Genetic isolation in an endemic African habitat specialist

The Chestnut‐banded Plover Charadrius pallidus is a Near‐Threatened shorebird species endemic to mainland Africa. We examined levels of genetic differentiation between its two morphologically and geographically distinct subspecies, C. p. pallidus in southern Africa (population size 11 000–16 000) and C. p. venustus in eastern Africa (population size 6500). In contrast to other plover species that maintain genetic connectivity over thousands of kilometres across continental Africa, we found profound genetic differences between remote sampling sites. Phylogenetic network analysis based on four nuclear and two mitochondrial gene regions, and population genetic structure analyses based on 11 microsatellite loci, indicated strong genetic divergence, with 2.36% mitochondrial sequence divergence between individuals sampled in Namibia (southern Africa) and those of Kenya and Tanzania (eastern Africa). This distinction between southern and eastern African populations was also supported by highly distinct genetic clusters based on microsatellite markers (global FST = 0.309, GST′ = 0.510, D = 0.182). Behavioural factors that may promote genetic differentiation in this species include habitat specialization, monogamous mating behaviour and sedentariness. Reliance on an extremely small number of saline lakes for breeding and limited dispersal between populations are likely to promote reproductive and genetic isolation between eastern and southern Africa. We suggest that the two Chestnut‐banded Plover subspecies may warrant elevation to full species status. To assess this distinction fully, additional sample collection will be needed, with analysis of genetic and phenotypic traits from across the species’ entire breeding range.

The Chestnut-banded Plover Charadrius pallidus is a Near-Threatened shorebird species endemic to mainland Africa. We examined levels of genetic differentiation between its two morphologically and geographically distinct subspecies, C. p. pallidus in southern Africa (population size 11 000-16 000) and C. p. venustus in eastern Africa (population size 6500). In contrast to other plover species that maintain genetic connectivity over thousands of kilometres across continental Africa, we found profound genetic differences between remote sampling sites. Phylogenetic network analysis based on four nuclear and two mitochondrial gene regions, and population genetic structure analyses based on 11 microsatellite loci, indicated strong genetic divergence, with 2.36% mitochondrial sequence divergence between individuals sampled in Namibia (southern Africa) and those of Kenya and Tanzania (eastern Africa). This distinction between southern and eastern African populations was also supported by highly distinct genetic clusters based on microsatellite markers (global F ST = 0.309, G 0 ST = 0.510, D = 0.182). Behavioural factors that may promote genetic differentiation in this species include habitat specialization, monogamous mating behaviour and sedentariness. Reliance on an extremely small number of saline lakes for breeding and limited dispersal between populations are likely to promote reproductive and genetic isolation between eastern and southern Africa. We suggest that the two Chestnut-banded Plover subspecies may warrant elevation to full species status. To assess this distinction fully, additional sample collection will be needed, with analysis of genetic and phenotypic traits from across the species' entire breeding range.
Genetic differentiation most commonly emerges between populations divided as a result of geographical barriers. Over evolutionary time, this process can result in the formation of distinct species from a single ancestral population via allopatric speciation (Mayr 1942, Avise 2009). In the absence of geographical barriers, and particularly in highly mobile organisms such as birds, populations can remain connected by dispersal over thousands of kilometres, thereby remaining genetically homogeneous across their continental range (Clobert et al. 2004, Claramunt et al. 2012. The plovers (Charadrius) are shorebirds and include several species that exhibit high gene flow on a continental scale. For example, the Kentish Plover Charadrius alexandrinus of Eurasia remains genetically homogeneous across 10 000 km (K€ upper et al. 2012) and Kittlitz's Plovers Charadrius pecuarius exhibit high levels of genetic homogeneity both across continental Africa (7600 km; dos Remedios 2013) and within Madagascar (Eberhart-Phillips et al. 2015). For these species, genetic differentiation occurs only between populations separated by large bodies of water, such as between oceanic island and mainland populations (dos Remedios 2013, Almalki et al. 2017).
The Chestnut-banded Plover Charadrius pallidus is distributed exclusively in continental Africa. The most recent taxonomic analyses suggest the species is a member of the Kentish Plover superspecies complex (dos Remedios et al. 2015). Yet despite its mainland distribution, two phenotypically differentiated subspecies are currently recognized. The taxon C. p. pallidus is distributed patchily in southern Africa, breeding in both coastal and inland wetlands in Namibia, Botswana and South Africa, with a fluctuating population of 11 000-16 000 individuals (Delany et al. 2009). In contrast, C. p. venustus is a non-coastal species that inhabits a small number of saline lakes in the eastern African Rift Valley (Kenya and Tanzania), and includes approximately 6500 individuals (Simmons et al. 2007, Delany et al. 2009, Fig. 1a and1b). In C. p. pallidus, individuals are approximately 15% larger, with paler and greyer upperparts than C. p. venustus (Hayman et al. 1986, Fig. 1a). However, whether these phenotypic differences are matched by genetic differentiation is unclear. Charadrius plovers are typically highly mobile and often exhibit considerable gene flow over large geographical distances (Funk et al. 2007, K€ upper et al. 2012, D'Urban Jackson et al. 2017. Patterns of phenotypic and genetic differentiation do not necessarily match, making subspecies and species delineation in this taxonomic group often challenging (Rheindt et al. 2011, K€ upper & dos Remedios in press).
The Chestnut-banded Plover has previously been described as being 'overlooked' by the scientific community (Simmons et al. 2007). It is considered Near Threatened (BirdLife International 2016) due to its reliance on an extremely small number of saline wetland sites during both the breeding and the non-breeding seasons. In this study, we analyse nuclear and mitochondrial sequence data (six loci), as well as microsatellite markers (11 loci), to assess levels of genetic differentiation between the two subspecies of Chestnut- banded Plover for the first time. We assess the validity of the current taxonomic classification and discuss the evolutionary history of this species.
We captured individuals using mist-nets or funnel traps following standard procedures (Sz ekely et al. 2008). For blood sampling, we punctured the brachial wing vein with a hypodermic needle and used a capillary tube to transfer 25-50 lL of blood to an Eppendorf tube for storage in 1 mL Queen's Lysis Buffer (Sz ekely et al. 2008).
We selected putatively unrelated individuals, excluding known parent-offspring and sibling pairs, resulting in 73 samples for genetic analyses (21 from Namibia, 40 from Kenya and 12 from Tanzania). To extract DNA from these samples, we used an ammonium acetate precipitation method (Nicholls et al. 2000).

Microsatellite analyses
We utilized microsatellite primers developed for the Kentish Plover (K€ upper et al. 2007) and initially tested 18 markers in eight individuals. Four loci (Calex-1, Calex-13, Calex-17, Calex-34) were excluded from further analyses because of their low variability. We selected 14 polymorphic loci for which at least two different alleles were detected and designed two multiplexes using MULTIPLEX MANAGER v1.2 (Holleley & Geerts 2009).
We conducted multiplex PCRs for all study individuals on a DNA Engine Tetrad 2 Peltier Thermal Cycler (Bio-Rad, Hercules, CA, USA) in 2-lL reactions, including 10-20 ng DNA, 1 lL Qiagen Multiplex PCR MasterMix and 1 lL primers diluted in ddH 2 O (final primer concentration 0.2 mM; forward primers labelled with fluorescent dye), sealed with mineral oil to prevent evaporation. We carried out PCR amplification under the following conditions: 95°C for 15 min, 35 cycles of 94°C for 30 s, 56°C for 90 s, 72°C for 60 s, and finally 60°C for 30 min. We visualized amplicons on an ABI 3730 automated DNA analyser and scored fragment lengths using GENEMAPPER software version 4.1 (Applied Biosystems, Waltham, MA, USA). Full genotype data are included in Table S1.
We ran 10% of samples twice (four samples from Kenya and two each from Namibia and Tanzania) to check consistency in allele scoring. Scoring was fully consistent across markers. We assessed heterozygosity in each population using CERVUS v3.0.3 (Kalinowski et al. 2007). Subsequently, we excluded three more loci (Calex-8, Calex-11, Calex-23) from further analysis due to a high frequency of null alleles (> 0.2) and deviation from Hardy-Weinberg equilibrium across populations, leaving 11 loci in the final marker set (Table 1).
Although sampled individuals were putatively unrelated based on field observations, evaluation of microsatellite data in ML-RELATE (Kalinowski et al. 2006) suggested high levels of genetic relatedness r between some samples. We therefore excluded those with r > 0.5 from further analyses, leaving a sample size of 45 individuals (17 from Namibia, 18 from Kenya and 10 from Tanzania).
We estimated global fixation indices (F ST , G 0 ST and D; Weir & Cockerham 1984, Hedrick 2005, Jost 2008) using the 'diveRsity' package in R ('diff-Calc' function; Keenan et al. 2013) and inbreeding coefficients (F IS , i.e. deficit of heterozygosity relative to a randomly mating population; Weir & Cockerham 1984) using FSTAT 2.9.3 (Goudet 1995). We conducted analysis of molecular variance (AMOVA) and assessed pairwise fixation indices (F ST ) using ARLEQUIN 3.5.2 (Excoffier & Lischer 2010) to identify levels of genetic variance among sampling sites and subspecies. Lastly, we used two methods to assess and visualize genetic structure between populations based on microsatellite data: principal component analysis (PCA; 'princomp' package in R (v3.0.3; R Core Team 2013)) and a Bayesian clustering method implemented in the software STRUCTURE 2.3.4 (Hubisz et al. 2009). The latter was implemented with allele frequencies correlated for 500 000 Markov chain Monte Carlo repeats, after a burn-in of 100 000, using an admixture model. We used STRUCTURE HARVESTER (Earl & vonHoldt 2012) to evaluate results, and produced plots with CLUMPP (Jakobsson & Rosenberg 2007) and DISTRUCT (Rosenberg 2004).
We conducted PCRs for three individuals from each sample site. Reactions of 10 lL total volume contained 4 lL Qiagen Multiplex Mix, 0.1 lM of each primer and 20-30 ng DNA, and were PCRamplified on a DNA Engine Tetrad 2 Peltier Thermal Cycler. PCR conditions were as follows: 95°C for 15 min, followed by 42 cycles of 94°C for 30 s, T a (58°C for COI and FGB7; 62°C for ND3 and MB2; 64.5°C for ADH5 and RAG1) for 30 s, 72°C for 30 s and a final extension of 72°C for 10 min. We tested a 3-lL aliquot of the PCR products on a 1.5% agarose gel to determine amplification success. We purified PCR products using 2 lL 109 diluted ExoSAP-IT (GE Healthcare Chicago, IL, USA) according to the instructions of the manufacturer. Cycle-sequencing was performed by GenePool Laboratory, Edinburgh, UK, on an ABI 3730 DNA analyser (Applied Biosystems) using the BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems).
Sequences were attained for all nine individuals across all loci. We aligned manually edited sequences in CODONCODE ALIGNER 3.7.1 (Codon-Code Corporation, Centerville, MA, USA) using the ClustalW algorithm. We submitted sequence data to GenBank (for accession numbers see Table 2 and Table S1). We used DNASP v5 (Librado & Rozas 2009) to assess nucleotide variation across sampled populations. Lastly, we carried out phylogenetic network analyses using POFAD software (Phylogeny of Organisms From Allelic Data; Joly & Bruneau 2006) based on three datasets, the first incorporating sequences from all six loci, the second including only mitochondrial loci (COI and ND3) and the third including only nuclear loci (ADH5, FGB7, MB2 and RAG1). Uncorrected pairwise genetic distances were initially calculated for each locus using MEGA6 (Tamura et al. 2013) and POFAD was then used to generate a single unstandardized distance matrix for each of the three multi-locus datasets. Networks were constructed using the Neighbor-Net algorithm in SPLITSTREE4 v4.14.4 (Huson & Bryant 2006).

RESULTS
Our analyses of both microsatellite and sequence data indicated that the Chestnut-banded Plovers of Kenya and Tanzania form a distinct genetic cluster separate from those of Namibia.
For microsatellite loci, measures of heterozygosity and allele ranges are included in Table 1. The mean number of alleles per locus was four (mean expected heterozygosity H e 0.548). AMOVA attributed 38.93% of molecular variance at microsatellite loci to differences between the Namibian individuals and those of Kenya/Tanzania, whereas there was no molecular variance detectable between Kenya and Tanzania (À0.41%, Table 3).
The results of STRUCTURE analysis strongly supported the presence of two genetic clusters ( Fig. 1c; K = 2, Delta K = 1622.6 vs. K = 3, Delta K = 20.6) corresponding to southern and eastern African populations. Individuals were assigned with high confidence to each cluster (membership coefficients > 0.955; mean 0.992). Similarly, these two genetic clusters were clearly identifiable based on PCA, with the first principal component (PC 1) explaining 34.87% of total variance and the second (PC 2) explaining 7.56% of variance at microsatellite loci (Fig. 1d).
Based on sequence data, fixed nucleotide substitutions were identified between eastern Africa (Kenya/Tanzania) and Namibia at mitochondrial loci (COI, 12 fixed substitutions across 615 sites; ND3, nine fixed substitutions across 398 sites), but not at nuclear loci (ADH5 693 sites, FGB7 817 sites, MB2 679 sites, RAG1 911 sites). In total, 3100 sites were analysed across four nuclear loci and just three substitutions were present among Tanzanian individuals. Average sequence divergence at mitochondrial loci was 2.36% (COI 1.95%, ND3 2.76%) compared with < 0.01% at nuclear loci.
Results of phylogenetic network analyses based on mitochondrial sequence data supported assignment of Namibian individuals to a distinct genetic cluster from those of Kenya and Tanzania (Fig. 2). This clustering was also reflected in the combined mitochondrial and nuclear analysis. However, extremely low sequence variation at nuclear loci led to limited genetic clustering based on nuclear sequences (Fig. 2), with Tanzanian individuals Table 2. GenBank accession numbers for sequence data generated from Chestnut-banded Plovers. exhibiting the only polymorphism, as described above.

DISCUSSION
We identified strong genetic divergence between the two currently recognized subspecies of Chestnut-banded Plover, C. p. venustus of eastern Africa and C. p. pallidus of southern Africa, based on microsatellite genotyping and mitochondrial sequence analyses. Genetic structure and estimated levels of inbreeding can be affected by including related individuals in the analysis (Rodr ıguez-Ramilo & Wang 2012). To test the robustness of our results, we carried out all microsatellite analyses with both the full dataset and a conservative dataset taking into consideration the probability of relatedness. Results were qualitatively similar in both cases (full dataset included in Table S1), indicating that strong genetic differentiation is indeed present between the two subspecies of Chestnutbanded Plover and is not an artefact of sample selection. In the conservative dataset, inbreeding coefficients (F IS ) indicated a significant heterozygote deficit among Kenyan individuals, suggesting that levels of inbreeding may be higher at the Kenyan site in contrast to the Namibian and Tanzanian sites.
intraspecific and interspecific divergence (mean divergence in COI sequences was 0.24% within species and 7.95% between sister species; Johnsen et al. 2010). In the Chestnut-banded Plover, we identified COI sequence divergence of 1.95% and overall divergence at mitochondrial genes (COI and ND3) of 2.36%. This is greater than the divergence between other closely related Charadrius species (dos Remedios 2013), the Kittlitz's Plover and St Helena Plover Charadrius sanctaehelenae (0.49% mitochondrial divergence), and similar to that between the Kittlitz's Plover and Madagascar Plover Charadrius thoracicus (2.54% mitochondrial divergence), but also comparable to subspecieslevel divergence among island and mainland populations of the Three-banded Plover Charadrius tricollaris and White-fronted Plover Charadrius marginatus (1.05 and 1.68%, respectively). In line with clear morphological differences, our analyses of microsatellite loci also confirmed that the two Chestnut-banded Plover subspecies belong to distinct genetic clusters, with genetic differentiation between subspecies far greater than between populations of the same subspecies.
Based on these diagnostic genetic differences as well as the documented morphological differences (Hayman et al. 1986), elevation to full species status may be warranted for C. p. pallidus and C. p. venustus. Our current results indicate that dispersal of Chestnut-banded Plovers between eastern and southern Africa is limited and, with continued isolation, these taxa are likely to diverge further. This strong genetic divergence is surprising given that for most other Charadrius species studied to date, dispersal and gene flow between populations occur over much greater continental distances (K€ upper et al. 2012, K€ upper & dos Remedios in press). For example, the Kittlitz's Plover exhibits high population connectivity and genetic homogeneity across Africa and its range overlaps that of the Chestnut-banded Plover (D'Urban Jackson et al. 2017). We suggest that behavioural factors such as habitat specialization and mating behaviour may be responsible for maintaining isolation between Chestnut-banded Plovers in southern and eastern Africa, and that unlike other Charadrius plovers, these behaviours may lead to non-oceanic barriers restricting dispersal in this species.
The Chestnut-banded Plover is a habitat specialist usually found within a short distance of saline or alkaline water (< 50 m during the breeding season, Hockey et al. 2005; < 1 km during the non-breeding season, Simmons et al. 2007), with a preference for areas devoid of vegetation (Johnsgard 1981). Populations of C. p. venustus in eastern Africa are resident with a strongly localized distribution that reflects this habitat specialization. Records suggest that wanderings are confined to the eastern Rift Valley, with rare cases of breeding by vagrants in ephemeral wetlands no more than 300 km south of the breeding population at Lake Manyara. This is far short of the closest population of the nominate subspecies, 2000 km away in northern Botswana (Nata delta; Simmons et al. 2007). Individuals of C. p. pallidus from Namibia, Botswana and South Africa exhibit greater movement and may be resident, nomadic or migratory (Delany et al. 2009). Yet, although non-breeding vagrants have been recorded as far north as coastal Mozambique, south of the Zambezi River (Hockey et al. 2005), no suitable breeding habitat has been recorded in Mozambique, Malawi, Zambia or Zimbabwe (N. Baker unpubl. data). The lack of genetic admixture found in our study suggests that dispersal is unlikely to enable mixing with eastern African C. p. venustus populations, although breeding populations of C. p. pallidus from the east coast of South Africa should to be sampled to confirm this.
Mating behaviour may also facilitate genetic differentiation relative to some other plover species. The Chestnut-banded Plover is monogamous (Hockey et al. 2005) and therefore likely to exhibit a high degree of philopatry (Saalfeld & Lanctot 2015, D'Urban Jackson et al. 2017. This behaviour is in contrast to the behaviour of sequentially polygamous plovers, in which one parent deserts their brood shortly after hatching to remate, sometimes in a new location, leaving their partner to provide care for the offspring (K€ upper et al. 2012). Mating behaviour and associated breeding dispersal may have a large impact on genetic differentiation. Overall, greater genetic structure has been identified among monogamous than polygamous plover populations (D'Urban Jackson et al. 2017) and this may contribute to the lack of genetic mixing between populations of the Chestnut-banded Plover.
With a behavioural tendency towards sedentariness, geographical isolation is likely to decrease population connectivity over a smaller scale than for more widely dispersing species. Subspecies C. p. venustus avoids both freshwater and coastal areas such that today, the inland saline wetlands of the East African Rift Valley offer the only suitable breeding habitat for 2000 km between central Tanzania and northern Botswana (N. Baker unpubl. data). In addition, these wetlands are bordered by some of the highest mountains in Africa, as well as some of the deepest freshwater lakes. To understand the evolutionary history of Chestnutbanded Plover populations and how they reached their current state of isolation, it is important to consider the influence of long-term climatic and geographical changes.
Although it is not possible to date the divergence of the two subspecies without further molecular clock analyses, studies in a range of African bird species have highlighted the importance of palaeoclimatic events and habitat changes on patterns of dispersal, vicariance and speciation (Fjelds a & Bowie 2008, Voelker & Light 2011, Oatley et al. 2012, Kahindo et al. 2017. Open savannah and grassland habitats have dominated much of sub-Saharan Africa (with the exception of the Guinea-Congolian forests) for the last few million years . During this time, dramatic landscape changes have occurred in eastern Africa. First, habitats here have been affected by longterm aridification with episodic periods of extreme humidity (Maslin et al. 2014, Voelker et al. 2014, Kahindo et al. 2017. Secondly, mountain ranges and deep lake basins continued to form as the central plateau of Africa uplifted. This resulted in the formation of the modern rift escarpments of the Magadi-Natron basin within the last 2 million years (Foster et al. 1997, Trauth et al. 2005, now home to C. p. venustus. In contrast, south-western Africa, home to C. p. pallidus, has remained relatively stable both geologically and tectonically for the last 3.5 million years, providing refugia for many taxa (Maslin et al. 2012). It seems inevitable that these factors would have influenced the evolutionary history of Chestnut-banded Plovers and their present-day distribution.
The elevation of subspecies C. p. venustus and C. p. pallidus to full species should result in a further review of the conservation status of each taxon. Chestnut-banded Plovers are currently considered to be Near Threatened (BirdLife International 2016) and are especially vulnerable as they depend on an extremely small number of sites. As many as 87% of individuals are reported to congregate in just three locations during the non-breeding season (Walvis Bay and Sandwich Harbour in Namibia, and Lake Natron in Tanzania; Simmons et al. 2007) with only eight sites holding more than 1% of the population. Our results confirm that Chestnut-banded Plovers disperse only short distances and therefore may not be capable of re-locating elsewhere should any of these sites become untenable. They would therefore be highly vulnerable in the face of global threats to wetland habitat (Davidson 2014, Dixon et al. 2016. The inclusion of key sites in the Ramsar List of Wetlands of International Importance (Walvis Bay, Sandwich Harbour and Lake Natron) should aid the protection of the species in these locations, yet without full protection of the surrounding areas, continued conservation efforts will be needed to maintain the quality of habitats in these biologically diverse regions.
Before reclassifying the two subspecies of Chestnut-banded Plover as distinct species, we recommend the collection and analysis of genetic samples from across the full southern African range of C. p. pallidus, including breeding populations not covered by our work. Augmenting our current genetic analyses with more broad-scale sequencing at additional loci would not only clarify the taxonomic status of these populations but also allow full delineation of geographical distributions and the estimation of divergence times in their evolutionary history. Additionally, we suggest the examination of morphological data, detailing phenotypic variation in plumage, body size, ecology, life history and behaviour across the range of the species. Regardless of their taxonomic status, the genetic differentiation identified between C. p. pallidus and C. p. venustus indicates that each should be managed independently as a distinct evolutionary unit rather than a combined Africa-wide population.
We thank Nelli R€ onk€ a, Rauri Bowie and two anonymous reviewers for their valuable comments on drafts of the manuscript. For collection of samples, we thank Bernard Amakobe, Sylvester Karimi, Christine Ochieng and the National Museums of Kenya. Sequence analysis was supported by NERC-Biomolecular Analysis Facility at the University of Edinburgh and at the University of Sheffield (NBAFE/S547). This study was funded by the NERC as part of the PhD research of N.dR. C.K. was supported by a Marie Curie IEF Postdoctoral Fellowship. Permits for fieldwork and collection of samples were provided by the National Council for Science and Technology, Kenya, the Namibian government, Ministry of Environment and Tourism, and the Tanzania Commission for Science and Technology (COSTECH) and Tanzania Wildlife Research Institute (TAWIRI).

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article: Table S1. Microsatellite genotype data for all individuals (14 loci; n = 73) and GenBank accession numbers for sequence data (six loci; n = 9).