In the rainforest of southern Cameroon ecological divergence between two molecular forms of the malaria mosquito Anopheles gambiae along urbanization gradients  is postulated to be the outcome of niche expansion by the M form, a process driven by disruptive selection and local adaptation to marginal habitats . Historical records indicate that the core of highly urbanized settings constitute an unfavourable habitat for the persistence of An. gambiae populations due to the paucity of suitable unpolluted water bodies necessary for the development of its aquatic larvae [41, 42, 53, 54]. In the two major urban agglomerates of Cameroon, Yaounde and Douala , and in other large cities lying on the Gulf of Guinea [55–57], larvae of the M form appear to have evolved the ability to survive in polluted breeding sites. The physiological, behavioural, and other life-history traits that are involved in this process are, however, to date largely unknown. Starting from the observation that in the two major urban agglomerates of Cameroon the M form occurs at times in the same breeding sites of Culex quinquefasciatus, a mosquito pest which is notoriously abundant in waters enriched with nutrients, we postulated that the M form might have developed enhanced tolerance to environmental stressors ensuing from the decomposition of organic matter.
Ammonia is a toxicant resulting from the decomposition of decaying organic matter often contaminating mosquito larval habitats [41, 43, 46, 47, 58, 59]. Polluting human activities contribute to the generation and accumulation of this contaminant especially in unplanned urban environments in tropical Africa, where accessible water collections containing large amounts of decaying organic matter can be plentiful. Accordingly, by acute toxicity assays we have demonstrated quantitative differences in tolerance to ammonia between the urban population of the M form in Yaounde compared to the parapatric population of the S form occurring in neighbouring rural settings. Higher ammonia tolerance in the M form presumably constitutes an adaptive trait given that this toxicant occurred on average at higher concentration in An. gambiae urban larval habitats compared to rural ones. Moreover, urban breeding sites were on average more alkaline than rural ones, a physico-chemical property that enhances ammonia toxicity [45, 52]. To confirm the adaptive value of tolerance, however, it will be necessary to demonstrate that individuals with greater ammonia tolerance have a higher relative fitness in the polluted larval habitats of the urban environment compared to individuals with lower tolerance. Additionally, it will be necessary to verify the heritable basis of ammonia tolerance demonstrated in other dipterans  to support the notion that disruptive selection has produced the quantitative differences between the M and S forms in this trait. Adaptive phenotypic differentiation can in fact also result from adaptive phenotypic plasticity without underlying genetic differentiation, which is a pre-requisite for local adaptation.
The quantitative differences that we have measured between the M and S forms concern mainly the shape and the variance, more than the mean, of the tolerance response. The M form exhibited a larger variance in the distribution of individual tolerances, whose consequence is that the M population comprised a greater proportion of individuals with a higher tolerance threshold compared to the S population. Under a mode of polygenic inheritance of ammonia tolerance, this pattern can be interpreted to be the outcome of several, not mutually exclusive, processes. First, greater variability in the distribution of thresholds could indicate that the M form may have been subjected to relatively recent disruptive or directional selection for ammonia tolerance, in agreement with results from artificial selection experiments whereby the phenotypic variance of selected lines is generally increased . Second, greater phenotypic variance is expected under selection with assortative mating . If ammonia tolerance is selected along the urbanization gradient according to increasing selection intensities from the rural towards urban localities, on average individuals with greater tolerance will be selected for in the urban environment, and individuals with lower tolerance will be selected for in the rural environment. If mating occurs in the original environment before dispersal, this process will produce some assortative mating correlated to ammonia tolerance. The ensuing gametic-phase disequilibrium in loci influencing ammonia tolerance will result in increased additive genetic variance that will be reflected, if environmental and other genetic components of variance do not decrease, in greater phenotypic variance.
A concurrent interpretation hinges on the homogenising effect of gene flow, which counteracts diversification by selection during local adaptation . Gene flow between the M and S molecular forms is strongly reduced but—apparently—not nil [64–66]. Asymmetric introgression of maladaptive alleles present in the rural S population into the urban M population may hinder greater phenotypic divergence between the two forms, unless the genes underlying this phenotype are linked to genes controlling reproductive isolation between the two forms—a condition that would favour ecological speciation if ammonia tolerance is adaptive. Asymmetric introgression of M-form nuclear genes into the S-form genome (a pattern opposite to that predicted by our observations) has been demonstrated in Guinea Bissau, a relatively restricted geographical area of high hybridization between the two forms [64, 65]. In a similar way, neighbouring demes of the larger M-form metapopulation occurring in forested southern Cameroon , connected by dispersal to the Yaounde deme, may introduce maladaptive tolerance alleles if these allochthonous demes do not share the same local adaptation mechanisms of the Yaounde deme. Maladaptive alleles would contribute towards greater phenotypic variance in tolerance thresholds observed in the Yaounde M population and could persist in the urban environment because of the presence of breeding sites with lower ammonia concentration, as indicated by the greater variance in ammonia concentration observed in urban compared to rural natural larval habitats.
In the context of recent niche expansion, it is relevant to ask whether the differences in ammonia tolerance observed between the M and S forms from Yaounde predate or are subsequent to the process of ecological divergence and local adaptation to the novel urban habitat. In the absence of historical data it is quite difficult to answer this question. The experimental designs typical of evolutionary toxicology, a discipline that investigates the impacts of environmental pollutants on the genetics of natural populations, include comparison of matched reference and polluted sites , which is the approach we have adopted in this study. However, any conclusion derived from this approach is confounded by the ancillary question of whether local adaptation to the urban environment predates or arose subsequent to initial lineage splitting between the molecular forms. Hints to answer these questions are likely to come from studies investigating the nature and degree of tolerance in other M and S geographical populations with varying exposures to naturally occurring pollutants. Despite the limitations of this correlational approach, a match between exposure and degree of tolerance in replicated demes will add credit to the hypothesis that increased tolerance has evolved in response to higher levels of contamination. Further hints will come from a more detailed understanding of the phylogeography of the M and S forms across their distribution range; ultimately, both approaches will shed light on the population structure and history of adaptations in An. gambiae.
The difference in ammonia tolerance observed between the M and S forms in Yaounde reflects a process of local adaptation whose generality and limits should be verified in other cities of their distribution range, particularly in regions where populations are akin genetically to those occurring in the rainforest eco-climatic zone of southern Cameroon (e.g. carrying homosequential standard karyotypes like the M and S populations in Yaounde ). It is already reported that in some West African cities in the savanna, An. arabiensis— another member of the An. gambiae sensu lato (s.l.) species complex—is extending its distribution southwards into urban environments [68, 69], and has developed the ability to breed in larval habitats contaminated by waste waters [70, 71]. In East Africa, members of the An. gambiae s.l. complex occur in polluted urban larval habitats in Dar-es-Salaam , Kisumu, and Malindi . These geographical differences outline the fact that adaptation to urban pollution is happening repeatedly and independently within the species complex, and it is presumably very recent given that urbanization is a relatively novel phenomenon; furthermore, they suggest that lineage splitting between the M and S forms predates adaptation to uban habitats, in agreement with the estimated age of their divergence that is situated well before the Neolithic revolution .
In our acute toxicity bioassays, the difference in tolerance between the two molecular forms appeared only at ammonia concentrations well above those measured in natural larval habitats in and around Yaounde. It must be noted, however, that the potency of a toxicant depends also on the duration of exposure, which in our case was the shortest usually employed by ecotoxicologists when testing ammonia tolerance of macro-invertebrates. Longer exposures usually engender lower lethal concentrations. In the case of the shrimp Litopenaeus vannamei, for example, the median lethal concentration of ammonia at 96 hours was approx. 33% of that after 12 hours exposure . From our preliminary analysis, the median lethal concentration after 36 hours exposure in the S form was 65% of that scored at 12 hours, and showed the same kind of decay with time observed in other invertebrates . Nevertheless, we chose to score mortality after 12-hours exposure, and consequently shifted the range of test concentrations towards higher—ecologically less-relevant—values to avoid the influence of external factors like changes in pH or hunger, as evidenced by increased mortality in the control batches.
Ammonia tolerance, moreover, might reflect a general physiological response of detoxification and excretion following the same metabolic mechanisms involved in nitrogen homeostasis and osmoregulation . Ammonia is one of several nitrogenous compounds that can be found in nutrient-enriched waters (e.g. those resulting from nitrification of organic matter, such as nitrites and nitrates, or fertilisers, such as urea), whose accumulative toxicity may act additively or synergistically with that due to ammonia . Longer exposure to ammonia and other nitrogenous compounds in natural larval habitats might therefore select for a differential tolerance pattern between the M and S forms akin to that observed in our bioassays in response to short-term exposure to ammonia alone.
If differential tolerance to ammonia contributes to explain habitat segregation of the two molecular forms along the urbanization gradient in Yaounde, however, it is unlikely that it does so solely by selective effects engendered by acute toxicity. Sub-lethal or (sub-)chronic toxicity effects are also likely to operate in this context. It is not yet known, for example, how acute toxicity in An. gambiae relates to chronic effects on life history traits affecting fitness. In Drosophila melanogaster, fruit-flies artificially selected for ammonia tolerance developed faster and expressed higher viability on ammonia-supplemented media . The aim of this work was to highlight intrinsic physiological differences in tolerance between the M and S forms in their response to toxic levels of ammonia rather than to assess the biopotency of ammonia or look at its chronic sub-lethal effects under longer exposure. This is a matter for future associative and experimental approaches once the ecological processes underlying the cline in the M and S populations abundance along urbanization gradients will be further clarified. Further, it is possible that local adaptation may be associated to differential responses not only to nitrogenous compounds, but also to other environmental stressors characteristic of An. gambiae breeding sites [57, 72, 73, 78, 79], like temperature, organic and inorganic suspended solids, dissolved substances, emulsions and colloids (including e.g. heavy metals, detergents, oils, or other kinds of nutrients), or biological factors like micro-organisms , competitors [81–83], predators [35, 84], and their interactions (e.g. [85–87]); these effects need to be investigated at both the population and community level.
In this study we used late-instar larvae collected in the field to test for differential ammonia tolerance between the two molecular forms. Given the spatial segregation of the two forms along the urbanization gradient and the difference in average ammonia concentration between urban vs. rural breeding sites, the differences observed might have resulted from prior selection of more tolerant individuals in natural larval habitats before experimental exposure to ammonia in our bioassays. However, if the distribution of tolerance thresholds in the larval population before selection had been the same in both molecular forms, natural selection due to field exposure to greater concentrations of ammonia in the M-form is expected to reduce the variance in tolerance, because only the more tolerant individuals would survive to the third- and fourth-instar stage of development. The fact that we actually observed greater—not reduced—variance in tolerance by the M form is not consistent with this scenario, unless the distributions of ammonia concentrations in urban vs. rural larval habitats largely overlapped in the lower range of concentrations, an outcome not manifest in our test samples (cf. Figure 1). Future ammonia tolerance bioassays performed on F2 progeny of field-collected mosquitoes could help resolve this issue and avoid further confounding factors due to maternal effects.
Despite the limitations of this study, it is clear that ecological divergence between the two molecular forms of An. gambiae in Yaounde is associated to differences in a physiological response that potentially affect the survival of larvae in natural breeding sites. Development of tolerance may engender costs leading to trade-offs in other fitness-related life-history traits, as observed for East African populations of An. gambiae in response to increased tolerance to heavy metals . We speculate, therefore, that development of ammonia tolerance might have fostered niche expansion into the urban environment by forest populations of the M form in Yaounde, while also conferring a competitive disadvantage in the unpolluted larval habitats typical of rural settings in the rainforest block of Cameroon.