Phylogeography of the Cape velvet worm (Onychophora: Peripatopsis capensis) reveals the impact of Pliocene/Pleistocene climatic oscillations on Afromontane forest in the Western Cape, South Africa


Savel R. Daniels, Department of Botany and Zoology, University of Stellenbosch, Private Bag X 1, Matieland 7602, South Africa. Tel.: +27 21 8083230; fax: +27 21 8082405; e-mail:


Habitat specialists such as soft-bodied invertebrates characterized by low dispersal capability and sensitivity to dehydration can be employed to examine biome histories. In this study, the Cape velvet worm (Peripatopsis capensis) was used to examine the impacts of climatic oscillations on historical Afromontane forest in the Western Cape, South Africa. Divergence time estimates suggest that the P. capensis species complex diverged during the Pliocene epoch. This period was characterized by dramatic climatic and topographical change. Subsequently, forest expansion and contraction cycles led to diversification within P. capensis. Increased levels of genetic differentiation were observed along a west-to-south-easterly trajectory because the south-eastern parts of the Cape Fold Mountain chain harbour larger, more stable fragments of forest patches, have more pronounced habitat heterogeneity and have historically received higher levels of rainfall. These results suggest the presence of three putative species within P. capensis, which are geographically discreet and genetically distinct.


Climatic ameliorations are one of the major abiotic driving forces of cladogenesis in plants and animals (Goldblatt & Manning, 2002; Linder, 2003; Daniels et al., 2006; Gouws et al., 2010; Schnitzler et al., 2011). Historically, these climatic oscillations have resulted in significant habitat fragmentation causing temporal and spatial shifts in species distribution patterns, promoting isolation and speciation (Hewitt, 2003). Whereas the impact of historical climatic changes is well documented and understood in the northern hemisphere, limited comparative studies have been conducted in the southern hemisphere. Forest biome biogeography appears particularly neglected, and evolutionary affinities among them remain poorly understood in southern Africa. In southern Africa and more specifically the south-western Cape (or Cape Floristic Region), the early Miocene was characterized by the presence of subtropical rainforest (Coetzee, 1993). A dramatic climate shift from warm, tropical conditions to drier, more seasonal conditions occurred during the Pliocene and prompted the development of drought-resistant vegetation (Linder, 2003; Cowling et al., 2009). This climatic shift was attributed to the development of the Benguela upwelling system along the west coast of southern Africa during late Miocene (Siesser, 1980). Furthermore, this was a period when the Cape Floristic Region (CFR) underwent a phase of geomorphic evolution with tectonic uplift (Partridge & Maud, 2000). Dramatic geotectonic uplift (600–900 m) in eastern southern Africa increased the east–west rainfall gradient, reinforcing western CFR aridity by intercepting a greater proportion of rainfall from the warm Agulhas current (Tyson & Partridge, 2000). This effectively divided the CFR into a western ‘winter rainfall zone’ (WRZ) and an eastern ‘year-round rainfall zone’ (YRZ). In the western CFR, seasonal drought provided an environment conducive to regular lightning-induced fires. The CFR fire regime is suggested to have become established < 6–8 million years ago, during the late Miocene (Bytebier et al., 2011). This promoted the evolution of the pyrophytic Cape fynbos vegetation (Mucina & Rutherford, 2006; Cowling et al., 2009; Swart et al., 2009). Aridification became increasingly more marked during the Pliocene/Pleistocene epoch with moderate marine transgression and regression. These developments resulted in considerable habitat fragmentation among forested areas, restricting forest habitats to high-lying areas.

Within the CFR, the Afromontane forests represent the smallest biome and are comprised of two subtypes, namely southern Afrotemperate and southern coastal forest (Castley & Kerley, 1996; Mucina & Rutherford, 2006). These forest patches generally occur below 1000 m, in areas where rainfall exceeds 600 mm (Rutherford & Westfall, 1986; Mucina & Rutherford, 2006). The impacts of historical climatic amelioration on forested areas remain largely unexplored in the absence of phylogeographic studies on forest-dwelling taxa. However, it would be reasonable to assume that the increased aridification experienced during the Miocene/Pliocene has fragmented forest habitats. It has been demonstrated that habitat specialists such as soft-bodied invertebrates (e.g. land planarians and springtails) can be effectively employed to reconstruct the biogeographic patterning of forested areas (Garrick et al., 2007; Carnaval et al., 2009; Álvarez-Presas et al., 2011). Their dependence on stable microenvironments and highly restricted dispersal ability make them suitable for reconstructing biome affinities (Garrick et al., 2007; Álvarez-Presas et al., 2011). Onychophora, commonly known as velvet worms, are habitat specialists, restricted to moist environments like closed-canopy forests where they typically inhabit saproxylic environments (Hamer et al., 1997; Daniels & Ruhberg, 2010). Their habitat specificity and general physiological intolerance of desiccation coupled with their low dispersal capabilities render these organisms ideal to test the contractions and expansions of their habitats (Hamer et al., 1997; Álvarez-Presas et al., 2011). The Cape velvet worm, Peripatopsis capensis Grube, 1866, has a relatively wide distribution in the Western Cape of South Africa where conspecific populations are confined to discontinuous, Afromontane forest areas and adjacent fynbos along the Cape Fold Mountains (Brinck, 1957; Hamer et al., 1997). This taxon provides the ideal template organism with which to explore the impact of climate and topography on Afromontane forest in the Western Cape. Daniels et al. (2009) demonstrated that P. capensis comprises three distinct clades corresponding to three distinct biogeographical regions in the Western Cape. The latter study further suggested a historical link between the south-western Cape and the Cape Peninsula forests based on the close phylogenetic relationships of the respective sampling localities; however, limited geographical coverage of the species distribution precluded biogeographic inferences.

In the present study, P. capensis was extensively sampled from Afromontane forests throughout the Western Cape Province of South Africa. The following three hypotheses are explored: (i) due to its habitat specificity, P. capensis will display a genetic history that mirrors the palaeogeography of CFR forests in the light of climatic and geological perturbations, (ii) populations of P. capensis are isolated due to the inhospitality of low-lying coastal plains and the absence of low-lying forests and (iii) increased levels of genetic differentiation should be observed along a west-to-south-easterly trajectory because the south-eastern parts of the Cape Fold Mountain chain harbour larger fragments of forest patches and more pronounced habitat heterogeneity and have historically received higher levels of rainfall.

Materials and methods

Taxon sampling

A total of 104 Peripatopsis capensis specimens were collected from 21 localities in the western and south-western Cape regions of the Western Cape Province of South Africa (Fig. 1; Table 1). Sample sizes ranged between a minimum of one to a maximum of ten. Samples were hand-collected from forest understorey including rotten logs, leaf litter, moss and beneath ravine rocks. Locality coordinates were recorded with a hand-held global positioning system (Garmin-Trek Summit). Samples were killed by submergence in absolute ethanol. Specimens were preserved in absolute ethanol and stored at 4 °C in a refrigerator. Samples were identified using the dichotomous key provided by Sherbon & Walker (2004).

Figure 1.

 Map of the Western Cape (WC), South Africa (SA), showing the localities where Peripatopsis capensis was sampled. Sample localities: 1, Myburgh Ravine; 2, Orangekloof; 3, Cecelia Forest; 4, Skeleton Gorge; 5, Newland’s Ravine; 6, Rhode’s Memorial; 7, Rondevlei Nature Reserve; 8, Jonkershoek; 9, Bergriver Dam, Franschoek; 10, High Noon; 11, Dappat se Gat; 12, Kogelberg Biosphere Reserve; 13, Fernkloof Nature Reserve; 14, Caledon; 15, Grootbos Private Reserve; 16, Napier; 17, Greyton; 18, Oubos; 19, De Hoop Nature Reserve; 20, Marloth Nature Reserve; and 21, Grootvadersbosch Nature Reserve.

Table 1.   Samples of Peripatopsis capensis collected from the Western Cape, South Africa, during the present study and by Daniels et al. (2009). Pop N corresponds to the population number as indicated on the map (Fig. 1).
Pop NSample localityn*nCoordinates
  1. *Present study.

  2. Daniels et al. (2009)

 1Myburgh Ravine 4 –34°00′ 19.28″S 18°22′ 41.96″E
 2Orangekloof 4 –34°00′ 15.69″S 18°22′ 58.64″E
 3Cecelia Forest 4 333°59′ 75.00″S 18°24′ 99.00″E
 4Skeleton Gorge 3 133°58′ 48.00″S 18°25′ 12.00″E
 5Newland’s Ravine – 333°57′ 15.50″S 18°26′ 41.03″E
 6Rhode’s Memorial 6 –33°57′ 09.73″S 18°27′ 04.66″E
 7Rondevlei Nature Reserve 1 –34°03′ 37.50″S 18°29′ 59.76″E
 8Jonkershoek 4 –33°58′ 05.76″S 18°56′ 24.65″E
 9Berg River Dam 4 –33°54′ 25.14″S 19°03′ 04.48″E
10High Noon – 433°54′ 29.23″S 19° 16′ 56.05″E
11Dappat se gat 5 –34° 13′ 25.84″S 18°50′ 24.16″E
12Kogelberg Biosphere Reserve 2 –34°19′ 58.51″S 18°57′ 05.34″E
13Fernkloof Nature Reserve 4 134°23′ 37.00″S 19°16′ 34.00″E
14Caledon 1 –34°12′ 43.00″S 19°30′ 23.00″E
15Grootbos Private Reserve 4 534°55′ 05.00″S 19°41′ 37.00″E
16Napier 6 –34°29′ 23.72″S 19°43′ 20.25″E
17Greyton – 334°02′ 01.52″S 19°36′ 57.32″E
18Oubos 4 –34°04′ 34.33″S 19°49′ 43.76″E
19Potberg-De Hoop Nature Reserve 5 534°17′ 35.00″S 20°29′ 19.00″E
20Marloth Nature Reserve 7 333°59′ 51.84″S 20°27′ 27.68″E
21Grootvadersbosch Nature Reserve 3 533°58′ 55.00″S 20°49′ 23.00″E
 Total number of specimens7133 

DNA sequencing

DNA was extracted from tissue samples using a Qiagen DNEasy kit (QIAGEN, Austin, Texas, USA), following the manufacturer’s protocol. Extracted DNA was stored in a refrigerator until required for PCR. Prior to use, 1 μL DNA was diluted in 19 μL water. Two partial gene fragments were selected for this study. These were the mtDNA cytochrome c oxidase one subunit (COI) and the 18S rRNA nuclear DNA gene locus. Selection of these loci was based on the fact that they exhibit varying mutational rates, represent both the mtDNA and nDNA genomes, respectively, and have been used with great success to reconstruct evolutionary relationships among velvet worms (Allwood et al., 2010; Gleeson et al., 1998; Trewick, 2000; Daniels et al., 2009, Daniels & Rudberg, 2010). The following primer pairs were used: Folmer et al. (1994) LCOI-1490 and HCOI-2198 for a partial fragment of the COI locus and the primer pair 5F and 7R for 18S after the study of Giribet et al. (1996). Polymerase chain reactions (PCRs) were performed on a GeneAmp PCR System 2700 Thermocycler (Applied Biosystems, Foster City, CA, USA). For COI, standard PCR conditions were employed using 1–3 μL template DNA (Daniels et al., 2006, 2009; Daniels & Ruhberg, 2010). Standard protocols were used to cycle sequence-purified PCR products. Sequencing was performed on an ABI 3730 XL automated machine. All specimens were sequenced for COI, whereas for 18S, a single representative specimen for each locality was sequenced.

Phylogenetic analyses

Sequence Navigator (Applied Biosystems) was used to check for base ambiguities and compute a consensus sequence from forward and reverse strands. Sequence alignment was performed in CLUSTAL X (Thompson et al., 1997). A haplotype network was constructed for the COI data set using TCS 1.21 (Clement et al., 2000). A phylogeny for the COI haplotypes was estimated using a Bayesian inference approach. mrbayes 3.0b4 (Ronquist & Huelsenbeck, 2003) was used to investigate optimal tree space using Bayesian inferences for the COI haplotype data set. Akaike information criteria (AIC) (Akaike, 1973) were used to determine the best-fit maximum-likelihood score. modeltest (Posada & Crandall, 1998) was used to determine the best-fit substitution model for each gene locus for the Bayesian analysis. For each analysis, ten Markov chain Monte Carlo (MCMC) simulations were run, starting from a random tree for five million generations, sampling from every 1000th tree. A 50% majority rule consensus tree was generated from the trees retained. After burnin, trees were discarded. Posterior probabilities (pP) for each node were estimated by the percentage of time the node was recovered. Posterior probability values < 0.95 were regarded as poorly resolved. Similarly, a Bayesian analysis was conducted on the reduced, combined mtDNA and nDNA (COI and 18S) data matrix again using a single representative sample per locality and taxon. In addition, an ML analysis was conducted on the reduced, combined mtDNA and nDNA (COI and 18S) data matrix using a single representative sample per locality and taxon for both genes. The phylogenetic support for nodes recovered from the ML analysis obtained from a bootstrap analysis of 1000 pseudo-replicates of data sets (Felsenstein, 1985). Bootstrap values > 75% were treated as strongly supported. Uncorrected sequence divergence values were calculated for the COI locus using PAUP (Swofford, 2002). COI and 18S sequences generated during the present study were deposited in GenBank.


Phylogenetic data indicate that Peripatopsis (Pockock, 1894) is monophyletic (Daniels et al., 2009). P. moseleyi Wood-Mason, 1879 and P. sedgwicki Purcell, 1899 are sister species to P. capensis (Daniels et al., 2009). Hence, two P. segdwicki specimens from Diepwalle (Western Cape) and Port Elizabeth (Eastern Cape) and two P. moseleyi specimens from Karkloof (KwaZulu-Natal) and Hogsback (Eastern Cape), respectively, were selected as outgroup taxa. The outgroup sequences were downloaded from GenBank.

Phylogeographic analysis

Population genetic structure analysis was performed exclusively on the COI mtDNA locus using arlequin version (Excoffier, 2010). This is the most rapidly evolving marker used in the present study for which the most comprehensive geographical sequencing was undertaken (Brower, 1994). Standard diversity indices, including number of haplotypes (Nh), haplotypic diversity (h) and nucleotide diversity (π), were used. To examine hierarchical population structure, analysis of molecular variance (amova) was performed by pooling the sample localities from different locations into geographical clades evident from the preliminary phylogenetic analyses. Deviations in allele frequencies were investigated using Fu’s F statistic (Fu, 1997). This test has proven to be especially sensitive to a departure from population equilibrium as in the case of a population expansion (Excoffier, 2010). GeoPhylobuilder version 1.0 (Kidd & Liu, 2007) for ArcGIS was used to test the geographical concordance of the COI haplotype phylogram.

Divergence time estimation

To infer divergence times between the Peripatopsis capensis clades, the complete COI data set was used. Divergence time estimation was performed using a Bayesian framework; this employs a probabilistic model to define rates of molecular sequence evolution of lineages over time and uses the Markov chain Monte Carlo (MCMC) method to derive clade ages as executed in the programme beast version 1.5.1 (Drummond & Rambaut, 2007). A relaxed molecular clock was employed (Drummond et al., 2006). For COI, the suggested mutation rates of 1.5–2.3% per million years for arthropods were considered (Brower, 1994; Farrell, 2001; Trewick & Wallis, 2001; Boyer et al., 2007). Hence, a mean mutation rate of 1.9% per million years was used. For P. capensis, a multiple coalescent model was used (Heled & Drummond, 2010). modeltest was used to obtain the most likely substitution model and parameters for the combined COI locus. Twelve independent MCMC chains were run for 10 million generations with sampling carried out every 1000 generations. The convergence of the 12 combined chains was determined by EES for each parameter in Tracer after appropriate burnin cut-off. The trees comprising the 12 chains were combined using Logcombiner and were assessed using TreeAnnotator. FigTree was used to construct a chronogram (version, Rambaut, 2009). We are cognizant of the use of a single marker to infer divergence time estimations. However, mtDNA remains widely used to infer divergence time estimations Zakharov et al., 2004; Daniels, 2011).


Phylogenetic relationships

A 638-bp fragment of the COI locus was amplified and sequenced for 104 ingroup specimens and four outgroup specimens. The GenBank accession number for the ingroup COI locus is JN 798075–JN 798141. A total of 54 COI haplotypes were retrieved for the 104 sequences using the TCS analyses. Using the AIC, the substitution model for the 54 haplotypes was TIM + I + G (−1nL = 2136.69; AIC = 4501.38) (Akaike, 1973). The base pair frequencies were A = 24.31%, C = 12.49%, G = 17.64% and T = 45.56%. Similar results have been reported for velvet worms in other studies (Gleeson et al., 1998; Daniels et al., 2009; Daniels & Ruhberg, 2010). The rate matrix was R(a) [A-C] = 3.87, R(b) [A-G] = 34.99, R(c) [A-T] = 3.87, R(e) [C-T] = 19.05, R(d) [C-G] = R(f) [G-T] = 1.00, and the proportion of invariable sites was 0.33, with a gamma-shaped distribution of 0.39. The Bayesian analyses retrieved a statistically well-supported monophyletic group for P. capensis (1.00 pP). Three geographically discrete clades were retrieved with strong nodal support (1.00 pP) (Fig. 2). The three clades were the Cape Peninsula (clade A), Overberg (clade B) and the Theewaterskloof-Overstrand (clade C). Marked uncorrected sequence divergence values between the three clades were retrieved for the COI locus. For example, clade A (Cape Peninsula) differed from clades B (Overberg) and C (Theewaterskloof-Overstrand) by 8.93%, respectively. Furthermore, clades B and C displayed a maximum sequence divergence value of 7.84%. Sequence divergence within clades was comparably low with 2.35% for clade A, 3.91% for clade B and 3.13% for clade C. Superimposition of the Bayesian phylogram on the geography of the sample area using GeoPhyloBuilder revealed two low-lying barriers to dispersal (Fig. 3).

Figure 2.

 A Bayesian phylogram derived from the analyses of the COI haplotypes among the 21 Peripatopsis capensis sample sites across the Western Cape, South Africa. The values at each node represent the posterior probability (pP) value derived from the Bayesian inference analyses. Clades are labelled A (Cape Peninsula), B (Overberg) and C (Theewaterskloof-Overstrand).

Figure 3.

 A map indicating the geographical concordance of clades. Dropnodes from the Bayesian phylogram correspond to various sample localities. Clades correspond to three geographical regions, namely Cape Peninsula (Red), Theewaterskloof-Overstrand (Green) and Overberg (Purple).

Population genetics

The haplotype network for the COI locus corroborates the phylogenetic analyses and retrieved three clades (Fig. 4). Clades A (Cape Peninsula) and B (Overberg) each comprised ten haplotypes, whereas clade C (Theewaterskloof-Overstrand) was comprised of 34 haplotypes (Appendix S1). The Cape Peninsula displayed the most shared haplotypes indicating gene flow between localities. For example, haplotypes five and two were each found at three different localities in clade A (Appendix S1). According to the amova, clade A had the highest levels of genetic variation within localities followed by clades C and B, respectively. Clade B displayed the highest levels of genetic variation among localities followed by clades C and A, respectively. In summary, amova revealed limited genetic variation among localities on the Cape Peninsula, with the highest levels among the Overberg localities (Table 2). FST values within clades were statistically highly significant, indicating substantial genetic structuring (Table 2, Appendix S2). Nucleotide diversity (πn) and haplotype diversity (h) within clades are summarized in Table 3. The highest nucleotide diversity (πn) was retrieved for the Theewaterskloof-Overstrand clade followed by the Overberg clade. The highest h-values were found for Theewaterskloof-Overstrand, with the Cape Peninsula and Overberg having lower but similar values. Fu’s Fs were not statistically significant, thus limiting any inferences from this analysis.

Figure 4.

 Haplotype networks derived from all samples sequenced for COI demonstrating three distinct haplogroups corresponding to the phylogram in Fig. 2. The numbers inside the circles correspond with the haplotypes in Appendix S2. The closed black circles represent missing or unsampled haplotypes.

Table 2.   Results of the amova for the mtDNA analysis of the three clades detected within Peripatopsis capensis. All highlighted values are statistically significant (< 0.001).
CladesFST(Va)(Vb)Fu’s Fs
Clade A (Cape Peninsula)0.6329863.30%36.70%1.27
Clade B (Overberg)0.8268382.68%17.32%0.84
Clade C (Theewaterskloof-Overstrand)0.8095280.95%19.05%0.40
Table 3.   Diversity measures for Peripatopsis capensis with population numbers (Pop N) corresponding to those in Fig. 1, sample size (N), number of haplotypes (Nh), number of polymorphic sites (Np), haplotype diversity (h) and nucleotide diversity (πn).
Pop NLocalitiesNNhNphπn
 1Myburgh Ravine 43 70.8333 ± 0.22240.005486 ± 0.004195
 2Orangekloof 42 40.5000 ± 0.26520.003135 ± 0.002630
 3Cecelia Forest 73 60.7143 ± 0.12670.004926 ± 0.003318
 4Skeleton Gorge 43140.8333 ± 0.22240.011755 ± 0.008312
 5Newland’s Ravine 31 01.0000.0000
 6Rhode’s Memorial 62 00.3333 ± 0.21520.0000
 7Rondevlei Nature Reserve 11 01.00000.0000
 8Jonkershoek 43 20.8333 ± 0.22240.001567 ± 0.001553
 9Berg River Dam 42 10.5000 ± 0.26520.000784 ± 0.000972
10High Noon 42 20.5000 ± 0.26520.001567 ± 0.001553
11Dappat se gat 52 10.6000 ± 0.17530.000940 ± 0.001030
12Kogelberg Biosphere Reserve 22 31.0000 ± 0.50.004702 ± 0.005430
13Fernkloof Nature Reserve 54 90.9000 ± 0.16100.007210 ± 0.004971
14Caledon 11 01.00000.0000
15Grootbos Private Nature Reserve 97160.9167 ± 0.09200.008447 ± 0.005098
16Napier 63 50.8000 ± 0.12170.004180 ± 0.002980
17Greyton 33 91.0000 ± 0.27220.009927 ± 0.008070
18Oubos 43 70.7000 ± 0.21840.004389 ± 0.003241
19Potberg-De Hoop Nature Reserve102 10.5556 ± 0.07450.000871 ± 0.000876
20Marloth Nature Reserve105170.7556 ± 0.12950.010658 ± 0.006205
21Grootvadersbosch Nature Reserve 83 20.4643 ± 0.20000.000784 ± 0.000842

Combined DNA analyses (COI mtDNA and 18S rDNA)

The combined sequence data yielded a total of 1052 bp of which 18S included 414 bp. The GenBank accession number for the 18S locus is JN 798142–JN 798162. Using the AIC, the GTR + G (−1nL = 2204.02; AIC = 4426.04) model was selected for COI, whereas the JC + I (−1nL = 665.52; AIC = 1333.05) (Jukes & Cantor, 1969) model was selected for 18S (Akaike, 1973). Both analytical methods (ML and BI) produced near-identical tree topologies (Fig. 5). The tree topology retrieved a statistically well-supported monophyletic group for P. capensis (97%/1.00 pP). Clade A (Cape Peninsula) formed a monophyletic group with strong nodal support (99%/1.00 pP). Furthermore, the monophyly of both clades B and C were also well supported (99%/1.00 pP and 80/1.00 pP, respectively).

Figure 5.

 Bayesian inference topology for the combined DNA analyses (COI 18S) among Peripatopsis capensis sampled in the present study. The values above each node represent bootstrap values (%) for maximum likelihood (ML), whereas values below each node represent posterior probability (pP) value derived from the Bayesian inference (BI) analyses.

Divergence time analyses

Cladogenesis within P. capensis occurred 3.14 Mya (95% confidence interval: 2.13–4.38 Mya). Divergence between clade B (Overberg) and clade C (Theewaterskloof-Overstrand) occurred 2.39 Mya (95% confidence interval: 1.55–3.3 Mya). These results suggest a Pliocene/Pleistocene divergence among the three P. capensis clades. Divergence times within clade A range from 0.59 to 0.02 Mya. Clades B and C are characterized by divergence times that range between 0.92 and 0.75 Mya and between 1.16 and 0.28 Mya, respectively.


Divergence time and biogeography

The Cape Peninsula (clade A) is an isolated mountainous area surrounded by the Atlantic Ocean in the west and the Cape Flats to the east, thus isolating the velvet worm fauna. This conforms to the mainland–island metapopulation pattern where, as a result, genetic differentiation among populations is low in comparison with the two remaining clades (B and C) (Mosblech et al., 2011). In contrast, genetic divergence (FST) among localities in the latter clades is high, indicating considerable genetic structure, and provides further evidence for low dispersal among localities (Mosblech et al., 2011). Peripatopsis capensis is generally restricted to forested palaeorefugia occupying sheltered ravines and gorges on the Cape Peninsula. Contemporary and historical forest habitat in this region could be the main driver of the observed phylogeographic pattern. Regionally, the prevailing climate during the Pliocene/Pleistocene would have marked a dynamic period for the Cape Flats–Cape Peninsula boundary. Divergences within the Cape Peninsula clade ranging between approximately 600 000 and 10 000 b.p. suggest a response to fluctuating climatic patterns. Climatic fluctuations periodically created islands of differing vegetation types, leading to repeated periods of habitat expansions, followed by allopatric diversification (Tolley et al., 2006). Schalke (1973) describes at least two extensions of Podocarpus forest on the Cape Flats between 50 000 and 20 000 b.p. Sea-level dynamics played a large role in the onshore conditions of the CFR. Linder (2003) suggests that low sea levels may have led to a loss of rainfall at mountain regions. These marine regressions are associated with lower absorption of incoming solar radiation due to a decrease in the area covered by the oceans. This results in decreased evaporation and, consequently, decreased precipitation with an overall effect of increased aridity on land (Theron, 1983). Lower temperatures further detract from past (Miocene) tropical conditions (Deacon, 1983a; Daniels et al., 2001). Forest habitat loss was further exacerbated by prehistoric pastoralists (1500–2000 b.p) who used fire to manage the vegetation (Deacon, 1983a,b; Brain & Sillen, 1988). Hence, the increasingly limited forest cover has a reduced capacity to maintain high levels of genetic diversity within the P. capensis Cape Peninsula clade.

Biogeographically, the Cape Peninsula (clade A) is separated from the Theewaterskloof-Overstrand region (clade C) by a phylogeographic barrier of approximately 60 km, the Cape Flats. The Cape Flats is assumed to have emerged after the closure of the ‘Cape Strait’, which once united Table Bay and False Bay (Schalke, 1973). The Cape Flats is characterized by the deposition of aeolian sands of marine origin (Walker, 1952; Siesser & Dingle, 1981; Adelana et al., 2010). Thus, the Cape Flats is characterized by nutrient-poor and calcareous alkaline soils supporting the shrub-dominated Cape Flats Dune Strandveld (Mucina & Rutherford, 2006). Several phylogeographic studies on mountain-living invertebrates have identified the Cape Flats as a barrier to gene flow between the Cape Peninsula and the Theewaterskloof-Overstrand region (Daniels et al., 2001; Wishart & Hughes, 2001, 2003; Gouws et al., 2004, 2010). Interestingly, the Rondevlei locality falls within the Cape Flats phylogeographic break. Yet, the Rondevlei specimen clustered with the Theewaterskloof-Overstrand clade despite being geographically closer to the Cape Peninsula. More specifically, the phylogenetic trees retrieved indicated that it has a closer genetic affinity with Jonkershoek taxa, suggesting that a historical corridor of suitable habitat may have connected these localities in the past. This can be explained by the Pleniglacial periods between 33 000 and 45 000 b.p during the Pleistocene, where mixed Podocarpus forests were present in the central Cape Flats region (Schalke, 1973; Chase & Meadows, 2007).

The highest levels of genetic variation were observed within the Overberg clade (B). High genetic variation and a lack of shared haplotypes (Fig. 4) suggest that dispersal is limited among localities at the Overberg. However, within clade B, two specimens from Marloth were nested within the Grootvadersbosch subgroup, suggesting the possibility of a historical connection between the two localities. Both these localities occur along the sheltered slopes at the middle and upper reaches of the west–east-trending southern Langeberg mountain range (McDonald et al., 1996). These areas receive between 600 and 1200 mm of rainfall annually. According to McDonald & Cowling (1995), the latter lies within the nonseasonal rainfall zone of the CFR where rainfall is associated with circumpolar westerly fronts and post-frontal conditions related to the advection of cool moist air above the warm Indian Ocean. Gene flow between Potberg-De Hoop and the former localities is limited by an area colloquially known as the ‘Rûens’, which is an undulating landscape straddled by the Potberg-De Hoop Mountain in the south and the Langeberg Mountains in the north. The Rûens is characterized by a highly dissected landscape, relatively low rainfall and shallow.

Soils supporting Rûens Silcrete Renosterveld (Schloms et al., 1983). Divergence within clade B is dated between 0.75 and 0.92 Mya in the mid-Pleistocene. The late Pleistocene period was characterized by lower sea levels (100–160 m below current sea levels) and a drier climate (Deacon, 1983a,b; Cowling & Lombard, 2002).

Clades B and C are separated by the Breede River Valley Basin. According to Theron (1983), the elevated coastal topography in the southern Cape facilitated the establishment of larger rivers being deeply entrenched in steep-sided valleys as a result of marine regressions. The Breede River Valley lies between the Riviersonderend Mountains in the south and the Langeberg Mountains in the north (Kirchner et al., 1997). The valley displays unique abiotic characteristics relative to the latter montane regions. It is characterized as semi-arid and receives approximately 270 mm of rainfall per annum. The valley comprises a variety of soil types (sandy, aeolian, acidic, alluvial, clay and loam) supporting different variants of fynbos (Mucina & Rutherford, 2006). These abiotic factors are unsuitable for the establishment of forests.

Clade C (Theewaterskloof-Overstrand) represents an area that forms the boundary between the winter rainfall zone (WRZ) of the westernmost CFR and the year-round rainfall zone (YRZ) in the south-east. In addition to high climatic variability, the Theewaterskloof-Overstrand subregion also forms a large part of the Cape Fold Mountains and hence represents a landscape characterized by high topographic heterogeneity. As a result, clade C displayed three unique haplotypes and had the highest haplotypic (h) and nucleotide (πn) diversity. These findings suggest that clade C has had a long evolutionary history in large, stable populations (Fitzpatrick, 2009).

Evolutionary and taxonomic considerations

The biogeographical conclusions of this study have important implications for the taxonomy of the Cape velvet worm species, Peripatopsis capensis. Multiple lines of independent evidence exist for the recognition of three evolutionary units within the P. capensis. All three clades are geographically discrete and exclusive, with no dispersal between them. Furthermore, all three clades were characterized by marked sequence divergence values, > 7% for the COI locus. This value is higher than what has been reported for sister velvet worm species. For example, Hebert et al. (1991) employed a 3.3% sequence divergence in the Jamaican velvet worm species, Plicatoperipatus jamaicensis, which comprised two distinct lineages. Similarly, Rockman et al. (2001) obtained sequence divergence values ranging from 1.1 to 11.6% for morphologically distinct Planipapillus species. Despite the comparably high divergence values retrieved for P. capensis, we exercise caution before basing our classification on a single molecular marker. The inclusion of the nuclear 18S rDNA marker to create a combined topology was congruent with the COI data, providing additional corroborative evidence for the genetic distinction of the three clades. We prefer to use the phylogenetic species concept as a starting hypothesis for species description (Cracraft, 1989). Preliminary gross morphological and scanning electron microscopy (SEM) analyses have revealed several potential diagnostic characters. Collectively, these results provide evidence for the recognition of two novel lineages within P. capensis because the species was originally described from the Cape Peninsula. A study is currently underway to describe these two novel lineages. The present study shows that velvet worm diversity associated with CFR forests was underestimated. The use of Peripatopsis capensis as a model organism to make biogeographical extrapolations of its habitat adds to a growing body of literature that demonstrates the utility of invertebrate habitat specialists in understanding forest biogeographic patterning and highlights the conservation value of these forest patches because our study further suggests the presence of undocumented biodiversity (Boyer et al., 2007; Garrick et al., 2007; Álvarez-Presas et al., 2011). Our study yielded significant novel insight into the biogeographic patterning of Afromontane forest habitat and provided a unique glimpse into the phylogeographic structure of an enigmatic endemic invertebrate group.


This research was supported by a National Research Foundation (NRF) grant to Savel Daniels. The NRF and the Lawrence Memorial Grant (via the Zoological Society of South Africa) is thanked for providing a bursary. Anzio Abels and Franscois Van Zyl are thanked for help with sample collection. Ilse Kotzee is thanked for help with GIS and map design. Hanlie Engelbrecht is thanked for constructive comments on the manuscript. Cape Nature and South African National Parks (SANParks) are thanked for issuing collection permits. Two reviewers are thanked for constructive comments that improved the quality of the manuscript.