Land snail biogeography and endemism in south-eastern Africa: Implications for the Maputaland-Pondoland-Albany biodiversity hotspot

RESEARCH ARTICLE Land snail biogeography and endemism in south-eastern Africa: Implications for the Maputaland-Pondoland-Albany biodiversity hotspot Sandun J. PereraID1¤a*, David G. Herbert2¤b, Şerban Procheş1, Syd Ramdhani3 1 School of Agricultural, Earth and Environmental Sciences, University of KwaZulu-Natal, Durban, South Africa, 2 KwaZulu-Natal Museum, Pietermaritzburg, South Africa, 3 School of Life Sciences, University of KwaZulu-Natal, Durban, South Africa ¤a Current address: Department of Natural Resources, Sabaragamuwa University, Belhuloya, Sri Lanka ¤b Current address: Department of Natural Sciences, National Museum Wales, Cardiff, United Kingdom * sandun.perera@appsc.sab.ac.lk, sandun.perera@gmail.com Abstract Invertebrates in general have long been underrepresented in studies on biodiversity, bioge- ography and conservation. Boundaries of biodiversity hotspots are often delimited intuitively based on floristic endemism and have seldom been empirically tested using actual species distributions, and especially invertebrates. Here we analyse the zoogeography of terrestrial malacofauna from south-eastern Africa (SEA), proposing the first mollusc-based numerical regionalisation for the area. We also discuss patterns and centres of land snail endemism, thence assessing the importance and the delimitation of the Maputaland-Pondoland-Albany (MPA) biodiversity hotspot for their conservation. An incidence matrix compiled for relatively well-collected lineages of land snails and slugs (73 taxa in twelve genera) in 40 a priori oper-

ational geographic units was subjected to (a) phenetic agglomerative hierarchical clustering using unweighted pair-group method with arithmetic means (UPGMA), (b) parsimony analy- sis of endemicity (PAE) and biotic element analysis (BEA). Fulfilling the primary objective of our study, the UPGMA dendrogram provided a hierarchical regionalisation and identified five centres of molluscan endemism for SEA, while the PAE confirmed six areas of ende- mism, also supported by the BEA. The regionalisation recovers a zoogeographic province similar to the MPA hotspot, but with a conspicuous westward extension into Knysna (towards the Cape). The MPA province, centres and areas of endemism, biotic elements as well as the spatial patterns of species richness and endemism, support the MPA hotspot, but suggest further extensions resulting in a greater MPA region of land snail endemism (also with a northward extension into sky islands—Soutpansberg and Wolkberg), similar to that noted for vertebrates. The greater MPA region provides a more robustly defined region of conservation concern, with centres of endemism serving as local conservation priorities. PLOS ONE PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 1 / 26 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Perera SJ, Herbert DG, Procheş Ş, Ramdhani S (2021) Land snail biogeography and endemism in south-eastern Africa: Implications for the Maputaland-Pondoland-Albany biodiversity hotspot. PLoS ONE 16(3): e0248040. https://doi. org/10.1371/journal.pone.0248040

Editor: Geerat J. Vermeij, University of California, UNITED STATES Received: July 4, 2020 Accepted: February 18, 2021 Published: March 4, 2021 Copyright: © 2021 Perera et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Funding: SJP was funded by the South African National Research Foundation (NRF; https://www. nrf.ac.za/) grants through ŞP (African Origins platform doctoral bursary, and incentive funding for rated researchers), a University of KwaZulu- Natal (UKZN; https://www.ukzn.ac.za/) doctoral research grant, Gay Langmuir bursaries for Conservation Biology (2010, 2012) and a UKZN

Introduction Terrestrial molluscs (Mollusca: Gastropoda; hereafter land snails) with a global species rich- ness estimated at over 25,000 [1, 2] are not distributed evenly across the globe. Although sev- eral island land snail faunas are reported to have exceptional species richness and high degrees of endemism [3, 4], and are thus recognized as hotspots for molluscan conservation, similar continental hotspots are not widely recognized. Global biodiversity hotspots [5, 6] have been delimited based on floristic endemism, often with the assumption of congruent distribution patterns for animals, but such congruence of endemism between plants and land snails is yet to be well established. Invertebrates in general have long been underrepresented in studies on biodiversity, biogeography and conservation, primarily because much of their diversity is still to be documented, while their taxonomy is often not sufficiently robust and the knowledge on their distribution is incomplete [5, 7–11]. Sub-Saharan Africa has a rich terrestrial malaco- fauna with an estimated 6,000 species [12]. Land snails in South Africa (SA) are taxonomically better resolved and distributionally better understood compared to many other groups of invertebrates, at least in eastern SA [13]. Over 650 land snail species are recorded for southern Africa [14], while 525 species (with c. 90% endemism) occur in SA alone [15]. However, their biogeography has never been systematically studied in southern Africa, except for qualitative

studies [13, 14]. The terrestrial malacofauna of eastern SA includes relict Gondwanan lineages and speciose radiations of more recent origin exhibiting high levels of local endemicity [13]. The indigenous land snail fauna of eastern SA includes well over 270 species, with at least 173 (63%) species being endemic or near-endemic [13]. With high levels of diversity and ende- mism, land snails make an excellent indicator group to start exploring the invertebrate zooge- ography of south-eastern Africa (SEA). Narrow-range distribution patterns commonly observed in land snails, stemming from their low dispersal ability, make them a suitable candi- date group to study regional endemicity and historical biogeography [4, 16, 17], and to deter- mine priority areas for conservation [18, 19], at least for the best studied groups. Boundaries of biodiversity hotspots have seldom been empirically tested on actual species distribution ranges and not for endemic invertebrate fauna. The Maputaland-Pondoland- Albany (MPA) biodiversity hotspot of SEA has been delimited based on intuitive phytogeo- graphical units such as the Maputaland-Pondoland region and the Albany centre of floristic endemism [20]. Possessing more than 1,900 species of endemic higher plants, this hotspot spans the eastern coast of southern Africa from Maputo in Mozambique to Port Elizabeth in Eastern Cape province of SA and extends inland to reach an altitude of 1,800 m a.s.l. along the south-eastern portion of the Great Escarpment [21]. The MPA hotspot has recently been fur-

ther explored in terms of its vertebrate endemism, documenting 62 endemic and 60 near- endemic species, and further recognizing a greater MPA region of vertebrate endemism [22– 24]. The greater MPA region with 166 endemic vertebrate species is an expansion of the MPA hotspot’s area by 59%, to include 168% more endemic species, indicating a higher animal endemicity in areas adjacent to the original flora-based hotspot [24]. Here we numerically analyse the zoogeography of the terrestrial malacofauna in SEA for the first time, also assessing the importance and the delimitation of the MPA biodiversity hot- spot for land snail conservation. This is achieved through numerical biogeographical analyses and mapping of species richness and endemism patterns followed by comparison of the results with the current boundary of the MPA hotspot. Biogeographical analyses are presented with three numerical approaches, our primary objective being a phenetic agglomerative hierarchical clustering of incidence data to propose a land snail-based biogeographical regionalisation and secondly to identify centres of land snail endemism, for which consensus and congruency is PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 2 / 26 post-doctoral scholarship (2013). SJP received study leave and a travel grant from the University Grants Commission and the Sabaragamuwa University of Sri Lanka to conduct this study in South Africa and a travel grant from Unitas

Malacologica to attend the World Congress of Malacology 2013 in Azores, where an early version of this work was presented. DGH’s work on the terrestrial molluscs of South Africa was funded by the NRF (South African Biosystematics Initiative – GUN 61261; http://www.sassb.co.za/sabi.htm). SR acknowledges the UKZN for research support. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

sought from two alternative approaches, a parsimony analysis of endemicity and a biotic ele- ment analysis. Materials and methods Study area & operational geographic units The study area is southern Africa, south-east of 22˚S and 23˚E and south-west of 31˚S and 23˚E (Fig 1). This includes the MPA hotspot, the greater MPA region of vertebrate endemism as well as the SEA zoogeographical dominion for vertebrates [24]. Furthermore, the entire Cape Floristic Region (CFR) biodiversity hotspot was also included in order to clarify the uncertain western limits of the SEA dominion. Land snail distribution data for the area could still be too incomplete for a numerical zoo- geographical analysis at a fine-grain such as the quarter-degree-squares (QDSs) scale. There- fore, a set of coarse-scale eco-geographical units (zoogeographical units for SEA [22], redrawn according to QDS borders; see Fig 1) were used as the operational geographic units (OGUs) for the present analyses. The use of these natural OGUs (delimited through qualitative overlay- ing or range maps in identification of biochoria for endemic vertebrates; see Perera et al. [22] for more details) for recording the incidence of selected land snail genera (a) avoids the caveats of incompleteness of distribution data at a finer grain of analysis, and (b) retains the natural geographical boundaries of the resulting zoogeographic entities that would have been distorted had a coarse equal-area OGU (e.g. 1˚ or 2˚ grid cells) been used. The OGUs for the western

section of the study area, beyond Knysna were defined based on the CFR and Succulent Karoo biodiversity hotspots [5, 6], and the bioregions of SA [25], resulting in a total of 40 a priori OGUs. Species incidence data matrix Twelve genera (monophyletic lineages) of land snails and slugs were selected for the study, all of which have recent species-level taxonomic revisions [2, 27–37]. We included all species described to date in each genus from the study area, with a few exceptions as detailed below. Hence the analysis included 73 taxa, comprising 70 species and three species complexes of two species each. These genera were selected due to, (a) their high endemicity within the study area, (b) they are comparatively well sampled and collected within the study area, and (c) they have a relatively recent well resolved and robust taxonomy. The selected genera represent three families, viz. Bothriembryontidae, Rhytididae and Urocyclidae. All genera occurring within the study area were included from the first two families, while the family Urocyclidae, was represented by the most speciose monophyletic clade of the family Sheldonia sensu lato [13, 27, 28], of which the taxonomy was recently resolved with morphological and molecular support recognising five genera from the study area; viz. Kerkophorus, Microkerkus, Ptilo- totheca, Selatodryas and Sheldonia sensu stricto [28, 29]. The family Bothriembryontidae is rep- resented in the area by three species in the genus Prestonella of the newly validated subfamily

Prestonellinae, of which only two species were included in this analysis, excluding P. quadin- gensis due to the unavailability of distribution data except the type locality and absence of any specimens or recent observations; [30–33, DGH pers. obs.]. Land snails and slugs of the family Rhytididae are represented in the area by nine species of the hunter slug genus Chlamyde- phorus [13], three genera of larger cannibal snails: Afrorhytida, Capitina, Natalina (four, two and six species, respectively) [34, 35] and two genera of dwarf cannibal snails: Nata (six spe- cies) and the monotypic genus Natella [36, 37], of which all species were included in the cur- rent analysis except for treating Nata erugata and N. vernicosa as a single entity. While the delimitation of these two species relied heavily on molecular data, Nata erugata-vernicosa is PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 3 / 26

believed to represent a species complex with additional unresolved taxonomic issues and hence ambiguities exist in their distribution records [36, DGH pers. obs.]. The family Urocy- clidae is represented in the analyses by the genera Kerkophorus, Microkerkus, Ptilototheca, Sela- todryas and Sheldonia. All species of Microkerkus (10), Ptilototheca (one) and Sheldonia (13) found in the study area were included, while Sheldonia hewitti was considered as a synonym of Fig 1. The study area showing the Operational Geographic Units (OGUs). The Maputaland-Pondoland-Albany (MPA) biodiversity hotspot is demarcated by the dashed line and the OGUs considered here as MPA units are indicated in dark grey, while the greater MPA region of vertebrate endemism [24] is in light grey. The OGUs are labelled as: ACB–Albany coastal belt, AWB–Amatola-Winterberg, CBV–central Bushveld, CFR–Cape Floristic Region except Knysna transition zone, DBP– Drakensberg plateau, DEE–Drakensberg-Eastern Cape escarpment, DKE–Drakensberg-KwaZulu-Natal escarpment, HUK–Highveld-upper Karoo, INH–Inhambane, KBV–Kalahari-Bushveld, KNY–Knysna transition zone (shared between the Cape Floristic Region and the greater MPA region), KUK–Kalahari-upper Karoo, LKR–lower Karoo, MLV–Mozambique lowveld, NBV–northern Bushveld, NCB–Natal coastal belt, NDH–northern dry Highveld, NGO–Ngoye, NMD–Natal Midlands, NME– northern Mpumalanga escarpment, NMH–northern mesic Highveld, NMO–northern mopane, NMP–northern Maputaland, NMV–northern Middleveld, NNT–northern

Natal, PND–Pondoland, SDH–southern dry Highveld, SKR–Succulent Karoo, SME–southern Mpumalanga escarpment, SMH–southern mesic Highveld, SMO–southern mopane, SMP–southern Maputaland, SMV–southern Middleveld, SNB–Sneeuberg, SPB–Soutpansberg, STR–southern Transkei coastal belt, TMD–Transkei Midlands, UKR–upper Karoo, WLB–Wolkberg, WTB–Waterberg. See text for further details. The MPA hotspot boundary is reprinted from Hoffman et al. [26] (public domain). https://doi.org/10.1371/journal.pone.0248040.g001 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 4 / 26

S. aloicola [2, DGH pers. obs.]. The genus Kerkophorus was represented in the area by 20 spe- cies and is included in the analysis as 18 definitive species plus the Kerkophorus piperatus-vit- tarubra complex, while K. phaedimus was considered to be a synonym of K. corneus [2]. The K. piperatus-vittarubra complex was treated as such due to the difficulty in morphologically separating taxa at some locations due to their close relatedness and even a possible hybrid pop- ulation in Nkandla [28]. The genus Selatodryas being an endemic lineage in the study area with a narrow geographical range and represented by two morphologically similar species that can only be reliably identified through dissection of the genital tract was also included in the analysis as the Selatodryas luteosoma-roseosoma complex. These three species complexes, with the known localities for the two described species in each combined, are treated as three spe- cies hereafter (given with their specific epithets combined by a hyphen). The species incidence matrix was prepared for the 73 species above by scoring their pres- ence/absence (1/0) in 40 a priori OGUs (S1 Appendix). A species was recorded as present for a given OGU, even with a single occupied QDS, as long as this QDS was not along the margin of the OGU. A species occupying only marginal QDSs for a given OGU was scored as present only if the species occupied more than 10% of the OGU. A species occupying marginal QDSs with <10% coverage of the OGU was considered present only when the species was absent in

the neighbouring OGU, or the relevant QDS was in the coastal margin of its range. Species dis- tribution data were sourced from the KwaZulu-Natal Museum Mollusca database (more than 2300 records, with field visits to all OGUs), after removal of dubious records. Data analyses Zoogeographical regionalisation. Numerical analysis of the land snail incidence matrix, following the phenetic approach [38] of agglomerative hierarchical clustering of OGUs was used to fulfil the primary objective of the study, i.e. to propose a land snail based biogeographi- cal regionalisation (see Kreft & Jetz [39] and Morrone [40] for recent reviews on methods). The Jaccard’s coefficient of similarity was used to compare each OGU with every other based on their respective species assemblages [41–44], while the unweighted pair-group method using arithmetic averages (UPGMA) algorithm was used to convert the similarity matrix into a distance based dendrogram [39, 43–46]. Mouillot et al. [47] and Perera et al. [24] provides a discussion on the above methods. The analyses were conducted using FreeTree ver. 0.9.1.50 [48]. The resulting dendrograms were visualised using TreeView ver. 1.6.6. [49]. The subre- gional hierarchy of zoogeographical entities i.e., dominions, subdominions, provinces, subpro- vinces and districts [40, 50–52] (see Perera et al [24] for more details) were determined based on phenon lines [38, 53] placed on the dendrogram in order to generate geographically contig- uous clusters of OGUs, while OGUs that were not placed in such geographically contiguous

clusters were dissolved in to the geographically nearest and ecologically closest biogeographical entity derived from the same dendrogram [23, 24]. Centres of endemism. Centres of endemism (COEs: areas where endemic species con- centrate, usually having more endemics in comparison to the surrounding areas) [54–56], were also derived from the UPGMA cluster dendrogram. The COEs and centres of narrow endemism (CONEs) are presented based on decreasing range size and number of endemics: (a) COEs: clusters identified above as biogeographical districts that harbour four or more endemic land snail species and (b) CONEs: single OGUs or geographically narrow clusters of OGUs defined by a higher similarity than the districts, harbouring two or more narrowly endemic species. Characteristic and narrow endemics for each COE were identified following Williams et al. [54], Crisp et al. [57] and Procheş & Ramdhani [43]. Characteristic endemics are defined here as species occupying more than two-thirds of OGUs in the centre, and/or PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 5 / 26

distributed over more than half of the centre’s extent, hence whose range edges roughly coin- cide with the boundary of the centre, while the narrow endemics occupy much narrower ranges within the COE. Areas of endemism and biotic elements. Areas of endemism (AOEs: areas with congru- ent distribution of at least two species of restricted range) [55, 58] (see Perera et al. [24] for a discussion on AOEs and COEs) were identified through the parsimony analysis of endemicity (PAE) [59]. However, we did not attempt a regionalisation based on the PAE as done is some studies [24, 60, 61] as the resulting area cladogram revealed only five clusters, although not hierarchically structured into larger AOEs, among which only four harboured at least two endemic species with sympatry, in addition to two terminal OGUs each with two narrow endemics with congruent ranges. The PAE was conducted using PAUP� 4.0b10 [62], where a full heuristic search was per- formed with tree bisection-reconnection (TBR) branch swapping, after the species present or absent in all OGUs (similar to constant characters in phylogenetics) and species found only in a single OGUs (i.e. parsimony uninformative characters; autapomorphies in phylogenetics) were excluded. A strict consensus tree was constructed from 1000 most parsimonious trees. Geographically contiguous clusters of OGUs in the strict consensus tree harbouring at least two endemic species with sympatry were recognized as AOEs. In the description of PAE meth- odology Morrone [59] recommends to “draw quadrats on a map of the region to be analysed,

considering quadrats only where at least one locality of one species exists” (p. 438), i.e. to remove any OGUs with zero presence of species in our dataset, and as such these OGUs are similar to outgroups and the parsimony analysis effectively remove multiple outgroups. Fur- thermore, during the parsimony analysis in phylogenetics, autapomorphies are removed from the dataset as these are considered uninformative. Similarly, during the PAE in biogeography, parsimony uninformative species (which are restricted to a single OGU) are removed from the analysis. However, such narrowly endemic species that are of conservation importance were included and considered in the delimitation of COEs. However, the PAE area cladogram did not show hierarchical structuring so we further tested our AOEs for consensus and congruency with those recovered from a biotic element analysis (BEA). The BEA proposed by Hausdorf & Hennig [63] is an alternative to AOEs, in identification of vicariant events during the history of an area. We performed a BEA which included a parametric bootstrap test for clustering of distribution areas [64] using the pro- gramme PRABCLUS in R studio [65]. Spatial patterns of land snail richness and endemism. The following spatially-based quantitative measures of land snail richness and endemism were calculated for each OGU using the species incidence matrix: (a) land snail species richness, (b) SEA dominion ende- mism, (c) narrow endemism, and (d) weighted endemism (WE). The measure (a) was calcu-

lated for the entire study area, while the latter parameters (b–d) were calculated only for the SEA dominion, delineated by the regionalisation above. Here a narrow endemic species was defined as one restricted to a single OGU, a CONE or occupying less than half the range of a COE recognized above. Based on this definition, Prestonella nuptialis, Afrorhytida trimeni and Selatodryas luteosoma-roseosoma were considered narrow endemic even though they occupy two OGUs each, on the basis that the OGUs concerned are narrow in extent and they each occupy less than half the area of the OGUs combined. On the other hand, Afrorhytida kraussi and Sheldonia phytostylus although present only in the Knysna OGU were not considered as narrow endemics on the basis that the Knysna OGU was too broad and the species are distrib- uted in more than half of its extent. The WE of each OGU was calculated as the sum of the reciprocals of the total number of OGUs in which each species of the respective OGU was found [56]. Further, the WE was corrected for the species richness in respective OGUs (= PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 6 / 26

corrected WE [56]). Values obtained for all measures were normalized to stand between 0 and 1. Each of the above parameters were mapped using ArcMap10.1 [66] with a graduated grey scale for a maximum of five classes determined by natural breaks calculated using Jenk’s opti- misation, so that the patterns inherent in the data are best revealed. Results Biogeography of land snails in south-eastern Africa The first ever numerical regionalisation for land snails in south-eastern Africa is presented in Fig 2 and Table 1. An initial division of the UPGMA dendrogram (Fig 2A) separates a very Fig 2. Proposed zoogeographical regionalisation for land snails of south-eastern Africa. (a) Dendrogram of hierarchical relationships between operational geographic units (OGUs), based on UPGMA clustering with Jaccard’s coefficient of similarity. The hierarchy of zoogeographical entities were established using phenon lines of increasing similarity: Do = dominion; Sd = subdominions; Pr = provinces; Sp = subprovinces and Ds = districts. Centres of endemism (dark grey) and centres of narrow endemism (light grey) are also recognized from the same dendrogram, and mapped in Fig 4. Three letter codes for OGU names follow Fig 1. (b) The proposed hierarchical zoogeographical regionalisation of the South- Eastern Africa (SEA) dominion (thick dark red boundary; Lowveld and Cape dominions respectively in north-east and south-west are also indicated). The coloured phenon lines in panel (a) defines subdominions, provinces, subprovinces and districts of the SEA dominion, listed in

Table 1, with respective codes for provinces (A-C), subprovinces (A1, A2, etc.). Entities labelled on the map are subprovinces (Sp) and districts (Ds); S = South, N = North and KZN = KwaZulu-Natal. (c) Malacofaunal provinces within the two subdominions of SEA dominion (separated by thick purple boundary): MPA-Highveld subdominion with three provinces—MPA-Knysna province (light green), Highveld province (pale orange) and Bushveld province (olive green); and southern escarpment subdominion with no subordinate provinces (pale maroon). OGUs shown here in white (and marked with × on the dendrogram) do not form geographically contiguous clusters, and were hence dissolved into the geographically nearest and ecologically closest biogeographical entity (see discussion for more details). MPA hotspot boundary is reprinted from Hoffman et al. [26] (public domain) in blue to show its recovery as a valid zoogeographical entity (MPA-Knysna province), despite an extension towards Knysna in south-west. https://doi.org/10.1371/journal.pone.0248040.g002 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 7 / 26

tested our AOEs for consensus and congruency from a BEA. The test for clustering of distribu- tion areas indicated that the distribution ranges of land snails in south-eastern Africa show a clustering tendency (Fig 6A), as the test statistic t = 0.429, the ratio between the 25% largest and smallest distances [64, 69] for our dataset, is smaller than expected by the null model (Sim- ulated t = 0.451; ranging from 0.355 to 0.627), even though with a low level of statistical signifi- cance (p = 0.33). However, when mapped, the BEA yields nine biotic elements, seven of which supporting the six AOEs derived from PAE, and congruent with those five COEs derived from Table 1. Proposed regionalisation hierarchy for land snail zoogeography in south-eastern Africa (SEA). Dominion Subdominion Provinces Subprovinces Districts South–Eastern Africa Maputaland-Pondoland- Albany-Highveld A Maputaland-Pondoland- Albany-Knysna A1 Maputaland-Natal Extended Maputaland Natal A2 Extended Pondoland A3 Extended Albany Extended Amatolas Albany-Knysna B Bushveld B1 Sky Islands (Soutpansberg-Wolkberg) B2 Bushveld-Mpumalanga Escarpment Central Bushveld Mpumalanga Escarpment C Highveld C1 Northern Mesic Highveld & Escarpment Northern Mesic Highveld & Mpumalanga Escarpment KwaZulu-Natal Escarpment Drakensberg Plateau C2 Southern Mesic Highveld Southern Escarpment Sneeuberg Lower Karoo The SEA dominion is placed within the southern Africa subregion of the Afrotropical region (see Fig 2A for the dendrogram, 2B for the map, and the text for further details).

https://doi.org/10.1371/journal.pone.0248040.t001 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 8 / 26

Table 2. Land snail species selected for the study (n = 73; representing 12 genera and three families) and their endemism within the Maputaland-Pondoland-Albany (MPA) biodiversity hotspot, the MPA-Knysna province, the greater MPA region of land snail endemism, and the south-eastern Africa (SEA) dominion. Family Species Endemic to MPA- hotspot MPA-Knysna province Greater MPA region of land snail endemism South-eastern Africa dominion The study area Bothriembryontidae Prestonella bowkeri p p Bothriembryontidae Prestonella nuptialis NE p p p p Rhytididae Chlamydephorus bruggeni NE p p p p p Rhytididae Chlamydephorus burnupi p p p Rhytididae Chlamydephorus dimidius p p p p p Rhytididae Chlamydephorus gibbonsi p p p p p Rhytididae Chlamydephorus lawrencei p p p Rhytididae Chlamydephorus parva NE p p p p p Rhytididae Chlamydephorus purcelli p Rhytididae Chlamydephorus sexangulus p p p p Rhytididae Chlamydephorus watsoni p p p p p Rhytididae Afrorhytida burseyae p p p p Rhytididae Afrorhytida knysnaensis p p p p Rhytididae Afrorhytida kraussi p p p p Rhytididae Afrorhytida trimeni NE p p p p p Rhytididae Capitina calcicola p Rhytididae Capitina schaerfiae NE p p p p Rhytididae Nata aequiplicata NE p p p p p Rhytididae Nata dumeticola Near- endemic Rhytididae Nata tarachodes p Rhytididae Nata vernicosa-erugata p Rhytididae Nata watsoni p p p p Rhytididae Natalina beyrichi p p p p p Rhytididae Natalina cafra p p p p p Rhytididae Natalina inhluzana NE p p p p p

Rhytididae Natalina quekettiana p p p Rhytididae Natalina reenenensis NE p p p p p Rhytididae Natalina wesseliana p Rhytididae Natella viridescens p p Urocyclidae Kerkophorus ampliatus NE p p p p p Urocyclidae Kerkophorus bicolor NE p p p p Urocyclidae Kerkophorus cingulatus p p p p Urocyclidae Kerkophorus corneus p p p Urocyclidae Kerkophorus inunctus p p p p p Urocyclidae Kerkophorus knysnaensis p p p p Urocyclidae Kerkophorus melvilli NE p p p p p Urocyclidae Kerkophorus perfragilis NE p p p Urocyclidae Kerkophorus perlevis p p p p Urocyclidae Kerkophorus piperatus- vittarubra p p p Urocyclidae Kerkophorus poeppigii p p Urocyclidae Kerkophorus pumilio NE p p p Urocyclidae Kerkophorus puzeyi NE p p p p p Urocyclidae Kerkophorus russofulgens NE p p p p p (Continued) PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 9 / 26

the phenetic clustering. The remaining two biotic elements are supporting another argument we propose later in the present study, i.e. the identification of Knysna as a transitional OGU between the greater MPA region and the greater CFR (already recognised for plants by Born et al. [70]) in land snail biogeography (Fig 6B). Therefore, our COEs for land snails in south-eastern Africa derived from the phenetic clus- ter dendrogram are supported by both PAE and BEA. The characteristic and narrow endemic Table 2. (Continued) Family Species Endemic to MPA- hotspot MPA-Knysna province Greater MPA region of land snail endemism South-eastern Africa dominion The study area Urocyclidae Kerkophorus scrobicolus NE p p p p p Urocyclidae Kerkophorus terrestris NE p p p p p Urocyclidae Kerkophorus vandenbroeckii NE p p p Urocyclidae Kerkophorus vitalis p p p p p Urocyclidae Kerkophorus zonamydrus p p p p Urocyclidae Microkerkus arnotti p Urocyclidae Microkerkus burnupi NE p p p p Urocyclidae Microkerkus chrysoprasinus p p p Urocyclidae Microkerkus fuscicolor p p Urocyclidae Microkerkus leucospira p p p p p Urocyclidae Microkerkus maseruensis NE p p p Urocyclidae Microkerkus pondoensis NE p p p p p Urocyclidae Microkerkus sibaya NE p p p p p Urocyclidae Microkerkus symmetricus p p Urocyclidae Microkerkus transvaalensis p p p Urocyclidae Ptilototheca soutpansbergensis NE p p p Urocyclidae Selatodryas luteosoma- roseosoma NE p p p p Urocyclidae Sheldonia aloicola p p p p p Urocyclidae Sheldonia asthenes p p p

Urocyclidae Sheldonia caledonensis p Urocyclidae Sheldonia capsula p Urocyclidae Sheldonia cotyledonis p Urocyclidae Sheldonia crawfordi p p p Urocyclidae Sheldonia fingolandensis NE p p p p p Urocyclidae Sheldonia hudsoniae p p p p Urocyclidae Sheldonia monsmaripi NE p p p p p Urocyclidae Sheldonia natalensis p p p p Urocyclidae Sheldonia phytostylus p p p p Urocyclidae Sheldonia trotteriana p p Urocyclidae Sheldonia wolkbergensis NE p p p Total endemics 30 37 57 63 72 Total species 53 60 69 70 73 Species endemicity (%) 56.6 61.7 82.6 90.0 98.6 Species endemicity is given for each geographical entity. NE denotes species considered as narrow endemics within the SEA dominion (see text for criteria). https://doi.org/10.1371/journal.pone.0248040.t002 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 10 / 26

taxa of COEs are presented in the Table 3. Twenty-nine species are restricted to COEs, with 21 (72%) of them being narrow endemics. Those species restricted to COEs are dominated by urocyclids (72%) showing an adaptive radiation in narrow ranges (76% of them being narrow endemics) representing almost all centres. They are followed by rhytidids (28%) with one spe- cies of Natalina each from Natal and extended Pondoland centres, one species each of Afrorhy- tida, and Nata from the Albany-Knysna centre, one species of Capitina from the Knysna CONE and one species of Chlamydephorus each from the Albany-Knysna centre and the Mid- dleveld CONE. A spatial comparison of COEs with biomes [25] and the protected area net- work of SA is presented in Fig 4D and 4E. Spatial patterns of land snail species richness for SEA and measures of endemism within the SEA dominion are mapped in Fig 7. Fig 3. Land snail species richness, endemism and percentage endemicity within the MPA hotspot, MPA-Knysna province and the greater MPA region of land snail endemism. https://doi.org/10.1371/journal.pone.0248040.g003 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 11 / 26

A greater Maputaland-Pondoland-Albany region of land snail endemism The COEs as well as the patterns of land snail endemism confirm the hotspot status of MPA for malacofauna (see Figs 4–7). Furthermore, the MPA-Knysna province (Fig 2) is recognized as a numerically derived zoogeographical entity for land snails. Nevertheless, the land snail endemism as indicated by the COEs (Fig 4) and high levels of weighted endemism (Fig 7), extends beyond this province towards the south-eastern escarpment (Drakensberg-KwaZulu- Natal escarpment, Drakensberg plateau, and Sneeuberg OGUs) and the eastern escarpment (northern & southern Mpumalanga escarpment, Wolkberg and Soutpansberg OGUs), Fig 4. Centres of endemism (COEs) and centres of narrow endemism (CONEs) for land snails of south-eastern Africa derived from the phenetic cluster dendrogram (Fig 2A). Centres are shaded on a grey scale based on number of endemics, given below for each in brackets (darker shades denote high endemism). See Table 3 for lists of endemic species. (a) COEs: A–Sky islands (4), B–Extended Maputaland (4), C–Natal (6), D— Extended Pondoland (7), and E—Albany-Knysna (7). (b) CONEs: a—Soutpansberg (2), b—Wolkberg (2), c—Northern Middleveld (2), d—Natal coastal belt-Ngoye (3), e—Natal Midlands, and f—Knysna (3). The whole (blue) line indicates the Maputaland-Pondoland-Albany (MPA) hotspot boundary while the dashed (black) line delimits the greater MPA region of land snail endemism. COEs and CONEs are spatially compared with (c)

South African biomes and (d) the protected area network of South Africa (SA); protected areas in Lesotho (L), Swaziland (S) and Mozambique (M) are not included. The MPA hotspot boundary, biomes and protected areas of SA are reprinted from Hoffman et al. [26], Open Knowledge Foundation [67] and the Department of Environmental Affairs, Forestry and Fisheries [68] (public domain), respectively. https://doi.org/10.1371/journal.pone.0248040.g004 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 12 / 26

comprising many OGUs recently recognised as individual centres of endemism [29, 71, 72]. The incorporation of these OGUs into the MPA-Knysna province designates the greater MPA region of land snail endemism (see Figs 4–7), with 57 endemic species (82.6% endemicity; Table 2) from selected land snail genera. The greater MPA region for malacofauna, being 46.7% larger in area compared to the MPA hotspot, harbours a remarkable 78.1% greater number of endemic land snail species (57 endemic species in the greater MPA region com- pared to only 32 in the MPA hotspot; Fig 3). Moreover, spatial pattern of species richness fur- ther justifies the greater MPA region with OGUs such as southern Maputaland, Natal coastal belt, Natal Midlands, Pondoland, southern Transkei coastal belt, Albany coastal belt, Amatola- Winterberg and the Knysna transition zone (from North to South) been identified as hotspots of high species richness, while northern Mpumalanga escarpment, northern Middleveld, Ngoye, Drakensberg-KwaZulu-Natal escarpment and the Sneeuberg are also fairly speciose in land snails (Fig 7A). Discussion Land snail biogeography of south-eastern Africa The study provides a preliminary assessment of the land snail biogeography in SEA, based on a selection of the most well-known genera. The delineation of the SEA dominion separated from the Cape in south-west, Lowveld in north-east and especially the Karoo and Kalahari in the west is driven mainly by the climatic factors separating xeric areas from relatively mesic

ones, as evident in previous regionalisations for vertebrates [24, 45, 57, 73–76], higher plants [45] and bryophytes [77]. The subsequent separation of the MPA-Knysna province from more inland OGUs seems to have ecological as well as historical explanations related to the greater habitat heterogeneity Fig 5. Delineation of areas of endemism (AOEs) for land snails in south-eastern Africa based on the strict consensus tree of parsimony analysis of endemicity (PAE). Four of the five clades in the area cladogram harbour more than two sympatric endemic species, and hence identified as AOEs, together with two OGUs each with two narrow endemics with congruent ranges. All six AOEs are shaded in the map on a grey scale according to the number of endemic species, given below for each in brackets (darker shades denotes high endemism; see Table 3 for lists of endemic species): A— Soutpansberg (2), B—Wolkberg (2), C—Extended Maputaland (4), D—Natal (6), E—Extended Pondoland (7), and F— Albany-Knysna. The MPA hotspot boundary is reprinted from Hoffman et al. [26] (public domain). https://doi.org/10.1371/journal.pone.0248040.g005 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 13 / 26

along the coastal belt and the cyclical expansion and contraction of forests over geological time [57, 78–80], contrasting with the more homogeneous savanna in Bushveld province, grass- lands in Highveld province, and the lower Karoo in the southern escarpment subdominion. The refugial role played by the patchy distribution of forests during the Pleistocene climatic cycles with corresponding expansions and retractions [78, 80, 81] are presumed to have shaped the present-day malacofauna of the area (see Fig 4D). The repeated isolation of once linked forest patches may have contributed to rapid diversification and increased narrow endemism Fig 6. Biotic elements of land snails in south-eastern Africa. (a) First two dimensions of the nonmetric multidimensional scaling of the distribution ranges of land snails of south-eastern Africa based on their incidence in pre-defined operational geographic units (Fig 1). Biotic elements found by PRABCLUS are denoted by 1–9; N = noise component. (b) Distribution maps of the nine biotic elements of the land snail fauna of south-eastern Africa. Four different shadings from light to dark indicate the areas where >30%, >45%, >60% and >100% of the species of each element are present. The MPA hotspot boundary is reprinted from Hoffman et al. [26] (public domain). https://doi.org/10.1371/journal.pone.0248040.g006 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 14 / 26

Table 3. Range-restricted species endemic to centres of endemism (COEs), centres of narrow endemism (CONEs) and areas of endemism (AOEs) within the south- eastern Africa dominion (see Fig 4), also showing their nested hierarchy, verification as and areas of endemism (AOEs) and corresponding biotic elements (see text for details). COEs a CONEs d ��also verified as AOEs b �� also verified as AOEs b and supported by biotic elements c COE (no. of endemics) List of centre endemics � and, species in the same biotic element CONE (no. of endemics) List of narrow endemics Sky islands (4) Characteristic Endemics: None ��Soutpansberg (2) Kerkophorus pumilio, Ptilototheca soutpansbergensis Narrow Endemics: ��Wolkberg (2) Kerkophorus perfragilis, Sheldonia wolkbergensis Four narrow endemics restricted to the Wolkberg & Soutpansberg CONEs Additional species in the same biotic element: None ��Extended Maputaland (4) Characteristic Endemics: Chlamydephorus watsoni Middleveld (2) Sheldonia monsmaripi, Chlamydephorus bruggeni Narrow Endemics: Two narrow endemics restricted to the Northern Middleveld CONE and Microkerkus sibaya (restricted to narrow southern Maputaland) Additional species in the same biotic element (shared among the OGUs NMP, SMP, NMV & SMV): Chlamydephorus lawrencei Natalina wesseliana �� Natal (6) Characteristic Endemics: None Natal Midlands (3) Natalina inhluzana, Kerkophorus bicolor, Microkerkus burnupi Narrow Endemics Natal Coastal Belt- Ngoye (3) Kerkophorus ampliatus, K. melvilli, K. russofulgens

Six narrow endemics restricted to the Natal Midlands & Natal Coastal Belt-Ngoye CONEs Additional species in the same biotic element: None �� Extended Pondoland (7) Characteristic Endemics: No CONEs are nested within the Extended Pondoland COE Natalina beyrichi, and Kerkophorus vitalis (widespread in both OGUs of the COE) Narrow Endemics: Microkerkus pondoensis, Kerkophorus puzeyi, Kerkophorus terrestris (restricted distributions in both OGUs of the COE), Kerkophorus scrobicolus (restricted to narrow Pondoland OGU) and Sheldonia fingolandensis (restricted to narrow southern Transkei coast OGU) Additional species in the same biotic element: None �� Albany-Knysna (7) Characteristic Endemics: Knysna (4) Capitina schaerfiae, Kerkophorus knysnaensis Nata aequiplicata (narrowly endemic within the broad Knysna OGU) Afrorhytida kraussi and Sheldonia phytostylus (widespread in the broad Knysna OGU), S. aloicola and S. hudsoniae (found in both Albany and Knysna OGUs) Narrow Endemics: Chlamydephorus parva restricted to Albany and four species narrowly endemic within the broad Knysna OGU. Additional species in the same biotic element (shared among the OGUs ACB, KNY, SNB & LKR): Nata watsoni Prestonella bowkeri S. natalensis S. trotteriana a Districts defined from the UPGMA dendrogram (Fig 2A) with four or more endemic species. b Geographically contiguous clusters of operational geographic units (OGUs), or individual OGUs in the PAE area cladogram (Fig 4A) with at least two endemic species. c Biotic elements found by PRABCLUS (Fig 6B).

d Individual OGUs or geographically narrow clusters of OGUs defined by a higher similarity than districts form the UPGMA dendrogram (Fig 2A), harbouring at least two narrow endemic species. � Characteristic endemics: species occupying more than two-thirds of OGUs in the centre, and/or distributed over a half of the extent of the centre, hence whose range edges roughly coincide with the boundary of the centre; Narrow endemics: species restricted within a single OGU, a CONE or less than half the range of a COE; Additional species in the same biotic element: Species shared among other OGUs indicated for the corresponding biotic element(s) as given in Fig 6. https://doi.org/10.1371/journal.pone.0248040.t003 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 15 / 26

[82], also leaving ancient Gondwanan geographical relicts in isolated forest patches due to the relative stability in climate within them [13, 34, 83]. The role of ecological stability of forest habitats in promoting radiation of relictual taxa has been recognised [84], while this phenome- non has also been documented for land snails in forest refugia along the Albertine rift valley of Africa [85], which can be attributed to the ecophysiology of land snails, including their prefer- ence for habitats with high micro-habitat complexity, high humidity, low temperature and damp soil with decaying organic matter [86–88]. The southern escarpment subdominion, possibly a partial recovery in the present analysis corresponds to the lower Karoo bioregion. The MPA-Knysna province indicates an extension of MPA elements into the Cape, which is distinct from the original delimitation of the MPA hotspot. The Knysna unit spatially represents the eastern section of the CFR biodiversity hot- spot, characterized by the Fynbos biome. The similarities of the malacofauna of the Knysna OGU (eastern CFR) with further eastward OGUs are associated with the presence of forest patches in them, despite the Knysna OGU being dominated by Fynbos. This can be attributed to the post-glacial expansion of Afromontane forests to the Knysna-Tsitsikamma area [80, 89]. Overall similarities in the present regionalisation to that for vertebrates [24], despite fine scale differences likely caused by life history attributes such as body size and dispersal ability [90],

suggest a common zoogeographical pattern for both invertebrates and vertebrates in SEA, while biogeographic patterns available for plant assemblages in southern Africa also support similarities [91]. This opens up avenues for future studies incorporating more invertebrate taxa, where there is adequate spatial data and a sufficiently robust taxonomy, and for cross taxon analyses involving both a wide suite of animal and plant lineages. Furthermore, the con- sensus for an extension of the MPA into the Knysna OGU in both vertebrate and land snail Fig 7. Spatial patterns of land snail fauna in south-eastern Africa (SEA). Species richness for the whole study area and all spatial measures on endemism for the SEA dominion (see Fig 2). (a) species richness, (b) SEA dominion endemism, (c) narrow endemism, (d) weighted endemism and (e) corrected weighted endemism. See text for details on above calculations. The whole (blue) line indicates the Maputaland-Pondoland-Albany (MPA) hotspot boundary while the dashed (black) line delimits the Greater MPA region of land snail endemism. The MPA hotspot boundary is reprinted from Hoffman et al. [26] (public domain). https://doi.org/10.1371/journal.pone.0248040.g007 PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 16 / 26

regionalisations also suggest further investigations on the limits of a Knysna transition zone between the greater CFR and the greater MPA region. Concerning land snails, the proposed regionalisation permits further refinement when a more complete data set is available at a finer scale equal-area geographical unit, for more genera, to support a much more rigorous analysis, while the data becoming newly available supports the patterns identified here (e.g. [92, 93]). Land snail endemism in south-eastern Africa Quantitative measures mapped for land snail species richness and endemism reveal similar patterns (Fig 5), with higher values along the south-eastern coastal belt and on the south-east- ern escarpment, especially its eastern aspects towards the coastal belt, similar to those for milli- pedes [94, 95], vertebrates [24], bryophytes [96] and higher plants [20]. The Highveld and Bushveld provinces are generally species- as well as endemic-poor, although the Bushveld sup- ports more land snail endemics compared to the Highveld (Fig 5E), which could possibly be attributed to the comparatively higher habitat heterogeneity in the Bushveld that provide shel- tering microhabitats into which snails can retreat, supported by the higher spatial area covered by Bushveld, and again to the supposedly stable and milder climate in the Bushveld (see the role of forests discussed above) compared to that of the Highveld characterized by drought, frost, waterlogging and wildfire causing its treelessness [97]. The northern Mpumalanga escarpment, northern Middleveld, southern Maputaland,

Ngoye, Natal coastal belt, Natal Midlands, Drakensberg-KwaZulu-Natal escarpment, Pondo- land, southern Transkei coastal belt, Amatola-Winterberg, Albany coastal belt and Knysna OGUs (from North to South) are established as strongholds of both species richness and ende- mism of land snails (Fig 5A and 5B), while narrow endemism is concentrated within the extended Pondoland and Natal districts (Fig 5C). Measures of weighted endemism (Fig 5D) further emphasize the importance of the land snail districts of Soutpansberg-Wolkberg, north- ern Middleveld, southern Maputaland, Drakensberg-KwaZulu-Natal escarpment, Natal, extended Pondoland, and Albany-Knysna (from North to South) for malacofaunal conserva- tion, strongly supporting the concept of a greater MPA region of land snail endemism. The corrected weighted endemism (Fig 5E) further highlights the Drakensberg plateau and the lower Karoo as important for land snail conservation in addition to the above districts. Recov- ery of similar endemism peaks from both the UPGMA and PAE clustering as well as the BEA re-confirms the importance of the greater MPA region for restricted-range land snail radiations. Pondoland, a prominent centre of plant endemism, is well recovered here for land snails, even though it showed a comparatively lower endemicity for vertebrates [24]. Nevertheless, the Natal centre, well established for vertebrates, was also found to be an important centre for land snails, even though it is not widely acknowledged for plants. In contrast to the western