Aliens on Boats? The Eastern and Western Expansion of the African House Gecko

Invasive species disrupt relations between endemics and their ecosystem and are an increasing biodiversity conservation problem. The Hemidactylus genus comprises the most successful invasive reptile species, including the worldwide-distributed Hemidactylus mabouia. In this study, we used 12S and ND2 sequences to taxonomically identify and tentatively determine the diversity and origin of these invaders in Cabo Verde while also clarifying this for several Western Indian Ocean (WIO) populations. By comparing our sequences to recently published ones, we showed, for the first time, that Cabo Verde individuals belong to the H. mabouia sensu stricto lineage and that both of its sublineages (a and b) occur there. Both haplotypes are also in Madeira, which indicates a connection between these archipelagos, possibly related to the past Portuguese trading routes. Across the WIO, results clarified the identity of many island and coastal populations, showing that this likely invasive H. mabouia lineage is widespread in the region, including northern Madagascar, with important conservation implications. Colonisation origins were difficult to access due to the wide geographical spread of these haplotypes; thus, several possible scenarios were outlined. The introduction of this species throughout western and eastern Africa may threaten endemic taxa and needs to be closely monitored.


Introduction
Since the industrial revolution, the displacement of native species by invasive species and their impact on the introduced ecosystems has been a problem for biodiversity conservation [1]. Hasty anthropogenic globalisation is leading geographically distant areas to become closer than they have ever been. The establishment of trade routes allowed an easy and faster way for species to relocate and enabled the establishment of populations in ecosystems that were otherwise unreachable [2]. Invasive species have been found to disrupt relations between endemic species and their ecosystem, for instance, by competition for resources, predation, and transmission of diseases [3,4]. These invaders can present several attributes that provide them with an advantage over native taxa, such as the possibility of repeated transport or the absence of natural predators and parasites [5]. Native (often endemic) species are at a disadvantage by not being adapted to compete with the invader. In this regard, reptiles are among the most affected groups since they are widely introduced, both unintentionally and through illegal trade, while there is a lack of studies and models to predict such invasions [6]. This problem increases in importance in insular ecosystems, which are often species-poor and disharmonic (i.e., with a biased taxa Across several WIO islands and eastern African coastal localities, the systematics and distribution of Hemidactylus are still particularly unclear. Based on previous works [20] that used different markers from the currently most comprehensive study [10], the Aldabra Archipelago (at the northwest of Madagascar) harbours a set of endemic haplotypes closely related to the most widespread Malagasy ones, assigned to H. mercatorius. However, the Comoros Archipelago (Mayotte), Seychelles (Mahé), Mascarenes (Reunion), several eastern African coastal localities throughout Tanzania (including Zanzibar and Pemba islands), and South Africa harbour haplotypes closely related to the ones assigned to a divergent lineage of H. mercatorius from North Madagascar [26,27]. Individuals from Uganda and the West African Gulf of Guinea islands, such as S. Tomé, Príncipe, and Annobón, also belong to this latter clade, which is hypothesised to reflect human-aided dispersals [10,28].
Taking advantage of the most recent data [10], the main goal of this work was to place these islands' populations in this new detailed phylogenetic context for this speciescomplex, finding their closest phylogenetic affinities, as well as characterise their diversity and identify the possible geographical origin(s) of H. mabouia in Cabo Verde using molecular tools. Importantly, we compared them to populations from several West African and WIO locations, intending to definitely clarify their species-identity (H. mabouia vs. H. mercatorius). By sequencing several new samples from Cabo Verde and re-sequencing several representatives of the main haplogroups previously identified across the WIO and comparing them to the recently available phylogenetic data [10], we aim to answer the following questions. Are the introduced populations of H. mabouia in Cabo Verde derived from single or multiple events? What is the genetic variation of the species throughout the archipelago? To which of the 20 species-level lineages previously described do they belong? Where do the previously described H. mercatorius-mabouia lineages fall within the current phylogenetic framework for this group? With this work, we have clarified the taxonomic identity and origins of several island populations from both the western and eastern sides of the African continent.

Data Collection
For Cabo Verde, we collected a total of 11 individuals in the field morphologically assigned to H. mabouia from 2006 to 2012. We collected and preserved their tail tips in 96% ethanol and successfully amplified and sequenced the mtDNA 12S ribosomal RNA marker for ten individuals. We further sequenced the mitochondrial gene ND2 in two of those specimens. To further clarify their taxonomy, we re-sequenced representatives of the main H. mercatorius-mabouia complex haplogroups previously identified in the WIO region for the ND2 marker [20] to enable data to be directly compared to other studies [10]. We sequenced five individuals from clade A-I, representing five different 16S haplotypes and localities (H1/MY46; H3/TZ12; H5/SA20; H7/MA5; H8/Z40; check [10]), plus three individuals from clade III (endemic to Aldabra Archipelago; check [10]), also from different islands (see details in Table S1 in Supplementary Material). All samples included in these analyses with available geographical coordinates are shown in Figure 1.
We extracted DNA from tissue samples using a standard high-salt protocol [29]. We amplified the 12S marker following a previous study [30]. We amplified the ND2 fragment using the primers MetF1/L4437 and H5934 [31]. For the PCR, we used an initial activation temperature of 95 • C for 15 min, a denaturation temperature of 95 • C for 30 s, annealing at 60 • C for 35 s, extension at 72 • C for 45 s for 35 cycles, and finally, 10 min at 72 • C.

Alignment and Phylogenetic Inference
For ND2, from a first visual examination of a larger alignment inclu representatives of the different species-level lineages identified in the previous study it was clear that all our individuals were most similar to those identified as H. mabo and H. mercatorius (only the ones from Aldabra Archipelago). Thus, for further ana we used only the most similar ND2 sequences to ours, 892 base pairs plus a coup outgroup sequences (see below). For 12S, we collected from GenBank all ava Hemidactylus 12S sequences that aligned with our fragment of 349 base pairs. In tota retrieved 63 12S and 133 ND2 sequences and added them to our dataset. Det information about the locality and GenBank accession codes is presented in Table S1 visualised all 12S and ND2 sequences using Geneious [32] and aligned them using MA v7.505 [33] with the -auto option, which automatically selects the appropriate align strategy according to the data characteristics.
In detail, first, we built phylogenetic trees for the ND2 dataset with our 10 seque plus 127 H. mabouia ss sequences, three H. mercatorius, and one Hemidactylus c.f. ma sp. 3 plus two Hemidactylus sp. SA1 as outgroups. All sequences but ours are from most recent and complete study on this group [10] (see details in Table S Supplementary Material). These were collapsed into 46 haplotypes using ALTER [ only different haplotypes were kept except for the individuals sampled in this study used the IQ-TREE 2 [35] to infer a Maximum-Likelihood (ML) tree with 1000 boot replicates. The best model (inferred through ModelFinder [36]) for the ML analyse the TIM2 + F + G4. Then we used MrBayes 3.2 [37] to construct a phylogenetic tree u Bayesian inference (BI) using the same model. We ran BI analyses twice for 10 7 genera and a sampling frequency of 10 3 to produce 10 4 trees. After examining the loglikeli curves, we discarded 20% of the initial samples as burn-in. We considered ≥70% boot  Table S1.

Alignment and Phylogenetic Inference
For ND2, from a first visual examination of a larger alignment including representatives of the different species-level lineages identified in the previous study [10], it was clear that all our individuals were most similar to those identified as H. mabouia ss and H. mercatorius (only the ones from Aldabra Archipelago). Thus, for further analyses, we used only the most similar ND2 sequences to ours, 892 base pairs plus a couple of outgroup sequences (see below). For 12S, we collected from GenBank all available Hemidactylus 12S sequences that aligned with our fragment of 349 base pairs. In total, we retrieved 63 12S and 133 ND2 sequences and added them to our dataset. Detailed information about the locality and GenBank accession codes is presented in Table S1. We visualised all 12S and ND2 sequences using Geneious [32] and aligned them using MAFFT v7.505 [33] with the -auto option, which automatically selects the appropriate alignment strategy according to the data characteristics.
In detail, first, we built phylogenetic trees for the ND2 dataset with our 10 sequences, plus 127 H. mabouia ss sequences, three H. mercatorius, and one Hemidactylus c.f. mabouia sp. 3 plus two Hemidactylus sp. SA1 as outgroups. All sequences but ours are from the most recent and complete study on this group [10] (see details in Table S1 in Supplementary Material). These were collapsed into 46 haplotypes using ALTER [34]-only different haplotypes were kept except for the individuals sampled in this study. We used the IQ-TREE 2 [35] to infer a Maximum-Likelihood (ML) tree with 1000 bootstrap replicates. The best model (inferred through ModelFinder [36]) for the ML analyses was the TIM2 + F + G4. Then we used MrBayes 3.2 [37] to construct a phylogenetic tree under Bayesian inference (BI) using the same model. We ran BI analyses twice for 10 7 generations and a sampling frequency of 10 3 to produce 10 4 trees. After examining the loglikelihood curves, we discarded 20% of the initial samples as burn-in. We considered ≥70% bootstrap proportion and ≥0.95 posterior probability as evidence for highly supported branches. For H. mabouia ss, as divergence was low, we further built a haplotype network with the 134 ND2 sequences to best visualise the relationships between all individuals using TCS [38] with the connection limit set to 40 steps, so that all haplotypes were connected as in the previous reference study [10]. Finally, we also constructed a haplotype network with 72 sequences for the 12S fragment using TCS with default parameters. We used FigTree v1.4.4 (https://github.com/rambaut/figtree, accessed on 12 July 2022) and tcsBU v1.21a [39] to visualise the results.

Results
ML and BI phylogenetic trees for the ND2 fragment had largely congruent topologies. The ML tree is shown in Figure 2 with support values from both methods. Both trees with detailed support values can be found in the Supplementary Materials ( Figure S1). As in the previous study [10], they showed a clear distinction between the two major clades, H. mercatorius and H. mabouia ss, which were well-supported clades by both methods. Then two groups were present within H. mabouia ss: a highly supported one corresponding to H. mabouia ss a [10] and the other corresponding to H. mabouia ss b (unsupported; 52% bootstrap at ML). The two individuals from Cabo Verde sequenced for this marker (SVH03 and SVH04) are both well-nested into the H. mabouia ss b clade, exhibiting identical haplotypes. The WIO individuals fell in different clades: the ones from Aldabra Archipelago clearly grouped within H. mercatorius samples from Madagascar, most closely related to the sample from north-western Madagascar, while all the others (from Seychelles, Comoros, Tanzania, and South Africa) fell both within H. mabouia ss a and b. proportion and ≥0.95 posterior probability as evidence for highly supported branches. For H. mabouia ss, as divergence was low, we further built a haplotype network with the 134 ND2 sequences to best visualise the relationships between all individuals using TCS [38] with the connection limit set to 40 steps, so that all haplotypes were connected as in the previous reference study [10]. Finally, we also constructed a haplotype network with 72 sequences for the 12S fragment using TCS with default parameters. We used FigTree v1.4.4 (https://github.com/rambaut/figtree, accessed on 12 July 2023) and tcsBU v1.21a [39] to visualise the results.

Results
ML and BI phylogenetic trees for the ND2 fragment had largely congruent topologies. The ML tree is shown in Figure 2 with support values from both methods. Both trees with detailed support values can be found in the Supplementary Materials ( Figure S1). As in the previous study [10], they showed a clear distinction between the two major clades, H. mercatorius and H. mabouia ss, which were well-supported clades by both methods. Then two groups were present within H. mabouia ss: a highly supported one corresponding to H. mabouia ss a [10] and the other corresponding to H. mabouia ss b (unsupported; 52% bootstrap at ML). The two individuals from Cabo Verde sequenced for this marker (SVH03 and SVH04) are both well-nested into the H. mabouia ss b clade, exhibiting identical haplotypes. The WIO individuals fell in different clades: the ones from Aldabra Archipelago clearly grouped within H. mercatorius samples from Madagascar, most closely related to the sample from north-western Madagascar, while all the others (from Seychelles, Comoros, Tanzania, and South Africa) fell both within H. mabouia ss a and b.  Table S1). The designation of H. mabouia ss a/ b follows a recent study [10]. Bootstrap (≥70%) and posterior probability (BI) support (>95%) are represented at nodes (solid black dots if there was ML support and open white dots if there was both ML and BI support). The two basal branches were shortened for graphical purposes, and Hemidactylus sp. SA1 was used as an outgroup [10]. See Figure 1 and Table S1 for detailed locality data and GenBank accession numbers of the new sequences.  Table S1). The designation of H. mabouia ss a/b follows a recent study [10]. Bootstrap (≥70%) and posterior probability (BI) support (>95%) are represented at nodes (solid black dots if there was ML support and open white dots if there was both ML and BI support). The two basal branches were shortened for graphical purposes, and Hemidactylus sp. SA1 was used as an outgroup [10]. See Figure 1 and Table S1 for detailed locality data and GenBank accession numbers of the new sequences.
The ND2 network (Figure 3) represents the relationships within the two H. mabouia ss clades in greater detail. The two major haplotypes widespread across central and western Africa reaching the Americas are clearly identifiable as the most frequent (and roughly central) of each clade. For this marker, only one haplotype was detected in Cabo Verde, precisely the most frequent one of H. mabouia ss clade b, in individuals that also shared the same 12S haplotype (Figure 4). For the individuals from the WIO, the situation is slightly different: several haplotypes were detected, belonging to both clades, but none corresponded to the most frequent and widespread of each clade. The sample from Zanzibar that clustered within H. mabouia ss a (Z40) exhibited a haplotype also found in Burundi and the Democratic Republic of Congo (DRC), a few mutation steps from the central one. The remaining samples fell within clade ss b. The South African sample (SA20) shared its haplotype with samples from Burundi, Uganda, and DRC; the Seychelles sample (MA5, from Mahé Island) exhibited a haplotype also found in Uganda, which is also the case of the sample from Tanzania (TZ12). A sample from Comoros Archipelago (MY46, from Mayotte Island) presents a unique haplotype, more or less equally distant from the latter two.
We were able to amplify and sequence a larger number of samples from Cabo Verde for 12S than for ND2, and thus a 12S haplotype network is represented in Figure 4. It further shows the presence also in Cabo Verde of the two major H. mabouia ss haplotypes, the same ones spread across the large areas of West Africa and the Americas, as well as in Madeira.
The ND2 network (Figure 3) represents the relationships within the two H. mabouia ss clades in greater detail. The two major haplotypes widespread across central and western Africa reaching the Americas are clearly identifiable as the most frequent (and roughly central) of each clade. For this marker, only one haplotype was detected in Cabo Verde, precisely the most frequent one of H. mabouia ss clade b, in individuals that also shared the same 12S haplotype (Figure 4). For the individuals from the WIO, the situation is slightly different: several haplotypes were detected, belonging to both clades, but none corresponded to the most frequent and widespread of each clade. The sample from Zanzibar that clustered within H. mabouia ss a (Z40) exhibited a haplotype also found in Burundi and the Democratic Republic of Congo (DRC), a few mutation steps from the central one. The remaining samples fell within clade ss b. The South African sample (SA20) shared its haplotype with samples from Burundi, Uganda, and DRC; the Seychelles sample (MA5, from Mahé Island) exhibited a haplotype also found in Uganda, which is also the case of the sample from Tanzania (TZ12). A sample from Comoros Archipelago (MY46, from Mayotte Island) presents a unique haplotype, more or less equally distant from the latter two.
We were able to amplify and sequence a larger number of samples from Cabo Verde for 12S than for ND2, and thus a 12S haplotype network is represented in Figure 4. It further shows the presence also in Cabo Verde of the two major H. mabouia ss haplotypes, the same ones spread across the large areas of West Africa and the Americas, as well as in Madeira.  Table S1) as well as the accession numbers/ tissue codes which are given next to the nodes or in boxes. The circle size represents the haplotype frequency. The grey boxes in the network delimit H. mabouia ss a (ss a) and H. mabouia ss b (ss b), following the ML tree and a previous study [10].  Table S1) as well as the accession numbers/tissue codes which are given next to the nodes or in boxes. The circle size represents the haplotype frequency. The grey boxes in the network delimit H. mabouia ss a (ss a) and H. mabouia ss b (ss b), following the ML tree and a previous study [10].  Table S1) as well as the accession numbers/ tissue codes which are given next to the nodes or in boxes. The circle size represents the haplotype frequency.

Discussion
Molecular phylogenetic methods are of great utility to precisely identify cases of cryptic species and lineages and tentatively infer colonisation or invasion routes, as is the case here. The close relationship between H. mabouia ss and H. mercatorius and the unclear identity of some populations was clearly shown in a previous study [20], to the point that some refer to specimens from Cabo Verde as H. mercatorius [21]. While these doubts remained for years due to the use of different gene fragments by different authors-16S rRNA was used [20], leading to incomparable datasets until now-the most recent study showed clearly that H. mabouia is a species-complex of circa 20 species-level lineages, as well as that the populations consensually called H. mercatorius (from central-south Madagascar) make H. mabouia paraphyletic [10]. Importantly, amongst the several H. mabouia lineages, they define H. mabouia ss as a sister taxon to H. mercatorius and a wellsupported clade of two sublineages (a and b) widely distributed across sub-Saharan Africa, the West African islands of S. Tomé, Príncipe, and Bioko, and several localities of the Americas. Phylogeographic patterns and biogeographic reconstructions support the origin of H. mabouia ss in the East African Valley region, with relatively recent expansions throughout its western range, including to the Gulf of Guinea Islands and South America. At least throughout this western range, this species is most likely invasive, given its clear preference for human-dominated habitats and its proven ability to outcompete and directly prey on both native and other invasive geckos [14,16,17,40].
This new phylogenetic framework provided the ideal opportunity to further clarify the identity of several island and coastal populations on both sides of the African continent, including the ones from Cabo Verde, while at the same time exploring possible colonisation routes and origins (native or introduced) of several island populations, which are of utmost importance for conservation management. As such, the phylogenies now generated by combining representatives of the haplogroups/ clades described in previous works [20] with the recently generated ND2 data provided us with important insights: populations from Cabo Verde are clearly assigned to H. mabouia ss, as well as all the representatives of the WIO populations, except for the ones of the Aldabra Archipelago, which are undoubtedly a lineage of H. mercatorius endemic to that archipelago. This allows us to say that the lineage referred to as H. mabouia-mercatorius clade A-I [20] are, in fact, H. mabouia ss, implying that these populations found in the islands of Mayotte (Comoros), Mahé (Seychelles), Reunion (Mascarenes) [41], Zanzibar, and Pemba  Table S1) as well as the accession numbers/tissue codes which are given next to the nodes or in boxes. The circle size represents the haplotype frequency.

Discussion
Molecular phylogenetic methods are of great utility to precisely identify cases of cryptic species and lineages and tentatively infer colonisation or invasion routes, as is the case here. The close relationship between H. mabouia ss and H. mercatorius and the unclear identity of some populations was clearly shown in a previous study [20], to the point that some refer to specimens from Cabo Verde as H. mercatorius [21]. While these doubts remained for years due to the use of different gene fragments by different authors-16S rRNA was used [20], leading to incomparable datasets until now-the most recent study showed clearly that H. mabouia is a species-complex of circa 20 species-level lineages, as well as that the populations consensually called H. mercatorius (from central-south Madagascar) make H. mabouia paraphyletic [10]. Importantly, amongst the several H. mabouia lineages, they define H. mabouia ss as a sister taxon to H. mercatorius and a well-supported clade of two sublineages (a and b) widely distributed across sub-Saharan Africa, the West African islands of S. Tomé, Príncipe, and Bioko, and several localities of the Americas. Phylogeographic patterns and biogeographic reconstructions support the origin of H. mabouia ss in the East African Valley region, with relatively recent expansions throughout its western range, including to the Gulf of Guinea Islands and South America. At least throughout this western range, this species is most likely invasive, given its clear preference for humandominated habitats and its proven ability to outcompete and directly prey on both native and other invasive geckos [14,16,17,40].
This new phylogenetic framework provided the ideal opportunity to further clarify the identity of several island and coastal populations on both sides of the African continent, including the ones from Cabo Verde, while at the same time exploring possible colonisation routes and origins (native or introduced) of several island populations, which are of utmost importance for conservation management. As such, the phylogenies now generated by combining representatives of the haplogroups/clades described in previous works [20] with the recently generated ND2 data provided us with important insights: populations from Cabo Verde are clearly assigned to H. mabouia ss, as well as all the representatives of the WIO populations, except for the ones of the Aldabra Archipelago, which are undoubtedly a lineage of H. mercatorius endemic to that archipelago. This allows us to say that the lineage referred to as H. mabouia-mercatorius clade A-I [20] are, in fact, H. mabouia ss, implying that these populations found in the islands of Mayotte (Comoros), Mahé (Seychelles), Reunion (Mascarenes) [41], Zanzibar, and Pemba (Tanzania), as well as the ones present in some (mostly coastal but not only) localities of both Tanzania and South Africa, are highly likely introduced through human-aided transport, which seems to be the predominant pattern of this species. This was, in fact, expected given their apparently restricted distribution and only recent detection across some of these locations (1905 and 2010 for Mahé and Reunion, respectively). With respect to the mainland locations where this species may have spread naturally vs. due to human aid, however, a much more detailed sampling throughout East Africa and higher resolution molecular data will be needed to properly clarify it. Populations from the Gulf of Guinea Island of Annobón (southwest of S. Tomé) [20] also belong to this clade. This is important information in terms of invasive species management, especially for those small islands where these introduced species may have a significant negative impact on native ones. Notably, these new results also allow us to infer that H. mabouia ss is present in northern localities of Madagascar (at least in Montagne des Français, Sambava, and Andapa-haplotypes H9, H11 and H37 from clade A-I [20]), corresponding to what has been erroneously interpreted as an H. mercatorious divergent lineage from that area [20,26,27]. Again, this adds to the list of invasive reptiles of these biodiversity hotspot islands and certainly has important conservation implications.
Regarding the genetic diversity of the introduced populations across these places, for Cabo Verde, it is now clear, taking together the 12S and ND2 results, that, similarly to what was found for Madeira and South America, both H. mabouia ss subclades are present, with individuals from subclade a (the most abundant one also in Madeira) present in S. Vicente, Santo Antão, and Brava and further populations assigned to subclade b in S. Vicente and Santo Antão for now. For ND2, only two individuals from Cabo Verde were successfully amplified (our samples were very old and most of them were too degraded to sequence), these exhibiting the most frequent and widespread haplotype from subclade b. However, for 12S, individuals from different islands were successfully amplified and sequenced, and one haplotype of each clade was found overall, and these were the same ones widespread across Madeira, South America, and also present in S. Tomé, Príncipe, and Bioko islands, clearly pointing to the expected scenario for this species of a relatively recent human-aided invasion of Cabo Verde. The fact that both haplotypes present in Cabo Verde are also present in Madeira indicates a possible connection between these archipelagos [23]. In fact, other studies have shown a possible colonisation route linking Madeira and Cabo Verde mediated by humans in mice Mus musculus [42]. This opposes previous studies, which claimed that H. mabouia found in Cabo Verde and Madeira all belong to one lineage [43] and that one of the haplotypes found in Madeira was exclusively related to South American individuals [23], and highlights the importance of a comprehensive sampling before any conclusions about dispersal patterns can be made, especially in species with cosmopolitan distributions, such as this one. Colonisation origins of the populations from Cabo Verde are still impossible to resolve due to the wide geographical spread of these haplotypes. Several scenarios are possible according to our results: a single south American origin from a place where both clades exist in sympatry, such as Brazil, as proposed for Madeira [23] or Puerto Rico [10]; a single colonisation from West Africa, again from a place where both clades co-habit; multiple colonisations both from South America or West Africa (or both); or a (single or double) colonisation from Madeira, such as hypothesised for M. musculus [42]. Further data with much higher resolution and a much more comprehensive sampling will be needed to distinguish these hypotheses and further clarify the phylogeographic and invasion patterns of this species.
In the WIO islands, the pattern was different. Although both H. mabouia ss subclades were again found in the region, none of the haplotypes found was the most frequent (and widespread) of each clade. On the contrary, there, we found low-frequency haplotypes some mutations away from the most frequent one, in most cases shared with East African individuals from Burundi, Cameroon, and Uganda from a previous study [10]. The sequenced representatives of H. mercatorius-mabouia clade A-I cover all the variation previously described for 16S within this group [20], which is completely unstructured, with many haplotypes shared between East African, Comoroan, and Seychelles populations.
Only five representatives were sequenced in this work. The 16S data in a previous work [20] provided a more complete picture of the extensive degree of haplotype sharing across these locations. Altogether, this pattern most likely reflects multiple origins along the east part of the East African Rift Valley for these WIO islands populations, from a much more diverse genetic pool that the one on the western African coast. However, precise origins for these introductions of the island populations and clarification of the native vs. introduced origin of continental East African populations must await a much more comprehensive sampling across its East African distribution and, again, higher resolution markers. Given the case that haplotypes are most often identical to other locations, there would be no power in the current data to estimate colonisation ages. For that, a population genetics approach with higher resolution markers should follow.
Importantly, these inferences are so far based on mtDNA only. For a thorough look into dispersal and differentiation patterns and processes and species delimitations across this complex of circa 20 species-level lineages, multiple nuclear markers should be taken into account. These may also be important to understand if any genetic impact was caused in native closely related populations due to hybridisation and/or introgression. However, that was outside the scope of this work.
The introduction of invasive species to both Cabo Verde and the WIO islands can threaten endemic taxa. Since H. mabouia is one of the most effective invaders of the Hemidactylus genus [10], they should be closely monitored to ensure that they do not outcompete the native species. This is especially important, knowing that the other introduced Hemidactylus species in Cabo Verde, H. angulatus, has been displacing the native Hemidactylus boavistensis (Boulenger 1906) [44]. Additionally, in the WIO Mauritius Island (Mascarenes), food competition was demonstrated between the introduced Hemidactylus frenatus and the native diurnal gecko Phelsuma ornata [45], as well as its effect in displacing the nocturnal endemic Nactus species [12]. Overall, throughout the world, invasive Hemidactylus have been constantly associated with the decline in native species, especially other geckos, both through site competition and direct predation (e.g., [46][47][48]) For this reason, and especially now knowing their specific identity and its usual invasive behaviour, efforts should be made to clearly assess their current distribution and abundance across these different islands, and avoid their spread within and across them. Considering the noticeable fauna of endemic diurnal and nocturnal geckos across all these islands, their impact is expected to be considerable.

Conclusions
In summary, we clarify the identity of populations from Cabo Verde as H. mabouia ss, describe their current genetic variation across the archipelago, confirm the distinctiveness of the Aldabra Archipelago population as an endemic lineage of H. mercatorius, and assign other populations from the WIO islands, including populations from the north of Madagascar also to H. mabouia ss. Although precise colonisation sources could not be inferred due to the widespread distribution of identical haplotypes, this preliminary data shows that Cabo Verde is related to West African, Madeira, and South American populations, while WIO island populations have clear affinities with East African populations. Detailed sampling, particularly in East Africa, and higher resolution molecular data are still needed to infer with more detail the biogeographic history and spread of this species.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/genes14020381/s1, Table S1: Details of the Hemidactylus samples used in this study. The taxon (Hm SA1, H. sp. SA1; Hm sp3, H. cf. mabouia sp. 3; Hr, H. mercatorius; Hm, H. mabouia sensu stricto), tissue code, GenBank accession number (GenBank), country of origin (STP, São Tomé and Príncipe; DRC, Democratic Republic of the Congo), locality, geographical coordinates, and colour group of the sequences used for the two markers in this study are given. The * stands for new samples from this study; Figure S1: Phylogenetic trees. (A) Maximum likelihood tree, and (B) Bayesian inference for the ND2 fragment showing the relationships between Hemidactylus from Cabo Verde and WIO region and their relatives; File S1: Portuguese version of the abstract.