correspondence Wim Van Bortel, Department of Parasitology, Prince Leopold Institute of Tropical Medicine, Nationalestraat 155, 2000 Antwerpen, Belgium. E-mail: firstname.lastname@example.org
Summary Elucidating the complex taxonomic status of the major malaria vector taxa and characterising the individual species within each complex is important for understanding the complexity of the vector system in the south-east Asian region and will allow to estimate the impact of vector control measures. This applies to countries such as Laos, Cambodia and Vietnam that spend about 60% of their malaria control budget on implementing vector control activities. We used isozyme electrophoresis to clarify the Anopheles minimus s.l. species composition in northern Vietnam and identify behavioural divergences of individual species. Using different collection methods, adult mosquitoes were caught at monthly intervals from June to November 1995 in four villages. An. minimus s.l. could be distinguished from closely related species, An. aconitus and An. jeyporiensis, at the Octanol dehydrogenase (Odh) enzyme locus. Significant positive Fis values gave clear evidence of nonrandom mating within the An. minimus s.l. population. The highest heterozygote deficiency was observed at locus Odh, which was diagnostic for 2 sympatric An. minimus species in Vietnam similar to the An. minimus A and C species known from Thailand. We found no evidence for restricted gene flow between monthly samples, villages, or collection methods in either of the two An. minimus species. They occurred in sympatry, but in different proportions depending on the collection site, and had dissimilar resting and biting behaviours. Thus a vector control strategy will have a nonuniform effect on the various components of this diverse vector system.
The original geographical distribution of the malaria vector Anopheles minimus sensu lato (subgenus Cellia, Myzomya Series, Minimus Group) extended from Uttar Pradesh in India eastwards to South-east Asia, China, Taiwan and Japan ( Harrison 1980). The species seems to have disappeared from Nepal and parts of India ( Parajuli et al. 1981 ) and has become rare in Taiwan ( Lien 1991). Its disappearance from these regions is primarily attributed to the use of DDT residual spaying. In other countries of Asia, the response of An. minimus to insecticide use was heterogeneous. In the plain regions of Thailand, where rice is cultivated, An. minimus s.l. was highly endophilic and responded favourably to residual spraying while in forested hilly and cleared forested foothill areas, malaria eradication projects faced difficulties in interrupting transmission. Poor results were attributed to variations in the behaviour of An. minimus s.l. and to the important role of An. dirus as a malaria vector ( Ismail et al. 1974 ). In Vietnam and Burma, DDT pressure induced the selection of exophilic An. minimus populations ( Myo Paing et al. 1988 ; Vu Thi Phan 1996). Several authors ( Ismail et al. 1974 ; Suthas et al. 1986) suspected the occurrence of at least two cryptic species within An. minimus to explain the heterogeneous responses to insecticides. This was confirmed by genetic studies of An. minimus s.l. from Thailand ( Sucharit et al. 1988 ; Green et al. 1990 ). Two species within the An. minimus complex are commonly accepted, informally designated An. minimus A and C ( Harbach 1994), and defined as electromorphs controlled by the enzyme locus Octanol dehydrogenase ( Green et al. 1990 ). Apart from the genetic studies from Thailand, and some studies of specimens from Japan and China ( Kanda et al. 1984 ; Yu Yuan in Zahar 1996) little is known of this complex in other parts of South-east Asia. The differences between the two species regarding their biting and resting behaviour, degree of anthropophily, longevity and ecological differences have hardly been studied. These factors determine the efficiency of a species as a vector.
Elucidating the nature of malaria vector species complexes and characterising the individual species within each complex will provide insight into the complexity of the vector systems in South-east Asia and assessments of the impact of vector control measures ( Coluzzi 1992). This is important for countries such as Laos, Cambodia and Vietnam that spend about 60% of their malaria control budget on vector control.
We studied An. minimus s.l. from northern Vietnam using isozyme electrophoresis to clarify its species composition in this part of South-east Asia. Variations in biting and indoor resting behaviour in relation to population genetic findings were examined. Misidentification using morphological characteristics to separate An. minimus s.l. from closely related species is common ( Harrison 1980). Therefore we searched for diagnostic loci for two other species, An. aconitus and An. jeyporiensis, which also occur in northern Vietnam.
Materials and methods
Mosquitoes were collected in four villages of Hoa Binh Province in northern Vietnam, located south-west of Hanoi. The area is characterized by plains, narrow valleys (< 100 m), hills and mountains. Annual rainfall, mainly from May to October, is between 1500 and 2500 mm. Average monthly temperatures range from 15 °C in the winter to about 30 °C in the summer. Humidity is very high all year round, rarely lower than 80%. Four collection sites were selected. Rong Vong village is situated on the border between two ecological systems, the Red River Delta plain and a hilly area. The site is located 40 km from Hanoi at an altitude of 30 m. Houses are built directly on the ground or on stilts not higher than 0.5 m, and separated cattle sheds are present. The villages of Co Phay and Xolo are situated at altitudes of 400 m and 380 m, respectively. Houses are built on stilts and cattle are kept under the houses at night. Co Phay and Xolo are surrounded by hills up to 600 m high. The fourth village, Khoi, is situated in a large U-shaped valley at an altitude of 330 m. Houses are comparable with those found in Co Phay and Xolo. Rice is cultivated around all four villages and houses are located near possible An. minimus breeding places. Khoi has a complex system of fishponds and rivers.
Malaria reached alarming proportions in Vietnam in 1991 and became a priority for the government ( Verléet al. 1998 ). In Hoa Binh Province transmission is currently low, but malaria epidemics remain a constant threat because people travel between this region and endemic areas, and potential malaria vectors are ubiquitous. During the last three years preceding our collections, no vector control measures had been applied in the study villages.
Adult mosquitoes were caught at monthly intervals from June to November 1995. Night landing collections on humans were done inside and outside three selected houses during three consecutive nights from 1800 h until 0600 h. Morning collections of indoor resting mosquitoes took place on two consecutive days in different houses from those selected for the human night landing collections. Adult mosquitoes were also captured in the vicinity of cattle by two persons for two consecutive nights from 2100 h until 2400 h.
Mosquitoes were identified morphologically using the identification key developed by the Institute of Malariology, Parasitology and Entomology ( IMPE 1987). Specimens belonging to An. minimus s.l. were scored for presence or absence of humeral and presector pale spots on the wings and separated into three morphotypes ( Table 1). This morphological variation is known to occur in An. minimus s.l. from Vietnam (Trung Ho Dinh, unpublished observation) and in An. minimus populations from Thailand ( Sucharit et al. 1988 ). We performed tests to determine if any relation exists between the 3 morphotypes and results of the population genetic analysis based on the isozyme electrophoresis. An. minimus s.l., An. aconitus, and An. jeyporiensis specimens were put into liquid nitrogen immediately after morphological identification.
Table 1. Definition of each morphotype, based on the presence or absence of the humeral and presector pale spots on the wings, and the percentage of each morphotype found in each of 2 different Odh forms of Anopheles minimus
Isozyme electrophoresis of 14 enzymes on cellulose acetate gels (Titan III, Helena Laboratories, U.K.) followed procedures described by Smits et al. (1996) . Enzymes and migration conditions were adapted to the Anopheles mosquitoes of Vietnam ( Table 2). LDH (Lactate dehydrogenase, E.C. number 188.8.131.52) and ODH (Octanole dehydrogenase, E.C. number 184.108.40.206) were tested on concentrated samples because of their low activity ( Smits et al. 1996 ). The banding patterns of SOD (Superoxide dismutase, E.C. number 220.127.116.11) were visible on the ODH gels.
Table 2. Enzyme systems and migration conditions for cellulose acetate electrophoresis adapted to three taxa of the Myzomia Series, in northern Vietnam
Mosquitoes collected in one village during one round by the different collection methods counted as a sample set ( Richardson et al. 1986 ). An. aconitus and An. jeporiensis can be identified morphologically when the classical characteristics are present. Mosquitoes with these diagnostic characteristics were used as standards for the detection of diagnostic loci separating An. minimus s.l., An. aconitus and An. jeyporiensis. These loci were used to withdraw misclassified An. aconitus and An. jeyporiensis from the sample sets of morphologically identified An. minimus s.l.
More than 2 species might exist within the An. minimus complex ( Zahar 1996), therefore no attempt was made to separate a priori the An. minimus s.l. sample sets by the criteria of Green et al. (1990) . The null hypothesis that random mating occurs within the morphologically identified An. minimus s.l. was tested by means of the F-statistics (Fis, Fit and Fst) ( Wright 1951), which enable the measurement of deviations from Hardy–Weinberg expectations at different levels. Any heterozygote deficit is classed into within and among population components. Fis is a measure of the ‘within population’ heterozygote deficit while Fst is a measure of the ‘among populations’ heterozygote deficit. Fit measures the global heterozygote deficit. This partition of the heterozygote deficit permits inferences about the levels of inbreeding and gene flow of the populations under investigation ( Goudet et al. 1994 ). The computer program FSTAT ( Goudet 1995) was used to obtain estimates of Fis, Fit and Fst, based on the hierarchical analysis of variance developed by Weir and Cockerham (1984), which explicitly accounts for sample size. The 95% confidence intervals (95CI) of the overall F-statistics were obtained by bootstrapping over loci ( Goudet 1995). Significance of the F-statistics was tested with the method of permutations (10000 perms per test) ( Goudet 1995).
Deviations from the Hardy–Weinberg expectation of the individual loci in each sample set were tested by an exact test using the program GENEPOP ( Raymond & Rousset 1995). This test is not adversely affected by small expected values and appropriate for small sample sets or when rare alleles are present ( Lessios 1992; Rousset & Raymond 1995).
Nei's unbiased genetic distance ( Nei 1978), which evaluates genetic similarities between populations, was calculated using BIOSYS ( Swofford & Selander 1981). To avoid type-1 errors resulting from multiple simultaneous tests, significance levels were adjusted through sequential Bonferroni procedures ( Hochberg 1988; Lessios 1992).
The Myzomya Series: species identification
An. minimus s.l., An. aconitus and An. jeyporiensis, belonging to the Myzomya Series, were collected in northern Vietnam. Two loci were found for MDH by cellulose acetate electrophoresis. Only Mdh-2, migrating cathodally on the cellulose acetate gels, displayed clearly interpretable band patterns. It could not separate the 3 taxa unambiguously. However, the Mdh-2100 allele could be considered typical for An. minimus s.l., whereas the Mdh-2135 allele was typical for the other two species. Only 1.2% (12/996) of Mdh-2100/135 heterozygotes were found in samples of the morphological identified An. minimus s.l. population. The Odh electromorphs were diagnostic for the three taxa. Four different Odh alleles were found in the An. minimus s.l. population: Odh100, 118, 133, 142. The Odh87 allele was diagnostic for An. aconitus and the Odh56, 73, 80 alleles were found in the morphologically identified An. jeyporiensis. Using the Odh and Mdh-2 loci, 3% (31/996) of the specimens were misidentified and withdrawn from the morphologically identified An. minimus s.l. sample.
An. minimus s.l: inter- and intra-specific variability
Eleven enzymes were suitable for further analysis. Mpi was excluded from analysis because interpretation of the zymograms was unreliable due to the high number of alleles. The Sod and Me loci were monomorphic and not used in the analysis of An. minimus s.l. Frequencies of the following alleles were pooled because they were difficult to distinguish when not on the same gel: Aat-1104 with Aat-1100, Acp112 with Acp100 and 6-Pgd104 with 6-Pgd100 and 6-Pgd96.
Table 3 gives the number of morphologically identified An. minimus s.l. per sample set. Those smaller than 5 were excluded from the analysis. The overall estimates (over sample sets and loci) of the F-statistics of An. minimus s.l. showed positive values of Fit, Fst, and Fis ( Table 4, criterion 1) significantly different from zero. The significant positive value of Fis denoted that nonrandom mating occurred within the sample sets, whereas the Fst value indicated a high heterozygote deficiency between the sample sets of the An. minimus s.l. population. Among the 13 enzyme loci Fis and Fst values were highest for Odh ( Table 5, criterion 1).
Table 3. Number of mosquitoes per sample set of Anopheles minimus collected in northern Vietnam in 1995
Table 4. Overall estimates of F-statistics over 13 gene loci for collections of Anopheles minimus from northern Vietnam. Each criterion, except criterion 1, defined an additional partition of each sample set. Tests of significance were performed by permutations, 95% confidence interval (95%CI) by bootstrapping over loci
Table 5. Fis and Fst values per locus for Anopheles minimus from northern Vietnam. Each criterion, except criterion 1, defined an additional partition of each sample set (criteria identical to Table 4). Tests of significance were performed by permutations
Four different Odh alleles were found in the morphologically identified An. minimus s.l. Based on the frequencies of heterozygotes, two distinct groups could be identified: (1) Odh form I with homozygotes Odh100/100 and heterozygotes Odh100/118; (2) Odh form II including genotypes Odh142/142, Odh133/142, Odh133/133 and Odh118/133 with Odh133/133 as predominant genotype. Eight heterozygotes between form I and form II, 1 Odh100/142 and 7 Odh100/133 (0.88% of total collection), were collected during this study. All were collected in Khoi, six of them in October. No specimens with genotypes Odh118/142 or Odh118/118 were caught. The number of mosquitoes per Odh form, and per sample set is shown in Table 3.
Dividing each sample set by Odh forms, excluding the heterozygotes Odh100/142 and Odh100/133, reduced the Fis value to 0.034 ( Table 4, criterion 2) but it was still significantly different from zero. Analysing the two populations of the An. minimus Odh forms separately showed that only form II had significant positive Fis and Fit values ( Table 4, criteria 3 and 4). No population structuring could be inferred for either An. minimus form because of the nonsignificant Fst values.
Mean Fis, Fit and Fst values of An. minimus form I were not significantly different from zero ( Tables 4 and 5, criterion 3). In the 17 sample sets of form I, none of the polymorphic loci, defined as a locus whose most common allele has a frequency of 99% or lower, showed significant deviation from Hardy– Weinberg expectations. The per-locus Fis of Odh form II ranged from − 0.038–0.497, and significant Fis values were observed for the Ldh and Gpi loci ( Table 5, criterion 4). These loci deviated significantly from Hardy–Weinberg expectations in sample set Khoi-October. Table 6 shows the frequencies of the most common allele per locus and Fis values of 13 enzyme loci for the pooled sample of An. minimus forms I and II.
Table 6. Number of alleles per loci, frequency of the most common allele per locus and Fis per locus for the pooled sample of each Anopheles minimus Odh form. Tests of significance were performed by permutations
Genetic distance between the species of the minimus group
Based on the 13 enzyme loci, Nei's unbiased genetic distance between the two An. minimus forms from Hoa Binh was 0.092. The genetic distance between An. aconitus and An. minimus forms I and II was 0.649 and 0.671, respectively.
Characterising the forms of An. minimus
All three morphotypes were found in both An. minimus forms ( Table 1). The relative importance of the morphotypes in the collections of both forms differed significantly between the monthly catches (exact test P < 0.001 for both forms). In collections of form I, morphotype 1 decreased significantly from June to November (χ2 for Trend; P < 0.001). Neither the humeral pale spot nor the presector pale spot could be used to discriminate between the two An. minimus forms ( Table 1). Based on the polymorphic loci, there was no heterozygote deficit among morphotypes within each An. minimus form (form I: Fst=− 0.002, P= 0.699; form II: Fst=− 0.001, P= 0.513).
Both An. minimus forms were found in sympatry in the four study villages of Hoa Binh. Form I was omnipresent while form II was mainly collected in Xolo and Khoi and almost absent in Rong Vong ( Figure 1). The distribution of both forms was not identical across collection months in Xolo (exact test P < 0.001) and Khoi (exact test P= 0.001). Despite seasonal changes in the proportion of An. minimus forms, their distribution between villages was significantly different (Mantel-Haenszel test, stratified by collection month, P < 0.001 for all combinations of villages). Based on polymorphic loci, there was no heterozygote deficit among villages within each An. minimus form (form I: Fst= 0.001, P= 0.212; form II, Rong Vong excluded because sample < 5: Fst=− 0.001, P= 0.542).
The distribution of An. minimus differed for each collection method across villages (exact p-value < 0.001) and collection month (exact p-value < 0.05). Odds ratios of collecting form I by each collection method adjusted for village and collection month were calculated by logistic regression. The effect of each collection method was compared to the overall effect. All collection methods yielded both An. minimus forms, but their frequencies in collection methods differed. Twice as many form II were collected by the outdoor human landing and the outdoor cattle collection, and about 5 times as many form I by the indoor resting collection. No difference in distribution was observed for the indoor human landing collection type ( Table 7). Based on polymorphic loci, no heterozygote deficit among collection methods was observed within either An. minimus form (form I: Fst= 0.000, P= 0.428; form II: Fst= 0.002, P= 0.225).
Table 7. Odds ratio of collecting Odh form I for each collection method obtained by logistic regression
Misidentification of Anopheles minimus s.l. using morphological characters could be avoided by the diagnostic enzyme loci identified during this study. The Odh locus was diagnostic for the three taxa of the Myzomya Series, An. minimus s.l., An. aconitus and An. jeyporiensis, found in northern Vietnam. In contrast to the findings of Green et al. (1990) , MDH could not separate unambiguously An. minimus s.l. from An. aconitus.
Significant positive Fis values provided clear evidence of nonrandom mating within the An. minimus s.l. population from northern Vietnam. The highest heterozygote deficiency within the sample sets was observed at locus Odh, which could be identified as a diagnostic locus for 2 sympatric forms of An. minimus in Vietnam. In Thailand, Green et al. (1990) also recognized the Odh locus as diagnostic for two isomorphic species within An. minimus, using a different gel system. Comparison of the Vietnamese specimens with Thai An. minimus mosquitoes indicated that form I is the same as the Thai species A and that form II can be equated with species C from Thailand (W. Van Bortel, unpublished data; R. Sharpe, personal communication). The two An. minimus species recognized by Green et al. (1990) were monomorphic for the Odh locus, while An. minimus species C from Vietnam (form II) was polymorphic for the same locus. The An. minimus species from Vietnam are genetically very similar. Nei's unbiased genetic distance was 0.092, which is slightly lower than the distance (0.134–0.172) found between the two isomorphic African malaria vectors An. gambiae and An. arabiensis ( Cianchi et al. 1983 ). The distance between these two An. minimus species and An. aconitus, a closely related species, was approximately 0.65.
Hybrids in nature are found in all major groups of higher organisms, and taxa that maintain their integrity despite this overlap have been classified as separate species ( Barton & Hewitt 1989). In our samples, 8 (0.88%) hybrids between the An. minimus species A and C were collected in Khoi. Six of them were collected in the sample set Khoi-October in which an heterozygote deficiency at loci Ldh and Gpi was observed in species C.
Sucharit et al. (1988) proposed the presence of the presector pale spot and humeral pale spot as diagnostic markers for the Thai An. minimus species C, while Rattanarithikul et al. (1995) used only the presence of the presector pale spot to distinguish species C from species A. We could not correlate these characters with one of the An. minimus species from Vietnam. Moreover, a change of the relative importance of the different morphotypes in each of the An. minimus species from Vietnam was observed during the study period (wet season to cool dry season). The use of these morphological characteristics to identify the two An. minimus species from Thailand led to 37% error ( Green et al. 1990 ) and to 33% in the Vietnamese samples.
No population structure could be inferred in either species populations from Hoa Binh; there was no evidence of restricted gene flow between monthly samples nor between the samples from different locations or between the collection methods.
Two closely related species occurring in sympatry are expected to be bionomically different. If these differences are epidemiologically relevant, they can increase the complexity of disease transmission patterns. Behavioural differences between cryptic species may also influence the effectiveness of vector control measures ( Coluzzi 1992). In our study both An. minimus species were attracted to man as well as to cattle. However, twice as many species C specimens were captured on humans outdoors and on cattle as species A, while similar proportions of both species were collected on humans indoors. This indicates that species C is likely to be more exophagic and zoophilic than species A. The most important finding was the highly endophilic behaviour of species A, which was 5 times more abundant in indoor resting collections than species C. The difference in resting catches between both species was even more striking considering the fact that cattle were kept under the houses during the night, and that mosquitoes of species C, prone to feed on cattle, could easily enter the houses to rest. Based on the endophilic behaviour, we may expect a lower impact of indoor spraying on species C than on species A.
In different regions of its distribution, An. minimus s.l. exhibits a wide range of responses to indoor spray campaigns with insecticides ( Harrison 1980; Parajuli et al. 1981 ; Suthas et al. 1986 ; Lien 1991). This can in part be explained by the presence of two cryptic species within An. minimus s.l. occurring in sympatry, but in different proportions depending on the location and dissimilarities in resting behaviour.
Introducing impregnated bednets in Assam, where the main vector, An. minimus s.l., is anthropophilic, endophagic and endophilic, reduced considerably the positive slide rate of malaria ( Jana-Kara et al. 1995 ). In Thailand, however, the impact of treated bed nets on malaria was variable and the poor results were mainly attributed to the exophagic and exophilic behaviour of the vector ( Somboon 1993). At this point it is not clear to what extent these variations could be explained by the presence of different An. minimus species. Differential impact of impregnated bednets on cryptic species is complex and extends beyond simply killing the mosquitoes. This was shown for two members of the An. punctulatus complex, An. koliensis and An. farauti, in Papua New Guinea. The survival rate of An. koliensis, which is more anthropophilic and endophilic than An. farauti, was affected after introduction of impregnated nets but this was not so for An. farauti. However, regularity and duration of the oviposition cycle was disturbed in An. farauti, which may shift the peak biting activity from postmidnight towards premidnight. This may increase the potential of the mosquito to transmit the parasite ( Charlwood & Graves 1987). Such a shift of the biting cycle was also observed in An. minimus s.l. from Thailand after indoor spraying ( Ismail et al. 1975 ).
In Vietnam impregnation of bednets is largely promoted in the framework of a comprehensive malaria control strategy, including disease management and prevention. This resulted in a decrease of malaria morbidity and mortality, particulary in northern Vietnam, over the past five years. But malaria transmission still occurs in remote areas. Indoor spraying is reserved to control these epidemics ( Verléet al. 1998 ). The relative role of the different cryptic species in the maintenance of these foci and their behavioural changes in relation to vector control should be further analysed.
Excellent technical support was provided by the entomology team of the Institute of Malariology, Parasitology and Entomology, Hanoi, Vietnam and by the staff of the provincial malaria centre of Hoa Binh Province. We are grateful to the Vietnamese Ministry of Public Health for facilitating this research. We thank F. Goudet who kindly provided the FSTAT computer programme and related literature. This work received financial support from the Belgian Administration for Development Co-operation, and the INCO-DC research project ERBIC18CT970211. Data analysis was supported by a grant of the Compagnie Maritime Belge.