Penrith’s (1984) morphological evidence for subspecific recognition involved shape differences of the pronotum and prosternal process (Figs 4–7). The pronotum is more strongly transverse in Onymacris unguicularis schulzeae, and its prosternal process is generally broader, featuring a blunt apex that is largely hidden in lateral aspect (Fig. 8). Conversely, the prosternal process in Onymacris unguicularis unguicularis is evident in lateral aspect, its apex often appearing as tooth-like projection (Fig. 9).

To reassess Penrith’s (1984) diagnostic morphological characters and explore additional aspects of phenotypic variation, we examined a series of pinned specimens from the Ditsong National Museum of Natural History (formerly Transvaal Museum), Pretoria, South Africa. Material included 93 Onymacris unguicularis unguicularis representing 11 populations and 30 Onymacris unguicularis schulzeae (all paratypes) representing four populations (Appendix). Two and five additional specimens of Onymacris unguicularis unguicularis and Onymacris unguicularis schulzeae, respectively (representing a portion of our genetic sample), were also examined. Specimens were photographed to provide dorsal and ventral images of each beetle. Images were made using a Visionary Digital Imaging System (Visionary Digital^TM, Richmond, VA).

Penrith (1984) quantified pronotal shape differences between subspecies as a ratio of pronotal length divided by pronotal width (PL/PW). We repeated these measurements as part of our morphological analysis, and also quantified differences in the prosternal process as a length/width ratio (Figs 5 and 7), using the software program Image J. Additionally, Penrith (1984) noted possible differences in elytral shape, which appears to be “less elongate, broader, and more abruptly tapered posteriorly” in Onymacris unguicularis schulzeae. We therefore used a geometric morphometric analysis to assess putative differences in dorsal (elytral) shape. Mindful that sexual dimorphism may contribute to elytral shape variation, we acknowledge its potential to confound signal attributable to subspecific variation. However, only one species of Onymacris, Onymacris plana (Péringuey, 1888), exhibits pronounced sexual dimorphism in elytral shape; in all others the female’s elytra are only “slightly broader than those of the male, ” with “much overlap” (Penrith 1975). In Onymacris unguicularis, frequency distributions of maximal elytral width, expressed as a percentage of elytral length, exhibit complete overlap between the sexes (Penrith 1975), and elytral shape in northern populations has been dismissed as being “scarcely dimorphic” (Penrith 1984). We should note that sexual dimorphism is evident in Onymacris unguicularis: males possess longer legs and, uniquely within Onymacris, bear setose brushes on the anterior femora (Penrith 1975). Given the species’ limited elytral dimorphism and our relatively small sample of Onymacris unguicularis schulzeae (n = 35), we elected to combine the sexes in our morphometric analysis.

We identified eleven dorsal landmarks—eight type 1 and three type 3 (Brookstein 1991; Fig. 10)—and used the programs tps-Util and tps-DIG2 (Rohlf 2012) to assemble dorsal (elytral) image files for analysis and score landmarks, respectively. Landmarks were aligned and scaled to size using the generalized least squares Procrustes superimposition method (Rohlf and Slice 1990), which removes information not relevant to shape (location, scale, and rotational effects). Relative warps were calculated with α set to zero, thus weighting all principal warps equally. Superimposition, calculation of relative warps, and calculation of centroid size were preformed using the program MorphoJ (Klingenberg 2011).

Sixteen beetles were captured, preserved (100% ethanol), and processed for DNA analysis; the specimens included 12 Onymacris unguicularis unguicularis, representing three geographic localities, and four Onymacris unguicularis schulzeae, representing two relatively close localities (Fig. 1; Appendix). Onymacris laeviceps Gebien, 1938 and Onymacris plana, identified as sister taxa to Onymacris unguicularis in a generic-level phylogeny (Lamb and Bond 2013), served as outgroups. The mitochondrial genes cytochrome oxidase I (cox1) and cytochrome oxidase II (cox2) were amplified using the primers and PCR conditions listed in Table 1.

Amplification products were cleaned using exoSAP-IT (USB Corp.) and sequenced on an Applied Biosystems 3130 capillary sequencer. Sequences were edited and assembled in Sequencher 4.9 (GeneCodes, Ann Arbor, MI) and aligned using ClustalX ver. 2.0 (Larkin et al. 2007), after which sequences were translated to ensure a correct reading frame.

We used Bayesian inference (BI) and maximum likelihood (ML) methods to analyze the concatenated gene (cox1-cox2) dataset. We used Kakusan 4 (Tanabe 2007) to select nucleotide substitution models for BI, partitioning protein-coding genes by codon position and assessing each gene/codon partition using the Bayesian Information Criteria (BIC4 criterion). BI analysis was conducted in MrBayes 3.1.2 (Ronquist and Huelsenbeck 2003) and involved two concurrent runs of four simultaneous Markov Chain Monte Carlo chains for 20, 000, 000 generations, with trees sampled every 1, 000 generations. Topologies in the first 25% of the posterior distribution were discarded as burn-in. Likelihood values for all post-analysis trees and parameters were evaluated for convergence and burn-in using the “sump” command in MrBayes and the computer program Tracer ver. 1.5 (Rambaut and Drummond; http://evolve.zoo.ox.ac.uk/software.html?id=tracer). Trees remaining after burn-in were used to calculate posterior probabilities using the “sumt” command. The ML analysis, executed in RAxML ver. 7.2.8 (Stamatakis 2006), comprised 1, 000 random sequence addition replicates (RAS) using the commands “-# 1000” and “–m GTRGAMMA.” Bootstrap support values were calculated using the same search parameters with 1, 000 replicates, and bootstrap results were applied to the best tree recovered in the RAS search.

Ratios generated for both the protonal and prosternal datasets differed significantly between subspecies (p < 0.0001), with minimal overlap for each character (Table 2). In the geometric morphometric analysis, the first two principal components based on the non-uniform components of dorsal (elytral) shape account for 78.54% of the variation between subspecies. An ordination plot of PC1 and PC2 revealed that the two subspecies are relatively well separated along the PC1 axis (Fig. 11). Dorsal shape separation probably reflects proportionally longer elytra in Onymacris unguicularis unguicularis, which become apparent in side-by-side comparisons with Onymacris unguicularis schulzeae (Figs 2–3). In light of these findings, we measured elytral length and width (at the midpoint of elytral length) for all specimens to determine whether a simple ratio (EL/EW) might reflect the subspecific separation observed in our geometric morphometric analysis. We also noted the position of greatest elytral width for each specimen, scored as midpoint, anterior to midpoint, or posterior to midpoint. Despite broad overlap, elytral ratios differed significantly (p < 0.0001) between subspecies (Table 2); elytral width is widest anterior to midpoint in both subspecies but is positioned closer to the pronotal suture in Onymacris unguicularis schulzeae. Of the three ratios, we consider that for the pronotum to be the strongest diagnostic metric.

Concatenated sequence data for the cox1 (1574 bp) and cox2 (680 bp) genes yielded eleven haplotypes among the 16 beetles surveyed: eight haplotypes for Onymacris unguicularis unguicularis and three for Onymacris unguicularis schulzeae (Genbank accession numbers KF835703-KF835721). No haplotypes were shared between subspecies. Mean haplotype divergence (calculated from uncorrected pair-wise distance values) was limited across geographic localities for both Onymacris unguicularis unguicularis (cox1 = 0.046%; cox2 = 0.027%) and Onymacris unguicularis schulzeae (cox1= 0.008%; cox2 = 0.014%) but differed substantially between the two subspecies (cox1 = 3.87%; cox2 = 3.01%). The BI (harmonic mean –ln = 4690.15) and ML (–ln = 4172.41) analyses generated topologically identical trees in which subspecies were shown to be reciprocally monophyletic (Fig. 12). Moreover, subspecific monophyly was strongly supported (Bayesian posterior probabilities = 1.0; ML bootstraps = 100%), in contrast to the marginal to moderate support observed for haplotype relationships of geographic localities within subspecies (Fig. 12).

We employed Braby et al.’s (2012) integrative approach to evaluate the polytypic status of Onymacris unguicularis and found support for each criterion in their template for subspecies delimitation. Onymacris unguicularis is an ultrapsammophile confined to major dune fields within the northern (Cunene, Skeleton Coast) and southern (Namib) sand seas. Separated by 300 km of unsuitable substrate, these populations are unquestionably allopatric, satisfying Braby et al.’s first criterion. Regarding criterion two, phenotypic distinctiveness, we confirmed qualitative differences in pronotal and prosternal shape and verified putative distinctions in elytral shape. Patterns in larval variation—the ninth abdominal tergum being shorter and broader in northern populations (Schulze 1964)—complement differences in adult morphology and augment the case for phenotypic distinctiveness. Support for Braby et al.’s third criterion, character difference correlations with genetic variation, involves a phylogeographic profile that is perfectly congruent with the north-south partition in phenotypic variation. More importantly, the reciprocal monophyly observed between northern and southern haplotypes (with associated levels of genetic divergence) identifies respective paths of evolutionary independence. These data demonstrate that northern and southern populations of Onymacris unguicularis are phylogenetically distinct under the general lineage concept. Thus, we endorse Penrith’s (1984) taxonomic interpretation in recognizing Onymacris unguicularis unguicularis and Onymacris unguicularis schulzeae as valid taxa.