Comparative phylogeography in a marine biodiversity hotspot provides novel insights into evolutionary processes across the Atlantic‐Indian Ocean transition

Intraspecific genetic variation is a key component of biodiversity, with higher diversity indicating greater resilience and population substructuring suggesting unique evolutionary histories. Comparative approaches, in which intraspecific genetic variation is assessed across multiple species, are powerful tools to identify evolutionary hotspots, but are still rarely applied at spatial scales relevant to conservation planning. Here, we use comparative phylogeography to understand patterns and potential drivers of genetic variation within a biodiversity and ocean warming hotspot.


| INTRODUC TI ON
With increasing human pressures and climate change, ecosystems worldwide face rapid modifications of environmental conditions resulting in biodiversity erosion (Ceballos et al., 2017;Panetta et al., 2018), with knock-on effects on human well-being (Mora et al., 2013). Representation and long-term persistence of biodiversity are central to conservation (Margules & Pressey, 2000;Moilanen et al., 2009), but such goals can only be achieved if organisms are able to respond to global changes with geographic range shifts, acclimatization or adaptation (Chevin et al., 2010). As such, conservation plans need to account for the processes shaping biological variability and species' adaptive potential, from ecosystems to genes Wagner & Fortin, 2013;Webster et al., 2017). As humanity enters the UN Decade of Ocean Science for Sustainable Development (2021)(2022)(2023)(2024)(2025)(2026)(2027)(2028)(2029)(2030), the importance of monitoring and maintaining genetic diversity within species to protect the adaptive capacity, community function and resilience within populations is widely acknowledged Thomson et al., 2021), although not necessarily fully accounted for at both national and international scales (Hillebrand et al., 2020). Intraspecific genetic diversity provides the basis for evolutionary processes that enable species to adapt to environmental changes (DeWoody et al., 2021;Schindler et al., 2015), and at the community level, genetic diversity has been positively associated with productivity, functioning and stability in various marine systems (DuBois et al., 2021;Phair et al., 2020;Raffard et al., 2019;Salo & Gustafsson, 2016;Sjöqvist & Kremp, 2016). Genetic diversity is impacted by effective population size (Kimura, 1985), and influenced by both species characteristics, such as dispersal (Ryman et al., 2014), reproductive strategies (Nunney, 1993) and historical demographic variations (Frankham, 1995), as well as landscape features (Selkoe et al., 2016;Wright et al., 2015).
The importance of genetic diversity for adaptation and short-term evolutionary potential is widely acknowledged (Hughes et al., 2008;Stange et al., 2021); yet, the key factors shaping genetic diversity remain ambiguous (Hewitt, 2000;Ludt & Rocha, 2015;Teixeira & Huber, 2021;Toms et al., 2014). For example, it has been suggested that biogeography, specifically core vs. edge effects (i.e. the abundant-centre hypothesis; Eckert et al., 2008), and environmental gradients and habitat stability over time (Hewitt, 2000) play critical roles in determining intraspecific genetic diversity. However, the drivers of intraspecific variation across species have largely been tested in isolation, and most of the few studies that compared multiple drivers of genetic diversity did so at spatial scales too large to be informative for conservation (Manel et al., 2020;Wyborn & Evans, 2021). Multi-species comparisons in particular have immense power to detect patterns in the spatio-temporal variation of genetic diversity at global scales (De Kort et al., 2021), although patterns tend to differ between regions and systems (Gratton et al., 2017;Manel et al., 2020;Miraldo et al., 2016). Inferences from multi-species studies may be limited by various factors including uneven sampling strategies, comparisons across taxa with different marker types or divergent evolutionary dynamics, but are likely to still resolve broad-scale patterns even where some species may have vastly different population genetic patterns (Page & Hughes, 2014;Wright et al., 2015). Yet, despite the potential power of multi-species analyses in shedding light on evolutionary population patterns and processes, several regions globally remain under-represented.
Population structure, as measured by the distribution of genetic variation, has been recorded across many marine species with a broad range of life-history characteristics, including broadcast spawners and direct developers; yet, few concordant genetic clines have been documented across multiple species Schiebelhut & Dawson, 2018;Stanley et al., 2018;Teske et al., 2011;Wares, 2002;Wright et al., 2015). Identifying multispecies patterns of genetic differentiation is particularly relevant for understanding regional processes that structure populations, and to disentangle the determinants of connectivity and gene flow in the marine environment at larger scales (Kuo & Avise, 2005;Manel et al., 2020). Furthermore, integrating community-level genetic diversity and differentiation into spatial conservation plans and management is central to ensuring the success of long-term conservation objectives (Nielsen et al., 2017;Phair et al., 2021;Thomson et al., 2021;von der Heyden, 2017).
Environmental and habitat characteristics, both contemporary and historical, are known to shape the spatial distribution of genetic diversity and differentiation (De Kort et al., 2021;Manel et al., 2020;Selkoe et al., 2016). In marine ecosystems, the main seascape features shaping spatial genetic patterns are likely ocean currents (which transport gametes and larvae, but which can also create barriers to dispersal) and habitat discontinuities (Pascual et al., 2017;Schunter et al., 2011;Selkoe et al., 2016;Toms et al., 2014). In particular, sea surface temperature (SST), depth and salinity have been identified as common drivers of genetic variation in various marine taxa (Riginos et al., 2016;Selkoe et al., 2016). Furthermore, including historical environmental variation, and identifying areas of stability in the period from the Last Glacial Maximum (LGM) to today, provides additional insights into the role of historical processes in shaping extant patterns of genetic variation (De Kort et al., 2021;Ludt & Rocha, 2015;Phair et al., 2019).
The southern African coastline, spanning the transition between the Atlantic and Indian Ocean biomes, offers an ideal system to study biogeographic hypotheses, as it is divided into several bioregions that are defined by distinct evolutionary histories and K E Y W O R D S conservation genetics, core-edge, evolutionary potential, genetic diversity, marine invertebrates, mtDNA, seascape genetics environmental characteristics (Griffiths et al., 2010). These bioregions broadly consist of the cool, nutrient-rich cool-temperate region on the west coast, and the warm, nutrient-poor warm-temperate and sub-tropical regions on the south and east coasts, with a region of overlap between the Atlantic and Indian oceans, along the South-West coast (Figure 1). The South African coastline encompasses a large variety of ecosystems, including rocky and sandy shores, kelp forests, estuaries and coral reefs, as well as contrasting environmental conditions (Branch & Branch, 2018) that drive high levels of coastal marine biodiversity. Four major phylogeographic breaks have been identified, and most coastal species, including organisms with high dispersal capacities, are divided into genetic lineages whose distributional ranges broadly match the bioregions (Figure 1; Teske et al., 2011;Wright et al., 2015), with evidence for temperature and salinity as important drivers of potential local adaptation Phair et al., 2019;Teske et al., 2021).
In this study, we compiled published data of genetic diversity from mitochondrial DNA (mtDNA) markers into the most comprehensive dataset for coastal marine species in South Africa generated to date to test multi-species patterns of evolutionary diversity. First, we used a biogeographic approach to investigate the spatial distribution of multi-species genetic diversity (haplotype and nucleotide diversity) and differentiation (Φ ST ) along the South African coast, and compare the levels of diversity and differentiation between biogeographic regions and species' habitat preferences. Secondly, we used a seascape genetics approach to test the effect of environmental and biogeographic variables on multi-species genetic diversity and differentiation. We predict that across multiple species, particularly for obligate rocky shore taxa, genetic diversity will be elevated in areas with long-term habitat and temperature stability, particularly around the south-western and southern coastlines (Phair et al., 2019;Toms et al., 2014), which would suggest shared evolutionary dynamics shaping present-day genetic variation. Not only does our work provide novel insights into the evolutionary dynamics of one of the world's most iconic marine systems but it also contributes towards improving protected area networks in the region.

| Genetic data selection
Mitochondrial DNA markers are useful tools to elucidate large-scale patterns on which to build further analyses , and are powerful datasets to perform multi-species analyses (Crandall et al., 2019;Nielsen et al., 2017;Selkoe et al., 2014). As mtDNA continues to be the most prevalent genetic marker employed in molecular studies of southern African marine biodiversity, and because it has proven useful to identify even fine-scale drivers of genetic differentiation and diversity Selkoe et al., 2014, F I G U R E 1 The spatial arrangement of the number of species included in the filtered dataset along the coastline, averaged within 30 × 30 km squares, shown for species inhabiting hard substrate (a) habitats and soft substrate (b) habitats. The range in sea surface temperatures (SST; c) and mean SST (d) over 2000-2014 (from Bio-ORACLE database; Tyberghein et al., 2012) is also shown, with the following bioregions overlaid onto the maps (d): Cool Temperate (CT), South-West Coast (SW), Warm Temperate (WT) Sub-Tropical (ST), and Tropical (T) 2016), we compiled mtDNA data of population structure for marine animal taxa in South Africa from the peer-reviewed published literature. We searched the ISI Web of Science using the following terms: South Africa AND ('mtDNA' OR 'mitochondrial') AND ('genetic structure' OR 'genetic differentiation' OR 'genetic diversity') in June 2018. This search returned 678 records, which we further screened to retain only articles that dealt with South African marine animals (i.e. discarding terrestrial and freshwater animals, as well as plant species such as seagrass and algae). Appropriate articles that were not recovered by the Web of Science search were supplemented by the authors' own knowledge of the field in early 2021. We refined the resulting dataset by discarding studies on pelagic species, as they tended to have less consistent sampling and are under different ecological and environmental pressures than coastal species.
We also excluded species which are not endemic to southern Africa, i.e. only native South African species were included, as species with global distributions will likely be under different evolutionary constraints than endemic species.
Summarized statistics were collated from compiled studies, in which we recorded published parameters of haplotype diversity (h) and nucleotide diversity (π), along with sample size, and spatial coordinates (longitude and latitude). To account for differences between molecular markers, we chose to only include data from the mtDNA cytochrome b and cytochrome oxidase c subunit I regions.
We retained only sites with at least eight individuals sampled per site, from at least nine sites total, with at least two sites per bioregion, to avoid bias due to low sampling size (Hale et al., 2012). We further removed species if they were not sampled in all bioregions within their geographic range (Table 1; Table S1). In addition to measures of diversity, we also categorized species into being associated with hard (predominantly rocky shores) or soft (predominantly sandy systems) substrates.
To assess genetic diversity patterns along the South African coastline, and how they relate to biogeographic patterns, we compared genetic diversity values between predefined biogeographic regions that were based on previously identified bioregions and phylogeographic breaks (Teske et al., 2011): Cool Temperate (CT),

South-West Coast (SW), Warm Temperate (WT), Sub-Tropical (ST)
and Tropical (T). The tropical bioregion was not included in the analyses, as there was insufficient sampling. Bioregions were included in subsequent analyses if at least five species were sampled within them, with each species needing at least two sample sites within that bioregion.
We also calculated Φ ST values within each bioregion to measure fine-scale population differentiation along the coastline. A different set of species were included in the Φ ST analyses (Table S2), as raw genetic data were needed to calculate within-bioregion genetic variation. The same set of filters as above were applied to the Φ ST analyses, with the additional filter of requiring species to have phylogeographic breaks aligned with biogeographic breaks (so as to not bias within-bioregion Φ ST estimates). We calculated withinregion Φ ST with analysis of molecular variance (AMOVA) tests, using  Nei, 1993). To estimate the degree of structure among groups within a region, we ran AMOVAs including all sample sites per bioregion in Arlequin v.3.5 (Excoffier & Lischer, 2010), using permutation tests with 10,000 permutations to determine significance using a p-value of .05.

| Predictors of genetic variation
We tested the effect of four contemporary and 12 historical climatic predictors expected to influence the spatial genetic patterns of marine organisms (Selkoe et al., 2016) on genetic diversity and differentiation. Including both contemporary and paleo-climatic data was considered appropriate because mitochondrial genetic variation is expected to reflect historical climatic change particularly well compared to more quickly evolving markers, such as microsatellites (Epps & Keyghobadi, 2015). The 16 predictors were obtained from remote sensing at a 5-arc-minute spatial resolution, and downloaded from the Bio-ORACLE platform using the R package 'sdmpredictors' As the position within a species' geographic range is expected to influence intraspecific genetic diversity (de Kort et al., 2021;, we also calculated each site's 'range position' as a predictor of genetic variation per species. The centre point of each species' range and the distance from each sample site to the range centre were measured along the coastline with the 'roadgraph' tool in QGIS (Hanson & Seeger, 2017).

| Statistical analyses
First, h and π were log-transformed to achieve normal distributions.
We fitted a linear model with genetic diversity (h and π) as a response variable, and bioregion or habitat (preferred substrate) as an explanatory variable. We used an analysis of variance (ANOVA) to evaluate the significance of the effect of each explanatory variable on haplotype and nucleotide diversity. We then performed a Dunn's test of multiple comparisons using ranks (Dunn, 1964)  We calculated pairwise Pearson's correlation between environmental predictors ( Figure S1). Within each dataset, autocorrelation was removed by eliminating one predictor when Pearson's correlation coefficient was higher than 0.7. A total of 11 variables were removed to account for autocorrelation (see Figure S1), with the final dataset containing four environmental predictor variables: mean sea surface temperature (SSTmean), range in sea surface temperature (SSTrange), mean sea surface salinity (SSSmean), and stability in sea surface temperatures from the present day to the Mid-Holocene (SSTstability_MH), as well as one biogeographic predictor variable: distance from the centre of the species' geographic range (Range_position).
The effect of the environmental predictors on genetic diversity and differentiation was tested using linear mixed models.
First, the environmental predictors were standardized using log- All models were fitted using the function 'lmer' of the lme4 R package version 1.1-18-1 (Bates et al., 2015). To select the most parsimonious models, we performed a backward elimination of randomeffect terms followed by backward elimination of fixed-effect terms in the full models. The effect of removing random-effect terms was estimated by likelihood ratio tests, and the effect of removing fixedeffect terms by an F-test using Satterthwaite's method for approximating degrees of freedom (Fai & Cornelius, 1996). Model selection was performed using the 'step' function of the R package lmerTest version 3.0-1.9003 (Kuznetsova et al., 2017).

| Literature review and genetic patterns
The ISI Web of Science search returned 697 articles of which 85 studied genetics of marine organisms. We retained 45 studies that fulfilled the criteria given in the Methods, and selected a final set of data from 13 published articles and one unpublished dataset that used mitochondrial markers (Table S1)  sample (min = 9 and max = 90) across all species and sites (Table   S1). Among bioregions, the average sample size ranged from 19.1 to 30.0 (Table 1).
Across the coastline, haplotype diversity ranged from 0.011 to 1 with a mean of 0.765 and a median of 0.85, and nucleotide diversity ranged from 0.0002 to 0.11 with a mean of 0.0082 and a median of 0.0047 (Figures 2, 3; Figure S2). The SW bioregion had the highest average h (0.882 ± 0.102) and highest average π (0.011 ± 0.0158).
The ST bioregion had the lowest average h (0.679 ± 0.250) and the CT bioregion had the lowest average π (0.0056 ± 0.0054; Table 1).
Bioregion had a significant effect on both h and π, and Dunn's posthoc test showed that the ST bioregion had significantly different values from those of SW and WT (as shown by the different shapes in Figure 2). While soft substrate species did show higher average diversity values of both h and π than hard substrate species (mean h of 0.804 ± 0.150 and mean π of 0.0166 ± 0.0278 for soft-substrate species; mean h of 0.758 ± 0.248 and mean π of 0.0056 ± 0.0044 for hard-substrate species; Table S1), the effect of substrate type was only significant on nucleotide diversity (Figure 2).
Within-region Φ ST data were calculated for 11 species (Table S2)

| Environmental and biogeographic drivers of genetic patterns
Most fixed-effect terms, including range position, mean sea surface salinity and temperature and SST stability since the Mid-Holocene, had no significant effect on either measure of genetic diversity (Table 2). For haplotype diversity, the most parsimonious model included a significant random-effect term for species, and significant fixed-effect terms for range in sea surface temperature (SSTrange). Haplotype diversity was positively correlated with SSTrange (Marginal R 2 = 0.052; Figure 5). No significant fixed-effect or random-effect term other than species was found in the π models (Table 2). Within the Φ ST models, the range in mean SST, as well as the length of the bioregion, had no significant effect on the genetic structuring within regions, with the most parsimonious model including only species as a random-effect term (Table 3).

| DISCUSS ION
Despite South Africa's status as a global marine biodiversity hotspot, the processes driving and maintaining the spatio-temporal patterns of evolutionary diversity of marine species across the Atlantic-Indian Ocean transition zone remain understudied. Using comparative analysis of the most complete dataset of genetic variation of a wide array  Table   S1), we investigated the environmental and biogeographic processes shaping coastal biodiversity in the region. We describe multi-species evolutionary hotspots, provide new insights into genetic structuring within bioregions and discuss the environmental variables that are highly correlated with community-level genetic diversity. Our findings not only add significantly to our understanding of the evolution of this important marine biodiversity hotspot but also provide new data relevant for biodiversity conservation in the region.

| A new state of knowledge of phylogeography for a global marine biodiversity hotspot: From patterns to conservation
The Atlantic-Indian Ocean transition zone has seen a steady increase in the number of population genetic studies, predominantly on single species, calling for critical cross-species inferences on the evolutionary processes shaping spatial genetic diversity in the region.
We recovered broadly similar patterns between h and π, which was expected, as they previously have been shown to co-vary (Goodall-Copestake et al., 2012); yet, our results also indicate some discrepancies between the two statistics. For example, π displays higher diversity along the WT region and lower diversity along the CT region compared to h (Figures 2 and 3). These differences could be driven by the fact that haplotype diversity per sample is strictly based on haplotype frequencies, while nucleotide diversity is dominated by mean sequence divergence (Nei, 1987). Hence, h incorporates population dynamics such as gene flow, making it more variable among samples than π, particularly when sample size is small. Consequently, h may reflect more recent population demographic events, while π displays the influence of older demographic fluctuations. This has important conservation relevance, as these two metrics may potentially lead to the identification of different priority areas if the objective is to protect multiple high diversity areas. The disparity between h and π patterns found here mirrors that found among five coastal species along just the CT region (Nielsen et al., 2017), and suggests that both metrics are needed to fully capture the breadth of genetic diversity when used to inform conservation and management (Beger et al., 2014).
Our results also suggest that mean genetic diversity values (i.e. averaged across sites and species) differ significantly between bioregions, with the ST region being the most distinct ( Figure 2).
The distinctiveness of that region could be explained by its specific oceanographic characteristics, with a narrow continental shelf and fast-flowing nearshore currents (Branch et al., 2016) and/or by historical factors . The ST region also shows higher levels of phylogenetic and functional diversity than other regions of the South African coast for sparid fishes , providing more evidence of its genetic distinctiveness for various marine organisms.
Notably, our analysis of genetic diversity confirms that the SW π compared to other areas. This region has also been shown to have increased species diversity, as well as endemicity, of coastal invertebrates (Awad et al., 2002). This is noteworthy because from both a habitat (particularly rocky shore) and temperature perspective, this region seems to have remained relatively stable from the LGM to present and provided a refugial area for marine species in the region. For example, Toms et al. (2014) showed that during the LGM and sea-level lows of up to 120 m, there were significant changes in coastline composition, with much of the present-day rocky shore lost. However, the region around the south-west Cape (as well as a small area further east), has consistent rocky shore habitat, which may well have acted as refugial site for obligate rocky shore species during sea-level fluctuations.
Temperature also plays an important role in determining habitat suitability of coastal marine species through time, with both Phair et al. (2019) and  showing that the SW bioregion remained suitable for a range of marine species.
As refugia have shown to enhance both species richness (Pianka, 1996;Tittensor et al., 2010) and genetic diversity (Hewitt, 2000;Provan & Bennett, 2008;Carnaval et al., 2009), this provides a suitable explanation for the patterns emerging from the SW re- gion. An additional factor could be that some areas of high genetic diversity are likely to be areas of admixture (i.e. where historically distinct evolutionarily significant units overlap; Maggs et al., 2008 (Maggs et al., 2008), hence increasing the genetic diversity in that region, although this pattern does not hold across all species spanning this biogeographic break (Teske et al., 2011;von der Heyden et al., 2008;Wright et al., 2015). From a conservation perspective, our results provide additional evidence for the uniqueness of this area, which is already recognized as the 'Seas of Good Hope' Ecologically and Biologically Significant Area (EBSA) driven by factors such as biological diversity, and highlighting the need for additional management measures that protect both biological and evolutionary diversity within the EBSA.
Interestingly, there were a number of divergent patterns of genetic diversity, supporting the need for further studies describing phylogeographic patterns of marine species. For example, higher levels of haplotype diversity were found in the SW for Bullia rhodostoma, Jasus lalandii, Palaemon peringueyi, and higher levels of nucleotide diversity were found in the same region for Acanthochiton garnoti, J. lalandii and Oxystele tigrina ( Figure S3). Some species showed decreasing diversity levels from west to east (Afrolittorina knysnaensis, J. lalandii, Parechinus angulosus, P. peringueyi and Siphonaria serrata), whereas others showed higher diversity in the ST region (Afrolittorina africana, Nassarius kraussianus and Siphonaria F I G U R E 5 Regressions are plotted for the best fitting model, including the relationship between haplotype diversity and the range in contemporary sea surface temperatures (SST range). Lines are coloured to represent regressions for each group of the random-effect term (species; see Table S1 for species abbreviations). As the most parsimonious models for π and Φ ST did not include any fixed-effect terms, the regressions are not shown  Figure S3). Those are typically species with the core distribution along the southern and eastern portion of South Africa's coastline (WT and ST bioregions) and for which the warm-temperate region marks the end of their range and may therefore display lower levels of genetic diversity.
In addition to genetic diversity, we also report the most comprehensive review on genetic differentiation within and across bioregions along the South African coastline. The most recent review on genetic structuring of marine species (Teske et al., 2011) identified phylogeographic breaks that are broadly concurrent with the boundaries between bioregions, but also noted several instances where the two do not overlap. Although structure between bioregions was recognized, patterns within bioregions were not accounted for. Here, we extend previous analyses by comparing genetic structuring within each bioregion, and found that the  (Nielsen et al., 2017.

| The effects of environment and geography on multi-species evolutionary patterns
Climatic variability and habitat complexity have long been considered important features shaping species diversity patterns (Gagné et al., 2020), but are not as often considered in understanding the processes driving molecular diversity. Recent multispecies studies have assessed these drivers of genetic diversity on a global scale, suggesting that environmental variables such as temperature (Manel et al., 2020) or position within a species' geographic range (De Kort et al., 2021) are important processes shaping genetic diversity. Here, we considered these processes on a more localized spatial scale, finding that neither mean temperature nor geographic range position significantly co-varies with intraspecific genetic diversity (Table 2). Instead, the linear mixed models revealed that the range in contemporary SSTs is positively correlated with intraspecific haplotype diversity values ( Figure 5). This distinction between drivers of genetic diversity highlights how global-scale analyses may not capture biodiversity trends occurring at more localized spatial scales (Sax & Gaines, 2003). Our results also contrast with previous studies within South Africa, which suggest that mean SST (Phair et al., 2019;Teske et al., 2019) and range position  are strong drivers of genetic diversity. However, these South African studies were conducted on single species, and thus cannot account for the large differences between species and the broad evolutionary processes shaping multi-species diversity. In our comparative multi-species analysis, it was not mean temperature, but range in temperature, that was significantly correlated with haplotype diversity. Range in temperature being an important driver of genetic diversity is not surprising, as environmental heterogeneity is regarded as one of the most important factors shaping species richness (Stein et al., 2014). This idea can be expanded to the level of genetic diversity by individuals within populations experiencing a higher range of SSTs requiring a larger range of both phenotypic and genetic variation to persist under this type of balancing selection (Banks et al., 2013). Evidence for this hypothesis has been shown by high climatic variability leading to higher levels of intraspecific diversity in bull trout in the Columbia River Basin (Kovach et al., 2015) and differential phenotypic expression in sticklebacks (Garduño-Paz et al., 2010).
In comparison with haplotype diversity, variability in nucleotide diversity did not correlate with any environmental or biogeographic predictor variables, suggesting that different diversity statistics reflect contrasting evolutionary histories (Epps & Keyghobadi, 2015), and highlighting the importance of comparing different metrics to understand evolutionary processes for conservation. Life-history traits may also render species more sensitive to changes in different molecular metrics, such as long-lived species being more prone to loss of allelic richness, and short-lived species more prone to loss of haplotype diversity (González et al., 2020), although this could not be tested here given the generally poor state of knowledge on the life histories of the marine species investigated here, especially the invertebrates that form the bulk of our study.
Range position was not a significant predictor of either h or π. This was unexpected, as distance from the centre of a species' distribution has been shown to be a strong driver of global genetic diversity across taxa (de Kort et al., 2021), as well as being a significant determinant of diversity for single species within South Africa . However, evidence for the core-edge hypothesis remains highly inconsistent (Dallas & Santini, 2020;Lira-Noriega & Manthey, 2014) and is probably largely influenced by life-history traits (Ellegren & Galtier, 2016). For example, many of the species included here have high dispersal capabilities and broad geographic ranges, making them less likely to display an abundant-centre pattern in genetic diversity (Ntuli et al., 2020).
Habitat (here broadly classified into soft and hard substrates) also had no significant effect on diversity in the mixed linear models, corroborating the results of Dalongeville et al. (2016) in the Mediterranean Sea, who found that habitat types (soft, rocky and seagrass beds) showed no significant correlation with the genetic diversity of 13 fish species. Habitat availability is thought to be a crucial component in genetic differentiation (Binks et al., 2019), but this is probably a less important factor in this system, as hard and soft substrates are evenly distributed throughout the region (Tolley et al., 2019) and habitat types are not always clearly differentiated. For example, approximately 42% of South Africa's coastline comprises 'mixed shore', that is, it has both rocky and sandy components.
The additional predictors 'bioregion size' and 'range in mean SSTs' had no effect on Φ ST , even though the latter was an important predictor for patterns of diversity. It is likely that a combination of life-history traits and near-shore oceanographic currents influences patterns of within-bioregion genetic structuring (Iacchei et al., 2013;Orsini et al., 2013;C. Sjöqvist et al., 2015), but these data are unfortunately largely unavailable for the region and many of the species. While the effects of life history, biogeography, temperature and oceanographic features on genetic diversity of South African marine species have been explored Phair et al., 2020;Teske et al., 2011), they are rarely combined into multivariate, multispecies analyses. These types of integrative studies are needed to capture the complex spatiotemporal interactions shaping the diversity along this dynamic coastline.
Our study suggests that factors shown to influence global genetic diversity patterns, mainly temperature (Manel et al., 2020) and range position (de Kort et al., 2021), may not be as strong of predictors of multi-species genetic variation on smaller spatial scales. However, our study does have some limitations which may have led to the lack of strong predictors of significant associations between genetic diversity and environmental and biogeographic variation. For instance, we were limited in only using data from mtDNA markers, as other marker types, such as genome-wide SNPs, do not exist for as wide an array of taxa within the region.
Furthermore, both h or π scale not only by population size, but also by mutation rate, which may be why we did not find simple relationships with genetic diversity and ecological variation. We were constrained by many studies having sampled few individuals at many sites, and after applying a set of filtering criteria in order to account for unbalanced sampling, we were left with smaller sample sizes. Understanding the amount of data sufficient to capture multi-species evolutionary patterns remains difficult, as power analyses used to model species-area relationships are not always suited to determine sufficient sampling genetic diversity assessments (Phillips et al., 2019). Our limitations in available molecular information, species and regions sampled and detailed life-history traits of study species, highlight important knowledge gaps in the region, which should be prioritized to further characterize the phylogeography of this marine biodiversity hotspot.

| The future of molecular research on southern Africa's marine realm
Ten years ago, Teske et al. (2011) provided the first comprehensive overview of genetic studies conducted on southern African marine organisms, where they recognized the link between species-level biogeographic transition areas and infraspecific phylogeographic breaks. They also noted that the CT and tropical bioregions had been poorly studied at the time. In our analyses, we found that genetic information available for the CT bioregion is now much improved, but that the tropical bioregion remains in need of further phylogeographic study. Indeed, the tropical bioregion was not included in this study due to a lack of suitable data. Additionally, studies with sampling domains spanning both South Africa and the neighbouring countries (Namibia in the west and Mozambique in the east) are limited, but will be vital to identify many of the potential intraspecific lineages that may remain undescribed (Beheregaray, 2008). Teske et al. (2011) also reported the lack of phylogeographic research conducted on commercially important species, which this study corroborates, with only 37% of the total studies being on exploited species. Yet, there has been notable progress with regard to both the number of species and the number of molecular markers included in recent population genetic studies in South Africa, with 38 studies including multiple markers at the time of the literature search, and many also including multiple species. Arguably, the most significant advancement in molecular research is the advent of genomic analyses; yet, the number of available genome-wide datasets in South Africa is a magnitude lower than genetic datasets, and rarely exists for multiple species (Tolley et al., 2019). Going forward, we expect to see an increase in the use of genomic markers, as these not only have much greater power to detect genetic structure by means of selectively neutral markers but can also be used to identify gene-environment interactions and highlight intraspecific adaptive signals (Helyar et al., 2011). As neutral and adaptive genetic patterns have been shown to differ (Boulanger et al., 2021;Liggins et al., 2020;Milano et al., 2014;Torrado et al., 2020), comparing adaptive signals across multiple taxa would provide an interesting addition to the present study, and could allow identifying environmental and climatic features causing local adaptation of marine species. For example,  showed that genomic vulnerability under climatic change differs between sympatric species requiring alternative solutions for protecting intraspecific diversity.
South Africa is one of the leading countries to integrate molecular data into conservation planning frameworks (Nielsen et al., 2017;Rouget et al., 2003;Scoble & Lowe, 2010;von der Heyden, 2009). We believe that the uniquely situated South African coastline will continue to supply novel insights into the genetic and genomic consequences of global climate change in the temperate southern hemisphere, with data informing conservation policies in this highly biodiverse marine realm.

ACK N OWLED G EM ENTS
At the time of the research, Alicia Dalongeville was funded by a postdoctoral fellowship of the Claude Leon Foundation.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13534.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data and code are publicly available on the following GitHub re-