Habitat segregation and ecological character displacement in cryptic African malaria mosquitoes

Abstract

Understanding how divergent selection generates adaptive phenotypic and population diversification provides a mechanistic explanation of speciation in recently separated species pairs. Towards this goal, we sought ecological gradients of divergence between the cryptic malaria vectors Anopheles coluzzii and An. gambiae and then looked for a physiological trait that may underlie such divergence. Using a large set of occurrence records and eco-geographic information, we built a distribution model to predict the predominance of the two species across their range of sympatry. Our model predicts two novel gradients along which the species segregate: distance from the coastline and altitude. Anopheles coluzzii showed a 'bimodal' distribution, predominating in xeric West African savannas and along the western coastal fringe of Africa. To test whether differences in salinity tolerance underlie this habitat segregation, we assessed the acute dose-mortality response to salinity of thirty-two larval populations from Central Africa. In agreement with its coastal predominance, Anopheles coluzzii was overall more tolerant than An. gambiae. Salinity tolerance of both species, however, converged in urban localities, presumably reflecting an adaptive response to osmotic stress from anthropogenic pollutants. When comparing degree of tolerance in conjunction with levels of syntopy, we found evidence of character displacement in this trait.

Keywords: Anopheles coluzzii; Anopheles gambiae; cryptic species; ecological character displacement; ecological speciation; habitat segregation; malaria vector; molecular forms; niche partitioning; saltwater tolerance; spatial ecology; species distribution modelling; urban pollution.

Figure 1

(A) Map of the study area showing major eco-regions (Olson 2001) and the sampled localities included in the data set (dark grey dots). Light grey area falls outside the limits of distribution of An. gambiae (Sinka et al. 2010) and was not included in species distribution modelling. The outer dashed line delineates the approximate limits of the area of sympatry of An. gambiae and An. coluzzii. The inner dashed line delimits the study area in Central Africa where the two species were tested for salinity tolerance. (B) The study area in Central Africa showing localities from which larvae were collected for salinity tolerance testing. The pies show the relative frequency of the two species in larval samples from each locality (blue: An. coluzzii; red: An. gambiae). The size of the pie is proportional to the size of the sample. Geographic coordinates of localities are given in Table S1. Toponyms referring to the ID number of each locality shown inside the pies are presented in the legend in Fig.2.

Figure 2

Relationship between the size of each locality shown in Fig.1B, expressed as number of inhabitants and the relative abundance of An. coluzzii and An. gambiae. Ellipses identify urban centres with respect to their position relative to the Cameroon Volcanic line (CVL): westwards (red), eastwards (blue) or on the CVL (green). On the opposite sides of the CVL, urban centres are dominated by one or the other of the two species, while localities on the CVL identify a ‘contact zone’. Exceptions to this pattern are Bata (N°1) and Malabo (N°15) in Equatorial Guinea, Libreville (N°12) in Gabon and Tibati (N°25) in Cameroon. The latter is situated on/east of the CVL, but it lies at >800 m altitude in the forest/savanna mosaic, outside of the rainforest ecozone. In Bata and Libreville, the two species are in apparent parapatry (data not shown). Malabo is on the island of Bioko, on the CVL, under the influence of insular biogeography (e.g. Deitz et al. 2012).

Figure 3

(A) Relative probability of occurrence predicted for An. gambiae and An. coluzzii (dark blue: 1 An. coluzzii, and 0 An. gambiae; dark red: 1 An. gambiae, and 0 An. coluzzii) across their sympatric range. The inset shows an enlarged view of the area delimited by a black line in the map where high altitudes are encountered close to the coastline due to the presence of Mount Cameroon and the Cameroon Volcanic line; (B) Frequency of An. coluzzii relative to An. gambiae plotted as a function of distance from the coast. The red line depicts the second-order polynomial regression curve fitted to the observation records, represented by grey dots. Black circles and associated standard error bars illustrate the mean frequency of An. coluzzii for arbitrary classes of distance plotted for a visual assessment of goodness-of-fit.

Figure 4

Functional relationship between electrical water conductivity and salinity tolerance in individual populations of An. coluzzii and An. gambiae from Central Africa. Trend lines show competitive regression models fitted to the data, whose parameters are presented in Table7. The solid continuous black curve represents a loess smoother fitted to the data in order to extract the general shape of the functional relationship between the two variates. The blue asymptotic curve is the most parsimonious model based on the AIC. The other competitive models are shown as a dashed black (linear model) and horizontal grey (null model) lines.

Figure 5

Evidence for ecological character displacement in populations of An. gambiae (red) and An. coluzzii (blue) from Central Africa. The Gaussian curves show the probability density functions of tolerance thresholds, according to habitat, location and degree of syntopy. The dotted red curve in all panels depicts the tolerance of inland rural forest populations of An. gambiae, which are postulated to represent the ancestral state of the trait. Panel (A) includes localities where both species extensively share the same breeding sites, whereas in (B), both species share the same locality but not the same breeding sites due to spatial segregation (Libreville, Gabon). The dotted blue line refers to coastal urban populations of An. coluzzii in allopatry, where An. gambiae does not occur (Bonaberi/Douala, Malabo, Tiko). In (C), both species share the same habitat but not the same localities, each locality being exclusive of one of the two taxa; An. gambiae urban populations are represented by towns located in the savanna (Tibati, Cameroon) or in the forest westwards of the Cameroon Volcanic line (cf. Figs1–1 and Table S1).

Figure 6

A theoretical framework of functional ecological interactions between An. gambiae and An. coluzzii in the forest ecozone of Central Africa. Panel (A) represents the zero isoclines of population growth in relation to different environmental stressors: blue (An. coluzzii) and red (An. gambiae) lines set the limits for combinations of osmotic and resource stress beyond which population growth becomes negative, leading to no stable persistence of either species. On the abscissa, osmotic stress is assumed to increase along the inland versus coastal, as well as the rural versus urban environmental gradients. On the ordinate, stress from lack of resources – particularly nutrients and suitable larval habitats – is supposed to increase in the reverse order along the same environmental gradients. The ‘N.A.’ region delimited in state space by a dotted curve is presumably not available in the study area (i.e. regions with abundant resources in presence of very low levels of osmotic stress). The ‘C’ and ‘G’ letters in state space define regions where An. coluzzii and An. gambiae, respectively, occur alone, whereas ‘C+G’ represents the region where both species can coexist. However, rural coastal areas have less resources and more variable osmotic stress than urban inland areas, accounting for their different location in state space and differences in carrying capacity K_i shown in panel (B). Panel (B) depicts graphically the outcome of competition generated by the classic Lotka–Volterra equations applied to the An. coluzzii versus An. gambiae species pair in the ‘C+G’ region of state space in panel (A). The axes refer to the population size of either An. coluzzii (blue) or An. gambiae (red). As carrying capacities increase moving from small rural to larger and larger urban habitats (K_a < K_b < K_c, direction of change indicated by arrows), the equilibrium point of joint population size changes from stable coexistence (cases a, b, continuous lines) to competitive exclusion (case c, dotted lines). For further discussion, see text.