Mosquitoes of the complex of Anopheles culicifacies Giles are widespread in tropical ... (GLC), with the advantage that material can be characterized long after.
Bull. ent. Res. 76, 529-537 Published 1986
529
Differentiation of Anopheles culicifacies Giles (Diptera: Culicidae) sibling species by analysis of cuticular components P. J. M. MILLIGAN, A. PHILLIPS and D. H. MOLYNEUX
Department of Biological Sciences, University of Salford, Salford, M5 4WT, UK S. K. SUBBARAO
Malaria Research Centre, Indian Council for Medical Research, 22 Sham Nath Marg, Delhi-1100054, India G. B. WHITE*
Department of Entomology, London School of Hygiene & Tropical Medicine, London, WC1E 7HT, UK
Abstract
Cuticular wax was extracted from adult females of Anopheles culicifacies Giles and analysed by gas liquid chromatography. The flies were taken from pure cage stocks of three sibling species: 56 individuals of species A, 43 of species B and 51 of species C. The three cytospecies were found to be significantly different in their cuticular hydrocarbon composition by a multivariate analysis of variance. Using discriminant analysis, each cytospecies was characterized by the amounts of C25-C33 hydrocarbons extracted from the cuticle, allowing criteria to be established for the identification of flies. By these criteria, specimens from stocks of known identity were allocated to the correct group with a high success rate. These results may reflect ecological differences among the sibling species.
Introduction
Mosquitoes of the complex of Anopheles culicifacies Giles are widespread in tropical Asia and are the principal vectors of malaria in many parts of the Indian and Arabian sub-continents. Cytotaxonomic studies on the nominate taxon A. culicifacies from various localities have revealed the existence of at least three sibling species, known provisionally as species A, B and C (Green & Miles, 1980; Subbarao et al., 1983; Vasantha et al., 1982, 1983; Akoh et al., 1984). These biologically distinct species often occur sympatrically, and all three are endemic at the type-locality of A. culicifacies in Hoshangabad, Madhya Pradesh, India (Subbarao et al., 1983). Malaria sporozoites have been found repeatedly in wild-caught females of species A in northern India (Subbarao et al., 1980; Subbarao, 1984), in southern India (Suguna et al., 1983; Reuben et al., 1984) and in Pakistan (Mahmood et al., 1984), indicating that this member of the A. culicifacies complex is a widespread vector of malaria. Species B has also been implicated as a malaria vector in southern India (Suguna et al., 1983), and laboratory* Present address: Public Health Section, Insecticides and Fungicides Department, I d , Fernhurst, Surrey, GU27 3JE.
530
P. J. M. MILLIGAN, A. PHILLIPS, D. H. MOLYNEUX, S. K. SUBBARAO and G. B. WHITE
reared species B females from villages around Delhi developed sporozoites when experimentally fed on blood containing Plasmodlum vivax and P. falciparum (Mahmood et al., 1984). Species B is presumably the vector in Rameswaram Island, Madras, India (Sabesan et al., 1984) and Sri Lanka where A. culicifacies s.l. has been incriminated, but neither species A nor species C has been found. Under some circumstances, however, species B is less anthropophilic than species A (Tewari et al., 1984). Species C has not yet been implicated as a malaria vector. For distinguishing between specimens of species A, B and C in the course of field studies, it may be necessary to develop the use of various taxonomic techniques, as reviewed in Newton & Michal (1984). In this paper, we report the results of preliminary investigations of the cuticular hydrocarbon profiles of A. culicifacies adults from inbred laboratory strains of species A, B and C. The method has the sensitivity needed for reliable determination of individual insect specimens, or fragments of them, analysed by gas liquid chromatography (GLC), with the advantage that material can be characterized long after the specimens have died and become desiccated. GLC analysis may therefore be a practical method for distinguishing between sibling species such as members of the A. culicifacies complex in cases where the condition of specimens makes them unsuitable for the application of conventional morphological, genetic or other biochemical identification techniques.
Materials and methods
Mosquitoes Specimens of A. culicifacies sibling species for chromatographic analysis were desiccated adult females from colonies maintained at the Malaria Research Centre, Delhi. These reference strains were cytotaxonomically identified by means of the specific banding sequence of polytene chromosomes prepared from ovarian nurse cells (Green & Miles, 1980; Subbarao et al., 1983). The strains originated from the following localities in India: Species A. A strain homozygous for vermilion-eye (Adak et al., 1983) and isolated from a colony established in 1980 with species A parent females collected from Basantpur, Haryana (Subbarao et al., 1980). Species B. Established in 1977 with mosquitoes collected from Okhla, Delhi (Subbarao et al., 1980). Species C. Established in 1982 with species C parent females collected from Serula, Surat District, Gujarat. Chromatography The total hexane-soluble cuticular component of each fly was collected by covering the individual with hexane (20^1) in the bottom of a Reacti-vial (Pierce Chemical Co. Ltd), and leaving for 10 min. This allowed the total cuticular wax to be extracted and standardized the amount of internal lipid also included in the extract. The sample was then evaporated to dryness, and resuspended with 1 [A of hexane containing 10 p.p.m. of an internal standard (pentadecane); I fi\ was injected directly onto the column, which was a 12M CPsil 5 CB capillary column with an internal diameter of 0-32 mm and a phase thickness of 15 jum. The gas chromatograph (GC) was a Hewlett Packard 5790A equipped with a flame ionization detector and connected to a 3390A recording integrator. The GC run began at 120°C for 2 min and rose at 7°C/min to 310°C. The integrator recorded the chromatogram and gave the retention time and integrated area of all peaks. The area under each peak was divided by the area for the internal standard to compensate for differences in injection volume and machine response; these area ratios formed the raw data for statistical analysis. Tentative peak identification could be made by comparison with known standards.
DIFFERENTIATION OF ANOPHELES CULIC1FACIES SIBLINGS
531
Statistical approach
Statistical analysis of the data presents three problems: (a) how to determine whether there are significant differences between species A, B and C, taking all cuticular :omponents into account; (b), if the species are different, how to characterize the differences and draw up criteria for classifying individuals on the basis of their cuticular :omponents; and (c), how to test the reliability of the classification criteria. Differences in cuticular components distinguish populations of sibling species if the variability between species is significantly greater than that within species. A multivariate analysis of variance tests for whether or not this is the case, by determining whether there is a significant difference between the mean vectors for the species groups. Classification criteria need to be as powerful yet as simple as possible. Establishing such criteria is complicated by the fact that some cuticular components, though not significantly different when considered singly, are useful discriminators when combined with others. At the same time, many single components are found in significantly different amounts in one species compared to another, but often there is redundancy because they are intercorrelated. We therefore used discriminant analysis, which determines which components are most useful in distinguishing groups by finding linear combinations of variables (discriminant functions) which maximize the ratio of between- to within-groups variance (see Morrison, 1976). The absolute values of the discriminant function coefficients (when they have been standardized to compensate for differences in scale and variability among the original variables) indicate the relative importance of each variable in discrimination. The discriminant functions can be used to classify new individuals of unknown identity. Using the functions to classify flies whose identity is known provides a test of the reliability of the technique. However, when the number of components is large compared to the number of flies, discriminant analysis, and particularly discriminant analysis with stepwise inclusion of variables, can give misleadingly good results by capitalizing on chance differences between groups; given enough variables, it will always be possible to find a discriminant function which separates the groups well. Such bias is best eliminated by using jack-knifed estimators of the proportion of correct classifications; each individual is classified using discriminant functions which are formed using all the data except the data for that individual. The proportion of correct classifications is then an unbiased and relatively efficient measure of the reliability of the technique (Kendall et ai, 1983, chapter 44). Rare peaks which occur in only a few individuals were removed from the analysis, and transformations were sought to stabilize variance and promote marginal normality in the remainder, with a view to improving multivariate normality within groups. Power transformations of the form: X' = (Xk - l)IX
for X 4= 0
X' = In X for X = 0 were used, the value of X found by the maximum likelihood method of Box & Cox (1964), giving values for X of around 0-5. Results
One hundred and fifty individuals were analysed, 56 of species A, 43 of species B and 51 of species C. Some differences between species can be seen on inspection of the chromatograms, typical examples of which are shown in Fig. 1. The components numbered 3 and 4 in the diagram are fatty acids, 6-35 are hydrocarbons. The hydrocarbons ranged from C15 compounds (15 carbons) to C44. A multivariate analysis of variance showed there to be a highly significant difference between the species as measured by an approximate F ratio derived from Wilks' lambda (A = 0-094, F = 5-19 with 90 and 206 d.f.,P
