Morphological differentiation between seven Brazilian populations of Haemagogus capricornii and Hg. janthinomys (Diptera: Culicidae) using geometric morphometry of the wings

One of the most important genera of mosquitoes capable of infecting and transmitting the wild yellow fever virus (WFV) is Haemagogus Williston, 1896, which is considered a biological vector and responsible for maintaining the natural cycle of this zoonosis in forested areas of the Americas¹. Mosquitoes of this genus are restricted to the Americas and almost all species have a Neotropical distribution, except for Hg. equinus Theobald, 1903, which can even be found in some southern parts of the Nearctic region². These are mainly wild, diurnal, and acrodendrophic mosquitoes inhabiting primarily dense forest and gallery areas³.

Haemagogus is very diverse; it includes twenty-eight species of which nine are found in Brazil¹. Some of these are epidemiologically important in the transmission of the wild-type yellow fever virus¹. Among the nine known vector species, five stand out for the efficiency of their transmission in Brazil: Hg. albomaculatus Theobald, 1903, Hg. leucocelaenus Dyar and Shannon, 1924, Hg. spegazzini Bréthes, 1912, Hg. capricornii Lutz, 1904, and Hg. janthinomys Dyar, 1921. Larvae and females of Hg. capricornii and Hg. janthinomys species are currently morphologically indistinguishable, their differentiation being based primarily on characteristics of the male genitalia. Their identification is carried out based on the following: the presence (Hg. janthinomys) or absence (Hg. capricornii) of notable spiculosity on the ventral face of the aedeagus and the existence of a medial process, with a hooked shape, near the apex of the paraproct in Hg. janthinomys. These structures are small and only distinguishable by well-trained personnel and misidentifications can be frequent.

Although traditional morphometry contributed to the identification of these species, a more robust approach is necessary⁴. Geometric morphometry is a powerful, low-cost tool that addresses issues in taxonomy, ecology, and morphology, particularly in insects and especially in the family Culicidae, which possesses wings⁵. These bi-dimensional structures are eminently suitable for morphometrical description⁶. Geometric morphometry makes it possible to identify morphological variations and to explore their causes both within and between populations⁷. In Diptera, it has been widely used to answer questions mainly related to population studies⁶. A recent study of Culex mosquitoes from the state of Rio de Janeiro showed the effects of seasonal variations on phenotypic variations using this tool³.

Considering the difficulties in the identification of Hg. capricornii and Hg. janthinomys females, the poor knowledge about them, the partial overlap of their geographical distribution, and their eco-epidemiological importance, attention must be paid to the evaluation of old reports of infection of these and similar species, especially in Brazil¹. In this context, the aim of this study was to determine the phenotypic variability in Hg. janthinomys and Hg. capricornii females at species and population levels, using the geometric morphometry of the wings. For this purpose, we included populations of the two species that have a large proportion of their geographic distribution in Brazil.

A total of 108 right and left wings of females belonging to Hg. capricornii and Hg. janthinomys from Brazil were used in this study (Figure 1A and Figure 1B; Table 1). The Haemagogus populations came from ecological and epidemiological studies carried out by the Diptera Laboratory team and from the Entomological Collection at the Oswaldo Cruz Institute, Fiocruz, Brazil. Species were identified by direct observation of morphological characters using an optical microscope (Leica DMD108® – Morrisville, United States of America – USA) according to Arnell (1973)¹. Once identified, the wings were extirpated and later photographed according to Alencar et al. (2016)³.

Fourteen type-1 landmarks were selected and included in the analyses⁸. We used coordinate data and the isometric estimator centroid size (CS) to compare overall wing sizes between species and populations. The Mann-Whitney test was applied to comparisons of CS between species and populations. The shape variables (partial warps and uniform components) were obtained using the generalized Procrustes analysis superimposition algorithm. Mahalanobis distances derived from the shape variables were used to explore shape proximity between the species and populations. Statistical significance was determined by permutation tests (1,000 runs each) and corrected by the Bonferroni method.

We represented the Mahalanobis distances between species and populations in neighbor-joining (NJ) trees. The percentage of phenotypic similarity between species and populations was calculated using the cross-check test of discriminant analysis. Shape variables were regressed onto CS by multivariate regression analysis to detect allometry. The correlation between geographic and Mahalanobis distances was determined by a Mantel test (1,000 permutations) using straight-line geographic distances between collection sites as described by Rosenberg and Anderson (2011)⁹.

The geometric coordinates of each landmark were digitalized using the program tpsDig version 2.09 (available at http://life.bio.sunysb.edu/ee/rohlf/software.html). Centroid size generalized Procrustes analysis, Mahalanobis distances, permutation tests, and allometry were performed using the modules VAR, MOG, PAD, and COV respectively, included in the CLIC98 package, according to Dujardin 2008¹⁰. The correlation between geographic and Mahalanobis distances was determined by Mantel tests using the PASSaGE 2 software (available at http://www.passagesoftware.net/).

For interspecific comparison, the size variable revealed that the centroid size of Hg. capricornii was significantly smaller (Mann-Whitney test, P = 0.01) than Hg. janthinomys. The permutation test based on the Mahalanobis distances revealed significant differences for shape variables between the two species (P = 0.01). The “cross-checked classification” of Hg. capricornii and Hg. janthinomys individuals showed that 81% and 67%, respectively, of all specimens were correctly assigned.

For intraspecific comparison, the size variable, revealed that all populations of Hg. janthinomys were significantly different among themselves and bigger than Hg. capricornii (Mann-Whitney test, P = 0.01). However, the analysis of populations of Hg. capricornii showed no significant differences among them (P = 0.06). The permutation test based on the Mahalanobis distances revealed significant differences for shape variables among some populations (Table 2). The contribution of the canonical factors resulted from 38%, 26%, and 15% for the first, second, and third factors, respectively. A “cross-checked classification” of individuals of the seven populations of Hg. capricornii and Hg. janthinomys showed acceptable and heterogeneous reclassification scores. Hg. capricorniipopulations showed low reclassification scores (30-42%). Although very heterogeneous, populations of Hg. janthinomys had better reclassification scores, from low (22% Bahía), to medium (42% Rio de Janeiro), to high (77% Atalaia, Rio de Janeiro). The NJ tree based on the distances of Mahalanobis showed that the two populations of Hg. capricornii (Hgc_RJ, Hgc_Man) were the most similar, followed by the Hg. janthinomys(Hgj_RJ) population, and morphologically different from the population of Campina Verde (Hgj_CaV) (Figure 1C). In addition, the NJ tree showed that the Hg. janthinomys populations (Hgj_Ata, Hgj_Jac) were different from the Bahia population (Hgj_BA). The Mantel test revealed a positive and significant association between the geographic distances and distances of Mahalanobis (r = 0.467; P = 0.01). A multivariate regression analysis of shape variables on the size variable showed no significant effect (test after 1000 permutations, P = 0.11).

Our results based on wing geometric morphometry of Hg. capricornii and Hg. janthinomys indicate that both size and shape variables can differentiate at the interspecific level. However, at the intraspecific level, the results show a relative differentiation. The two populations of Hg. capricornii had a smaller centroid size with no significant difference between them, whereas all Hg. janthinomys populations showed significant differences. The shape variables were able to separate the two Hg. capricornii and Hg. janthinomys populations, except for the two originating in Bahia, which were not statistically different.

The importance of taxonomy in biological sciences is undeniable. Biodiversity mapping should focus on limited groups so that research that is more thorough can be carried out effectively. Our results are congruent with the hypothesis that suggests Hg. capricornii and Hg. janthinomys may constitute a complex of species whose morphological differentiation is complex. To help identify these cryptic species and to study the relationship between them, new tools, such as molecular biology and biochemistry, have been used in addition to morphological methods, such as classical morphology, scanning electron microscopy, and morphometry¹¹.

Modern molecular tools are available to discriminate between sister species living in sympatry¹². However, they are expensive to use and require specialized training. Geometric morphometrics have been shown to be highly informative, fast, and affordable. With minimal training, geometric morphometry can be used to answer ecological or taxonomic questions⁶. This study demonstrates that geometric morphometry can discriminate with considerable success Hg. capricornii and Hg. janthinomys females that cannot be identified by traditional morphological criteria.

Although centroid size is not a good measure to use in species identification because it is affected by environmental factors, our results show that this size variable was able to differentiate between the two species¹³. Thus, conformation is a reasonably good feature to solve identification problems and is merely affected by the environmental factors¹⁴. Our study was able to differentiate between the two species, as well as between some populations. The correlation analysis between centroid size and the shape variables for Hg. capricornii and Hg. janthinomys did not show a common allometric slope. The association between the shape of the kites and the geographic distance between the populations suggests that the morphological variation could fit a distance isolation model.

Our study had some limitations. The results were obtained from a limited number of individuals and samples were more abundant for populations of Hg. capricornii than Hg. janthinomys. This type of problem is frequent in works that present data that involve field collections.

Our results support the use of geometric morphometry in the morphological discrimination of Hg. capricornii and Hg. janthinomys females. Proper identification of species is the fundamental basis for building knowledge of biodiversity, ecology, and other areas of biology. Failures in species identification may lead to the diffusion and amplification of conceptual and methodological errors in other areas, with implications not only for our knowledge of nature, but also for ecosystem structure functioning, management decisions, and human health vector control programs¹⁵. Correct species identification using geometric morphometry could contribute to improving vector control strategies.