Phylogenetics of the freshwater crab (Potamonautes MacLeay, 1838) fauna from ‘sky islands’ in Mozambique with the description of a new species (Brachyura: Potamoidea: Potamonautidae)

Patterns and processes of cladogenesis among taxa living on the Mozambique ‘sky islands’ remain poorly studied. During the present study, we report on a new freshwater crab species from Mount Lico, an inselberg and ‘sky island’ in the Zambezia Province of Mozambique. Phylogenetic analyses using three mitochondrial DNA sequence loci (12S rRNA, 16S rRNA and COI) were used to determine the evolutionary placement of the freshwater crab specimens from Mount Lico. The freshwater crab specimens from Mount Lico were retrieved sister to Potamonautes choloensis. The new species, Potamonautes licoensis sp. nov., is described and compared with other southern African freshwater crab species. Divergence time estimations for the Mozambican freshwater crab species suggest a Miocene / Plio–Pleistocene diversifi cation. Some endemic ‘sky island’ species form an early branching and are sister to other predominantly East African species, while other ‘sky island’ species are more recently derived and nested within a predominantly southern African clade. The present study presents the description of the fourth endemic freshwater crab species from Mozambique and suggests that the species diversity in the country is likely highly underrepresented, reiterating the call for renewed systematic surveys. An argument for the conservation of these mountainous ‘sky islands’ is presented.

bat (Taylor et al. 2012). Collectively, these results reiterate the renewed call for continuous biodiversity surveys of the neglected 'sky islands' in the interior of Mozambique. Such studies are important to fostering a better understanding of the evolutionary and biogeographic relationships of taxa in the region and aid the conservation management of these species.
A recent expedition, organized and led by the last author, to Mount Lico, an inselberg in the interior of the Zambezia Province of Mozambique, was undertaken to document its biodiversity, since the region has historically been unsampled. Mount Lico is approximately 1100 m above sea level (a.s.l) and has sheer rock walls of up to 700 m a.s.l., with old growth rain forest on top of a granite crater, and is uninhabited. The expedition resulted in the collection of a new freshwater crab species, which does not conform to the ten described species from Mozambique. The species occurs on the slope and atop of Mount Lico. In the present study, the new freshwater crab species is compared phylogenetically to the ten known species from Mozambique and described based on morphological and genetic data. In addition, we undertook a divergence time estimation of all of the described mountain-living species from Mozambique in order to understand the mechanism of speciation that are causal to cladogenesis. We compared the divergence time estimations obtained during the present study to those observed in previous studies to understand the phylogenetic relationship of the fauna of the Mozambique 'sky island' region.

Sample collection
Twenty-three freshwater crab specimens were collected from under stones by hand or using a dipnet in streams and rivers on the slopes and atop of Mount Lico, in the Zambezia Province, Mozambique. A handheld GPS was used to record the latitude and longitude, and freshwater crabs were preserved in either absolute ethanol or formalin.

DNA extraction, PCR and sequencing
Muscle tissue extracted from walking legs was subjected to DNA extraction using a Nucleospin kit (Macherey-Nagel, Duren, Germany), following the manufacturer′s protocol. Extracted DNA was stored in a refrigerator at -20 º C until required for PCR. Generally, a 1 μl DNA in 19 μl water dilution was performed prior to use. Three partial mtDNA gene fragments were selected for the present study; these included cytochrome oxidase subunit one (COI), 12S rRNA and 16S rRNA. These three loci were selected because each has a different mutational rate and they have been successfully used for reconstructing evolutionary relationships among freshwater crabs (Daniels et al. 2002a(Daniels et al. , 2002b(Daniels et al. , 2006(Daniels et al. , 2015Phiri & Daniels 2013Gouws et al. 2015;Wood & Daniels 2016;Daniels & Klaus 2018). Primer pairs are outlined in Daniels et al. (2002bDaniels et al. ( , 2006Daniels et al. ( , 2015. Standard PCR conditions for amplifi cation and DNA sequencing protocols were followed (Daniels et al. 2006(Daniels et al. , 2015Daniels 2017). Sequences for all the described southern African freshwater crab species were downloaded from GenBank for each of the three mitochondrial gene regions and included to understand the phylogenetic placement of the Mozambique freshwater crab species (Daniels & Bayliss 2012;Daniels et al. 2014Daniels et al. , 2019Peer et al. 2015Peer et al. , 2017Wood & Daniels 2016;Daniels 2017). We used four species of Liberonautes Bott, 1955(L. lugbe Cumberlidge, 1999, L. latidactylus (De Man, 1903, L. nimba Cumberlidge, 1999 andL. rubigimanus Cumberlidge &Sachs, 1989) as outgroups, since the latter genus is sister to Potamonautes (Daniels et al. 2015).

Phylogenetic analyses and divergence time estimation
Sequence Navigator (Applied Biosystems) was used to compute a consensus sequence from forward and reverse strands for each of the three gene fragments. No insertions or deletions were evident for the protein-coding COI locus and sequences for this locus were aligned manually. The 12S rRNA and 16S rRNA loci were aligned using Clustal X ver. 2.1 (Thompson et al. 1997). Since all three partial fragments occur on the mitochondria and are linked, we combined the DNA sequence data for the three fragments into a single data matrix and conducted all analyses on the combined data set. Maximum Likelihood (ML) and Bayesian approaches were used to estimate evolutionary relationships. jModelTest (Posada 2008) was used to obtain the best-fi t substitution model for each of the three gene loci (results not shown). These substitution models were used in the partitioned Bayesian analyses. The best-fi t substitution models were chosen using the Akaike Information Criteria (AIC) (Akaike 1973), since this reduces the number of parameters that contribute little to describing the data by penalizing more complex models (Nylander et al. 2004;Posada & Buckley 2004). Maximum likelihood analysis was conducted on the concatenated data set in RAXML ver. 7.2.7 (Stamatakis 2006). The robustness of branches of the best ML tree was assessed with 1000 bootstrap replicates using the CAT algorithm for fast bootstrapping, while the fi nal tree search was conducted under the GTR+G model for all three partitions as less complicated models are not implemented in RAxML. Bayesian inferences were used to investigate optimal tree space using the program MRBAYES ver. 3.2.6 (Ronquist et al. 2012). For each analysis, four Markov chains were run, with each chain starting from a random tree and run for fi fty million generations, sampling each chain every 10 000 th tree. This process was repeated four times to ensure that trees converged on the same topology using MRBAYES ver. 3.2.6. A 50% majority rule consensus tree was generated from the trees retained (after the burn-in trees were discarded -using likelihood plots) with posterior probabilities (pP) for each node estimated by the percentage of time the node was recovered. Posterior probabilities values < 0.95 pP were regarded as poorly resolved . Uncorrected 'p' distances were calculated for the COI locus among sister species pairs in PAUP ver. 4.0b10 (Swofford 2002).
To determine the divergence time estimations for the southern African potamonautid crabs, we applied a Yule tree prior and an uncorrelated lognormal relaxed molecular clock after initial test runs (investigating if the standard deviation of the uncorrelated lognormal clock approaches zero), using potamonautid substitution rates (and their standard deviations) as priors that originate from a fossil calibrated phylogeny of the family Potamonautidae Bott, 1970 (Daniels 2011;Daniels et al. 2015;Daniels & Klaus 2018): 0.81% per Ma for the rRNA loci (SD = 0.0013; linked clock models), 2.85% per Ma (SD = 0.005) for the COI locus. The maximum clade credibility tree was determined and annotated in TreeAnnotator ver. 2.4.1 (part of the BEAST package) after removal of 10% of the trees as burn-in.

Morphology
Preliminary phylogenetic analyses revealed the new species to be sister to P. choloensis Chace, 1953. Hence, we included 27 specimens of this species (Daniels & Bayliss 2012), deposited in the South African Museum of Natural History, Iziko Museums of Cape Town, for morphological comparison (accession numbers SAMC 46802-46807). Freshwater crab specimens of both species were divided into males and females, since freshwater crabs exhibit sexual dimorphism and characters for both females and males can thus be taken. The following measurements were taken with digital callipers: carapace length (CL); the carapace width at widest point (CWW); the width of the posterior margin of the carapace (CWP); the distance between the postfrontal crest and the anterior margin of the carapace (PFCD); the frontal width, measured between the medial margins of the orbits (FW); the distance between the exorbital teeth (CWA); the carapace height (CH); the length and width of the merus of pereiopods 2 and 5 (PML and PMW), the length of the propodus of the major cheliped (MCPL). All measurements are given in millimetres (mm). Samples have been deposited in the South African Museum of Natural History, Iziko Museums of Cape Town (SAMC). The new species from Mount Lico, Mozambique, was photographed in the fi eld and the laboratory with a Canon EOS 6D camera and Canon 100 mm f2.8 Macro USM lens. The structure of gonopods 1 and 2 for the new species and P. choloensis is compared. In addition, the major and minor cheliped of the new species were photographed with a 31 mm extension tube to the aforementioned setup to increase magnifi cation.

Phylogeny and divergence time estimation
A total of 1250 base pairs were included in the present phylogenetic analyses and included 600 bp, 450 bp and 310 bp fragments for the COI, 16S rRNA and 12S rRNA loci, respectively. The tree topologies from the BI and the ML analysis were nearly identical, hence, only the BI tree topology is shown (Fig. 1). Potamonautes MacLeay, 1838 was monophyletic (1.00 pP / 100 %) and comprised three deeply divergent clades. Clade 1 (1.00 pP / 98%) contained fi ve east African freshwater crab species, all occurring in Mozambique (but not all endemic to the country); the endemic P. namuliensis, sister to P. obsesus A. Milne-Edwards, 1868, in turn sister to a clade containing P. calcaratus Gordon, 1929 as sister to P. choloensis Chace, 1953 and P. licoensis sp. nov.
Using the COI locus, the sister species, P. choloensis and P. licoensis sp. nov. had an uncorrected 'p' sequence distance of 8.67%. The latter value is comparable to the uncorrected 'p' distances that have been reported for sister species. For example, the uncorrected sequence 'p' distances between P. dentatus and P. mhlophe was 10.89%, while between P. ngoyensis and P. ntendekaensis the uncorrected 'p' distance was 8.10%. Similarly, between the Indian Ocean Coastal Belt sister species, P. lividus and P. isimangaliso the COI uncorrected 'p' distance was 7.90%, while the uncorrected 'p' distance between these species and P. danielsi was 8.67%. The uncorrected sequence distance between P. licoensis sp. nov. and its sister species, together with its morphological distinction from P. choloensis leads us to recognize it as a distinct species.  2B). Cheliped: dactylus (moveable fi nger) slim, highly arched, enclosing oval interspace, with three larger teeth interspersed by smaller teeth along length; propodus (fi xed fi nger) with four larger teeth interspersed by smaller teeth along length ( Fig. 2A-C); carpus inner margin distal tooth large, pointed, proximal tooth reduced to granules (Fig. 3A); medial inferior margin of merus lined with series of small granules terminating distally at small, low distal meral tooth, lateral inferior margin smooth. G1 terminal article: ⅓ rd length of subterminal segment; fi rst third straight in line with longitudinal axis of subterminal segment, middle part directed outward at 45°, widened by raised rounded ventral lobe, tip curving sharply upward (Fig. 3A-B).

Etymology
Named for Mount Lico, from where the species was fi rst collected.

Colour in life
Carapace dark to light brown in living specimens (Fig. 5B).

Habitat
Primary rain forest streams on top of Mount Lico, Zambezia Province, Mozambique. Frequently found under small stones in fi rst order streams (Fig. 5A).

Distribution
Known from Mount Lico, Zambezia Province, Mozambique. More recently, the species has also been collected from Mount Nallume (Daniels pers. obs.).

Remarks
Phylogenetically, P. licoensis sp. nov. was retrieved as sister to P. choloensis (Fig. 1). Both species are characterized by a small but prominent epibranchial tooth on the anterolateral margins of the carapace and a arched right dactylus. However, the two species can easily be distinguished morphologically.
Potamonautes choloensis is a large-bodied species (CL > 42.00 mm) (Fig. 6A-C; Table 2), with a swollen carapace (CH/CL = 0.50), while P. licoensis sp. nov. is small-bodied (CL > 20.00 mm) and has a more fl at carapace (CH/CL = 0.44). Potamonautes licoensis sp. nov. shows adaptations to living under stones in fast fl owing mountain streams. Gonopods 1 and 2 of P. choloensis and P. licoensis sp. nov. are different in appearance. Gonopod 1 is nearly straight (Fig. 7A-B) in P. choloensis, while arching in P. licoensis sp. nov. (Fig. 3A-B). The terminal segment of gonopod 2 is fi lamentous in both species, however, in P. choloensis it is less curved (Fig. 7C) while it exhibits a signifi cant curvature in P. licoensis sp. nov. (Fig. 3C-D). Furthermore, the two species are characterized by an uncorrected 'p' distance for the COI locus similar to what has been recorded in other sister species pairs (Daniels & Bayliss, 2012;Daniels 2017;Daniels et al. 2019). Similarly, Gouws et al. (2015) found deep genetic divergence, ranging from 9.20 to 11.80% between lineages within P. sidneyi, leading to the recognition of P. danielsi. In addition, P. choloensis and P. licoensis sp. nov. are ecologically distinct. While both species are associated with forest streams at high altitude, Potamonautes choloensis is known to occur between 1000-2000 m a.s.l. and has been collected from Mounts Inago, Mulanje, Mabu, Cholo, the Zomba Plateau and more recently from Mount Socone (Fig. 8). The species is present in both Malawi and Mozambique (Daniels unpubl. data;Bayliss pers. com.;Chace 1953;Daniels & Bayliss 2012). Potamonautes licoensis sp. nov. is known from a maximum of 700-900 m a.s.l. and is present on Mounts Lico and Nallume, and is endemic to Mozambique. More extensive sampling is required to confi rm the distribution of P. licoensis sp. nov.
Superfi cially, P. licoensis sp. nov. resembles two other freshwater crab species that are also present in Mozambique, to whom it is phylogenetically related ( Fig. 1; clade 1). Potamonautes obsesus and P. calcaratus both have a single small, near spine-like tooth on the anterolateral margin of the carapace. However, both of the latter species are large-bodied and characterized by near swollen carapaces, chelipeds and gonopods, morphologically distinct from P. licoensis sp. nov. (Reed & Cumberlidge 2004). In addition, P. obesus and P. calcaratus are both burrowing, semi-terrestrial species, and live in ephemeral pans in low-lying areas, where they dig into soft soil to locate the water table during the dry   season (Daniels et al. 2002a;Reed & Cumberlidge 2004). Both species are widespread; P. calcaratus is present in northeastern South Africa and into Mozambique, while P. obsesus is present in Somalia, Kenya, Tanzania, Malawi, Mozambique and Zimbabwe (Reed & Cumberlidge 2004). Both species show deep genetic differentiation, a pattern typical of burrowing species, and both might potentially represent a species complex (Daniels & Bayliss 2012). Potamonautes licoensis sp. nov., in contrast, lives at high altitude in rain forest streams. Potamonautes namuliensis, endemic to Mount Namuli in Mozambique, has a smooth anterolateral margin with an arched dactylus of the major pereopod and can easily be differentiated from P. licoensis sp. nov. Potamonautes montivagus (Chace 1953) and P. bellarussus are both large bodied species with CL > 43.8 mm and 39.29 mm, respectively, and lack dentition on the anterolateral carapace margins (Chace 1953;Daniels et al. 2014). In P. montivagus the terminal segment of gonopod 1 is at a 90 o angle, very distinct from both P. bellarussus and P. licoensis sp. nov. Potamonautes montivagus is distributed in Malawi where it is present on Mounts Cholo, Mulanje (formerly spelled Mlanje), Chiradzulu, Ntchisi, Zomba Plateau and Dedza (Chace 1953). Potamonautes bellarussus is only known from two localities, Mounts Mecula and Yao in the Niassa Province of Mozambique, although specimens have recently been collected from neighbouring Malawi and Tanzania N. Cumberlidge and M. Genner pers. com.). Potamonautes sidneyi is present in Mozambique and widespread in southern Africa. It is a large bodied species mainly found in rivers and lacks dentition on the anterolateral carapace margin (Peer et al. 2017). Phylogenetically, P. sidneyi is not closely related to P. licoensis sp. nov., ( Fig. 1; clade 2). Potamonautes sidneyi is a species complex and comprises several genetically distinct lineages and is in need of a taxonomic revision .
Potamonautes bayonianus is present in Mozambique, however, it is a large-bodied riverine species, with a sharp and prominent tooth on the anterolateral carapace margin and distantly related to P. licoensis sp. nov. Finally, P. gorongosa, endemic to Mozambique, is phylogenetically and morphologically distinct from P. licoensis sp. nov. The former species lacks a dentition on the anterolateral carapace margin and does not have the highly-arched dactylus ( Fig. 1; clade 2) of P. licoensis sp. nov. (Fig. 1; clade 1). Potamonautes licoensis sp. nov. bears a superfi cial resemblance to P. parvispina (Stewart 1997). The latter species is also mountain-living, has a small tooth on the anterolateral carapace margins and a  Fig. 1; clade 2). All the remaining South African mountain-living freshwater crab species (P. clarus, P. depressus, P. brincki, P. parvicorpus and P. tuerkayi) lack dentition on the anterolateral carapace margin and are phylogenetically distinct from P. licoensis sp. nov.

Discussion
Our results clearly demonstrate the presence of a new freshwater crab species from Mount Lico, from the Zambezia Province in Mozambique. These results affi rm the call for renewed sampling of the interior of Mozambique, which contains several unsampled mountain areas (Fig. 8). Our phylogenetic tree (Fig. 1) indicates that the fi ve 'sky island' species are distributed between the three major clades and have a tendency to be narrow endemic species, with the exception of P. choloensis that is more widespread.
In clade 1, divergence in the basal 'sky island' lineage was initiated during the Miocene. However, the divergence between P. licoensis sp. nov. and its sister species, P. choloensis (present on several 'sky islands', Fig. 8), occurred more recently at the onset of the Pleistocene, while the divergence between P. namuliensis, a 'sky island' endemic from Mount Namuli, and P. obsesus occurred at the end of the Miocene (Fig. 1). In clade 2, the divergence of P. mulanjeensis, from Mount Mulanje, from its sister species occurred at the end of the Pliocene. In the latter clade, most of the species are the products of a Plio-Pleistocene divergence. Finally, in clade 3, the divergence of P. bellarussus, present on Mounts Mecula and Yao, from its east African sister species occurred at the end of the Miocene. The fact that these clades are distantly related suggests biogeographical infl uences at different time periods and from different aspects. The lack of monophyly of the 'sky island' freshwater crab fauna, and its distribution across three clades suggests that the divergence among the freshwater crab fauna from these 'sky islands' has occurred at various time intervals during the Mio / Plio-Pleistocene periods. While clade 1 is exclusive to Mozambique, clade 2 has species that predominantly occur in South Africa. In clade 3, P. bellarussus is the only Mozambican 'sky-island' species present in a largely East African clade that occurs further north, predominantly in Tanzania where it has also recently been recorded (N. Cumberlidge and M. Genner pers. com.), suggesting that there is a biogeographical divide between these mountains and those further south, which host species from clades 1 and 2. This supports the theory that these mountains are biogeographically distinct and may represent a distinct ecoregion, as suggested by Bayliss et al. (2014). These results suggest that the diversifi cation and colonization history of freshwater crabs in the region is biogeographically complex. The question now arises is what possible abiotic factors drove the observed cladogenesis among freshwater crabs on the high-altitude mountains of this region of Africa ('sky islands').
Our dated phylogeny indicates both ancient Miocene and a more recent Plio-Pleistocene diversifi cation for the freshwater crabs from the 'sky islands' (Fig. 1). The Miocene consisted of several short term, wet-dry cycles, with a general trend towards an increase in aridifi cation. This resulted in the contraction of lowland forest, the establishment of open grassy landscapes and the general restriction of forested regions to high altitude mountains with sustained precipitation (Rose et al. 2016). The timescale of the cladogenesis suggests that certain ancient, palaeoendemic lineages, such as P. namuliensis and P. bellarussus, were retained within refugial areas, while others, such as P. licoensis and P. choloensis, underwent a more recent diversifi cation within refugia and likely refl ect Plio-Pleistocene climatic shifts. The refugial areas likely represent areas of climatic and habitat stability for species and might be a pattern that persists across several taxa. However, the extent of this is unknown in the absence of dated phylogenetic studies for other 'sky island' taxa. There is considerable evidence for progressive vegetation shifts from C 3 to C 4 plants in the East Africa Biodiversity hotspot (Bonnefi lle 2010; Uno et al. 2016). The vegetation shifts can be associated with increased aridity and the progressive rifting and tectonic uplift in East Africa (deMenocal 2004), a result that is also supported by climatic modelling (Uno et al. 2016).
Several of these episodes of increased aridifi cation and vegetation shifts appear to be very localized, making it diffi cult to extrapolate this scenario to the broader East African Rift Valley system, and its general applicability to the 'sky islands' is limited by the lack of climatic modelling and palynological studies in these areas (deMenocal 2004). The phylogenetic studies conducted on the recently described 'sky island' fauna lack divergence time estimations, rendering direct comparison of cladogenic drivers diffi cult (Branch et al. , 2019. In this respect, multitaxon sampling across a broad spectrum of taxa is needed to properly understand factors causal to the observed pattern. However, in neighbouring South Africa, several animal taxa are thought to have diversifi ed during the Miocene / Plio-Pleistocene epochs as a consequence of shifts in climatic regime and their impact on vegetation, as can be observed in, e.g., tortoises (Hofmeyr et al. 2017), chameleons (Tolley et al. 2008), a forest dwelling gecko , frogs (Nielsen et al. 2018) and velvet worms (Barnes et al. 2020).
Several of the novel 'sky island' species are confi ned to very small patches of pristine forest habitat, suggesting that species need formal conservation assessment. Most of the 'sky island' freshwater crab species are narrow endemics, making them susceptible to extinction, suggesting that a conservation management plan should be developed to conserve the biodiversity of these 'sky islands'. In the interior of Mozambique there are several areas that remain unsampled. These include, for example, Mounts Chiperone, Ribáuè and Njessi -future surveys of freshwater systems should focus on these regions and are likely to reveal additional alpha taxonomic diversity for a number of poorly sampled animal groups. For example, we are aware of at least two, as yet undescribed freshwater crab species, that require additional collection, specifi cally that of male specimens (Fig. 8). In addition, collectors need to sample and preserve animals in ethanol to render them useful for DNA based evolutionary studies.