, this evidence suggests that all P. exigua specimens collected in this study represent a single, morphologically variable species.

The validity and identification of Parvulastra dyscrita
Parvulastra dyscrita and P. exigua show a clear separation, based on morphology, anatomy and genetics, unambiguously confirming them to be two distinct species. The differences in characteristics of the two species supports those defined in earlier taxonomic work (Table 5) with gonopore position, abactinal surface texture and abactinal surface spinulation playing a major role. SIMPER results confirm that spinulation is a main delineator between these species.
The clear separation of these two species confirms Dunbar's (2006) results, but that study also highlights the ease with which specimens can be misidentified. This is understandable, especially in the case of large P. exigua, that can sometimes look morphologically very similar to P. dyscrita (personal observation), an example being the specimen in Fig. 4 that causes the 75% similarity circles to overlap. Based on the position of the gonopore, and several other characteristics, this specimen is definitely P. exigua, but is one of the largest specimens collected overall, had spinulation characteristics similar to those of P. dyscrita and was collected from the front of Strandfontein rockpool, which could be considered a subtidal location. All these characteristics facilitate misidentication in the field and caution that more detailed examination of spinulation and gonopore position are needed to confirm identification of ambiguous specimens.

Species resolution in P. exigua
All P. exigua specimens collected exhibited oral gonopores, with only slight morphological differences in oral plate spines and abactinal surface spinulation, but no separation great enough to indicate the presence of a cryptic species. This was confirmed by molecular analyses; the haplotype network shows little genetic variation within P. exigua sampled at different localities, tidal height or specimen group. Mertens (2012) conducted genetic analyses on a further 177 P. exigua specimens collected ~800 km of the west coast of South Africa (from Kommetjie to Port Nolloth) and did not find evidence of any cryptic species. However, it is important to note that no intermediate reddish-orange color morphs were collected in the present study and only a few Kommetjie specimens were analyzed genetically. All of this evidence suggests that if an undescribed Parvulastra species exists, it occurs in very low numbers, or in a very narrow geographic range, as suggested by Dunbar (2006). This warrants further investigation, and perhaps an extensive future sampling survey at Kommetjie. In addition, the slight intraspecific variation and observed colour polymorphism in P. exigua may be maintained by temperature and/or predation, as suggested by Dunbar (2006), and warrants more study. Future work should also include further comparisons (both morphological and molecular) of P. exigua and P. dyscrita specimens across the entirety of their currently known distributions, as well as the potential hybridization between South African Parvulastra species, which has not been investigated.

Gonopore position
After no P. exigua specimens from our original samples were found to have aboral gonopores, a further 200 P. exigua individuals were collected from Mouille Point, a location where a museum specimen with supposed aboral gonopores had been collected previously and examined by Hart et al. (2006). These specimens were also found to all have oral gonopores. In addition, Dunbar (2006) also found an examined subset of her P. exigua specimens to only have visible oral gonopores. On enquiry, it was determined that Hart et al. (2006) only examined museum specimens externally (Michael Hart and Maria Byrne, pers. comm.) and it is suggested that they mistook abactinal dimples (or the lack of visible oral gonopores) for aboral gonopores. This proposal is supported by the fact that P. dyscrita, a species with aboral gonopores, has no easily identifiable external gonopore position, making it hard to confirm gonopore position without dissection. A similar situation was faced with regards to the few P. exigua specimens examined in this study, as these displayed no oral gonopores and had to be dissected in order to reveal that the gonads were indeed orally directed. Finally, Langebaan Lagoon P. exigua specimens have deep abactinal grooves, and as some lack visible oral gonopores unless dissected, they could easily be mistaken as having aboral gonopores. It is also important to note that  did not include P. dyscrita in their molecular analysis and so some of the samples included in their analysis that depicts cryptic diversity in South Africa, could include P. dyscrita misidentified as P. exigua (Michael Hart, pers. comm.). Thus, this study suggests that P. exigua specimens in South Africa with aboral gonopores do not exist.