In addition to H. frenatus, six other species of house geckos, mostly small gray or brown, largely facultatively anthropophilic animals, inhabit the historically uninhabited remote islands of the Indian Ocean. None of the volcanic islands of Mascarenes and Comoros, nor the continental fragments (Madagascar and the granitic Seychelles), much less the numerous atolls of the western Indian Ocean (Figure 3), have been connected to any continent for upward of 60 million years, and therefore either natural or human-mediated over water dispersal are necessary to explain their distribution.

Of the seven, two, Hemidactylus platycephalus and the H. mabouia-mercatorius complex are Afro-Malagasy species (Rocha et al., 2010), which historically were merged within the polyphyletic umbrella species ‘H. mabouia’ (Louette et al., 2004; Rocha et al., 2010; Agarwal et al., 2021). Of these, only H. mercatorius seems to have spread beyond the immediate area (Madagascar, Comoros, Aldabra group, Farquhars), to the granitic Seychelles and, unexpectedly, to Platte, but only recently via human agency (Gerlach, 2007; Rocha et al., 2009a). For example, although Boulenger (1909) reported “H. mabuia” on Mahé in 1905 (specimen confirmed by Cheke (1984) as H. mercatorius), it probably did not establish a population since it was not detected again there until 1995, and in 2002 on Frégate (Gerlach, 2007). Platte, a low coral island, is infrequently visited by zoologists, and it is possible that H. mercatorius may have been long established. H. parvimaculatus was initially misidentified as H. mercatorius (Vinson and Vinson, 1969) in the Mascarenes, but true mabouia/mercatorius was absent in the Mascarenes (Cheke and Hume, 2008; Cole, 2008) until discovered in Réunion in 2010, with molecular data suggesting their recent introduction from either the Comoros, the Seychelles or the East African coast (Sanchez et al., 2012), the most likely of these being Mayotte (Comoros), also still a French territory.

Of the other five species, one is Pacific, and four are Asian in origin, but with differing dispersal histories. Lepidodactylus lugubris is a parthenogenic Pacific species with apparently endemic Indian Ocean clones (Ineich, 1999), which suggests a long-established natural dispersal via drift. This species is also the least associated with humans, and only occurs on the most easterly of the islands including Rodrigues, the Chagos group, and Coëtivy (Gerlach, 2007; Cheke and Hume, 2008; Cole, 2009). Boulenger (1909) reported specimens of L. lugubris collected in 1905 from both Mahé and Praslin in the granitic Seychelles, but it has not been reported since (Gerlach, 2007), and no specimens exist in the United Kingdom Natural History Museum collections (Cheke, 1984), so this record may have been made in error. L. lugubris does occur sympatrically in houses in Rodrigues with three other gecko species (ASC pers. obs.), and it has also recently been found on St. Brandon/Cargados Carajos (Nik Cole, pers. comm. to ASC, 2020), directly in the path of the South Equatorial Current (New et al., 2005).

Parthenogenic Hemiphyllodactylus typus, only found in the Mascarenes, is not usually associated with buildings (Grégory et al., 2007) and was first reported in 1948 (Vinson and Vinson, 1969; Deso et al., 2020). Two studies (Grégory et al., 2007; Deso et al., 2020) however, suggested that given the difficulty in observing this species, it may have escaped notice and be native in the Mascarenes, especially since it appears to favor the native forest habitat. This narrative would require oceanographically difficult cross-equatorial drifting. Another Asian gecko species, Gehyra mutilata, is widespread in the Indian Ocean, but lacks sufficient genetic variation in this region to identify its geographical origins (Rocha et al., 2009b). G. mutilata is strictly commensal and is unlikely to have survived on islands where the human presence was only fleeting. Therefore, even if cross-ocean mariners introduced the species unintentionally, a temporary human stopover would not be sufficient for the species to maintain a presence. In fact, this species failed to establish a population on Aldabra until there was a permanent human presence at a research station (Cheke, 1984; Gerlach, 2007).

The final two species, Hemidactylus frenatus and H. parvimaculatus, are both facultative commensals, with distributions, histories and genetic signatures that warrant further investigation. Both of these species, common in southern India and Ceylon/Sri Lanka (Daniel, 2002; Das and de Silva, 2005), have long-term established populations in the Maldives where H. frenatus was likely introduced on multiple occasions (Agarwal et al., 2019). When the first faunal collections were made in the Maldives by the Stanley Gardiner expedition of 1899-1900, H. frenatus was widespread throughout, but parvimaculatus (as ‘H. gleadovii’) was only collected on Hulule, Male Atoll (Laidlaw, 1902). People from Sri Lanka are known to have settled on numerous occasions, so the presence of H. parvimaculatus is not unexpected, and H. frenatus could also have arrived from long-standing trading (Maloney, 1980; Litster, 2016) with ports to the east.

The Maldives are long-inhabited islands, possibly from 1400 BCE (Maloney, 1980; Jaufar, 2019), though Litster (2016, 2020) is more cautious about BCE dates, suggesting probable early stopover use, if not settlement. The oldest confirmed direct radiocarbon date is remarkably late, 249–393 CE, evidently after the cowry-shell (Monetaria moneta) trade for currency was well-established. While they could be the source of gecko populations elsewhere, it is the presence of geckos on further-flung islands first encountered by humans more recently, in the last 1,500 years, mostly the last 500, that concern us here.

Hemidactylus parvimaculatus is part of the Indo-Asian H. brookii complex, and was only recognized as a separate species in 2010; the natural range of the species being the island of Sri Lanka and also southern India (Bauer et al., 2010). Although not reported in Mauritius until 1818 (Cheke and Hume, 2008) (as ‘H. brookii’), and still absent from the granitic Seychelles (Gerlach, 2007), it was collected on then uninhabited isolated sand cay of Coëtivy in 1803. Coëtivy, discovered in 1771 (Lionnet, 1972), was first settled in 1811 (de Froberville, 1848), though it had been visited on various occasions by Seychellois and French/Mauritians (Toussaint, 1967) seeking free coconuts and turtles. This specimen was illustrated in a remarkable watercolor of seven geckos by Baudin expedition artist Charles-Alexandre Lesueur (Collection Lesueur, Muséum du Havre MS 78-115; reproduced in Cheke (2009); Figure 1); the species was also collected in 1905 on the remote atoll of Desroches in the Amirantes (Boulenger, 1909; as H. brookii), and was still present in 1981 (Gardner, 1986).

Its presence on these isolated low islands is anomalous, and is unlikely to result from natural drift (it would be a unique cross-equatorial case of sea drift in the Indian Ocean). It is also unlikely to have been brought from Mauritius, whence the Seychelles and their outer islands were settled, as it would have shown up on the much more populated main granitic islands, where it is still absent (Gerlach, 2007; Rocha et al., 2009a). It is therefore possible that it arrived via Austronesian cross-ocean mariners traveling via Sri Lanka or the Maldives - the species does not occur in these travelers’ source areas (Sumatra, Borneo). The species was not found on Desroches in 2005 despite targeted searches (Rocha et al., 2009a) - only frenatus, absent in 1905, (Boulenger, 1909 contra Gerlach, 2007) but present in 1981 (Cheke, 1984; Gardner, 1986) was identified, nor has it been rediscovered on Coëtivy (Cheke, 1984; Gardner, 1986). A more likely source is the Maldives: Maldivian folklore recounts tales of boats lost at sea ending up in remote islands, so, although the islands are not specifically named (apart from Chagos; Romero-Frías, 1999; Romero-Frias, 2012), this could be the source of H. parvimaculatus on Desroches and Coëtivy.

One story, Hoïïavai (= Chagos in Maldivian; Litster, 2016), was based on a 17th century BCE case of a fishing vessel wrecked in the Chagos, and its crew rescued to tell the tale (Romero-Frias, 2012). Although some Portuguese ships had been wrecked on the northern Chagos banks in the mid-late 16th century, they were traders heading for the East Indies and back (Wenban-Smith and Carter, 2016), so would have been unlikely to have carried parvimaculatus. The same applies to the first known deliberate landing on Egmont in 1605 by an English party traveling east (Wenban-Smith and Carter, 2016). There is no evidence of any European landing on Diego Garcia until much later. H. parvimaculatus also occurs on Moheli and Anjouan (Comoros) (Vences et al., 2004; Rocha et al., 2005; Hawlitschek et al., 2011), and all three Mascarenes (Vinson and Vinson, 1969; as ‘H. mercatorius’; Vences et al., 2004), but not Madagascar (Glaw and Vences, 2007).

Two clades are present in the Mascarenes (Vences et al., 2004; Rocha et al., 2005), and one extends to the Comoros, potentially indicating invasion events from different source populations. Although Austronesian and subsequent pre-European cross-ocean trading from c.800 BCE (Cheke, 2010; Boivin et al., 2013) is a possible source of the Comorian clade (with back-spread to Réunion later), a post-European origin of both clades is probably more likely since the Comoros was formerly administered colonially from Réunion. There is no reason to suppose that Hemidactylus spp. in the Mascarenes arrived other than accidental transport post-European contact since there is no evidence of pre-European landings in this group (Cheke and Hume, 2008; Cheke et al., 2017).

The most complex and interesting case is that of the common house gecko Hemidactylus frenatus. Although its almost universal presence on Indian Ocean islands suggests a generalized, probably anthropogenic distribution, there are numerous anomalies within this apparently uniform pattern. Rocha et al.’s (2010) statement that this species is ‘present throughout Indian Ocean islands without any signs of geographical structure’ overlooked the fact that some structure is evident in both this and previous papers (Vences et al., 2004; Rocha et al., 2005) as discussed below.

House geckos tend only to be reported by naturalists, and not by mariners or explorers, so it is unsurprising that the first Indian Ocean record is not until c.1770 when Philibert Commerson’s artist Paul Jossigny drew a Hemidactylus frenatus from Mauritius; the previously unpublished image is preserved in the Muséum National d’Histoire Naturelle in Paris (Cheke and Hume, 2008), reproduced by Bour (2015); Figure 2. A report (Hoffstetter, 1946) identifying some gecko subfossils as H. frenatus was made in error; the bones were those of the osteologically similar endemic gecko Cyrtodactylus (now Nactus) serpensinsula (Arnold, 2009), extinct on the Mauritian mainland, and only discovered alive on offshore islets after Hoffstetter’s study. Despite intensive fossil collection across different Indian Ocean islands (Aldabra, Reunion, Mauritius, Madagascar and Seychelles), where other gekkonid species have been identified (Cheke and Hume, 2008; Hume, 2004; Hume et al., 2018) there are no known subfossils of Hemidactylus spp. in the Indian Ocean, suggesting that a natural and old (pre-human) colonization is unlikely.

The interesting aspects of frenatus are its curious distribution on outlying atolls and sand cays of the Seychelles (Figure 3) – the species is widespread on these islands which have only small agricultural (coconut) or, more recently, tourist-related settlements. Despite this, the species was absent on the much more populated, visited and biologically known granitic islands until the 1990s (Gerlach, 2007), barring a doubtful record from “the Seychelles” (island unspecified) in 1863 (Peters and von der Decken, 1869).

H. frenatus occurs, or occurred, on many low islands and atolls visited by the 1905 Stanley Gardiner/Percy Sladen Trust expedition (Boulenger, 1909; Cheke, 1984). There are no earlier records since, other than Rémire (Amirantes; collection in 1882; Günther, 1884), no-one had previously collected lizards on these islands. The expedition found them on Bird (central Seychelles), Poivre, Rémire and Desroches (Amirantes), Sirène and ‘Establishment’ [ = Raphael] (St. Brandon), Diego Garcia, Salomon group, Peros Banhos group and possibly Egmont (Chagos), but none were identified on the granitic islands. This species has subsequently been found on several other low islands: Platte, Alphonse, Desnoeufs, African Banks (Cheke, 1984; Gerlach, 2007), Agalega (Cheke and Lawley, 1983; Webster and Cadinouche, 2013), and belatedly reconfirmed on St. Brandon (Nik Cole, pers. comm. 2020). The species was, and remains (Cheke, 1984; Gerlach, 2007), absent on more westerly atolls closer to Madagascar, where H. mercatorius was, and is, present (Farquhars, Aldabra, Cosmoledo, Astove) (Boulenger, 1909), and where the currents favor transport northwest from Madagascar, and at times effectively north/north-east via the SECC (Peng et al., 2015; Hawlitschek et al., 2017).

While commonly found in houses, H. frenatus is found living in rocks and trees on islands inhabited only by seabirds in the Amirantes’ (Gardner, 1986), leading him to agree with Cheke (1984) that this distribution was due to natural sea-drift from the east. However, that does not explain why the drift missed the granitic islands (but hit Bird further north), and carried the geckos further west to the Amirantes. The fullest flow of the South Equatorial Current runs south of the Seychelles Bank (New et al., 2005), which might have impacted the Amirantes more than the granitics, which would also be ‘protected’ (in a sense) by the shallow bank diverting the flow around it. Unlike many of the atolls, the granitic islands have other native lizards including geckos, but pre-existing reptiles have never been known to prevent its establishment elsewhere, and H. frenatus is known to aggressively out compete other similarly sized geckos (e.g., Cole et al., 2005; Dame and Petren, 2006; Newbery and Jones, 2007; Csurhes and Markula, 2009).

Hemidactylus frenatus is also widespread in Madagascar (Glaw and Vences, 2007) and is present on all four Comoro islands (Louette et al., 2004; Hawlitschek et al., 2011). Unlike the oceanic high islands and atolls (Maldives excepted), these have been inhabited for well over a millennium (Crowther et al., 2016; Anderson et al., 2018b), but reptile collections, apart from one Malagasy chameleon, only date from the mid-19th century BCE (Andriamialisoa and Langrand, 2003; Hawlitschek et al., 2011). Hence, there is no historical basis for insight into the means or date of arrival of the species there. Some consider three, or all four, species of Hemidactylus in the Comoros to have been introduced (Vences et al., 2004; Hawlitschek et al., 2011) but given the distribution and habitat preferences (Hawlitschek et al., 2011, 2017), the evidence for mercatorius and especially platycephalus remains ambiguous.

In summary, from the seven largely commensal house gecko species that today inhabit Indian Ocean islands, only the early records of both H. parvimaculatus and H. frenatus across some isolated islands suggest the possibility of early (pre-European) colonizations or even natural dispersals. Of these, H. frenatus is undoubtedly the widest distributed and most abundant across the Western Indian Ocean (and globally), and we therefore tried to gather a comprehensive widespread set of samples that allowed for an exploration of the geographic distribution of its genetic variation, and possibly shed some light into this species’s (early or modern) dispersal patterns.

Several recent studies have attempted to resolve colonization patterns across parts of the range of H. frenatus using mtDNA sequences. Vences et al. (2004) analyzed partial 16S rRNA sequences from individuals from Madagascar, Grand Comoro, the Andaman islands, Mauritius, Rodrigues and Sri Lanka and considered this to show “low genetic differentiation … with no recognizable phylogeographical structure, indicating recent colonization or introductions.” Rocha et al. (2005) focused on the Comoro islands, and likewise, did not identify any notable structure with the same 16S rRNA marker.

Using partial 12S rRNA and Cytochrome b (CYTB) gene sequences to examine the origin of South American populations, Torres-Carvajal (2015) found no variation across Ecuador, Colombia, Hawaii and Papua New Guinea, although these were highly distinct from lineages in Myanmar/Burma and India. Tonione et al. (2011) also identified high diversity within the tiny island of Moorea, French Polynesia, based on Cytochrome Oxidase 1 (CO1) sequences, and Agarwal et al. (2019) determined the probable origin of H. frenatus from the Maldive Islands to be from Southeast Asia. Both these latter studies proposed multiple introductions, and the need for improved sampling of potential source populations to unravel colonization patterns.

At least five lineages have now been identified within H. frenatus, and geographic structure is present. In addition to the Indian and Sri Lankan lineages, the worldwide (Indian, Pacific and Atlantic Ocean islands) samples analyzed here fit clearly into three well-supported clades. One (red in Figure 3) is present in both Myanmar and the Western Indian Ocean (WIO) islands of Madagascar and Comoros (Moheli), another (green) is found both in southeast Asia (Vanuatu, Philippines, Malaysia, Myanmar) as well as across some WIO islands (Diego Garcia, the Seychelles -Mahé- and Madagascar) and also in the Atlantic Ocean island of St. Helena. The third (yellow) is also widespread across the Pacific and Indian Ocean islands.

Since we sequenced samples from the main 16S rRNA haplogroups in the Western Indian Ocean, we were able to match these three lineages of CYTB to the different haplogroups described in Rocha et al. (2010). Although structure was, unsurprisingly, previously less obvious using the more conserved locus than with CYTB, these three lineages are also reflected in 16S data. In this way, the yellow clade corresponds to 16S haplotypes H4-H7 and H10, the red clade to Comoros and Madagascar H8 and H9, and the green clade, to remaining haplotypes from Mahé (represented by H2 - sample 2MA21).

We did not sample additional South American localities, but extrapolating from Torres-Carvajal (2015), South American localities from Colombia to Ecuador (including Galapagos) all belong to the yellow lineage, which is further present also at least in Myanmar (Tonione et al., 2011). In addition, though not all lineages can be unambiguously associated with the ones in Agarwal et al. (2019), some correspondence is evident (e.g., lineage A from (Agarwal et al., 2019) with the blue lineage here, both solely in Sri Lanka). The remaining lineages are not comparable, highlighting the diversity harbored by this taxa, especially in Southeast Asia.

Given this, and the number of lineages identified in other studies with other mitochondrial loci (e.g., Tonione et al., 2011), it seems likely that additional lineages remain unidentified from Southeast Asia. With highly divergent lineages present in India and Sri Lanka, supporting their Indian origin (Bansal and Karanth, 2010), at least three of the lineages spread eastward and westward, giving rise to their modern global distribution. Contrary to its colonization across continental South America, which seems to have been recent and involved mostly a single lineage (Case and Bolger, 1991; Case et al., 1994; Torres-Carvajal, 2015), Tonione et al. (2011) identified three lineages within Moorea, a small island in French Polynesia, and similarly, Western Indian Ocean islands overall currently harbor at least three different lineages of this species.

As with many other widespread species of Hemidactylus assessed with genetic methods, including H. mabouia (Agarwal et al., 2021) and H. fasciatus (Wagner et al., 2014), H. frenatus appears to be a species complex. Variation between the samples from mainland India and Sri Lanka and the island populations ranges up to 14 and 11%, respectively, which is as high as the divergence between pairs of reptile (Harris, 2002) and mammalian (Allen et al., 2020) species as measured using CYTB sequences. Variation between the main lineages within the Indian Ocean islands ranges between 7.0 and 8.5%, which in some small archipelagos such as Diego Garcia (and in the Seychelles) co-occur, highlighting the intricacy of the phylogeography of H. frenatus.

These current patterns highlight both the high levels of diversity, particularly within the native range, and suggest a high number of colonization events required to explain their current distribution. All five primary lineages are currently found in India, Sri Lanka and/or Myanmar, but without a comprehensive sampling across this area and an increased number of markers with various levels of resolution, the timings of the different Western Indian Ocean island colonizations remain unknown. While there is yet no clear evidence for natural and old colonization events (e.g., differentiated lineages endemic from certain islands or groups), it does appears is possible that this species has been transported in multiple waves, perhaps initially in pre-European times, but that these few early dispersals are now predominantly obscured by a much greater number of recent and ongoing translocations associated with modern transportation networks.

The patterns revealed here related to the three primary lineages identified allow for the development of provisional and testable hypotheses. The widespread ‘yellow’ lineage (Figure 3) occurs on scattered atolls of the coralline Seychelles. H. frenatus, however, was not present on the islands from which these atolls were colonized when they were first sampled in the 19th century. In addition, the yellow lineage still does not occur in Mauritius, and H. frenatus was only confirmed in the granitic Seychelles (Mahé) in the 1990s. Since this lineage occurs widely, both on the mainland and in the islands of Wallacea and the Pacific, it is likely to have been present in Sumatra and Borneo when the Austronesian voyagers colonized Madagascar and the Comoros. No archeological remains have been found to confirm that they crossed mid-ocean (Anderson et al., 2018a). Geckos could still have been dispersed either via shipwrecks that did not result in an archeological signature, or via lost mariners who strayed from the Maldives (Agarwal et al., 2019).

The ‘red’ lineage’s presence in the Indian Ocean only on Madagascar and the Comoros (Rocha et al., 2010), could match the profile of Indian trade with East Africa and Madagascar/Comoros in the medieval and early modern period (Boivin et al., 2013). Specifically, the geographic distribution could be correlated with the distribution of the introduced Indian civet Viverricula indica found in Madagascar, Comoros, the East African islands of the Zanzibar group, and Socotra (Gaubert et al., 2017), but nowhere else in the western Indian Ocean. The third (green) lineage could correspond to more modern trade and travel (notably to formerly isolated Diego Garcia), which may also be responsible for the yellow lineage’s dispersal to Australia and the Pacific.

Higher resolution datasets, through either greater number of targeted markers such as SNPs or the retrieval of complete mitochondrial and nuclear genomes, are required to further investigate the lineage splits and chronology of these different dispersals, but the patterns presented are suggestive, and the use of these commensal species as tracers for early human movements merits further consideration. This approach would be significantly improved though denser sampling of mainland southeast Asia and the other suggested sources of all the lineages.

Furthermore, determination of dispersal patterns based on any single molecular marker, such as mtDNA, needs to be treated with caution. In Southern Europe, patterns of minimal variation in mtDNA in two unrelated gecko species conflicted with nuclear markers, suggestive of a “selective sweep” that could obscure phylogeographic history based on mtDNA (Rato et al., 2010, 2011). A higher than usual nDNA/mtDNA diversity pattern was also identified in H. frenatus (Tonione et al., 2011), which certainly warrants further investigation and highlights the need to interpret cautiously the phylogeographic patterns based on mtDNA.

Finally, it appears from the current data that some gecko species and lineages are significantly more amenable to human translocation, or are more efficient at establishing colonies thereafter. From the greater than 165 currently recognized species of genus Hemidactylus, ten have intercontinental distributions, with two that are widely present in Western Indian Ocean islands, H. mabouia-mercatorius complex and H. frenatus, among the five most widespread and invasive Hemidactylus (Rocha et al., 2010; Weterings and Vetter, 2018).

As noted earlier, geckos have specific characteristics that make them efficient colonizers, but the characteristics that make some species especially effective remain unknown. Several studies have suggested that thermal physiology may play a role. Thus, an investigation of thermal tolerances in widespread species, relative to those with more restricted ranges may be revealing. Clearly H. frenatus is able to colonize what is thought to be less optimal habitat (McKay and Milenkaya, 2020), and there is some evidence for increased cold tolerance in invasive populations (Lapwong et al., 2021). On the other hand, H. frenatus showed no evidence of increased boldness which is often (although not necessarily) associated with invasive species (Nordberg et al., 2021).

The identification of likely “cryptic species” within both H. mabouia species-complex and H. frenatus, however, complicates the situation, since it becomes unclear if the same lineages have been compared, or if some lineages have unique characteristics. Modeling approaches, to determine potential ranges, and the impact of climate change (e.g., Rödder et al., 2008), as well as analyses of diet, behavior and other ecological variables, may benefit from determining which lineages within these diverse species are being compared. The sequence dataset presented in this study may form a useful comparative framework for such future assessments.