Considerable advances have been made in recent years with regards to the development of transgenic Anopheline mosquitoes refractory to the malaria parasite (Alphey 2009). However, before such an approach can be widely implemented, it is critical to evaluate risks associated with GMM through rigorous scientific investigation and to demonstrate proof of principle of such technologies in this system. Oceanic islands, in particular, have been highlighted as potentially suitable sites for such trials, as they should be isolated from mainland sites due to the natural barrier created by water. Here, we genetically evaluated the suitability of the Bijagós archipelago and the Comoros Islands, for GMM releases.
We assessed three islands of the Bijagós archipelago located off the coast of Guinea Bissau in West Africa, where there is active and ongoing malaria transmission (WHO 2011). Anopheles gambiae s.s. was found to be present alongside An. melas, at all sites in Guinea-Bissau (Petrarca et al. 1983; Jaenson et al. 1994). On the islands of Bubaque and Orango, the relative abundance of these two vectors is difficult to assess, as collections were largely comprised of larvae which were specifically collected from typical An. gambiae s.s. larval habitat sites. However, An. gambiae was the dominant species on the mainland and the island of Formosa where only indoor resting adult mosquitoes were collected (Table 1). For a GMM release site, this is advantageous, as it is preferable for the target species to be the dominant malaria vector (Malcolm et al. 2009).
The Bijagós islands we assessed are separated from mainland Guinea-Bissau by 52–93 km. However, despite being separated by distances exceeding the natural dispersal capabilities of An. gambiae s.s. (7 km), we found no evidence of genetic sub-division between the island and mainland sites in either the M or S forms, suggesting there is considerable gene flow (Fig. 1). Consistent with this, in the S form we found no evidence of a reduction in genetic diversity which might be expected for a genetically isolated island population. In the M form, diversity was lower on Formosa; however, this site was represented by <10 samples, and thus is likely a sampling effect. We suggest that the genetic connectivity we observed in the Bijagós may be explained by ‘island hopping’ between the many islands of the archipelago, most of which are located <10 km apart, and some of which are located <10 km from the mainland (Fig. 1). Specifically, we hypothesize that rather than direct dispersal between the mainland and far islands (~50–90 km), there is dispersal between the mainland and closest island, and then the closest island and nearby islands, which are separated by much smaller distances (<10 km). Alternatively, or additionally, there may be inadvertent movement of mosquitoes via the high volumes of human local boat traffic between the mainland and island sites as well as amongst the islands themselves, which is important for trade, transportation and fishing. These findings are consistent with the low UNEP isolation index for the Bijagós islands we assessed (10 – Formosa, Bubaque; 17 – Orango), which reflects not only the distance of an island to the continent but also to the other islands. Overall, we suggest the lack of isolation between the Bijagós archipelago and mainland Guinea-Bissau makes them an unsuitable site for a GMM release. In addition, the presence of both the M and S forms as well as the atypical high rates of hybridization between them, present unique challenges for implementation of an isolated GMM trial at this site. Specifically, previous genetic studies have shown that hybridization results in highly asymmetric introgression from the M form into the S form (Marsden et al. 2011). It is uncertain how modified genes would move between the M and S forms under these patterns of gene flow. However, it is possible that if only the S form was modified, genetically modified genes would be unlikely to spread into the M form, whereas if the M form was modified, genes would spread into the S form (Marsden et al. 2011).
A number of vector borne diseases are transmitted on the Comoros Islands including dengue fever, (Gautret et al. 2010), chikungunya virus (Sergon et al. 2007), rift valley fever (Sissoko et al. 2009) and malaria (Rebaudet et al. 2010). We found vectors for these diseases on each of the islands (Table S1–S4), including the first record of Ae. albopictus on Anjouan (vector of Chikungunya and Dengue fever virus). However, ongoing malaria transmission presents the highest health burden on the islands (WHO 2011). We detected two primary malaria vectors, An. gambiae s.s. and An. merus. Unfortunately, the relative abundance of these vectors is difficult to assess due to the bias in sampling efforts towards atypical peridomestic, but known An. gambiae s.s., larval habitat sites (Julvez and Mouchet 1996).
In comparison to the Bijagós archipelago, the Comoros Islands are considerably more physically isolated (~700–800 km vs <100 km) resulting in a higher isolation index (49, UNEP 2010). We found An. gambiae s.s. from the three islands (Grande Comore, Moheli, Anjouan) to be derived from different genetic populations of mosquitoes from mainland Tanzania, and found no evidence of recent migrants between the mainland and islands amongst our samples (Fig. 2), which is consistent with isolation of the islands.
The presence of multiple private ND5 haplotypes on the Comoros islands (5/6 full-length sequences, 2/4 trimmed sequences) was also suggestive of isolation between the islands and mainland. Furthermore, as expected for isolated populations, genetic diversity estimates based on SNP data (Table 2) and ND5 sequences (Table 3) were lower on the islands than the mainland, with the exception of SNP based HE estimates which were similar on the mainland and islands despite lower allelic richness on the islands, which in itself is suggestive of a recent bottleneck (Table 2). Lastly, the unique morphological features of some taxa recorded during the mosquito community sampling on the Comoros Islands is also consistent with isolation of these islands (e.g. unknown Anopheline species, wing pattern divergence of An. pretoriensis, Tables S4–S7). Together, these data suggest that the Indian Ocean presents a significant barrier to gene flow between the mainland and Comoros islands and that human-assisted movement of mosquitoes is likely to be minimal. As such, in terms of isolation, the Comoros Islands appear to be good potential candidate sites for GMM release trials as our data suggests there is a low risk of GMM escapees reaching mainland Africa.
Table 3. Tanzania and Comoros ND5 diversity based on full-length sequences (811 bp)
|Population||Number of haplotypes||Haplotype diversity (h)||Sequence diversity (π)|
| Mainland-Tanzania |
|Dar-es-Salaam (n = 32)||8||0.843||0.0040|
| Comoros Islands |
|Grand Comore (n = 62)||2||0.500||0.0006|
|Anjouan (n = 30)||3||0.384||0.0017|
|Moheli (n = 59)||3||0.161||0.0002|
Despite this evidence of restricted movement, FST values were not indicative of complete genetic isolation between the Comoros islands and mainland (FST = 1). This could be the result of FST not reflecting the true level of isolation. For example, if the Comoros islands were colonized relatively recently, there may have been insufficient time for differentiation to occur despite the absence of any ongoing gene flow (non-equilibrium population). This is consistent with ND5 haplotype data which shows the presence of both unique and shared haplotypes. Or, levels of divergence may have been underestimated due to the absence of Comoros samples in the SNP ascertainment panel, which would have resulted in under-representation of polymorphisms unique to the Comoros. Another explanation is that an FST value of <1 may reflect that there is ongoing gene flow between the mainland and islands sites. Indeed, population bottlenecks can cause rapid divergence in allele frequencies resulting in elevated FST values despite high gene flow as the population is not at equilibrium. In these situations, methods that do not assume mutation-equilibrium, such as STRUCTURE, are more appropriate. As stated above, STRUCTURE analyses and the distribution of ND5 haplotypes showed strong genetic structuring between mainland and island sites, indicating gene flow is restricted. Moreover, amongst our samples we found no evidence of recent migrants. As such, our data do not appear to be consistent with a hypothesis of elevated FST despite high gene flow due to a bottleneck. However, there may well be a low level ongoing migration which our sampling regime failed to detect. Additional sampling from the Comoros, sites along coastal East Africa (e.g. Northern Mozambique, ~300–450 km from the islands) and Madagascar (400–450 km), as well as sampling from transport locations (airports and ports) would be useful to quantify any migration, particularly from the Comoros islands.
Given that An. gambiae is always found alongside human populations which may inadvertently transport individuals, there will always be some risk of migration and thus complete isolation of field site is unrealistic. General guidelines concerning acceptable levels of risk have not been developed as these will be specific to the species, trial conditions (e.g. fitness of the GMM) and technology being applied: A low level of migration would pose a smaller risk for self-limiting approaches (versus self-propagating see Introduction), such as are available for An. gambiae. For this reason, mathematical models would be useful in assessing the potential risk posed by migration on a case by case basis (e.g. Marshall and Hay 2012). Moreover, risk could be reduced by implementing measures to prevent human-assisted movement of vectors (insecticide spraying of aeroplanes) and a monitoring programme to detect escapees (e.g. fluorescent markers, James 2005).
If a GMM trial were to be conducted in the Comoros, a single island would need to be selected. The question of which of the islands is the most suitable for a GMM release requires consideration of genetic isolation amongst and within the islands. Flights, ferries and local boats result in the movement of people, and thus potentially mosquitoes, between the three islands, which are separated from each other by between 80 and 145 km. We found the greatest genetic differentiation between Grande Comore and Anjouan (FST = 0.126), indicating that human-assisted movement is lowest between the islands which are located the furthest apart (145 km; Fig. 2). Interestingly, this is consistent with population structure analyses of Plasmodium falciparum in the Comoros, which found differentiation of parasite populations to be highest between this pair of islands (Rebaudet et al. 2010). We found differentiation to be lowest between An. gambiae s.s. from Moheli and Anjouan (FST = 0.093), which were shown to constitute a single genetic population (Fig. 2). Overall, Grande Comore was the only island found to be genetically distinct from the other islands, although it should be noted that there was evidence of some ongoing gene flow between Grand Comore and Moheli and Anjouan (Fig. 2). In terms of a GMM release, evidence that gene flow between Grande Comore and the other islands is restricted is advantageous, as it reduces the risk of spread of escapees to other islands, as well as the immigration of mosquitoes onto the island.
Population structure within a release site, that is, an island, has the potential to inhibit GMM programmes as genetic sub-division may impede the spread of genes across the target population if a genetic drive system is linked to an effector gene so to facilitate the rapid spread of modified genes through the population. However, in taxa such as An. gambiae, where genetic drive technology is not yet available, or in instances where sterile insect technologies are used, inundative release (release of enormous numbers of modified mosquitoes) will be necessary (Benedict 2011). In these cases, population structure is advantageous as sub-populations can be used to enable a release programme to be conducted in sections (e.g. as recommended for An. arabiensis on Reunion island, Malcolm et al. 2009). We found no evidence of sub-structuring within the island of Moheli, and on Anjouan only a single site was assessed. However, on Grande Comore, the most north-eastern site (Bouni), showed evidence of elevated differentiation (Fig. S3). This may be related to topography (Fig. S4), and it is possible these naturally isolating barriers could be enhanced to create an isolated site within Grande Comore. Alternatively, the topography of the most southern section of the island, indicate that this region may be isolated (Fig. S4). Indeed, our surveys detected An. gambiae in this region (Malé), but samples were degraded and unsuitable for SNP analysis, and therefore genetic evaluations of mosquitoes from this area would be useful to assess isolation.
In addition to genetic considerations, other island characteristics, such as island size, political stability and logistical considerations, are important. Whilst a full evaluation of these factors is outside of the scope of this study and thus not specifically investigated here, it is noteworthy that the highly mountainous terrain and political instability of Anjouan, make this island less appealing as a release site. In contrast, the smaller size (290 km2) and relatively flat terrain of Moheli, make this island an attractive prospect if trying to conduct an island wide implementation.
Within the range of An. gambiae, there are only a small number of island sites potentially suitable for field trials, and only a few of these have been assessed (e.g. Pinto et al. 2002; Moreno et al. 2007). Nonetheless, our findings are broadly consistent with studies of a volcanic chain of four widely separated islands (≥160 km apart) in the Gulf of Guinea (Bioko, São Tomé, Príncipe and Annobón). Specifically, Bioko Island which is isolated to a similar degree to the Bijagós ~75 km off the coast of Cameroon and with an UNEP isolation index of 17, was found not to be genetically isolated from the mainland (Moreno et al. 2007). In contrast, the islands of São Tomé, Príncipe and Annobón, which are located 240–350 km from mainland Gabon and with similar UNEP isolation indices (39–45) to the Comoros (49), were found to be genetically isolated from mainland populations (Pinto et al. 2002; Moreno et al. 2007). Together these data suggest that island sites with an isolation index of <20, are likely not worthy of investigation as potential GMM field trial sites for An. gambiae.