INTRODUCTION
South Africa is bathed by the warm Agulhas Current on the east coast and by the cool Benguela Current along the west coast: it is an environment that is characterized by the presence of a number of clear biogeographical provinces characterized by distinct communities of marine organisms (e.g. Emanuel et al., Reference Emanuel, Bustamante, Branch, Eekhout and Odendaal1992; Gibbons et al., Reference Gibbons, Buecher, Thibault-Botha and Helm2010a). As a consequence, it supports a remarkably high diversity of marine life (Gibbons et al., Reference Gibbons, Abiahy, Angel, Assuncao, Bartsch, Best, Biseswar, Bouillon, Bradford-Grieve, Branch, Burreson, Cannon, Casanova, Channing, Child, Compagno, Cornelius, Dadon, David, Day, Della Croce, Emschermann, Erseus, Esnal, Gibson, Griffiths, Hayward, Heard, Heemstra, Herbert, Hessler, Higgins, Hiller, Hirano, Kensley, Kilburn, Kornicker, Lambshead, Manning, Marshall, Mianzan, Monniot, Monniot, Newman, Nielsen, Patterson, Pugh, Roeleveld, Ross, Ryan, Ryland, Samaai, Schleyer, Schockaert, Seapy, Shiel, Sluys, Southward, Sulaiman, Thandar, van der Land, van der Spoel, van Soest, Vetter, Vinogradov, Williams and Wooldridge1999). While the diversity of organisms associated with the benthos is substantially greater than that in the pelagos, the region typically supports a greater proportion of global species in the latter than former environments (Gibbons et al., Reference Gibbons, Buecher, Thibault-Botha and Helm2010a). But, whilst 57% of the world's planktic urochordates or euphausiids have been reported from the region only ~7% of the world's scyphozoans have been formally logged (Gibbons et al., Reference Gibbons, Abiahy, Angel, Assuncao, Bartsch, Best, Biseswar, Bouillon, Bradford-Grieve, Branch, Burreson, Cannon, Casanova, Channing, Child, Compagno, Cornelius, Dadon, David, Day, Della Croce, Emschermann, Erseus, Esnal, Gibson, Griffiths, Hayward, Heard, Heemstra, Herbert, Hessler, Higgins, Hiller, Hirano, Kensley, Kilburn, Kornicker, Lambshead, Manning, Marshall, Mianzan, Monniot, Monniot, Newman, Nielsen, Patterson, Pugh, Roeleveld, Ross, Ryan, Ryland, Samaai, Schleyer, Schockaert, Seapy, Shiel, Sluys, Southward, Sulaiman, Thandar, van der Land, van der Spoel, van Soest, Vetter, Vinogradov, Williams and Wooldridge1999). Given both their conspicuousness (most Scyphozoa are of a relatively large size and many bloom on a seasonal basis) and the fact that the region has been explored by many of the Great Expeditions, this is somewhat surprising.
Globally, however, our understanding of scyphozoan diversity is relatively poor and a total of only ~200 species have been described to date (Mianzan & Cornelius, Reference Mianzan, Cornelius and Boltovskoy1999). This number is probably artificially low, given their meroplanktic nature (Gibbons et al., Reference Gibbons, Janson, Ismail and Samaai2010b), and likely reflects the conserved nature of medusoid morphology (Hamner & Dawson, Reference Hamner and Dawson2009). Many of the original descriptions of scyphozoans are archaic, being based on a few subjective and qualitative diagnostic characters (Bolton & Graham, Reference Bolton and Graham2004; Dawson, Reference Dawson2005a), so the foundation on which our knowledge has been based is weak at best. To add to this confusion, some species are now known to display considerable phenotypic plasticity (Dawson et al., Reference Dawson, Martin and Penland2001; Dawson, Reference Dawson2005a), and crypsis is becoming more widely reported (Dawson & Jacobs, Reference Dawson and Jacobs2001; Schroth et al., Reference Schroth, Jarms, Streit and Schierwater2002; Holland et al., Reference Holland, Dawson, Crow and Hofmann2004; Dawson, Reference Dawson2004, Reference Dawson2005b). Although morphological descriptions are still essential when documenting diversity, there is a desperate need to revise them using more objective and quantitative features, supplemented where possible with molecular data (Dawson & Jacobs, Reference Dawson and Jacobs2001; Schroth et al., Reference Schroth, Jarms, Streit and Schierwater2002; Dawson, Reference Dawson2003, Reference Dawson2004, Reference Dawson2005b, Reference Dawsonc; Holland et al., Reference Holland, Dawson, Crow and Hofmann2004; McManus & Katz, Reference McManus and Katz2009). Here we describe a species of Crambionella (Scyphozoa: Rhizostomeae) from the St Lucia Estuary using just such an approach, in part to set a modern descriptive standard for the wider taxon and in part to allow identification of the present material.
The genus Crambionella is an inscapulate daktyliophore (Scyphozoa: Rhizostomae), with an intra-circular network of anastomosing canals that communicates with the ring canal (not stomach), occasionally with adjacent radial canals (Stiasny, Reference Stiasny1922; Kramp, Reference Kramp1961). Lappets are separated by a deep furrow and are free of any anastomosing canals (Stiasny, Reference Stiasny1922, Reference Stiasny1937; Rao, Reference Rao1931). This genus possesses three winged oral arms with terminal clubs, pyramidal in shape, lacking any whip-like filaments (Rao, Reference Rao1931; Kramp, Reference Kramp1961). Three species are presently known, and all are confined to the Indian Ocean. Crambionella orsini (Vanhöffen, Reference Vanhöffen1888) was first described from the Red Sea and is known to bloom seasonally in the north-west Indian Ocean (Billet et al., Reference Billett, Bett, Jacobs, Rouse and Wigham2006; Daryanabard & Dawson, Reference Daryanabard and Dawson2008). Crambionella stuhlmanni (Chun, Reference Chun1896) was originally described from the mouth of the Quilimane River, East Africa and C. annandalei Rao, Reference Rao1931 was first identified from material collected off of the coast of Chennai (formerly Madras), India. The three species can be separated primarily on the basis of the presence of conical projections on velar lappets, accessory orbicular mouth appendages on the oral arms as well as the proportion of terminal club length to oral arm length, and their respectively isolated geographical ranges within the Indian Ocean (Vanhöffen, Reference Vanhöffen1888; Chun, Reference Chun1896; Mayer, Reference Mayer1910; Stiasny, Reference Stiasny1922, Reference Stiasny1937; Menon, Reference Menon1930; Roa, 1931; Ranson, Reference Ranson1945; Nair, Reference Nair1946; Kramp, Reference Kramp1956, Reference Kramp1961, Reference Kramp1970).
MATERIALS AND METHODS
Forty-eight specimens of Crambionella (Figure 1) were collected by dip-netting from the St Lucia Estuary (28°0′0″S 32°30′0″E) during December 2005, and were preserved immediately in 5% ambient seawater–formalin. The St Lucia Estuary forms part of the iSimangaliso Wetland Park (UNESCO, 2008) which is a World Heritage Site situated on the north-east coast of South Africa. This estuarine system is the largest in Africa occupying an area of approximately 325 km2 (Fielding et al., Reference Fielding, Forbes and Demetriades1991), and is made up of three main lakes with an average depth of less than 1 m (Anonymous, 2008) that connect to the sea through a 22 km long channel (100–200 m wide) (Cyrus et al., Reference Cyrus, Vivier and Jerling2010). The wetland park lies between tropical and sub-tropical climatic zones, and is typified by warm moist summers and mild dry winters; mean annual temperatures routinely surpass 21°C (Anonymous, 2008). It is prone to seasonal flooding in summer and periods of drought leading to temporary mouth closures and associated fluctuations in salinity during winter (Fielding et al., Reference Fielding, Forbes and Demetriades1991; Cyrus et al., Reference Cyrus, Vivier and Jerling2010; Jerling et al., Reference Jerling, Vivier and Cyrus2010). By comparison with many estuarine systems in South Africa, that of St Lucia is considered to be fairly well understood (Pillay & Perissinotto, Reference Pillay and Perissinotto2008).
A photograph of a live specimen from St Lucia Estuary (Ricky Taylor, Ezemvelo KZN Wildlife).
Morphological data collection
After 22 months in preservation, thirty-six morphological features were measured under magnification from 44 specimens (Table 1; Figure 2) using Vernier callipers. After all bell measures were taken, the oral arms were removed and the radial canal system was injected with coloured latex to highlight its arrangement. Five un-dissected specimens were deposited at the Iziko South African Museum, Cape Town: accession number SAMCT H5110.
Morphological features measured on Crambionella specimens collected from St Lucia Estuary, on the north-east coast of South Africa, during December 2005. (A) A schematic diagram of a longitudinal section along the perradial axis of a specimen (adapted from Dawson, Reference Dawson2005e); (B) a schematic diagram of the oral disc, from a subumbrella view (adapted from Dawson, Reference Dawson2005e). Only two of the eight oral arms are shown in A and B, and the key to all features (letters) measured should be sought from Table 1. Photographs showing the subumbrella view of a Crambionella medusa illustrating various morphological measurements (C) and the intra-circular and extra-circular anastomosing canal networks, after injecting coloured dye latex (D).
Morphological features (S #) of Crambionella specimens used in data analyses. Specimens were collected from St Lucia Estuary, on the north-east coast of South Africa, during December 2005 and preserved in 5% formalin in ambient seawater. Figure references are given where applicable. As C. orsini material examined at the Natural History Museum, London, had to be studied in a non-destructive way some measurements had to be excluded; measurements taken on C. orsini specimens are indicated (*).
Comparisons were made between the measured variables of the St Lucia material and those of five good specimens of Crambionella orsini in the collections at the Natural History Museum, London (specimen numbers detailed below). Owing to the fact that this material had to be studied in a non-destructive way, not all measures could be repeated and discrepancies are indicated in Table 1.
Morphological data analyses
In order to determine the effect of individual size (external bell diameter, S1) on measured variables; Pearson's R correlations or Spearman's rank correlations (for those variables that failed tests of normality) were computed following log10 transformation of data (Zar, Reference Zar1999). Test results were considered significant, following Bonferonni corrections for Type I errors (Quinn & Keough, Reference Quinn and Keough2002). Some morphological measures were expressed as proportions following Chun (Reference Chun1896), Mayer (Reference Mayer1910), Menon (Reference Menon1930, Reference Menon1936), Rao, (Reference Rao1931), Stiasny (Reference Stiasny1937) and Kramp (Reference Kramp1961). These included: oral disc diameter (Ѕ 13) to external umbrella diameter (Ѕ1); length of the distal oral arm portion (Ѕ7) to length of the proximal oral arm portion (Ѕ6); length of terminal club (Ѕ11) to total oral arm length (Ѕ6 and S7); ostia width (Ѕ15) to inter-ostia width (Ѕ15) and umbrella height (Ѕ3) to external umbrella diameter (Ѕ1).
Given that many of the variables did change with individual size (see below), which complicates straightforward field comparisons, raw morphometric data were standardized by dividing them by external bell diameter (S1; except the aforementioned proportions), and log10 ratios were correlated with log10 size in an effort to eliminate size dependency. Comparisons between the standardized measurements of the St Lucia material and those of C. orsini were computed as above.
In order to test for morphological differences between individual logged ratios and between untransformed meristic measures, of the Crambionella specimens from the St Lucia Estuary and those of C. orsini two-tailed t-tests were computed, and the results inspected following Bonferroni corrections. All univariate statistical analyses were executed using STATISTICA v. 7.
In order to visualize and test for multivariate differences between the St Lucia material and that of C. orsini from the Natural History Museum, London, we have used a variety of the non-parametric routines within the PRIMER 6 & PERMANOVA+ software (Clarke & Gorley, Reference Clarke and Gorley2006; Anderson et al., Reference Anderson, Gorley and Clarke2008). A similarity matrix based on Euclidean distance was first constructed between the multivariate states (untransformed standardized measures) of all specimens. In order to maximize the number of individuals used, gaps were filled either by mean substitution (if there was no significant relationship of the considered feature with size) or from regression equations: invariant meristic features were excluded. Non-metric multi-dimensional scaling (NMDS) was used to visualize relationships in multivariate space (Clarke, Reference Clarke1993) and a one-way analysis of similarities (ANOSIM) test was used (a priori) to test the null hypothesis of no morphological dissimilarity between species (Clarke & Warwick, Reference Clarke and Warwick2001). This latter routine computes an R statistic that measures the average distance between every specimen within a group and contrasts it to the average distance between every specimen from the other group. ANOSIM then performs a series (999) of permutation tests, wherein variables from each group (species) are randomly distributed between groups, and the R statistic is recalculated. If the original R statistic is more extreme than 95% of the permutations the null hypothesis is rejected at a level of P < 0.05.
In order to determine which of the variables contributed the most to dissimilarity between species, we used the similarity percentages (SIMPER) routine in PRIMER 6 (Clarke, Reference Clarke1993). SIMPER determines the average dissimilarity between all pairs of inter-group specimens (Clarke & Warwick, Reference Clarke and Warwick2001). These averages are then disaggregated into the percentage that each variable contributes to overall dissimilarity amongst groups (Clarke & Warwick, Reference Clarke and Warwick2001).
Finally, we used the canonical analysis of principal co-ordinates (CAP) routine in PRIMER 6 & PERMANOVA + , which is analogous to a discriminant functions analysis, in order to determine what percentage of St Lucia and C. orsini specimens were allocated to respective species groups. The CAP routine seeks a set of axes that best discriminates amongst a priori groups in multivariate space (Anderson et al., Reference Anderson, Gorley and Clarke2008). Anderson et al. (Reference Anderson, Gorley and Clarke2008) describe the processes executed within this routine. Numerous matrices are generated to produce a set of canonical axes. Conventionally in a canonical discriminant analysis a subset of principal co-ordinate (PCO) axes are chosen manually, based on the number of variables in the original data matrix. However, in the present study, as the number of standardized morphometric features approached the number of specimens, Anderson et al. (Reference Anderson, Gorley and Clarke2008) suggest ‘leave-one-out’ diagnostics to determine the subset of PCO axes. The PCO axes determined are all orthonormal and are therefore independent of each other. Running parallel to this process is a matrix based on codes for groups identified by a factor associated with the Euclidean distance matrix, also orthonormalized. An additional matrix is then generated by relating the subset of PCO axes to an orthonormalized data matrix, yielding canonical eigenvalues and their associated eigenvectors which can be used to produce a CAP plot. These CAP axes, which are linear combinations of a subset of orthonormal PCO axes, were used to determine if predefined groups were correctly classified. The CAP routine was also used to test the null hypothesis of no differences in the positions of centroids among groups in a multivariate space through a series of permutation tests (Anderson et al., Reference Anderson, Gorley and Clarke2008). This routine makes no assumptions about the underlying distribution of variables rendering it suitable for non-parametric analyses (Anderson et al., Reference Anderson, Gorley and Clarke2008).
DNA analysis
Three specimens of Crambionella were collected from the St Lucia Estuary at Charters Creek during September 2008, and immediately preserved in 99% ethanol. Material was stored at –20 °C prior to analysis in the laboratory.
DNA was extracted from ethanol-preserved oral arm tissues using a phenol–chloroform based method. Samples were placed in separate Eppendorf tubes. Extraction buffer (SDS 0.5%; 50 Mm Tris; 0.4 M EDTA; pH 8.0) in quantities of 0.5 ml were pipetted over each sample. Tissue samples were then macerated. Proteinase K (20 mg/ml) in quantities of 10 µl was then added. Samples were vortexed and incubated at 55°C for a minimum of three hours until the majority of protein was digested. Samples were then mixed with 500 µl phenol:chloroform:isoamyl alcohol (24:24:1), finger vortexed, then centrifuged at low speed (5000 × g) for 10 minutes. Supernatants were removed and placed in new Eppendorf tubes, mixed with 500 µl chloroform:isoamyl alcohol (24:1) and finger vortexed. Solutions were then centrifuged at low speed (5000 × g) for 10 minutes. Supernatants were removed and placed in new Eppendorf tubes. DNA was precipitated with 45 µl Na acetate and 650 µl of ice cold ethanol and left to incubate at –18°C overnight. Samples were then centrifuged at high speed (13000 × g) for 10 minutes and supernatants were discarded. Eppendorf tubes were inverted and left to air dry for a minimum of an hour. Each DNA sample was finally resuspended in 50 µl TE buffer.
Cytochrome c oxidase subunit I (COI) was amplified using primers LCOjf (5′-ggtcaacaaatcataaagatattggaac-3′) and HCOcato (5′-ctccagcaggatcaaagaag-3′) (Dawson, Reference Dawson2005d) or HCO2198 (5′-taaacttcagggtgaccaaaaaatca-3′) (Folmer et al., Reference Folmer, Black, Hoeh, Lutz and Vrijenhoek1994). Internal transcribed spacer one (ITS1) was amplified using the primers jfITS1-5f (5′-ggtttcgtaggtgaacctgcggaaggatc-3′) and jfITS1-3r (5′-cgcacgagccgagtgatccaccttagaag-3′) (Dawson & Jacobs, Reference Dawson and Jacobs2001). Sequences were amplified through polymerase chain reaction (PCR) and PCR conditions were different for each fragment analysed. PCR conditions (adapted from Daryanabard & Dawson, Reference Daryanabard and Dawson2008) are summarized in Table 2. PCR products were purified and sequenced at the Central Analytical Facility, University of Stellenbosch. Electopherograms were checked visually, misreads corrected and poorly resolved terminal portions of sequences were discarded using Sequencher 4.9. Forward and reverse sequences were then aligned, using default settings, in Sequencher 4.9. Sequence identifications were verified by BLAST in GenBank.
Polymerase chain reaction (PCR) conditions used to amplify cytochrome c oxidase subunit I (COI) and internal transcribed spacer one (ITS1) from Crambionella specimens collected from St Lucia Estuary, on the north-east coast of South Africa, during December 2005 and preserved in absolute ethanol (adapted from Daryanabard & Dawson, Reference Daryanabard and Dawson2008).
Phylogenetic analyses were utilized to examine family level relationships using COI rhizostome sequences received from Professor M.N. Dawson, also available on GenBank. The following sequences were utilized: Cassiopea andromeda (samples from Bermuda, AY319463 and AY319465); C. frondosa (sampled from Panama, AY319469 and AY319470); C. ornata (sampled from Palau, AY31945); C. ornata (sampled from Berau, AY319472); Mastigias papua (sampled from Palau, EU363340 and AY902982); M. papua (sampled from Berau, AY903048 and AY903049); Phyllorhiza punctata (sampled from Australia, EU363341 and EU363342; Catostylus mosaicus (sampled from Australia, AY737184 and AY737216); Crambionella orsini (sampled from Iran, EU363343 and EU363344); Cephea cephea (sampled from Palau, EU363345; C. cephea (sampled from Kwajalein, EU363346); Acromitus flagellatus (sampled from India, EU363347 and EU363348) and Nemopilema nomurai (AB243416). Aurelia aurita (Ulmaridae) has been used as an outgroup, as the order Semaeostomeae are now recognized to form the subclass Discomedusae with Rhizostomeae (Collins, Reference Collins2002; Dawson, Reference Dawson2004; Marques & Collins, Reference Marques and Collins2004; Collins et al., Reference Collins, Schuchert, Marques, Jankowski, Medina and Schierwater2006). Sequence data for A. aurita (sampled from Korea) were downloaded from GenBank (EF010537). Prior to further analyses, all sequence lengths were edited in Sequencher 4.9. A parsimony analysis was performed under direct optimization in the program POY 4.1.1 (Varón et al., Reference Varón, Vinh, Bomash and Wheeler2009) which simultaneously optimizes nucleotide homology and tree costs, thereby reducing the set of assumptions throughout the analysis. Bootstrap analyses (1500 pseudoreplicates) were performed to assess support of branch nodes. Mean pairwise sequence differences, using uncorrected ‘P’, distances were calculated in PAUP* 10.4b.
RESULTS
Photographs showing the subumbrellar view of a Crambionella medusa collected in the St Lucia Estuary illustrating anastomosing canal network connections to both the rhopalial and inter-rhopalial canals, after injecting coloured dye latex (A) and inter-rhopalial canals that appear to extend beyond the ring canal (B, C). On closer inspection more than one canal originated from ring canal section and was thinner than canals that preceded the ring canal. Inter-rhopalial canals were therefore accepted to terminate at the ring canal.
Non-metric multi-dimensional scaling (NMDS) of standardized morphometric data illustrating the morphological dissimilarity between Crambionella medusa collected in the St Lucia Estuary (grey) and C. orsini (black) examined at the Natural History Museum, London. Stress value is indicated.
A consensus tree of Rhizostomeae (original sequence data received from Professor M.N. Dawson), and outgroup, based on 474 nucleotides from cytochrome c oxidase subunit I (COI). Analysed by direct optimization in POY. Bootstrap values are indicated.
A summary of all Crambionella specimen measures from St Lucia Estuary and C. osini specimens examined from the Natural History Museum, London. Highlighted variables (bold typeface) are logged variables that were significantly correlated, using either Pearson's R (P < 0.0014, accepting Bonferonni correction) or Spearman's rank correlation (P < 0.02, accepting Bonferonni correction) test, with external bell diameter (S 1) in the case of the St Lucia specimens. Excluded measurements are S 1, S 2, S 29 and S 31 (indicated by †). Max, maximum; min, minimum.
Summary of all ratios (morphological features divided by external bell diameter) of Crambionella specimens from St Lucia Estuary and C. orsini specimens examined from the Natural History Museum (NHM), London. Highlighted (bold typeface, left-hand side column) are those logged ratios that were significantly correlated with external bell diameter (P < 0.05, accepting Bonferonni correction) in the case of the St Lucia material. Excluded standardized measurements included in correlation analyses S2 and S3 as well as all proportions and meristic data except S35 and S36 (indicated by †). Highlighted P-values (right-hand side column) are the standardized variables that differed significantly (P < 0.05, accepting Bonferonni correction) between C. orsini and the specimens of Crambionella from the St Lucia Estuary, as determined using two-tailed t-tests.
A character matrix highlighting morphological features that differ among the three Crambionella species. (Vanhöffen, Reference Vanhöffen1888; Chun, Reference Chun1896; Mayer, Reference Mayer1910; Stiasny, Reference Stiasny1922, Reference Stiasny1923, Reference Stiasny1937; Menon, Reference Menon1930, Reference Menon1936; Roa, 1931; Ranson, Reference Ranson1945; Nair, Reference Nair1946; Kramp, Reference Kramp1956, Reference Kramp1961, Reference Kramp1970) and the Crambionella material under investigation. Recorded geographical ranges are also given for all species.
Standardized morphometric data that contributed most to the dissimilarity between Crambionella material collected from the St Lucia Estuary and C. orsini specimens examined at the Natural History Museum, London, as determined by SIMPER analysis.
SYNONYMY
Crambessa stuhlmanni Chun, Reference Chun1896: 10, figure 1, pl. I, figure 1; Stiasny, Reference Stiasny1922: 50, figure 3.
Catostylus stuhlmanni Mayer, Reference Mayer1910: 669; Crambionella stuhlmanni Stiasny, Reference Stiasny1921: 129; Stiasny, Reference Stiasny1937: 237; Ranson, Reference Ranson1945: 319; Kramp, Reference Kramp1961: 374.
COMPARATIVE MATERIAL EXAMINED
Five preserved specimens of C. orsini were examined from the Natural History Museum, London, all collected on the ‘Murray’ expedition in 1933: 1950.3.25.343 (Station 76; 29 November 1933; 2 m tow net, 2800 m wire out); 1950.3.25.346 (Station 71; 26 November 1933; otter net, 106 m); 1950.3.25.347 (Station 70; 25 November 1933; otter net, 199 m); 1950.3.25.356 (Anchorage Muscat; 22 November 1933; Hand net, surface); 1950.3.25.357 (Station 75; 28 November 1933; otter net, 210 m).
DESCRIPTION
Umbrella between 62 and 181 mm in diameter (Table 3), finely granular, hemispherical or dome-shaped; margin cleft into narrow velar lappets, separated by deep furrows. Eight oral arms, each divided into a naked proximal (ratio to bell diameter: 0.08; Table 4) and a three-winged distal (ratio to bell diameter: 0.21; Table 4) portion, latter almost three times longer than former (mean: 2.78, SD: 0.86; Table 3); distal portion with one adoral and two aboral rows of mouthlets and club-shaped appendages, adoral row originating proximal to and terminating distad of the two aboral rows; terminating in a naked pyramidal club, proportion of terminal club length to oral arm length low (mean: 0.17, SD: 0.04; Table 3). In life, exumbrella transparent-white; oral arms transparent-white, bearing light-brown mouthlets and appendages; terminal clubs transparent-white; gonads cream. In preservation, they are all transparent-cream.
The canal system with a continuous ring canal, and four perradial and four inter-radial canals extending to umbrella margin and eight adradial canals terminating at the ring canal. Intra-circular network of anastomosing canals originating from ring canal (~7 connections with the ring canal; Table 3), no communication with the gastric cavity except occasionally through the perradial and inter-radial canals (0.22 connections with adjacent radial canals; Table 3 & Figure 3), less dense (~17 connection points within intra-circular network; Table 3) than that of the extra-circular network, which does not extend into lappets.
Eight rhopalia (range: 6–10; Table 3), situated in pits with radiating furrows, flanked laterally by ocular lappets that are smaller than, and slightly dorsal to, velar lappets. Twelve velar lappets per octant (range 4–29; Table 3), each with a row of small conical projections (mode: 12, range: 1–19; Table 3) mid-dorsally. There are approximately 84 (range 40–111; Table 3) annular muscles on subumbrellar surface. Four crescent shaped ostia lead from the gonadal and gastro-vascular cavity; ostial and inter-ostial widths approximately equal (mean: 0.61, SD: 0.16; Table 3). Gonads at the time of sampling were either immature and thin or mature and plump. Maturity in specimens was reached when external bell diameter reached ~100 mm. Of the 48 medusae examined 26 had developed gonads.
VARIATION
The majority of the meristic measurements taken were found to be size dependant (Table 3), although some were not. These features are highlighted in Table 3, as they can be useful in species-level comparisons. Adradial canals sometimes appear to extend beyond the ring canal but on closer inspection extensions are thinner, and seem to be more subdivided than those of the rhopalial canals (Figure 3).
REMARKS
A comparison of the key morphological and meristic features that can be used to distinguish the three recognized species of Crambionella, together with the appropriate data from this study are shown in Table 5. From this it can be seen that the number of velar lappets in each octant of the present material was similar to that of C. stuhlmanni (Chun, Reference Chun1896). The presence of conical projections on the dorsal median line of each lappet was also consistent with observations for C. stuhlmanni, and this feature can be used to distinguish the material from C. orsini (Menon, Reference Menon1930, Reference Menon1936; Stiasny, Reference Stiasny1937) but not from C. annandalei (Rao, Reference Rao1931; Stiasny, Reference Stiasny1937). However, the high ratio of terminal club length to the oral arm length as well as the ratio between distal winged portion and naked proximal portion of the oral arm separate C. annandalei (Menon, Reference Menon1930; Rao, Reference Rao1931) from the present material. Both C. annandalei and C. orsini possess accessory orbicular mouth appendages (Rao, Reference Rao1931; Menon, Reference Menon1936; Stiasny, Reference Stiasny1937; Kramp, Reference Kramp1961), which both the present material and C. stuhlmanni lack (Stiasny, Reference Stiasny1922; Kramp, Reference Kramp1961).
Although meristic differences (Table 3) between the present material and C. orsini are pronounced enough to allow ready separation, and generally agree with the literature (see Table 5) (Vanhöffen, Reference Vanhöffen1888; Chun, Reference Chun1896; Mayer, Reference Mayer1910; Stiasny, Reference Stiasny1922, Reference Stiasny1923, Reference Stiasny1937; Menon, Reference Menon1930, Reference Menon1936; Rao, Reference Rao1931; Ranson, Reference Ranson1945; Nair, Reference Nair1946; Kramp, Reference Kramp1956, Reference Kramp1961, Reference Kramp1970), there are also clear differences in some of the standardized morphometric data (Table 4). The results of the NMDS analysis (Figure 4) show that the two species are well separated, and even though the stress value is relatively high, the CAP was able to successfully categorize all specimens into the correct group (the permutation test results were significant: P < 0.001). For the canonical procedure a subset of three PCO axes were used based on the ‘leave-one-out’ diagnostics, which accounted for 66.74 % of the total variation in the species data. The first squared canonical correlation (δ12) was high: 0.56. Similar results were also obtained from the ANOSIM, where the Global R value of 0.67 was highly significant (P < 0.001). The variables contributing to the dissimilarities between species are highlighted in Table 6, foremost of which are the various characteristics of the oral arm.
For cytochrome c oxidase subunit I (COI) a maximum length of 660 nucleotides was amplified from three Crambionella specimens sampled in the St Lucia Estuary (GenBank accession numbers HM348770–HM348772) and compared to two C. orsini specimens (sequences downloaded from GenBank, accession numbers: EU363341 and EU363342). DNA sequence data from COI showed an average of 11.84% pairwise sequence difference between the material examined in this study and C. orsini. Dawson & Jacobs (Reference Dawson and Jacobs2001) suggest that differences of 10–20% between COI sequences set the standard for species level divergence. Phylogenetic analyses computed using COI sequence data demonstrate a monophyletic Crambionella clade (Figure 5). The consensus tree was supported by generally high bootstrap values, except at the branch that illustrated Catostylidae to be paraphyletic to the other rhizostome families represented. This is in contrast to previous molecular phylogenetic analyses executed on rhizostomes using COI (Daryanabard & Dawson, Reference Daryanabard and Dawson2008) and future work is needed to verify the findings in the present study. For internal transcribed spacer one (ITS1) a maximum length of 335 nucleotides was amplified from two Crambionella specimens sampled in the St Lucia Estuary (GenBank accession numbers HM348773 and HM348774); no comparative data for C. orsini were available.
Although on balance the material most closely resembles C. stuhlmanni, which is in agreement with its geographical distribution (Table 5), there was one feature at odds with previous descriptions. In the present specimens the intra-circular anastomosing canal network sometimes connected to both the rhopalial and inter-rhopalial canals (Figure 3), whilst in the original descriptions the anastomosing canals were only connected to rhopalial canals (Stiasny, Reference Stiasny1922). It is unlikely that these discrepancies reflect erroneous observations on the part of Stiasny; but rather it is probable that previous descriptions overlooked this rare feature due to small sample sizes examined. Scyphozoans often display considerable intra-specific morphological variation between geographically isolated or separated populations (Brewer, Reference Brewer1991; Bolton & Graham, Reference Bolton and Graham2004; Dawson, Reference Dawson2005a). Morphological variation is often as a result of phenotypic plasticity, a response to variable environmental conditions (Dawson et al., Reference Dawson, Martin and Penland2001; Bolton & Graham, Reference Bolton and Graham2004). Dawson (Reference Dawson2005b) highlights the importance of thorough geographical sampling, in combination with adequate sample sizes (as observed in this study), to get a more accurate representation of morphological variation.
Molecular analyses are increasingly being used in scyphozoan systematics (Dawson & Jacobs, Reference Dawson and Jacobs2001; Schroth et al., Reference Schroth, Jarms, Streit and Schierwater2002; Dawson, Reference Dawson2003, Reference Dawson2004, Reference Dawson2005a, b, c, d, e; Holland et al., Reference Holland, Dawson, Crow and Hofmann2004) and the decision about whether to use molecular or morphological analyses when describing species is subject to much debate (Dawson, Reference Dawson2005f). Molecular data increase the number of objective characters used, which enhances the likelihood of distinguishing taxa and permits phylogenetic reconstruction, free of impractical or inappropriate morphological features (Dawson, Reference Dawson2004). However, in some studies molecular analyses have failed to differentiate groups that showed significant morphological, behavioural and physiological differences (Dawson, Reference Dawson2005a). An approach which combines all data available is therefore required in scyphozoan systematics (Dawson, Reference Dawson2003, 2005f). Although this study did not utilize ecological or behavioural data, integrating molecular and morphological data is an important stepping stone for future work on this species.
ACKNOWLEDGEMENTS
We are grateful to Mr Martin Hendricks (UWC) for technical support and to staff at the Natural History Museum (London) for allowing us access to Crambionella orsini. We would like to thank Marine and Coastal Management and Ezemvelo KwaZulu Natal Wildlife for giving us permission to collect material from the St Lucia Estuary and to Mr A. Bali (Marine and Coastal Management, Cape Town) and Professor R. Perissinotto (University of KwaZulu Natal, Durban) for supplying the specimens. We are also grateful to Ms Annegret Bitzer and Professor Ralf Henkel (University of the Western Cape) for their translation of German papers. We acknowledge Professor Michael N. Dawson (University of Merced, USA) for supplying us with rhizostome sequence data. Particular thanks are also due to Professors André C. Morandini (University of São Paulo, Brazil) and Michael N. Dawson for their very useful comments on a previous draft of the manuscript, and to the anonymous referees who helped us to improve the text. This work was supported by the National Research Foundation (NRF grant number 61257); the NRF–Royal Society (London) SET Development grant in Zoology to the University of the Western Cape; and the Canon Collins Trust (S.N., grant number Nee 1500).
