Ethics and Permissions
Data collection received ethics clearance from Nelson Mandela University: Research Ethics Committee (Animal; A15-SCI-ZOO-007) as well as from the Ethics Committee of the Department of Zoology & Entomology, Rhodes University (RU-DZE-2017-10-028). Capture permits have been received from both Western Cape Province: Cape Nature (AAA041-00565), and Eastern Cape Province: Department of Economic Development and Environmental Affairs and Tourism (CRO55/17CR and CRO56/17CR). Birds were ringed with permission from SAFRING (Ringer no: 17059), and blood sampling technique was confirmed by Dr. Tarryn Fick (BVSc, Newton Park Animal Hospital, Port Elizabeth, SA).
Results
Sequence variation
In total we produced 74 cytb sequences (380 bp; C. frenatus N = 59, C. aurantius N = 15), 85 ND2 sequences (505 bp; C. frenatus N = 70, C. aurantius N = 15), and 55 RAG1 sequences (455 bp; C. frenatus N = 45, C. aurantius N = 10). For the cytb marker we found more genetic variation within C. aurantius(N = 5 haplotypes) compared to C. frenatus (N = 4 haplotypes). For the ND2 marker we found five haplotypes in C. aurantius , and 15 haplotypes for C. frenatus (see below for details). Uncorrected p -distances within C. frenatus varied from 0.0–0.002, within C. aurantius varied from 0.0–0.003, and between C. frenatus and C. aurantius varied from 0.0–0.043 (Table 2).
Tajima’s D for both the cytb and ND2 markers in C. frenatus were significantly negative (cytb: Tajima’s D = −1.70,p = 0.020; ND2: Tajima’s D = −1.77, p = 0.023), while significantly negative for cytb but not significant for ND2 inC. aurantius (cytb: Tajima’s D = −0.78, p = 0.028; ND2: Tajima’s D = 1.08, p  = 0.893).
Population genetic structure
We found no evidence for genetic differentiation among C. frenatus localities, with low fixation indices showing strong evidence for interbreeding among the various populations (cytb: FST = 0.035, DF = 6; ND2: FST = 0.057, DF = 6). Despite the proximity of C. aurantius sample populations to one another, we found little evidence for interbreeding among them, with higher fixation indices than those for C. frenatus (cytb: FST = 0.277, DF = 1; ND2: FST = 0.470, DF = 1). The phylogenetic networks did not show any clear pattern of genetic structure among the sky islands (Figure 2).
The haplotype network for cytb showed little variation or pattern, while we found a ”starburst” pattern for C. frenatus in the ND2 network (Figure 2). When visualized by frequency per locality, ND2 showed little evidence for spatial genetic structure (see inset Figure 3 below). The low number of haplotypes within C. frenatus at the cytb locus (N = 4) resulted in a haplotype network which did not have the variability to visualize frequency by locality
Overall, most of genetic variation was within populations of C. frenatus (cytb: 96.48 %, ND2: 94.26 %; see Table 3 for summary of AMOVA), with similar variation partitioned between speciesC. frenatus and C. aurantius (cytb: 93.86 %; ND2: 93.77 %). We also found that while some variation existed within individual populations of both C. frenatus and C. aurantius (cytb: 5.59 %; ND2: 5.35 %), there was little variation among populations (cytb: 0.55 %; ND2: 0.88 %). High fixation indices for both cytb and ND2 showed there was little evidence for interbreeding betweenC. frenatus and C. aurantius (cytb: FCT = 0.939, DF = 1; ND2: FCT = 0.938, DF = 1).
Our consensus tree based on ML inference resulted in stronger predictive weight of node probabilities compared to the Bayesian tree (see Appendix Figure A1 for Bayesian inference tree), and while we found no obvious structuring within C. frenatus , we did find strong support for two clear clades within family Chaetopidae being C. frenatus andC. aurantius (Figure 3).
Discussion
We found no evidence for genetic structure among populations ofC. frenatus . The results of population substructure among populations (F ST < 0.05) indicated only a slight to moderate genetic differentiation among the populations, similar to two other studies on genetic structuring in sky island endemic birds (Bech et al., 2009; Sittenthaler et al., 2018). Instead, our study seems to provide support for the hypothesis that sky island species with specialized niche requirements may experience selective pressure which limits their genetic diversity [see Orsini et al., (2013)]. We also found evidence for interbreeding among individuals from different mountain ranges, suggesting that birds may be able to effectively disperse between mountain ranges. This suggests C. frenatus is not currently experiencing inbreeding depression. However, the lack of structure may also indicate the chosen loci are under positive selective pressure and so are similar due to mutations relevant for alpine species. The shape of our mtDNA haplotype networks suggestC. frenatus experienced a bottleneck or founder effect in their recent genetic past. Tajima’s D also showed support for a past bottleneck event in C. frenatus at both mtDNA markers. In addition, overall low genetic diversity within C. frenatusindicates the potential for negative effects from a reduced effective population size.
Population variation, structure, and dispersal
Our finding that within C. frenatus most genetic variation was from within populations, as opposed to among populations, suggests some genetic structuring, and our haplotype frequency network implies this is not geographically based (Figure 3). The genetic composition of C. frenatus , along with the lack of geographic distribution of genetic populations, suggest that although mountain populations ofC. frenatus are separated from one another by unsuitable lowland habitat, the actual degree of separation does not hamper effective dispersal. The question then remains as to how C. frenatus are managing to effectively disperse across large tracts of unsuitable habitat, presenting an interesting avenue for future research. While in some sky island species low occurrence rates in lowland habitat implied low dispersal, as in Mexican Jays Aphelocoma ultramarine(McCormack, Bowen, & Smith, 2008), according to SABAPs 1 and 2 (1987–1991, and 2007–present, respectively) there is currently a complete lack of observations for C. frenatus in the lowland habitat separating the mountain ranges of the Cape Fold Belt (http://sabap2.adu.org.za; Harrison et al. 1997).
We also found evidence that while populations may be interbreeding,C. frenatus have low overall genetic diversity among populations for both cytb and ND2 markers (i.e. genetic variation <5 %). While we found no variation for our nDNA marker, we decided against the addition of a different marker; the little variation we found for both faster-evolving mtDNA markers suggests the addition of a second (slower-evolving) nDNA marker, or a faster-evolving nDNA marker (i.e. microsattelites) was unlikely to provide additional insight,. Moreover, our choice of nDNA marker (RAG1) remains relevant for determining population structure in birds (e.g. Dantas et al., 2019; Stervander, Ryan, Melo, & Hansson, 2019). Low genetic variation among populations may be from low diversity within the species, or close inbreeding within an otherwise diverse population (Ceballos, Joshi, Clark, Ramsay, & Wilson, 2018); since the results of the AMOVA suggest C. frenatuspopulations are interbreeding, there may be low diversity within the species itself. Low diversity can arise from a recent demographic expansion (e.g. bottleneck), here possibly tracking the mountain Fynbos expansion ~37 KYA. A bottleneck may have extirpated many individuals, reducing overall genetic variability. Because such a population expansion would be recent in evolutionary terms, there has not been time enough to accumulate measurable mutations. Alternatively, as the loci analyzed are protein-coding, natural selection may be purging all but a few advantageous alleles (e.g. strong stabilising selection due to habitat specialisation; Orsini et al., 2013), asTajima’s D ’s for the mtDNA loci were negative, indicating a higher frequency of rare alleles than would be expected under neutrality. Although C. frenatus had a greater number of ND2 haplotypes than C. aurantius (N = 15 and 5, respectively), the number is not proportionally greater considering the difference in distribution coverage (~750 km and ~20 km respectively).