Background
Anopheles pseudopunctipennis is a major vector of malaria in the Americas. Its geographical distribution extends from the United States (south of 40°N), Mexico, through Central America and the Andean countries of South America down to the northern part of Argentina (30°S), with an eastern extension into Venezuela and the Lesser Antilles. It is present in the foothills and mountainous areas of these countries, and above 600 m, it is often the only malaria vector. In Bolivia, it is present in the valleys of the Andean piedmont, up to an altitude of 2,800 m and transmits
Plasmodium vivax. Despite numerous studies on its taxonomic, genetic and vector status, as well as insecticide effects on control strategies since the appearance of DDT, very few data are available on
An. pseudopunctipennis ecology, and in particular on its host preference and host selection for blood feeding. Neither the Human Blood Index (HBI)(the proportion of blood meals taken on man), nor
a, the proportion of blood meal taken on human per mosquito per 24 h (
a =
HBI/
duration of the gonotrophic cycle) have ever been studied in details for
An. pseudopunctipennis, despite the major importance of these factors in the understanding of malaria transmission. These parameters appear in the computation of entomological indexes such as, for example, the vectorial capacity (
C), which expresses the potential for malaria transmission. The vectorial capacity (or at least some of its components) is an important parameter to estimate when malaria transmission is to be understood and controlled [
1,
2]. It was defined by Garret-Jones [
3] from formulas developed by Macdonald [
4] and may be defined as the daily rate at which future inoculations arise from a currently infective case. One of the possible equations used for measuring it is:
C = m.a 2·p
x
/(-Log(p)) (1)
Where
a = Proportion of bloodmeals taken on man by a mosquito per 24 hours
m.a = Number of bites per man per 24 hours (m = mosquito density)
p = daily survival rate of the mosquito
x = duration of the extrinsic cycle of the transmitted Plasmodium sp.
Because a mosquito has to take two bites on man to transmit malaria, (
C) is sensitive to changes in the man biting rate (
ma), and to
a, the proportion of blood meals taken on man in 24 h, which appears as
m.a2 in the formula. Because the vectorial capacity varies as the square of the frequency of feeding (
a2 in the
C formula), and likewise as the square of the HBI, small changes in the HBI may have a significant impact on operational control of transmission. The
a parameter may also be useful to compute the malaria stability index [
4]:-
a/
Log(p). This basic analysis of the Ross-MacDonald model of malaria transmission reveals that along with the probability that a mosquito will survive long enough to transmit, the probability of biting on human is the second most important parameter.
The estimation of
a (or the HBI) alone is always troublesome [
5]. However, blood meal identifications provide an integral part of epidemiological studies. The prevalence of malaria in an area is greatly influenced by the process of host selection by malaria vectors which, in turn, is influenced by many factors including host preferences of the vector [
6]. A good simple example of the influence of
a on potential for malaria transmission is given by the comparison of
Anopheles arabiensis with
Anopheles gambiae in Africa.
An. gambiae species complex is one of the most highly anthropophilic, while
An. arabiensis is an opportunistic species that feeds preferentially on humans in many parts of Africa, but can be diverted to domestic animals as their density increases. As such, in certain areas, this less antropophilic pattern leads to a low mean annual sporozoite rate of about 1–2% compared to
An. gambiae which shows a mean annual sporozoite rate of 6% [
7‐
9].
For
An. pseudopunctipennis, very few data exist on its host feeding patterns, and in general terms, studies lack rigorous sampling and data analysis. In Argentina, the HBI was found to be 50% [
10], in Venezuela, a small sample gave 2.2% [
11]. Indoor mosquitoes from Mexico gave an HBI of 67% [
12] and in Peru, the HBI ranged from 40–50% to 85% [
13,
14]. Unfortunately, all these estimates lacked data from the surrounding available host densities and as such, only point out the variability of the HBI within sampling areas. Others found that
An. pseudopunctipennis was indeed anthropophilic but as well was able to feed on animals [
15].
An. pseudopunctipennis from Mexico, El Salvador and Costa Rica took bloodmeals from bovids and equiids (39%-22%; 60%-12%; 78%-0% respectively), but because of inadequate sampling, data could not be used to derive any HBI [
16]. In villages of Mexico, 54% to 86% of engorged females resting inside houses were human-blood fed; in all captured resting females (inside and outside houses) humans and dogs were the more common blood sources and the HBI ranged from 29 – 55% [
17]. Comparing trapping methods, a horse-baited trap was more effective than human landing catches [
18]. Another study carried out in Mexico showed that the HBI ranged from 3.3 – 6.8% in DDT sprayed houses, because of irritability and repellence of the chemical which limited mosquito entrance in the houses [
19]. These few data only demonstrate the ability of
An. pseudopunctipennis to feed on a variety of blood sources, and to change its targets when ecological factors change.
An. pseudopunctipennis live in various ecological situations, with different habits according to regions [
20]. Because of the variability of environmental confounding factors that influence feeding patterns, it is often impossible to compute a general estimate of feeding preference to be used in the whole distribution range of the species. As such, the present study only claims to evaluate the feeding patterns of
An. pseudopunctipennis in a characteristic situation in the central Andes of Bolivia, and interpret the results in terms of malaria transmission risks in that area where similar conditions occur.
Discussion
Host choices are genetically based, mosquitoes responding to particular hosts cues [
29], but in nature, the expression of host preference depends also on the influence of many confounding factors. Characterizing mosquito host choice with the two parameters of host preference and host selection is useful but may be more complicated than that. Relevant parameters may be distinguished as follows. Determining parameters may be as follow. The probability for one type of host to be bitten can separated in two major components. The first one, named here "host availability", quantifies all the ecological, biological and behavioural factors that may modify the probability for one type of host to be bitten. The second one, called here "biting power" of the mosquito, groups all the intrinsic factors of the mosquito.
Availability can also be divided into two components. On one hand, the "accessibility", which is related to the host density in the area, and on the other hand, the "vulnerability", which depends on the interactions on hosts and mosquitoes and are often the consequence of their behaviour (vulnerability is, for example, correlated to the flight range of the mosquito). Practically, it may be difficult to distinguish between these two components but in general terms, they act at two different spatio-temporal scales. Variations in accessibility are usually on a season scale basis (i.e., migrations of hosts) while vulnerability factors vary at a smaller scale level (i.e., at the night level, inducing changes in host behaviour that may for example, move from the areas of high mosquito densities).
The other major component, the "biting power" of the mosquito, quantifies all the mosquito's intrinsic factors that make it more or less attracted to one type of host. These factors may have a genetic basis and are expressed through the physiology and behaviour of the mosquito. As such, when computing for example the HBI, results depend on (i) the number of humans present, (ii) the variations in their "availability" and (iii) the variations in the "biting power" of the mosquito species. The "biting power" for one type of host is composed of two different aspects. The first one is related to the mosquito host preference as defined by Boreham and Garrett-Jones [
22]. The second one is the ability of the mosquito to bite a specific host, and covers two notions that should be distinguished: the biting efficiency and the biting effectiveness.
The biting efficiency corresponds to the ability for one mosquito to encounter vulnerable hosts and is roughly related to the notion of host selection, while the biting effectiveness can be defined as the ability to bite vulnerable hosts and is more depending on the behaviour of hosts and the ability of the mosquito to change to alternative hosts when disturbed.
Based on these new concepts, the "host selection" depends on both the "host availability" and "mosquito biting power". So, mechanisms underlying the computation of the HBI (or other similar index) are numerous and often difficult to quantify. The consequence of the numerous sources of variability is that host-vector contact is far from randomly distributed [
6] and these interactions between genetic and environmental components make the estimation of feeding patterns difficult [
30]. Moreover, mosquito sampling techniques and sampling locations may introduce biais in computations [
31]. For example, fixed sampling places may capture biased subpopulations of mosquitoes [
32]. Sampling resting mosquitoes indoors may over-estimate an HBI. On the contrary, sampling resting mosquitoes too close to a corral with animals may under-estimate an HBI if the mosquito species is not strongly anthropophagic. Because the HBI of a population can only be assessed by field sampling, it is hardly possible to know whether the samples are fully representative.
In Mataral's samples (indoors, outdoors, cave, 200 l-drums etc.), the HBIs ranged from 14–60%, and the true HBI would then depend on a proper combination of the values in the various biotopes. In Mataral, the cave was located at one end of the village and was more favourable to shelter resting mosquitoes which have fed on wandering cattle than on man. On the contrary, mosquitoes collected inside houses during morning resting catches may overestimate the HBI. In the Mataral experiments, the HBI's least biased value may be that calculated on human landings, as mosquitoes were collected at various locations covering the whole village area, and not at fixed sites such as the cave or the 200 l-drums. In that case, the HBI is 44%, larger than the overall HBI of 26.6% which represents a kind of "average proportion fed on man". Taking into account the numerous factors of variation in HBI computation, the HBI estimate for Mataral would lie between 30–50%.
If the mosquito species is not highly anthropophilic, the HBI may vary from one location to another, depending on the proportion of humans and other possible hosts living in the areas and the availability of these hosts. It is not always possible to generalize an estimated value to other areas, even close. For example, in 3 villages of Venezuela HBIs varied 0–50% for each of the anopheline species studied [
33]. In Papua New Guinea HBIs for
Anopheles farauti or
Anopheles punctulatus ranged from 9 – 83% in villages within a 20 km radius [
34,
35]. For
An. pseudopunctipennis, old (and not very accurate) data give estimates ranging from 2 – 85% depending on countries (40–80% in Peru [
13,
14], 50% in Argentina [
10], 67% in Mexico [
12] and as low as 2% in Venezuela [
11]. Despite the lack of precision of these data, the underlying conclusion is that the HBI appears to be variable from one place to another. So it is better to refer to the HBI of a population than to that of a species as a whole.
In the range of distribution of
An. pseudopunctipennis in central Andes of Bolivia, villages are similar to Mataral and one can expect to encounter similar HBI values, i.e. estimates ranging from 30–50%. Even crude, this estimation is far below HBI estimates of efficient malaria vectors such as
An. gambiae or
An. funestus which are commonly >90% [
31]. As such,
An. pseudopunctipennis enters the "opportunistic" category of mosquitoes as opposed to "fixed" species as regard to host preference [
6], which include highly antropophilic species (such as
An. gambiae) or zoophilic species at the other end of the spectrum (such as
An. quadrimaculatus) [
36].
In terms of host preference, accurate interpretation of mosquito blood-feeding patterns cannot be made unless data are also available on densities of the various possible hosts present in the study area. Forage ratios enable converting such data to indices of host preference. They distinguish species which have "fixed" feeding preferences from those which exhibit "opportunistic" behaviour. They point out preferences (and avoidance), and also quantify the degree of opportunistic behaviour and/or the degree to which host specific feeding patterns are obligate. A strictly anthropophilic species would only have "humans" as preferred hosts while an opportunistic species (such as
An. pseudopunctipennis) will exhibit more than one preferred host. The number of significant "preferred" hosts is a measure of the opportunistic behaviour of the mosquito, and the value of the forage ratio quantifies the intensity of preference. The difficulty in using forage ratios is obtaining accurate population estimates of available hosts. However, even with rough population estimates, forage ratios are potentially more powerful indices than other specialized ones [
37] for studying mosquito feeding preferences. Forage ratios may be used to compare situations in different areas (for example the choice of humans by
An. pesudopunctipennis in various villages) or to compare various mosquito species.
In Mataral, experiments carried out with baited mosquito-nets show that "man" is seldom selected in first, except when only pigs, dogs and Chicken are the only other hosts offered (August 2002 experiment). An. pseudopuntipennis does not seem to be strongly attracted to humans. Pigs, chickens and dogs seem to be the less chosen hosts. On the other hand, sheep, goats and donkeys are the most appreciated (followed by "humans" and "cows"). So, as demonstrated by forage ratios, An. pseudopunctipennis make host choices at least in two categories: "preferred hosts" (goats, donkeys, humans, cows) and "avoided hosts" (dogs, chicken, pigs). If An. pseudopunctipennis do make host choices, the intensity of preference depends on what species of hosts are competing: If all the preferred species are absent, An. pseudopunctipennis will feed on whatever host is encountered.
The only experiment when "man" has been chosen as the first host choice was in May 2003 and curiously was opposed to the usually favourite "donkey" and "sheep/goat". This result points out the large variations that may exist in host choice and the likely role of seasonal environmental varying factors. Nevertheless, even in that last experiment, the proportion of mosquitoes attracted to man was only 51.4% (weighted proportion 45.1%) indicating that An. pseudopunctipennis is not very anthropophilic. The opportunistic behaviour of An. pseudopunctipennis is also illustrated by the variations of the HBI values within the sampling locations. In that case, HBIs reflect the availability of the closest preferred host and also that, amongst the "preferred" group, the opportunistic behaviour of An. pseudopunctipennis may be characterized as "first preferred host encountered, first bitten".
Because of the weakly anthropophilic behaviour of An. pseudopunctipennis, the dozens of goats and sheep raised by the villagers in the central Andes that are kept close to houses at night may act as a zooprophylaxis barrier to malaria transmission. De facto, malaria transmission is less active in the Mataral area than in the Yacuiba area (south of Bolivia) where less cattle is raised and where cattle do not sleep close to human habitations (pers. observations). The question is then to quantify in what proportion the ratio humans/animals has to change to significantly diminish the HBI and thus the malaria transmission.
A poor HBI is one of the cue factors that diminish the vectorial capacity of An. pseudopunctipennis. Moreover, the proportion of parous females attracted to humans is not higher than the proportion of nulliparous indicating that transmission risk does not increase when the mosquito moves from one physiological category to the other. Neither seems to increase the proportion of mosquitoes that came back to bite humans when they already have taken a bloodmeal on man during the preceding trophic cycle or during multiple feedings. This proportion is 44%, in the range of the estimated HBI. In that group, the proportion of parous females (i.e., that are potentially dangerous for malaria transmission) is not different in those that came back to humans than in those who bite animals. The population of An. pseudopunctipennis from Mataral does not seem to be separated into two groups, one anthropophilic and the other more zoophilic.
Seasonal shift from feeding on mammals to birds has been reported in some
Culex species [
38‐
42]. In Mataral, with time, there was a slight increase in the monthly HBI values from January to October. In the two sampling situations (cave and human landings), the rate of increase was the same (regression lines were significant and parallel). However these calculations were carried out with small numbers (30 to 70 mosquitoes for each month and for each situation), and so the conclusion may be wrong. If not, the tendency is not well understood, as there is no evident correlation between temperature, rainfall or other environmental parameter variations and the increase in the proportion of blood-feedings on humans rather than on other mammals.
Except with cows, multiple bloodmeals distribution and the high proportion of them are in accordance with the opportunistic behaviour of the mosquito. The over-representation of mixed meals taken on cows are a consequence of the combination of the host preference (high forage ratios) with the few animals present in Mataral. If mixed meals are not from immediate successive meals, their high proportion in Mataral and their high proportion in the sub-population of mosquitoes captured on human landings indicate that the duration of the gonotrophic cycle of
An. pseudopunctipennis may have a large variance and/or that this mosquito (or at least some individuals) may often need more than one bloodmeal to complete egg maturation. Mixed meals originated from interrupted feeding are highly frequent as 51% of the mosquitoes which are in Christopher's stage II or with fresh incomplete bloodmeals when captured had already a bloodmeal from an animal when coming to feed on volunteers. All proportions calculated with patent mixed meals may be under evaluated because cryptic meals were not taken into account in these computations (no attempt has been made to estimate exactly the proportion of cryptic meals). However, data on human landings suggest that maybe >40% of bloodmeals taken on humans could be cryptic. Cryptic meals are of epidemiological interest as they come from feedings on two or more hosts of the same species. Some argue that probing may in fact diminish the probability that a mosquito becomes infected (and then infective) as the total number of gametocytes ingested will be less for a partial meal, and the probability for the same mosquito to complete its meal on another gametocytemic person is low [
35,
43]. However, if the mosquito is infective, probing may increase the probability for one person to become infected. As such, cryptic meals increase the classic definition of vectorial capacity [
22]. The likely high proportion of cryptic meals on humans as calculated from human landings in Mataral is in accordance with other findings [
35]. Research is still needed to quantify the heterogeneity of cryptic meals in one group of hosts (humans for example). Heterogeneity may have a strong impact on transmission dynamics if, for example, mosquitoes are attracted more to one (e. g. gametocytaemic) category than to another one amongst the group [
44,
45].
In the central Andes, villages are dispersed and the human population as well as other mammals populations remain scarce.
An. pseudopunctipennis cannot survive in such habitats if it has a high specialized host preference (to humans for example). Moreover,
An. pseudopunctipennis is also found in unpopulated areas indicating that, without humans, it may feed on other (wild?) mammals. The recent historical settlement of the central Andes suggests that
An. pseudopunctipennis has been in contact with humans for only a few centuries and so still keeps its "opportunistic" (zoophilic-based?) habits. However, when humans are present the mosquitoes feed on them with an HBI of 30–50%. This range of values for
An. pseudopunctipennis in the central Andes area is similar to that of weak malaria transmitting species in other parts of the world [
16]. From the transmission viewpoint, this weakness may be compensated by other parameters such as high female densities that can enhance the vectorial capacity.
An. pseudopunctipennis is well adapted to its varying environment: it is ubiquist in the choice of its larval habitats, is active even at low temperatures and can quickly reconstitute high adult densities.
P. vivax is so far the only human malaria parasite circulating in the central Andes. The extrinsic cycle of the parasite is shorter than that of
Plasmodium falciparum and the parasite presents hypnozoites stages that can remain dormant in the liver of humans for extended periods (months or years) before reactivating and invading the blood. These characteristics enable a weak vector such as
An. pseudopunctipennis to activate a transmission cycle, even during short periods, as soon as ecological and population conditions are adequate.
The pair
An. pseudopunctipennis/P. vivax demonstrate a good level of co-adaptation in the central Andes. Weak transmission points of the mosquito (such as a low HBI) and fluctuating environmental conditions are "compensated" by intrinsic biological factors of the parasite. Malaria in the central Andes is then unstable in the sense of Macdonald [
4], with determining causes such as low vector antropophily and climatic conditions favourable for short periods of transmission. Due to difficulties in estimating all the components of the vectorial capacity (including the HBI), the risk of malaria transmission is often measured with more simple indexes for which the estimation of the HBI is not needed. For example, the entomological inoculation rate (EIR), defined as EIR =
m.a.s, where
m and
a are as usual, and
s is the sporozoite rate (proportion of mosquitoes with sporozoites) is often used. The product
m.a may simply be estimated by human bait collections and in itself may serve as a rough index of transmission risk. In combination with the proportion
P of parous females, the measurement of
m.a.P (the number of parous females biting per human per night) gives also an interesting index of transmission, as well as
m.a alone in some situations. In those cases, the direct estimation of
a is not necessary. However, the understanding of mosquito feeding habits, host choices and the computation of the HBI (even crude), are essential to know the conditions under which a mosquito may be distracted from its preferred hosts. Some vector control techniques such as zooprophylaxy [
4,
24,
46,
47] or inside residual insecticide spraying [
48] have an impact on the HBI. The degree of variation may be one of the possible parameters to investigate in the follow-up of transmission patterns. To achieve this, adequate sampling techniques should be implemented in time and space to estimate the HBI with precision and distinguish confounding factors that may affect its estimation [
49]. Samples should be taken from various sites, avoiding the introduction of bias due to collections too close from specific sources of blood. The geographic distribution of engorged resting females in the field is a cue factor. Unfortunately, this distribution is not random, even after some time following the bloodmeal enabling mosquitoes to move off the blood sources and disperse in nature. The HBI should then be computed using a weighted mean taking into account all the selected sites of the sample. Weighting should take into account what proportions of the blood-fed females of the species are resting in each type of habitat, which may be difficult to assess. For that reason, the estimated HBI for
An. pseudopunctipennis in the central Andes was estimated with a range and not as an exact figure.
Authors' contributions
LF Designed the study, carried out the field work, taught LP to process blood-fed mosquito samples with ELISA, analysed the data and wrote the manuscript.
LP Processed the blood-fed mosquito samples (ELISA)
BB Helped in designing the study and in collecting mosquitoes in the field.
CT Helped in the field work and in processing ELISA samples.
All authors erad and approved the final manuscript.