Biogeography of Angolan rodents: The first glimpse based on phylogenetic evidence

Assessment of phylogenetic diversity and biogeographical affinities of the rodent fauna from one of the most neglected areas in Africa.


| INTRODUC TI ON
The shortage of high-quality data on the distribution of organisms is one of the biggest challenges of conservation biogeography (Richardson & Whittaker, 2010). Many areas of the world remain heavily under-sampled for most taxa as a result of unfavourable conditions for field surveys, including political instability. Even in areas with good sampling coverage, distributional data are often affected by geographical sampling bias: much of our distributional data are limited to political geographical units (state borders) (Hughes et al., 2021) or lacking biological meaning (Farooq et al., 2021). Besides geographical sampling bias, the quality of the collected data is another factor limiting our knowledge about species diversity of a particular area. An essential source of data on species distribution consists of historical inventories (museum collections; Ferguson, 2020). In assessing diversity distribution, this kind of data should be used with caution mainly due to two drawbacks: (1) historical data do not necessarily reflect current species distribution, e.g., due to habitat loss; and (2) this type of information is often restricted to morphological description and does not sufficiently recognize cryptic diversity, i.e., the occurrence of distinct evolutionary lineages that are otherwise morphologically indistinguishable to a human observer. These are usually reliably detected only by genetic data obtained by molecular techniques (Chenuil et al., 2019).
In Africa, south-western Africa, represented roughly by Angola and adjacent parts of neighbouring countries, is, in general, one of the most understudied regions concerning biodiversity (Clark et al., 2011;Huntley et al., 2019). More than forty years of instability related to the colonial and civil wars in Angola (1961Angola ( -2002 prevented field studies and the collection of biological samples. Despite the revival of biodiversity research in some Angolan regions (e.g., Conradie et al., 2012;Mills et al., 2013) in the last two decades, genetic data reflecting evolutionary diversity is still scarce. For example, the only large-scale study of Angolan mammalian biogeography is solely based on records of morphologically identified species (Rodrigues et al., 2015).
Because of its geological, topographical and climatic diversity, Angola has seven biomes and 15 terrestrial ecoregions within its borders, ranking first and second in Africa, respectively (Burgess et al., 2004;Huntley, 2019), being therefore particularly interesting for biogeographers. Three major biogeographic regions of sub-Saharan Africa meet in Angola: the moist forests and savannahs of the Congolian region; the woodlands, savannahs and floodplains of the Zambezian region; and the arid savannahs, dwarf shrublands and desert of the Karoo-Namib region (Linder et al., 2012). Although there are several areas of great global conservation importance and putative centres of endemism in Angola, such as the Angolan Escarpment Zone, the evolutionary history of most biological groups occurring there is virtually unknown .
Rodents are well suited for detection of phylogeographic patterns owing to their short generation time, limited dispersal ability and strong association with particular habitats. For instance, periodic fragmentation of African forests during the Plio-Pleistocene caused by palaeoclimatic changes resulted in remarkable genetic differentiation of forest dwelling rodents (e.g., Bohoussou et al., 2015;Demos et al., 2014;Nicolas et al., 2011).
Moreover, their populations were not substantially modified by hunting as with large mammals, which have been significantly affected during the civil war until now. A few recent surveys of large mammals confirm that many species once common in Angola only persist currently in remote areas, usually having small and fragmented populations Monterroso et al., 2020;Overton et al., 2017).
Current knowledge of Angolan rodent fauna stands mostly on a foundation laid by the detailed studies by Crawford-Cabral (1966a, 1966b, 1989, 1998. According to the most recent species list of Angolan rodents (Beja et al., 2019), there are 10 rodent families in Angola, represented by 85 rodent species, 13 of which are considered endemic either to the country or to south-western Africa.
However, this compilation is exclusively based on morphological identifications and would greatly benefit from the inclusion of genetic data. This type of data allows to test for the distinctiveness of the currently recognised species, the presence of cryptic diversity and the existence of intraspecific phylogeographic lineages. In addition, such studies can also provide information about phylogenetic position, evolutionary uniqueness and biogeographical affinities if performed at a broader geographical scale and not limited by borders of a single country.
The aim of this study was to perform the first-ever phylogeographic assessment of Angolan rodent diversity based on cytochrome b sequences. It is based on four sources of genetic information from Angolan rodents: (1) Material collected in our recent survey conducted in three provinces of western Angola (Namibe, Huíla and Cuanza-Sul) supplemented with (2) other donated material; (3) museum samples and (4) publicly available sequences (GenBank, African Mammalia). The results provide the most comprehensive information about the evolutionary diversity and biogeographic affinities of Angolan rodents based on genetic data and indicate taxa worthy of further taxonomic studies. They also highlight areas with evolutionary uniqueness and local endemism, which might contribute to the identification of regions deserving formal protection.

K E Y W O R D S
DNA barcoding, endemism, mice, molecular operational taxonomic units, Muridae, phylogeography, rats, South African region, Zambezian region 2 | ME THODS

| Sampling
The trapping survey was conducted in July 2017 at nine localities in south-western Angola (Figures 1 and 2). Animals were trapped using snap traps and Sherman traps. To maximize trapping success and diversity of the captured rodents, we set the trap lines to cover a representative proportion of each habitat rather than using a fixed distance or grid. In addition, one cane rat (Thryonomys) was obtained from local people by the road to Sumbe, and one house mouse (M. m. domesticus) was captured in the city of Lubango (Table S1.1).
Standard body measurements (body weight and lengths of head +body, tail, ear and hind foot without claws) were taken from all specimens, and a tissue sample (spleen where possible) from each of them was preserved in 96% ethanol and stored at −20°C until DNA extraction. Voucher specimens were preserved in 70% ethanol for further taxonomic work and deposited at the University of South Bohemia, České Budějovice, Czech Republic, and in the ZC-IICA, Angola.
To cover a broader geographical area of Angola, we took tissue samples from 58 dry skins from the collection of ZC-IICA (collected in 1962-1974. Preferably toes were cut and stored dry in Eppendorf tubes until DNA extraction with the Invisorb Spin Forensic Kit (Stratec). In addition, we used sequences of eight specimens collected recently by Taylor in central Angola (Taylor et al., 2018) and two tissue samples from the collection of the Field Museum of Natural History, Chicago, U.S.A.
To reveal the phylogenetic position of Angolan taxa with a particular focus on their biogeographic affinities, we sequenced 674 rodent specimens from our previous expeditions to eastern Africa (GenBank accession numbers and other details are provided in Table S1.1). Additionally, 69 samples were obtained from various museum collections. For details about the collected specimens and origin of the museum samples, see Table S1.1. The obtained data were combined with sequences from GenBank (N = 1584) and the African Mammalia (Van de Perre et al., 2019) database (N = 92).

| Genotyping, phylogenetic analysis and biogeographical assessment
The captured individuals were classified to genera and tentatively to species based on their physical attributes and the measurements. At least one specimen per species ("morphotype") and locality was chosen for sequencing of the mitochondrial gene for cytochrome b (CYTB). For more information on the DNA extraction, polymerase chain reaction (PCR) and Sanger sequencing see Krásová et al. (2019).
In the material from ZC-IICA, we targeted specific short fragments of CYTB using high-throughput amplicon sequencing on the Illumina MiSeq platform (Illumina, San Diego, CA, USA) as described in Šmíd et al. (2020). Short fragments of CYTB from the additional 69 museum samples were obtained by 454-pyrosequencing on the GS Junior Roche platform (Roche, Basel, Switzerland) as described in Aghová et al. (2017).
Based on their CYTB sequences, the individuals were classified to genera using the BLAST tool (Altschul et al., 1990) at the NCBI GenBank website (www.ncbi.nlm.nih.gov/genbank). To put the Angolan specimens into an appropriate phylogenetic context, for each genus, we made a separate CYTB alignment comprising relevant sequences available in public databases (see above), as well as our previously collected and unpublished data (Table S1.1). In a few cases, we relied on unpublished data of other research groups (Table S1.1).
Every genus was subjected to a separate phylogenetic analysis, which followed the same protocol, wherever it was possible.
First, we conducted a preliminary analysis using FastTree (Price et al., 2009) and created a representative CYTB alignment covering the phylogenetic diversity of the genus (or species) in question as much as possible. Then, we obtained an unrooted tree with branch lengths in substitution units using Bayesian inference as implemented in MrBayes 3.2.6 (Ronquist et al., 2012). The same setting was used in all cases: uniform prior overall topologies and HKY nucleotide substitution model (Hasegawa et al., 1985) with Gamma distributed variation in substitution rates and a proportion of invariant sites and 12-3 partitioning according to the codon position. Three outgroups were included in every tree for their post F I G U R E 1 Map of Angola showing the nine localities that were sampled for the present study (white circles), localities of genotyped samples from ZC-IICA (red crosses) and genotyped specimens from an expedition to the Okavango basin (Taylor et al., 2018;black crosses) hoc rooting (Table S1.2). Four independent Markov Chain Monte Carlo simulations were conducted to test for the convergence, which was done visually in Tracer 1.7 (Rambaut et al., 2018). After discarding burn-in, the four posterior samples were pooled, the trees rooted and outgroups removed not to bias subsequent analyses. The Maximum Clade Credibility (MCC) tree, as defined by Drummond and Bouckaert (2015, p. 162), was then calculated to represent the pooled sample. Its branch lengths were based on the common ancestor heights (Drummond & Bouckaert, 2015, p. 92). The MCC tree was calculated in R (R Core Team, 2019) using packages ape (Paradis & Schliep, 2019), phangorn (Schliep, 2011) and custom functions written by one of the authors (OM; https:// github.com/onmik ula/mcctr ee_mrbayes).
The haplotypes included in the tree were classified into molecular operational taxonomic units (MOTUs), which can be considered as approximations of species or deeply divergent intraspecific (phylogeographic) clades. We used the maximum likelihood-based mPTP algorithm of Kapli et al. (2017), which partitions the tree into K + 1 components, one interspecific backbone and K MOTUs, each with its own exponential distribution of branch lengths. The resulting classification is fully reported in Table S1.1, but we focused on the Angolan taxa and their relatives. Only these focal MOTUs are, therefore, shown as distinct, while the others were collapsed into larger monophyletic lineages.
Where appropriate, the existing species (or intraspecific lineages)

| Trapping success and genotyping summary of Angolan material
In total, 307 rodents representing 16 genera were collected in 2017 during our fieldwork in Angola. Details about trapping success and genera captured at particular localities are summarised in Tables S1.3 and S1.4, respectively. In total, we sequenced the CYTB   (Table S1.1).

| Intrageneric diversity in a biogeographical context
In total, the genetic data allowed the identification of 44 taxa (i.e., MOTUs or potential species) of rodents in Angola, F I G U R E 2 Habitat pictures of the localities where trapping was performed.
(1) Arco Oásis, (2) Caraculo, (3) Bibala, (4) Tundavala, (5) Hombe, (6) Bicuar NP, (7) 20 km SW of Cassongue, (8) Kumbira Forest, (9) Conda belonging to 19 genera (Table S1.1). For the majority of them, CYTB sequences represent the first genetic information from the country. Specific accounts for particular genera, including phylogenetic trees, distributional maps and taxonomic comments are available in Data S1. We were able to assign names to three out of 12 taxa not listed in the most recent species list of Angolan rodents (Beja et al., 2019)  However, this will require further quantification of phenotypic differences and comparison with the type material. Two taxa are possibly new and require further taxonomic work (Table S2.1).
Poemys sp. indet. 13 (sensu Voelker et al., 2021) was found solely at Tundavala, while Rhabdomys sp. 1, was found exclusively at the Lungwebungu River locality. Mastomys angolensis recorded solely at the 20 km SW Cassongue locality was previously reported as Myomyscus angolensis (Crawford-Cabral, 1989). For some taxa F I G U R E 3 Distribution patterns of Angolan MOTUs. Distribution of all MOTUs present in Angola in this study (shown separately in Data S1), except of Angolan endemics, is visualised. MOTUs were collapsed into three groups according to their distribution affinities: MOTUs distributed in Angola and DRC (red crosses, N = 7), MOTUs distributed in Angola and eastern Africa (blue crosses, N = 13) and MOTUs with affinities to Southern Africa (green circles, N = 4). For assignment of MOTU to a particular group see column "Ecoregion" in Table S1.1 Table 1 (blue stars). Two ecoregions from Burgess et al. (2004) are highlighted: the Angolan Miombo woodlands (yellow) and Angolan montane forest-grassland (light red) Kumbira forest 20 km SW Cassongue Tundavala (Fukomys "Kinshasa-Quimbango", Graphiurus sp. indet. 1, 2, 3 and 4), we were not able to assign any particular species name because of the lack of systematic revisions based on molecular data (see more details in Data S1).

F I G U R E 4 Distribution of 12 MOTUs endemic to Angola listed in
We defined four major biogeographical patterns based on the distributional patterns of Angolan MOTUs in Angola and neighbouring countries and the analysed phylogenetic trees including clades related to the Angolan MOTUs (see phylogenetic trees and distributional maps for particular taxa in Data S1) (Figure 3 and Figure 4).

| Diversity of Angolan rodents
The most recent checklist of Angolan rodents presented in Beja et al. (2019) counts 10 families and 85 species, thirteen of them being endemic or near-endemic to the country. This checklist is largely a summary of morphological studies pre-dating the molecular revolution in taxonomy (e.g., Crawford-Cabral, 1966a). Pending future integrative taxonomic revisions, the present study records 23 species from the list and suggests the existence of up to 12 others, including two endemics (Table S2.1).
Given that rodent diversity has never been surveyed in a large part of Angola, it is not surprising that additional species were recorded. An updated checklist of Angolan rodents based on the findings of the present study could contain more than 90 species, which would be more than, for instance, the number of species found in the Angola. However, they are both better surveyed than Angola and, especially Ethiopia, has an extremely heterogeneous landscape with high mountains separated by substantial migration barriers such as the Rift valley, which may promote speciation (Huhndorf et al., 2007).
Even though our sampling represents just a snapshot of the rodent fauna in south-western Angola, we covered an important part of the country's rodent diversity. During only ten trapping nights, 26 rodent taxa were recorded, which corresponds to one-third of the total count (75)  Sciuridae, the groups not targeted by our trapping method. It is clear, therefore, that even such limited field studies are very valuable, especially in poorly documented areas like Angola.
Any assessment of biodiversity relies on the estimates of species boundaries and/or phylogenetic relationships. In the present study, we employed single-locus methods of phylogenetic and species delimitation inference and used sequences of the mitochondrial cytochrome b gene as our data (135 bp in museum samples, 1,140 bp otherwise). Thanks to its frequent use as a mammalian barcode marker (DeSalle & Goldstein, 2019), it allowed us to combine our sequences with the largest possible amount of data from public databases. The use of a single marker with matrilinear inheritance has its inherent limitations. We checked for the presence of unusually conservative sequences (Puechmaille et al., 2011) and the presence of stop codons (Song et al., 2008) to remove possible nuclear pseudogenes, but these signs may not always be apparent in short museum barcodes. Also, cytonuclear discordance can mislead single gene analyses, whether it is due to mitochondrial introgression or incomplete lineage sorting. Therefore, the kind of trees presented in Data S1 cannot be conclusive about the phylogenetic relationships in a particular genera, but it is known to provide a sound basis for biodiversity assessment (e.g., Boubli et al., 2018).

| Biogeographic affinities
Although an increasing body of papers focuses on the biodiversity of Angolan mammals, only a few have used genetic approaches Congolian and South African (summarised in Figure 3). Accordingly, it is possible to divide Angolan rodent taxa into the following major groups: 1. Taxa with affinities to the Zambezian region (summarised by blue crosses in Figure 3). This group consists of lineages  (White, 1983).
It is therefore not surprising that taxa associated with this ecosystem exhibit wide and continuous distributions. Their internal phylogenetic structure is relatively homogeneous, reflecting only a few existing significant biogeographic barriers, the most important being the Zambezi River and the large rift lakes (e.g., Krásová et al., 2019;Mazoch et al., 2018;McDonough et al., 2015;Mikula et al., 2016). The existence of gene flow between Angola and Zambia is also suggested by the distribution of many avian taxa, although this is not yet verified by molecular data (Beja et al., 2019).

Taxa with affinities to the savannah/forest mosaic in south-
western DRC (red crosses in Figure 3).  Figure 3, Figure S2.12). We hypothesize that this southward colonization occurred due to the presence of relatively moist gallery forests along the Kunene River during more humid periods. A similar pattern is known for several bird species (e.g.,

Dryoscopus angolensis and Apalis cinerea) that occur in western
Angola and have disjoint populations in eastern DRC and/or in Cameroon/Gabon (Diamond & Hamilton, 1980). Further sampling in poorly known areas of southern DRC would be necessary to test this hypothesis with genetic approaches. Figure 3). This group is represented by species adapted to arid  (Portik et al., 2011), but also other mammalian species such as Antidorcas marsupialis (Cain et al., 2004). More widely distributed taxa from Angola associated with the South African region (e.g., Saccostomus, Gerbilliscus) inhabit other types of open savannah-like habitats south of the Zambezi River (Data S1).

Taxa with affinities to the South African region (green circles in
Importantly, these affinities are apparent on different phylogenetic scales. While some of them likely reflect recent colonization events (i.e., Angola shares the same MOTUs with neighbouring countries), others have a more ancient basis (i.e., Angola and neighbouring regions) and are occupied by different but closely related MOTUs corresponding either to conspecific phylogeographic lineages or sister species. For instance, in rodents of savannah-like habitats, the affinity to the South African region is revealed by five shared MOTUs (Table 1) and two MOTUs endemic to Angola, but having the most closely related counterparts in southern Africa (Micaelamys namaquensis MOTU 2, Rhabdomys sp. 1).
The fact that the genetic data used in the present study revealed a clear biogeographic pattern is even more intriguing when compared with a similar study based on morphologically determined museum records. Rodrigues et al. (2015) used museum records of three mammalian groups (carnivores, rodents and ungulates) to develop a biogeographical regionalization of Angolan mammals. Four biogeographical subdivisions emerged from ungulate data: a northern region (Zaire-Lunda-Cuanza), a central region (Central Plateau) and two regions in the south (Namibe and Cunene-Cuando Cubango), whilst rodent and carnivore data were largely uninformative (Rodrigues et al., 2015).
The study hypothesised that this could be explained by the lower number of records available for rodents and carnivores or by a stronger association of ungulates with specific vegetation types. Our study does not support this, because a smaller dataset of rodent records was used, and revealed strong biogeographic divisions. This highlights the fact that the quality of data used for biogeographical assessments (i.e., data including genetic information) is crucial in determining spatial patterns of biodiversity. This also questions the broad use of morphologically determined species in biogeographic studies of rodents and other groups.

| Endemism in Angola
Numerous vertebrate species have been described as endemic or nearly endemic to Angola, although the precise number of Angolan endemics is difficult to estimate due to data deficiency and taxonomic ambiguities (Beja et al., 2019). Recently published studies have shown that especially the use of genetic material represents a valuable tool for identifying Angolan endemics and their phylogenetic position (Ceríaco, Agarwal et al., 2020;Ceríaco et al., 2018;Ceríaco, Tolley, et al., 2020;Hallermann et al., 2020;Krásová et al., 2021;Marques et al., 2019Marques et al., , 2020. The results presented in this study might serve as a starting point for a re-assessment of Angolan rodent fauna, including the level of endemism. Our phylogenetic analysis revealed the presence of 12 MOTUs endemic to Angola (Table 1, Figure 4). Despite the lack of detailed phenotypic and ecological data, seven of them can be safely considered distinct species. Aethomys bocagei represents a genetically distinct species sister to A. kaiseri from Eastern Africa. Fukomys bocagei is a small-bodied species more widespread than the large-

| Western Angola as a centre of diversity− conservation implications
Although our results do not allow to identify areas of high endemism and diversity, the observation of 12 unique MOTUs endemic to Angola justify further conservation efforts, even in the absence of up-to-date comprehensive surveys. Our data suggest that some parts of western Angola are especially important centres of biological diversity, crucial for the persistence of Angolan endemics.
The Tundavala site is located in the Humpata plateau in the southern region of the Angolan Escarpment and is the southernmost representative of the Angolan montane forest-grassland mosaic ecoregion (Burgess et al., 2004) (Figure 4) Chain (mostly 1,600-1,850 m a.s.l.) (Huntley, 1974). The habitat at this locality included grassland and degraded miombo. During one trapping night, ten species from six genera were captured ( Kumbira Forest is a localized patch of forest between dry coastal vegetation and the moist savannahs of the Angolan interior plateau located in the central part of the Angolan Escarpment (Mills, 2010). It represents a remnant of a forest with close affinities to the Guineo-Congolian biome (Gonçalves & Goyder, 2016). Recently, a new species of bushbaby (Galagoides kumbirensis) was described from the forest (Svensson et al., 2017), which is an endemism centre for many bird species (Cáceres et al., 2015). In a single trapping night,

| CON CLUS IONS
Our study brings two significant findings. First, Angola apparently hosts core populations of several phylogenetically distinctive rodent species. Even though Angola does not host palaeoendemics comparable, for instance, to the endemic genera from the Albertine Rift (Plumptre et al., 2007) or from the Ethiopian highlands , its rodent fauna bears an imprint of the country's geomorphological and climatic diversity, which allowed long-term survival of several unique deeply divergent lineages (e.g., in genera Graphiurus, Mastomys, Rhabdomys) and many local variants of more widely distributed rodent taxa. Nevertheless, a full appreciation of the uniqueness of Angolan rodents would require further sampling, as the rodent fauna has never been surveyed in many regions of Angola. Second, the remarkable diversity of rodent species is related to habitat diversity and the confluence of three major biogeographical regions (Linder et al., 2012), six biomes and 14 terrestrial ecoregions (Burgess et al., 2004;Huntley, 2019)  Attention should also be paid to seemingly indistinctive environments of low-and mid-altitudes, whose species may be seriously threatened by rapid and large-scale conversion of their natural habitats for agriculture. Currently, no region of the Angolan Escarpment falls into any protected area. Conservation efforts in this area are urgent as the local environment is threatened by progressively increasing human activities, especially logging and burning for charcoal production, and grazing of livestock (Baptista et al., 2018). We wish to thank P. Taylor, who kindly accepted to share his unpublished genetic data with us. We thank natural history museums and their curators for providing us with access to skin samples necessary for molecular barcoding. Namely, we acknowledge the contri- Kreislinger and B. Tichý.

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to declare. All fieldwork complied with legal regulations in Angola and sampling was carried out under permission of the Instituto Superior de Ciências de Educação, Lubango (see Acknowledgements).

PE E R 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.13435.

DATA AVA I L A B I L I T Y S TAT E M E N T
Details on all individuals used in the phylogenetic analyses (including the GenBank accession numbers of CYTB sequences) and details on all localities (including GPS coordinates) are provided in