An intensive, dedicated, large-scale systematic study into the diversity of South Africa's freshwater crab fauna was launched during the 1990s (Hart, Stewart & Bickerton 2001; Stewart 1997a, 1997b; Stewart, Coke & Cook 1995). Through this, seven new species were described (Daniels, Stewart & Burmeister 2001; Daniels, Stewart & Gibbons 1998a; Gouws, Stewart & Coke 2000; Gouws, Stewart & Reavell 2001; Stewart & Cook 1998; Stewart 1997a; Stewart et al. 1995) and one re-described (Stewart 1997b). At its conclusion, 13 species of Potamonautes MacLeay, 1838, the only genus of freshwater crabs occurring in South Africa, were recognised (Cumberlidge et al. 2009; Daniels, Phiri & Bayliss 2014; Gouws & Stewart 2001; Hart et al. 2001). More than 50% of these are endemic (Cumberlidge et al. 2009). Subsequently, ongoing research has identified and described four new species (e.g. Daniels et al. 2014; Peer et al. 2015; Phiri & Daniels 2014a) throughout the country.

Along with the Western Cape province of South Africa, the province of KwaZulu-Natal (KZN) was a substantial focus of this research programme. Previously, only two species were known to occur in KZN: Potamonautes sidneyi (Rathbun, 1904) and Potamonautes depressus (Krauss, 1843) (Gouws & Stewart 2001).

Potamonautes sidneyi is a widely distributed species, occurring across the eastern part of South Africa and extending northwards, at least into Zimbabwe (Gouws, Daniels & Stewart 2002). It is generally widespread across KZN, from the foothills of the Drakensberg Mountains through the lower-lying midlands to the coast (Gouws & Stewart 2001). Potamonautes depressus occurs in the upper, faster-flowing tributaries that drain the Drakensberg and its foothills (Gouws & Stewart 2001). A number of accounts of unexpected species or enigmatic records are known from the province, which likely reflect taxonomic uncertainty and/or the assignment of known (extralimital) species epithets to the then-undescribed species (see Gouws & Stewart 2001; Gouws et al. 2002).

With dedicated sampling and systematic investigation using morphometric and genetic approaches, three new species were described from KZN. Potamonautes dentatus Stewart, Coke & Cook, 1995 was described from sections of the Mngeni and Thukela rivers by Stewart et al. (1995). Gouws et al. (2000) described Potamonautes clarus Gouws, Stewart & Coke, 2000 from the headwaters of the Thukela River along the Drakensberg in the north-western part of the province. This species was delineated from P. depressus, which occurs along the Drakensberg to the south (Gouws & Stewart 2001). Daniels et al. (2003) later revealed P. clarus and P. depressus to be a complex of considerable cryptic diversity.

Potamonautes lividus Gouws, Stewart & Reavell, 2001, a morphologically distinct species associated with swamp forests dominated by Barringtonia racemosa and Ficus tricopoda, was described from a few localities in the north-eastern region of KZN (Gouws et al. 2001). More recently, Potamonautes isimangaliso Peer & Gouws, 2015 was described from the iSimangaliso Wetland Park in the north-eastern part of the province. This ecologically unique, burrowing species, occurring in ephemeral pans, was primarily delineated from the burrowing and morphologically similar P. lividus (Peer et al. 2015). Although the programme in KZN was underpinned by a substantial sampling effort, there were still notable sampling gaps (particularly in terms of specimens examined genetically), especially in the south and north-east of the province (Gouws & Stewart 2001). The iSimangaliso Wetland Park, an area of considerable biological, cultural and socio-economic importance, is found in the latter. This park is South Africa's first United Nations Educational, Scientific and Cultural Organization World Heritage Site and includes three Ramsar Wetlands of International Importance, namely the St Lucia estuarine lake, Lake Sibaya and Kosi Bay (Perissinotto, Stretch & Taylor 2013).

Although genetic approaches, specifically allozyme electrophoresis, were used to identify and delineate two of the species from the province (Gouws et al. 2000; Gouws et al. 2001), these studies were limited in their geographic or taxonomic focus (i.e. in terms of specific species compared to the putatively new species).

A wider geographic consideration of genetic diversity and structure amongst and within the species of KZN was attempted, again using allozymes (Gouws & Stewart 2001). However, it has been consistently recognised that allozyme variation in southern African potamonautid crabs is limited and that these markers may be too conservative to detect finer genetic diversity (Daniels 2003; Daniels, Gibbons & Stewart 1999; Daniels, Gouws & Crandall 2006; Daniels, Stewart & Gibbons 1998b; Daniels, Stewart, Ridgway & Florence 2001; Gouws & Stewart 2001; Gouws et al. 2001; Gouws et al. 2002).

Aside from studies of cryptic diversity and species delineation within the P. clarus – P. depressus complex (Daniels et al. 2003; Phiri & Daniels 2014b), DNA sequence data have generally not been used in studies of KZN crabs and there has been no published DNA-based survey of lineage diversity within the most widespread species in the province. The extent of divergence amongst mitochondrial DNA sequences generated for P. isimangaliso and published data for P. lividus (Daniels et al. 2002) provided initial evidence of the separation of these species (Peer et al. 2015). However, the phylogenetic relationships between these species, and between P. isimangaliso and other KZN species, were not considered.

The primary aim of this study is to provide a coarse-grained and broad-scale snapshot of mitochondrial DNA lineage diversity along the coastal zone in the KZN province for the most widespread species, P. sidneyi, and to consider aspects of its biogeography in the region.

Highlighting unique diversity within the province would be of interest to provincial conservation (Ezemvelo KZN Wildlife) and regional management (e.g. the iSimangaliso Wetlands Park Authority) agencies. Documenting undescribed diversity (novel species or unique genetic lineages) in the area around the iSimangaliso Wetland Park would be noteworthy and add to the conservation value of the park, considering the recent description of P. isimangaliso. This species is presently believed to be endemic to the park (Peer et al. 2015) and there is a growing number of cryptic taxa and unique lineages recently documented in this region (see Maake et al. 2013). The study also seeks to examine the relationship between P. isimangaliso, the morphologically similar P. lividus, and other species from the province and southern African region to support or refute its specific status.

Potamonautes sidneyi and P. isimangaliso were sampled between February 2012 and June 2013 from nine localities across KZN, South Africa (Figure 1), and were identified using published keys (Gouws & Stewart 2001; Hart et al. 2001; Peer et al. 2015). Four of these localities – Lake Sibaya (Site 1), Mpophomeni Stream (Site 2), Dukandlovu Pan (Site 3) and Hluhluwe (Site 4) (Figure 1) – were from the north-eastern part of KZN, within or in close proximity to the iSimangaliso Wetland Park. The remaining localities were distributed along a coastal transect extending to the south of the province.

Total genomic DNA was extracted from pereiopod muscle tissue from each specimen using a Thermo Fisher Scientific GeneJET Genomic DNA Purification Kit (Massachusetts, USA) following the manufacturer's protocols, except that extracted DNA was eluted in a final volume of 100 μL. A fragment of the mitochondrial cytochrome c oxidase subunit I (COI) gene was amplified by a polymerase chain reaction. Each 25 μL reaction contained 1 × buffer, 3 mM MgCl₂, 0.2 μM of each of Folmer et al.'s (1994) primers (LCOI-1490 and HCOI-2198), 0.2 mM of each deoxynucleotide, 0.5 U Taq-polymerase (Southern Cross Biotechnology, South Africa) and 3 μL template DNA, made up to the final volume with ultrapure water.

The thermocycling regime for the amplification of this gene fragment included an initial denaturing step of 2 min at 95 °C, followed by 35 cycles of denaturing (95 °C for 30 s), annealing (44 °C for 40 s) and extension (72 °C for 1 min). This was followed by a final extension step of 72 °C for 10 min. A fragment of the 16S ribosomal DNA (rDNA) was amplified from representative individuals (see below), using the primers (16Sar and 16Sbr) of Palumbi et al. (2007) and the above polymerase chain reaction recipe and thermocycling regime, but with annealing performed at 50 °C.

To confirm successful amplification, products were visualised on an ultraviolet transilluminator, following electrophoresis in 1% agarose gels stained with ethidium bromide. Amplicons were sent to a commercial sequencing facility (Macrogen Inc., South Korea), where they were purified, sequenced using standard BigDye v3.1 (Applied Biosystems, Austin, Texas, USA) terminator chemistry and analysed on an ABI 3730XL (Applied Biosystems) automated sequencer.

Sequences were checked against their respective chromatograms for ambiguities, misreads and sequencing errors using Chromas Lite (Technelysium Pty Ltd., South Brisbane, Australia). Alignment and further editing were performed using DNASTAR^® Lasergene SeqMan Pro 11 (Madison, Wisconsin, USA). The final alignments of the COI and 16S data sets were produced using ClustalX2 (Larkin et al. 2007) and MAFTT version 6 (Katoh & Toh 2008), respectively, with an iterative refinement strategy (L-INS-i) (Katoh et al. 2005) for the latter.

Data analyses proceeded in two stages. Firstly, to understand diversity within and amongst sampling localities of P. sidneyi and P. isimangaliso, and the diversity of lineages within P. sidneyi, analyses were conducted with the full set of COI sequences generated. To examine genealogical and geographic relationships, a 95%-probability parsimony network was generated using TCS1.21 software (Clement, Posada & Crandall 2000). A midpoint-rooted neighbour-joining tree (Saitou & Nei 1987), based on uncorrected sequence divergences, was generated in PAUP*4b10 (Phylogenetic Analysis Using Parsimony) software (Swofford 2002). Uncorrected sequence divergences within and amongst identified lineages were examined using PAUP.

In a second set of analyses, the phylogenetic placement of, and divergences amongst, the above lineages were contextualised and an attempt was made to confirm or assign species identifications to each. Representatives of the unique lineages identified were included in a matrix containing published data, sourced from GenBank^® (Table 1), of other southern and eastern African Potamonautes species. Data from the COI and 16S fragments were analysed as a combined data set and phylogenetic analyses were conducted by unweighted parsimony and Bayesian inference. For the parsimony analysis, a heuristic tree search was conducted in PAUP*4b10, using tree bisection and reconnection branch swapping of a tree obtained by a random stepwise addition of taxa, employing 1000 such iterations. Nodal support was evaluated by bootstrapping of the data set (Felsenstein 1985), using 1000 pseudoreplicates.

Prior to the Bayesian analysis, the optimal model of nucleotide substitution for each partition (COI and 16S rDNA) of the combined data set was identified using jModelTest 2.1.4 (Darriba et al. 2012) with the choice of model determined by the Akaike (1974) Information Criterion. Bayesian inference was conducted in MrBayes 3.1.2 (Ronquist & Huelsenbeck 2003). Four simultaneous, independent analyses, each employing three heated and one cold Markov chains, were run for 10⁷ generations, sampling every 2000 generations. The chosen models were specified for each partition, with Mr Bayes estimating the model parameters. Default, unlinked priors were used. The standard deviation of split frequencies was observed to monitor convergence of the analyses on a similar part of the posterior distribution. Stationarity and effective sample size (ESS > 200; such that sampling of the various parameters from the posterior distribution was sufficient) were determined using Tracer 1.5 software (Rambaut & Drummond 2009). A majority rule consensus tree was constructed from the 18 000 trees retained from the four analyses combined, once a 10% burn-in was discarded in each; the frequency of retrieval of particular nodes representing the Bayesian Posterior Probabilities (BPPs) for those relationships.

Liberonautes rubigimanus Cumberlidge & Sachs, 1989 and Sudanonautes floweri (De Man, 1901) were included as outgroups in these analyses. Previous multi-locus phylogenies (Daniels, Cumberlidge, Pérez-Losada, Marijnissen & Crandall 2006; Daniels et al. 2015) revealed Liberonautes Bott, 1955, and a Sudanonautes Bott, 1955 and Potamonemus clade to be sister taxa to the Potamonautes A. Milne-Edwards, 1887 clade that nests both Erimetopus Rathbun, 1894 and Platythelphusa A. Milne-Edwards, 1887. Representatives of Liberonautes and Sudanonautes have been used as outgroups in other phylogenetic studies of Potamonautes (Daniels & Bayliss 2012; Daniels et al. 2014; Phiri & Daniels 2013). Uncorrected sequence divergences amongst known species and the representatives of the respective lineages were calculated in PAUP.

Specimens of P. isimangaliso were sampled only from Dukandlovu Pan (Figure 1, Site 4), whilst specimens from all the other localities were positively identified as P. sidneyi.

All COI sequences were submitted to GenBank^® under accession numbers KT275869 to KT275936. The full COI alignment of all 69 specimens was 639 nucleotides in length. Upon translation to amino acids, there was no evidence of stop codons in the alignment and it was concluded that the data represented functional proteins and were of mitochondrial origin.

One haplotype was found amongst the four P. isimangaliso individuals from Dukandlovu Pan, including the individual sequenced by Peer et al. (2015). Nineteen unique haplotypes were identified within P. sidneyi (Figure 2). The distribution of haplotypes within this species indicated substantial genetic structuring amongst sampling localities. Eight individuals possessed unique haplotypes, while nine haplotypes were found in multiple individuals from the same locality. The remaining two haplotypes were shared amongst sampling localities and only amongst two localities in each instance: amongst the proximate Lake Sibaya (Site 1) and Mpophomeni Stream (Site 2) localities, and amongst the geographically more distant Hluhluwe (Site 3) and Entumeni (Site 5) localities, respectively.

The 20 haplotypes observed in this study were placed in 8 independent networks that could not be linked with 95% confidence. Five networks (Figure 2, Networks C, E, F, G and H), all but one consisting of single haplotypes, represented single localities. Two larger networks (Networks A and D) and one consisting of a single haplotype (Network B), found in both Lake Sibaya (Site 1) and Mpophomeni Stream (Site 2), represented multiple localities. Haplotypes of Network A were found in four localities from the north-eastern to central parts of the province: Mpophomeni Stream site (Site 2), Hluhluwe (Site 3), Entumeni (Site 5) and Siyayi (Site 6). Those of Network D were found in the localities to the south of KZN: Mhlanga (Site 7) and Oribi Gorge (Site 8). Several localities possessed haplotypes of more than one network: Entumeni (Site 5), Mpophomeni Stream (Site 2), Mhlanga (Site 7) and Mtamvuna (Site 9).

Uncorrected sequence divergences among haplotypes belonging to different networks ranged from 2.8% to 14.7% (with a mean of 10.7%). The two haplotypes of Network E were 0.2% divergent. Network A showed the greatest diversity (seven haplotypes), with divergences amongst the constituent haplotypes ranging from 0.2% to 1.3% (with a mean of 0.6%). Greater divergences were observed within Network D (from 0.6% to 2.5%; mean 1.5%).

The diversity and the geographic structure observed were substantiated by the neighbour-joining tree (Figure 2). Two distinct clusters were retrieved, with a single sample from Entumeni (NPP47) placed outside of these. The first cluster, corresponding to Networks A and B, contained samples identified as P. sidneyi from the north-eastern and central parts of the province: Lake Sibaya (Site 1), Mpophomeni Stream (Site 2), Hluhluwe (Site 3), Entumeni (Site 5) and Siyaya (Site 6).

The second cluster contained individuals identified as P. sidneyi, as well as the P. isimangaliso specimens from Dukandlovu Pan (Site 4). These formed separate sub-clusters and corresponded to separate networks (Networks D to G for P. sidneyi and Network H for P. isimangaliso). The P. sidneyi individuals in this cluster were from the three sites towards the south of the province: Mhlanga (Site 7), Oribi Gorge (Site 8) and Mtamvuna (Site 9).

For the phylogenetic analyses, a 16S rDNA fragment was sequenced for a single representative of each of seven networks (Networks A–C, and E–H) and three representatives of Network D (representing the three divergent groups of haplotypes in this network). These 16S rDNA sequences were lodged under GenBank^® accession numbers KT275860 to KT275868. This fragment was analysed in combination with the COI data for each of these specimens and data downloaded from GenBank^®.

Once sequences were trimmed to equal length, the COI and 16S alignments were 362 and 401 nucleotides in length, respectively, yielding a combined alignment of 763 nucleotides. The parsimony analysis, based on 207 parsimony informative characters (97 from the COI data and 110 from the 16S rDNA), yielded a single most parsimonious tree (Figure 3a: 1106 steps; consistency index = 0.326; retention index = 0.543; rescaled consistency index = 0.177).

Hasegawa, Kishino and Yano’s (1985) HKY model, with unequal base frequencies (A = 0.332, C = 0.206, G = 0.077 and T = 0.385), a transition to transversion rate of 4.826, a proportion of invariable sites (I = 0.487) and a gamma-distribution (α = 0.707) of rate variation was determined to be the most appropriate for the COI partition. The model chosen for the 16S partition included unequal base frequencies (A = 0.394, C = 0.078, G = 0.136 and T = 0.393), mostly independent transition and transversion rates (R_[A↔_C] = 0.364, R_[A↔_G] = 7.317, R_[A↔_T] = R_[G↔_T] = 1.000, R_[C↔_G] = 0.675 and R_[C↔_T] = 2.893), a proportion of invariable sites (I = 0.169) and a gamma-distribution (α = 0.443) of rate variation.

The most likely topology (-lnL = 7841.114) obtained through the Markov chain Monte Carlo searches across the four Bayesian analyses is presented in Figure 3b. Both phylogenetic trees showed certain similarities to published phylogenies (Daniels & Bayliss 2012; Daniels, Cumberlidge, Pérez-Losada, Marijnissen & Crandall 2006; Daniels et al. 2014; Phiri & Daniels 2013) relating to the species represented in the major clades retrieved, but many of the relationships within and among these clades were not consistent. Nonetheless, it was possible to resolve, in either of the two analyses, the placement of representatives of the unique lineages with high support.

Like the neighbour-joining tree of the COI data, the representatives of P. sidneyi were placed in two distinct and well-separated clades in both analyses. The representatives of the lineages from the north-eastern and central parts of the province (NPP65 Hluhluwe, representing Network A; and NPP70 Lake Sibaya, representing Network B) formed a clade with the published P. sidneyi sequence with high support (95% bootstrap; 1.00 BPP). NPP47 Entumeni (Network C) was well-supported as a sister-taxon to this group (0.99 BPP) in the Bayesian analysis. Although this specimen was not associated with this clade in the parsimony analysis, the alternative placement was not supported (Figure 3a). Regardless, these specimens were all placed in a larger clade containing a number of large-bodied riverine species (sensu Daniels et al. 2002; Potamonautes barbarai Phiri & Daniels, 2014, Potamonautes barnardi Phiri & Daniels, 2014, Potamonautes bayonianus [Brito-Capello, 1864], Potamonautes granularis Daniels, Stewart & Gibbons, 1998, Potamonautes perlatus [H. Milne-Edwards, 1837], P. sidneyi, Potamonautes unispinus Stewart & Cook, 1998 and Potamonautes warreni [Calman, 1918]) with strong support (98%; 1.00 BPP).

The second group of individuals identified as P. sidneyi (NPP13 Oribi Gorge, NPP15 Oribi Gorge and NPP21 Mhlanga, representing Network D, and the representatives of Networks E, F and G) from the south of the province were placed in a clade containing two published P. lividus sequences and the P. isimangaliso (Network H) representative, with 1.00 BPP. The P. isimangaliso specimen was consistently the sister-taxon to the sequence of P. lividus from South Africa, to the exclusion of the representative from Swaziland.

Sequence divergences amongst all included, recognised species and representatives of the lineages are presented in the Appendix, Table 1. Divergences amongst known species ranged from 2.3% (P. barbarai versus P. granularis) to 21.5% (Potamonautes calcaratus [Gordon, 1929] versus P. lividus from South Africa), with a mean of 13.5%. Those specimens of P. sidneyi from the north-eastern and central parts of the province, placed in a clade with P. sidneyi and the large-bodied riverine species, were 1.8% – 4.8% divergent from the published P. sidneyi sequence. These individuals (and P. sidneyi itself) were, in turn, 9.2% – 11.8% divergent from the representatives of the second clade of P. sidneyi individuals from the south of the province.

The divergences amongst these two groups were substantially higher than the divergences within (0.4% – 4.8%). This southern clade showed a marginally closer relationship to P. lividus and P. isimangaliso (6.7% – 10.0% divergence) than to the north-eastern P. sidneyi clade. Considering the combined COI and 16S rDNA data, the P. isimangaliso representative was 7.4% and 7.8% divergent from the Swaziland and South African representatives of P. lividus, respectively.

The analysis of the COI data reveals substantial genetic diversity and structure within P. sidneyi in KZN; 7 divergent, unlinked networks were retrieved (with an additional network representing P. isimangaliso) and only 2 of 19 haplotypes are shared amongst sampling localities.

For the most part, the data show a clear isolation and lack of gene flow amongst most sampling localities, even amongst those that are geographically proximate. This is contrary to expectations for a species that is assumed to be highly dispersive (Daniels 2003; Daniels et al. 1998a, 1998b; Daniels et al. 1999; Daniels, Gouws & Crandall 2006; Daniels, Stewart, Ridgway & Florence 2001) and follows more closely patterns observed in certain primary, obligate freshwater fishes from South Africa, for example Galaxias (Chakona, Swartz & Gouws 2013). The patterns observed reflect not only the isolation of individual drainages, but often the isolation of localities within the same system (e.g. amongst Hluhluwe and Mpophomeni streams, which flow into Lake St Lucia).

In the similarly widespread Potamonautes perlatus sensu lato, extensive sharing of haplotypes was observed amongst localities, but these were within the same catchments (Daniels, Gouws & Crandall 2006; Phiri & Daniels 2014a). Very few haplotypes were shared amongst currently isolated systems, much like the present case. The isolation of the individual localities and systems is perhaps facilitated by the hydrology of KZN. In the subtropical to tropical north-east of the province, drainages are generally mature, widely separated and flow over low relief floodplains (Skelton 2001). Geographic distance may prohibit dispersal amongst systems, despite the area being flood-prone and the presence of ephemeral wetlands (Skelton 2001), which should facilitate dispersal. Towards the south of the province, in the montane escarpment region, a dense network of shorter, high-gradient, deeply-incised catchments fragment the landscape (Skelton 2001). Dispersal may be hindered across higher drainage divides.

Two divergent phylogenetic lineages were found in P. sidneyi, with one of these showing a possible closer affinity to P. lividus and P. isimangaliso. These two lineages were restricted to the north-eastern and central, and southern parts of the province, respectively. The distribution of these lineages and the division amongst them reflects the two aquatic ecoregions mentioned above, as well as the separation of the Sibayi and Zululand from the Mngeni and Mzimkhulu aquatic bioregions (see Rivers-Moore, Goodman & Nkosi 2007); these regions defined by ecology, species diversity and endemism.

Previous studies considering the genetic diversity or structure in P. sidneyi in KZN provided conflicting patterns. An allozyme study, delineating P. lividus from P. sidneyi, found low levels of diversity and only shallow structure in the latter (Gouws et al. 2001). However, localities from which P. sidneyi was sampled were from the north-eastern region of KZN and may represent only one of the lineages identified in the current study. With slightly wider sampling, a study of morphometric and allozyme variation within and amongst all Potamonautes species in KZN provided some evidence of genetic structure and differentiation within P. sidneyi (Gouws & Stewart 2001).

Two distinct genetic clusters were identified: one from the central and north-east of the province and the other from the high-lying Drakensberg in the extreme west (Gouws & Stewart 2001). Whether these correspond to the two lineages of the present study or represent additional lineage diversity remains to be determined as no P. sidneyi populations from the Drakensberg were included in this study. Nonetheless, the genetic differentiation of P. sidneyi from the Drakensberg was confirmed in a subsequent study (Gouws et al. 2002).

These populations form a distinct genetic cluster separate to samples from elsewhere in the province; the latter showed closer genetic similarity to P. sidneyi samples from other provinces, that is, Mpumalanga, Eastern Cape and Limpopo (Gouws et al. 2002). Interestingly, the latter study showed genetic differentiation amongst populations of P. sidneyi from different reaches of the same river system, the Thukela. This reflects a longitudinal pattern of differentiation involving the phylogenetic separation (see Daniels et al. 2002) of high-altitude, small-bodied species (members of the P. depressus – P. clarus complex in this system) in the upper reaches from the robust species of the middle to lower reaches (P. sidneyi) and then possible longitudinal genetic structure within the latter.

The co-occurrence of multiple, genetically independent networks at single localities, particularly in the south of the province where drainage is steep and with potential for isolation along the river course, attests to the diversity in the species and may perhaps reflect such longitudinal or altitudinal genetic structure. As the present study focused on sampling in the coastal zone only, this requires further investigation in the Thukela River and these southern systems. Nonetheless, P. sidneyi clearly shows a complex pattern of lineage diversity and structure within KZN and beyond. This is currently being investigated through a much broader phylogeographic study of this species across its full South African distribution.

The taxonomic status of the two P. sidneyi lineages requires consideration. The sequence divergence amongst these lineages (9.2% – 11.8%) was more extensive than was observed amongst other species (P. barbarai, P. barnardi, P. bayonianus, P. granularis, P. perlatus, P. unispinus and P. warreni) in the group of large-bodied, riverine species from South Africa, and amongst these species and the reference P. sidneyi individual (2.3% – 8.3%).

Furthermore, the higher values from individual comparisons amongst the two lineages approached the mean divergence amongst all valid species included in the study. Potamonautes sidneyi in KZN shows comparable diversity to P. perlatus (see Phiri & Daniels 2014a), which, along with P. sidneyi, is one of the most widely distributed South African freshwater crab species (Daniels 2003; Daniels, Gouws & Crandall 2006; Gouws et al. 2002; Phiri & Daniels 2014a). However, a deeper divergence was observed between lineages of P. sidneyi in KZN than was observed amongst the lineages of P. perlatus. Comparisons amongst representatives of the two P. sidneyi lineages yielded sequence divergences of 6.2% – 9.3% for the 16S data, and 12.2% – 14.3% for COI, which exceeded the maximum values of 6% and 10% for 16S and COI, respectively, in P. perlatus (Daniels 2003; Daniels, Gouws & Crandall 2006).

Two new species (P. barbarai and P. barnardi) were delineated, using genetic evidence alone, amongst the three lineages in P. perlatus sensu lato (Phiri & Daniels 2014a). The extent of divergence, the phylogenetic placement of lineages identified as P. sidneyi in the present phylogenetic analyses, results from past analyses (Gouws & Stewart 2001; Gouws et al. 2002) and the precedent set by Phiri and Daniels (2014a) suggests that there may be additional species-level diversity with P. sidneyi, requiring recognition.

The identification and description of new species within P. sidneyi is deferred at present, for two reasons. The first is the uncertainty over which lineage or clade represents the ‘true’ P. sidneyi. Rathbun (1905), when describing the species, indicated the type locality as being ‘Natal’, without providing a more specific locality within the province. The species were described from material collected by Sarah Abraham in 1871 (Rathbun 1905) and archival research may be required to determine the type locality more accurately. Secondly, the wider phylogeographic study and a more rigorous consideration of morphological variation across the species’ range would provide greater insight into the taxonomic status of these and any other lineages.

While P. isimangaliso shows a clear resemblance to P. lividus (Peer et al. 2015), the lack of the distinctive colouration of the latter (Gouws & Stewart 2001; Gouws et al. 2001) aided the identification of P. isimangaliso specimens in the current study. However, recent collections of P. lividus (Daniels et al. 2014) suggest that colour may be variable in this species (S.R. Daniels, pers. comm.).

Phylogenetically, the representative P. isimangaliso specimen was associated with P. lividus from South Africa (KZN), with the representative from Swaziland sister to these (or even placed outside of a clade formed by these and one of the P. sidneyi lineages). The phylogenetic placement of P. isimangaliso suggests that it may be conspecific to P. lividus from KZN. However, the P. isimangaliso specimen was substantially (~ 7.5%) and roughly equally divergent from each of the P. lividus representatives.

Recently, Daniels et al. (2014) reported the occurrence of P. lividus from Dweza Forest in the Eastern Cape province. Together with the specimens from Swaziland (see Daniels & Bayliss 2012), this represents a significant range extension for the species. However, unpublished sequence data revealed almost no differentiation among P. lividus from KZN and from the Eastern Cape (Daniels et al. 2014). This suggests a genetic cohesion within P. lividus over a range of some ~ 750 km (Daniels et al. 2014) and indicates that both P. lividus from Swaziland and P. isimangaliso represent unique lineages, both warranting recognition at species level.

The description by Peer et al. (2015) and support for P. isimangaliso as a valid taxon are significant because Gouws et al. (2001) highlighted a number of specimens in museum collections that resembled P. lividus. However, identifications of these specimens could not be confirmed at the time as the seemingly diagnostic colouration (see above) was not preserved. Repeated sampling from these localities, which were not typical swamp forests, was unsuccessful. The possibility that these morphotypes may correspond to P. isimangaliso, P. lividus or additional taxa within a possible complex require further investigation.

The present study highlights the unique diversity within the north-eastern region of KZN, in and around the iSimangaliso Wetland Park. A unique lineage of P. sidneyi is found here, which is separate to a lineage occurring to the south of the province and possibly an additional lineage occurring along the Drakensberg to the west (Gouws & Stewart 2001).

Whether this north-eastern lineage is restricted to this area or has a more extensive distribution throughout the range of P. sidneyi remains to be determined. Allozyme invariance among populations from this area and from further afield, including other provinces, indicates that this may be the case (Gouws et al. 2002). However, the conservative nature of these markers in potamonautid crabs (see Introduction) may obscure true lineage diversity.

Potamonautes isimangaliso is currently endemic to the iSimangaliso Wetland Park (Peer et al. 2015), but possibly occurs at other localities in the north-east (Gouws et al. 2001), and clearly represents a unique species within the P. lividus – P. isimangaliso complex. These findings suggest a somewhat unique potamonautid fauna within this region, adding to the conservation value of this ecologically-, culturally- and socio-economically important system. The iSimangaliso Wetland Park was proclaimed a United Nations Educational, Scientific and Cultural Organization World Heritage Site in 1999 for three outstanding universal values, one of which is biodiversity (Perissinotto et al. 2013). The findings of this study confirm that the park truly represents a hotspot of biological diversity.