The Anopheles dirus complex: spatial distribution and environmental drivers

A map first shows the overall extent of the An. dirus complex distribution (Figure 2). A second map focuses on Southeast Asia, where the species diversity and the number of records are higher (Figure 3). In this area, the presence of An. dirus s.l. seems to correspond to presence of malaria cases. The additional file 1 allows better reading of the maps. It indicates the reviewed publication, the species discovered and the identification method used for each site. As it was not graphically possible to display individually the more than 500 sites, sites close to each other are represented by a single point on the map and a single reference number. The additional files (Additional file 2, 3 and 4) associated with the current article provide a description of the site, collection methods and the number of An. dirus s.l. collected, as well as the original location for each of the 500 sites organized by site number or by reviewed publication. The current article is also aimed at malaria control workers, and with this information, they can find the situation recorded at any of the collection sites in the past, as well as locate any collection site cited in the publications reviewed.

Species of the An. dirus complex can be found in forest and forest foothills from India to Taiwan and from the 30° north parallel to the Malaysian peninsular. The mean monthly temperature below 20° seems to limit the northern distribution of the complex to just beyond the border of India with Nepal and Bhutan. Rainfall is probably the limiting factor to the west with annual rainfall per year under 800 mm. The absence of species of the complex in large non-forested areas of Thailand, southern Vietnam and central India is probably linked to the lack of suitable habitat. However the species of the complex are also absent from the north of Vietnam. This is most peculiar, given that this area is still forested and that members of the complex occur at the same latitude in neighbouring countries. The possible occurrence of An. dirus s.l. in central India was investigated by Srivastava [21]. A predictive model identifying suitable factors for Anopheles survival and the distribution of wet evergreen and deciduous forest confirmed the presence of An. dirus s.l. in already well-known areas of occurrence, but also in a few suitable sites through central and North-Western India. This seems to correspond to historical collection sites in Uttar Pradesh, Kerala and Karnakata. Bhat [22] and Rajavel [7] also cite historical records in Kasauli and other regions now considered free of An. dirus s.l. Recent reports of the presence of An. dirus s.l. in the state of Jharkhand (Dr. Diwakar Dinesh, personal communication) call for further investigation into the actual distribution of the complex in India. Those areas are presented in Figure 2 as sites with presence suspected but not confirmed.

Regarding the species distributions within the range of the complex, there are no major geographic or topographic reasons that seem to justify the current species distribution pattern. Therefore, speciation probably does not result from adaptation to a specific environment but the species might have been isolated for some time. Their distribution most probably reflects geo-morphological changes which have occurred in the past and are today not evident [3]. The mountains of the Western Ghats in southern India are the most westward limit of the complex with the presence of An. elegans [10, 23] apparently isolated from the other species. The closest records to this area come then from the north eastern states in India [24] and the border of Nepal [25], and are considered to be An. baimaii based on identification in nearby Bangladesh [26]. An. baimaii seems to be the main species in Myanmar, occurring throughout the country, then giving way to An. dirus s.s. eastward from western Thailand. An. dirus s.s. is the only species recorded so far in Vietnam, Lao, Cambodia [27‐29] and Hainan Island [30]. An. crascens is confined to southern Thailand and Malaysia[31], and An. takasagoensis to Taiwan [32]. An. scanloni and An. nemophilous seem to have more patchy distributions. An. scanloni is restricted to western and southern Thailand, whereas the distribution of An. nemophilous closely follows monsoon forests. The south of Thailand and the Thai – Myanmar border present various sites of high sympatry for several species of the complex [3].

Some clarifications are still called for. Variations of ITS2 rDNA usually occur only between sibling species, thus intraspecific variations between populations of An. baimaii from Thailand and site 302 in China, and An. scanloni at sites 175 and 195 suggest the possible existence of two new species [15], with a less diverse population at site 195 indicating a period of isolation[33].

There is need for some investigations at the Thai-Cambodian border, the border between Lao and China and some areas in Myanmar. Indeed, although only An. dirus s.s seems to occur in Cambodia, two specimens of An. baimaii were earlier recorded near to the border in Thailand at site 241 [34] and specimens at the border have been observed to feed on monkeys in the canopy as would An. nemophilous in nearby Thailand [35]. Only An. dirus s.s. has been recorded in Lao although An. baimaii occurs nearby in China. Unexpected records such as An. dirus s.s. at site 49 [36] and 148 [37], and An. scanloni at site 126 [37], should be taken with caution and investigated further. This is also the case for records of An. baimaii at site 241, which are based on the collection of only two specimens. Sallum et al. [4] reported the presence of An. crascens in Sumatra but gave no reference.

Primary and secondary larval habitats of An. dirus s.l. have some constant characteristics: temporary, standing or slowly moving water under shade. While primary sites occur year round and tend to be associated with the stream system in deep forest, they vary in nature according to the season with drying pools in stream beds, pools connected with streams or meanders of slow moving streams in the dry season [38‐40], complemented in the rainy season by pools fed by underground seepage along streams, springs, rock pools in the beds of ravines, and in deep holes or pits which form ahead of gullies [32, 39, 41]. Such primary sites are often confined to the deep forest resulting in transmission of malaria mainly to forest workers. Some of them however occur in peri-domestic area: An. dirus s.s. occurs in gem pits [42] and in water pits fed by seepage along streams [43]. If wells are generally negative[39, 44], An. baimaii occurs in wells close to houses in a particular area of Myanmar throughout the year [36, 45‐47] even if densities are reduced in the dry season.

Secondary larval habitats occur in the rainy season and can be found closer to human settlement at the forest fringe. These are commonly small, shallow, temporary, shaded water-holding depressions [48]. Puddles in paths are commonly positive, whereas larger bodies of water such as ponds, large rivers, irrigation channels and marshy areas are generally negative. However, An. takasagoensis breeds in large permanent pools used by buffaloes for bathing if freshened by the rain [32]. An. dirus s.l. was also recorded in long marshy areas in the forest of Myanmar [41] and in two swamps and one rice paddy in Thailand[49]. Artificial containers [39] and natural containers are also generally negative but An. dirus s.s. is found reported in clay jars in Hainan[2], in a tin-hat[40], empty tins and in terracotta jars in Vietnam[50]. Positive natural containers include only bamboo stumps [5, 36, 40] and palm leaves [40]. Details on larval habitats and locations investigated are presented in Figure 4.

Immature stages of An. dirus s.s. and An. baimaii show adaptations to temporary habitats [39, 43]. The eggs survive on the moist floor of a drained pool up to one month after the rain stops, but not throughout the dry season, unless sporadic rains prevent the soil from drying or unless the eggs are on the wet soils of forest. Not only do the eggs survive between rainfalls but they can also partially mature on a moist surfaces thus reducing development times, with first stage larvae present in pools containing no water prior to rain [39, 43]. Larval stages can crawl to other pools in case of drying but they can also survive the desiccation of a draining pool and ant predation by getting covered in a coat of mud and becoming buried and motionless just before complete drainage. On re-flooding, only An. dirus s.l. reappears at the surface thus eliminating other competitors. An. dirus s.l. immature stages can feed on the larvae of other species and less frequently on its own larvae, in particular under crowded conditions [39]. Increased larval density results in higher larval mortality[36] and the production of smaller adults even with the same amount of food per larva, suggesting an influence on feeding efficiency rather than a shortage of food per se [51].

For a sibling species, larval habitats seem quite diverse according to location but Baimai [3] recorded a preference for vegetated limestone rock pools for An. scanloni and rocky-stony micro-environments for An. dirus s.s., An. crascens and An. baimaii. Sympatric populations of An. dirus s.s. and An. baimaii share identical larval habitat and even the same pools at site 204 [5]. An. baimaii was, however, more often found in pools that dried out in dry spells during the rainy season. This might reflect a seasonal variation in larval habitats or species, with An. dirus s.s. more abundant at the start and An. baimaii during the middle of the rainy season [3].

An. baimaii and An. dirus s.s. are extremely anthropophilic [1, 52‐55], even in the presence of numerous cattle [56], but less anthropophilic behaviour has been sporadically recorded for these two species [27, 57]. In particular, higher densities were recorded on cattle in some years [46] or every year for An. baimaii [58]. Some studies show that An. nemophilous feeds primarily on monkey, and An. dirus s.s. and An. scanloni feed more often on humans than An. crascens. These experiments were however based on few observations[3]. Monkeys are natural and winter hosts [2, 32, 59] but preference for monkey was not evident from human- and monkey-bait collections in the forest. More specimens were caught on human bait on the ground. As many were caught on monkey bait in the canopy as on human bait on the ground [35]. Anecdotal records report individual mosquitoes feeding on birds [56], dogs and pigs [41], as well as mixed feeding on bovines and humans suggesting that mosquitoes do not always come back to the same host[57, 60, 61].

Females are primarily exophagic but enter open shelters to feed [62]. Variations between regions are linked to housing facilities with the highest endophagy in largely open houses built directly on the ground [53]. Most open houses in the jungle show no significant biting differences between indoors and outdoors [1, 56, 57, 63, 64]. In some cases, indoor biting is even higher [38, 43, 65].

Resting places are difficult to find and occur mainly outdoors [58, 59, 66] in holes, wells[36, 58, 63]and vegetation such as bushes, tree holes [41, 52], thick grass, under the surfaces of leaves [36, 40], branches near the ground [40] and under the foots of trees[2]. In India, an intensive search collected none on the ground but 20 from the moist, dark crevices of large tree trunks, and 2 from creepers in the forest, the nearest at 150 m from a village[66]. Day-resting can be separated from night-resting[67]. Females rest early in the evening in outdoor vegetation, fences and wood stacked around houses [52, 56] but after feeding, they fly almost immediately back to the jungle[1], with none found even at dawn around dwellings [43, 52]. Eyles [35] caught more An. dirus s.l. in the canopy than on the ground and when Wilkinson [43] released hundreds of females in predawn darkness, they flew immediately upwards into trees suggesting that they might rest in the canopy of trees.

Biting can happen in daylight in the jungle[38, 59] but occurs mostly from dusk to dawn. From the feeding patterns reported in the literature and presented in Figure 5, it is not obvious whether or not An. dirus s.l. is a late night feeder or an early night feeder. Scanlon [62] emphasised that late night feeding is most common in Thailand but that occasionally local populations exhibit a striking early pattern. The peak feeding activity has also been recorded either late [52] or early [68] in the same location for different years. Species-specific patterns of outdoor biting were suggested for allopatric populations with a very early peak for An. scanloni, and An. crascens, around 22h for An. dirus s.s. and 02h for An. baimaii [3]. An. dirus s.s. and An. baimaii are, however, both often recorded as late night feeders [1, 2, 27, 35, 48, 69, 70], but peak activity was also recorded well before midnight for An. takasagoensis and An. baimaii [32, 41, 46, 57, 71]. The occurrence of a particular sibling species can thus not explain all the variation. Other influencing factors seem to include the moon phase, the season and the presence of DDT, which stimulates early biting [59, 63]. A two-year survey at site 12 recorded wide a repeated variation of the feeding pattern in accordance with the moon phase [59]. Variations have also been recorded between years and seasons [34, 36, 46, 59] with earlier peak activity in the dry season[27] or earlier peak activity in October than in June [63]. Late biting in post-monsoon in some regions might also be due to the late onset of dawn [60]. Compilation of data across years, seasons, moon phases or between indoor and outdoor biting can thus result in a wide variety of patterns that are difficult to interpret [64]. The Relative Risk (RR) of being bitten in the hour before 22h compared with an hour after 22h probably provides a better indicator of exposure. A large proportion of bites before 22h coincide with activities of people before bedtime when the bed-net protection is nil [53]. Indoor biting peaks occur often later than outdoor biting [59] but in the presence of very open houses no difference is observed [53]. Outdoor biting has been recorded as starting as early as 16h for An. scanloni [3], and 19h for An. baimaii, An. dirus s.s. and An. takasagoensis [1, 32, 38, 43, 46, 72].

The occurrence of An. dirus s.l. is mainly linked to rainfall, temperature and relative humidity [46]. Attacks started after the rain exceeded 50 mm at site 12 for An. baimaii [39]. Lower or different thresholds seemed to apply according to the soil type and topography with, for example, An .dirus s.l. in Hainan [73]. Heavy rains however, flush away larval habitats, impede mosquitoes from flying [3, 46, 74] and the flooded grounds can be unsuitable for weeks[43]. The pattern of rainfall might, thus, be more important than the amount of rainfall, and light rains, not too frequent, seem to be most favourable for larval development [32]. Kitthawee [51] observed that adult body size was positively correlated with rainfall 1–2 weeks before. She suggests that renewed rainfall may bring more particulate food onto the surface of the water where larvae feed. Conversely, excessive rain could dilute nutrients and lead to nutritionally stressed, smaller adults. The importance of the rainfall pattern can explain yearly fluctuations in populations recorded during different studies in the same region [28, 29, 64]. The month of peak densities is not constant but usually takes place during the rainy season, with up to 80 bites per man/night [75]. Variations of relative abundance of sibling species in sympatric sites show more An. crascens than An. baimaii at the beginning and less An. crascens than An. scanloni at the end of the wet season at site 175. At site 204, more An. dirus s.s. were found at the start and more An. baimaii towards the middle of the wet season [3]. Populations of An. takasagoensis also present high fluctuations in abundance. The species usually is fairly uncommon but is occasionally very abundant [32].

Temperature is rarely a limiting factor but it influences the longevity of the mosquito, the length of the sporogonic cycle and mosquito activity, thus influencing seasonally the vector status.An. dirus s.l. seems inactive when temperature falls below 15°C [27]. However, in the deep forest of northern Thailand, during the cool, dry season associated with important temperature fluctuations and minima around 10°C, An. dirus s.l. can still survive and even transmit malaria. [63]. The duration of the immature stages is reduced at higher temperatures but females are then smaller [36]. If size does not influence the number of oocysts [76], larger mosquitoes live longer and have thus a greater vectorial capacity [77]. Mean daytime temperature averages 25.4°C with up to 5°C variation in positive ground pools [43]. However, in colonised wells temperature remains almost constant (26.2°C), despite outside temperature oscillations from 19.5 to 33.1°C. The clay contained in the lateritic wall lining the wells might have a sustained cooling effect[47].

Topography, salinity, pH and shade also influence the availability and occurrence of larval habitat. Topography is a major element with sites found commonly in foothills where rain water can accumulate, next to streams or in the beds of ravines. Rosenberg observed that the appearance of waves of An. dirus s.l. after the rain is not systematic at site 12 and requires three concomitant elements: rapidly draining pools, intermittent, heavy rains and embryonated eggs [39]. If suitable pools are temporary, they should hold water for at least five to eight days [39, 66]. Suitability of a site might, thus, be a combination of the clay content of the soil with the appropriate amount and frequency of rain. In the sandy soils of site 12, suitable larval habitats occur only on the compacted soil of the path. The infiltration is too high elsewhere. Where An. baimaii breeds year-long in wells, the soil outside is not appropriate for larval habitat because it dries out very rapidly by percolation and evaporation. Even puddles in path are negative [46] and most wells dry up in summer [45]. Kitthawee [42] noticed that site such as 148 and 243 which presents the particular behaviour of year-long larval development in gem pits or wells also present a particular environmental setting with very high rainfall and infiltration rates.

Salinity, pH, shade and temperature have been extensively analysed in the context of An. bamaii in wells in Myanmar. Nitrate, iron, dissolved oxygen, sulphate, chlorine, ammonia and water hardness seem to have no influence, but larval density is negatively correlated to pH in ground pools and salinity in wells, with a threshold at 200 ppm NaCl [36, 45, 47, 49]. Larval density becomes very low when the distance between the well-water surface and the ground surface is less than 2.5 m, probably reflecting an effect of shade and temperature [36, 47]. Larvae gradually disappear if the shade is removed. Except for a few records [5, 69], most larval habitats are under the shade of trees which probably reduces the drying speed of pools, influences temperature and provides food through organic matter and leaves falling into the water. Positive wells have shade and shrub on inner walls, debris on the water surface and often abundant decaying leaves at the bottom, even if the water is clean [36]. Numerous larvae are even observed in pools often fouled by buffaloes [2], but An. dirus s.l. will not be encountered if the water is not freshened frequently by rains [48].

The most important environmental parameter is definitely land-cover. Throughout its distribution An. dirus s.l. is associated with forested foothills, forests or forest fringes. If availability of larval habitats and the presence of natural hosts, such as monkeys [35] are explanatory factors for this association, adults may also require the highest humidity and lower temperature of the jungle biotope for optimal survival [2]. Favourable environmental conditions of dense vegetation, humid soil, high relative humidity and shade, coupled with the presence of permanent suitable larval habitats or primary sites, appear to persist deep inside the forest during the dry season. Although rainfall in a tropical rain forest is unpredictable, the forest floor is humid. Even if the larval habitats loose their free water, the high humidity probably keeps the eggs viable for fairly long periods until the next rain comes[2]. The tree cover provides food for larvae, with leaves and other debris falling into larval habitats and assures stable micro-climatic conditions, even in the dry season. As the rainy season begins, conditions also become favourable at the margins of forest and An. dirus s.l. seems to spill over from the forest into secondary larval habitats.

Much larger densities of An. dirus s.s. are present in deep forest settlements or villages than in villages located at the edge of forest or in fragmented forests [78]. Variations in forest are difficult to interpret when mosquito densities are low near a village, very high at 1.5 km and moderate at 5 km[1, 35, 78, 79], Infected mosquitoes can be present at oviposition sites 1.5 to 3 km away from the village [59]. Terracotta jars provided for oviposition in the forest attracted the highest number of mosquitoes at 150 m from the village but hardly any 300 m away [29]. Specimens were captured up to 2 km away from a source during a mark-release-recapture study [80]. These complex and contradicting results show that all the parameters influencing the occurrence and density of the An. dirus s.l. populations in the forest are not totally understood but the distance to the potential host and suitability of oviposition site certainly play a major role. The association with the forest is high in any case and almost systematically results in high vector contact and malaria transmission.

Species such as An. nemophilous seem to be dependant on monsoon forest[3]. In Cholbury, Thailand, large populations of An. dirus s.l. were encountered by Scanlon [52] in 1964, but the site has been further deforested and the population has been considerably reduced. However, adaptation to other land-cover has been recorded, particularly to teak and rubber plantations and orchards. The mosquitoes can adapt to the edge of man-made clearings [39, 48, 81] and have once been recorded in rice fields [49]. However, it is not known,, if large forested areas are necessary for the survival of these vector populations or if fragmented forest or plantations might be sufficient. The particular behaviour of year-long oviposition in wells might be a consequence of an adaptation to a new type of larval habitat and resting places that provide a cooler and more constant temperature, corresponding to the forest biotope [47].

The role played by An. dirus s.l. in the transmission of malaria has only been assessed during the last 50 years. It is now considered as the most important vector in Southeast Asia. Several factors contribute to making species of the An. dirus s.l. complex an exceptionally efficient vectors: they are so long lived and highly anthropophilic that only small populations are necessary to maintain high malaria endemicity [81]. Their exophilic behaviour, early biting habits and insecticide avoidance undermine the efficiency of the most common vector control measures e.g. insecticide residual spraying and insecticide impregnated nets. High human/vector contact in the typical forest biotope inhabited by species of this complex can explain the extended occurrence of what has been called "forest malaria".

The forest activities of humans play a major role in the malaria epidemiology of Southeast Asia [82]. The colonization of new land for agriculture, logging, mining and other activities, as well as resettlement of populations in the forest, expose people to high transmission risks in the most favoured biotope of An. dirus s.l Overnight stays in the open for hunting and collecting fruits in forest increase the human/vector contact, and open temporary shelters and forest huts facilitate early indoor biting when people are not yet protected by bed-nets[40]. The invasion of the jungle by human settlers most probably increases the densities of these mosquitoes by providing hosts and the small transitory pools that are preferred for oviposition [39]. During the dry season, people are mainly getting malaria in the deep forest [63] where infected vectors [43] are commonly found all year round near permanent streams [39]. Malaria attacks occur in villages only during the rainy season, when An. dirus s.l. moves back to the valley and forest fringe and where secondary larval habitats become available [59]. People move between villages and semi-permanent huts in forest and migrations between infected areas in the forest and non-infected areas trigger the start of transmission in the forest fringe and its surroundings once conditions there become favourable.

The vector density peak occurs generally one month before the malaria incidence peak [40]. However, large populations of mosquitoes are, not required for maintaining a high level of transmission [1, 29, 61, 71]. Hence, small populations of mosquitoes might not be detected in short-term surveys as abundance and presence vary greatly between years, seasons and even from one week to another. In some regions, females feed late at night and may be missed unless night-long collections are made [62]. The sporozoite rates of An. dirus s.l. vary with season and location, with the highest rates recorded in October (7.8%) at site 17 [75] and rates up to 14% in forested site 22 [41]. Rosenberg [81] found high variation between villages 800 m away from each other with a sporozoite rate three to four times greater in the site of lower abundance. Sporozoites of P.vivax and P.falciparum have been commonly detected in An. baimaii and An. dirus s.s Baimai [83] reported sporozoites in An. scanloni and An. crascens with slight differences between species in relation to the parasite. An. dirus s.s. developed Plasmodium vivax and P. falciparum oocysts more readily than An. crascens and An. scanloni. An. elegans, An. nemophilous, and An. takasagoensis probably only transmit simian malaria [83].

Foci of chloroquine resistance have been commonly associated with An. dirus s.l. [1, 2, 70, 84]. Wilkinson [85] carried out an experiment in a highly endemic area for chloroquine resistant strains of P. falciparum and showed that 66% of An. dirus s.l. and 44% of An. minimus became infected when fed on the same infected patients. When comparing the infected mosquitoes, the number of oocysts was also higher in An. dirus s.l Trung [29] recorded P.falciparum, P.vivax-210 and P.vivax-247 circumsporozoite protein (CSP) in a single An. dirus s.l. mosquito. The great longevity of An. dirus s.l., its high susceptibility to Plasmodium infections and a tendency to develop high numbers of oocysts increases the risk of recombining parasite strains in the mosquito gut and as consequence the risk for a fast spread of multi-drug resistance.

Alternative methods to human landing collection should be used where multi-drug resistance is present. CDC light traps have been used successfully on other species in Africa [86, 87] and on An. dirus s.l. indoors in India [60] but other studies show less positive results when comparing various trapping method to human landing collections [88, 89]. Alternative methods to human landing collection such as CDC light trap should be further evaluated as possible tool for monitoring vector control programmes.

An. dirus s.l. is susceptible to DDT [36, 41, 43, 50] but due to exophilic behaviour, females avoid treated walls [43, 52, 63, 90] or even avoid the sprayed huts by biting more outdoors after residual spraying [59, 91] and high sporozoite rates may persist after application of DDT [2]. Insecticide impregnated bed-nets were proved to be effective [84] if kept in good condition [92], however, early biting habits in some areas exposes people to bites before bed time. Insecticide-treated hammocks and personal protection might thus be more effective. Alternative methods such as vegetation clearing are difficult to apply to such diffuse temporary larval habitats[2]. Treating the vegetation surrounding houses to target resting females would probably fail due to rapid loss of insecticide to rain and rapid vegetation growth [62].