Background
The main driver of early behavioural resistance in many malaria vectors globally was the extensive reliance on DDT-based indoor residual spraying (IRS). Widespread use of DDT led to modification of behaviour of several vector species that had previously taken blood meals indoors and rested indoors during egg development (endophagy and endophily, respectively), to mostly indoor feeding/outdoor resting (endophagy, exophily), to avoid insecticide exposure [
1,
2]. Over the past ~ 10 years, the most effective vector intervention has been long-lasting insecticidal nets (LLINs), which, alone or in combination with IRS, and together with rapid diagnosis and treatment and combination drug therapy, have reduced malaria such that elimination is being considered feasible [
3‐
5]. Despite these advances, primary reliance on LLINs and IRS for vector control has driven physiological resistance to insecticides [
6,
7], and behavioural resistance or resilience (e.g., increased exophagy and early evening or daytime biting). Such behavioural modifications have enhanced residual transmission, or transmission that persists despite the reduction of vector populations through control activities [
8], including outdoor and non-night-time transmission, in several malaria-endemic areas [
9‐
16], although there are counter-examples [
17‐
19].
It is unclear whether shifting biting behaviour after exposure to LLINs and/or IRS is the result of genetically differentiated populations of anophelines with different feeding behaviours, or of behavioural plasticity, the ability of individuals of the same genotype to adopt different behaviour in response to different environments. Insecticides commonly used for LLINs and IRS have been reported to exert spatial repellent as well as insecticidal effects on anophelines [
20]. If anophelines are deterred by the presence of insecticide from feeding successfully at their preferred time or location, they may continue to quest for a blood meal, shifting the overall biting behaviour of the population [
13]. If shifts in biting behaviour are instead caused by replacement by genetically different anophelines, these interventions may no longer be as effective against vector populations that have become behaviourally resistant [
12]. Evidence for genetic differentiation of anophelines by biting and resting behaviour has been inconclusive: studies have found chromosomal inversion frequency differences between exophagic and endophagic
Anopheles gambiae s.l. [
21,
22] and exophilic and endophilic
Anopheles funestus [
23]; yet studies comparing single nucleotide polymorphisms (SNPs) in putative circadian clock genes between exophagic, endophagic, early and late feeding
Anopheles arabiensis [
24], and whole genome SNPs between exophilic and endophilic
An. arabiensis [
25], detected no genetic differentiation. Understanding whether distinct sub-populations of malaria vectors with different biting behaviour exist could help predict responses to vector control measures and advise vector control programmes as to the most efficient use of their resources.
Except for Venezuela, which in 2015 accounted for an estimated 30% of malaria cases in the Americas [
26], Latin American countries have reduced the malaria case load substantially during the last 10 years [
26], mainly through rapid case detection and treatment. Generally in this region there is lower coverage of LLINs and IRS compared with endemic areas in Africa and Asia [
27‐
29]. In Peru, the northeastern Loreto Department reports most of the total malaria cases, with an estimated 80% of malaria cases caused by
Plasmodium vivax [
26,
30]. Transmission is seasonal (mainly rainy season, January to June), linked to river levels and mosquito abundance [
31‐
34]. Between 2005 and 2010 in Loreto, the Global Fund’s PAMAFRO (Spanish acronym) initiative strengthened malaria diagnosis and case management, and distributed LLINs, achieving high local coverage, estimated at 98.7% 1 year after distribution in a sub-set of targeted communities [
35] and a remarkable monthly malaria case incidence rate below 1/1000 in Peru in 2010–2011, compared to 6–7/1000 in 2005–2006 [
36]. However, between 2010 and 2015 (post-PAMAFRO period), there were no widespread LLIN distributions in Loreto, and overall case numbers and the proportion of
Plasmodium falciparum cases increased [
26,
29,
36]. Furthermore, due to the reduced malaria incidence rate, the Peruvian Ministry of Health (MOH) shifted its focus on vector control intervention towards new arboviral outbreaks following the PAMAFRO initiative [
36].
The main malaria vector in Latin America,
Anopheles darlingi, dominates several regions of the Amazon Basin, accounting for > 85% of the anopheline fauna feeding on humans, and much of the malaria transmission, particularly in frontier zones [
33,
37‐
39]. It has successfully invaded human-modified habitats, such as fish ponds, agricultural settlements, highways, mining sites and urban areas [
40‐
44]. This species is behaviourally very plastic, displaying mainly exophily with some reports of endophily (reviewed in [
45]), and both endophagy and exophagy (reviewed in [
46,
47]), depending on region, season and local environmental variables such as bed-net coverage, house type, and host number and availability [
48,
49]. In Amazonian Peru, there are regional records of both endophagic and exophagic behaviour in this species [
31,
33,
50]. A previous investigation in peri-Iquitos from 2011 to 2012 found many more exophagic than endophagic
An. darlingi, although rigorous longitudinal assessment of biting behaviour was not the main focus of that study [
34]. To investigate whether there was a modification in
An.
darlingi’s feeding behaviour following the end of the PAMAFRO initiative, the present study was designed to quantify the abundance of exophagic
versus endophagic
An. darlingi from 2013 to 2015, especially during the 6-month transmission season (~ January–June). Additionally, to test the hypothesis that there are distinct sub-populations of
An. darlingi with different biting behaviour in peri-Iquitos, a sub-set of collected
An. darlingi were genotyped using nextRAD (nextera-tagmented, reductively amplified DNA) genotyping-by-sequencing to compare individual exophagic and endophagic mosquitoes biting at different times.
Discussion
This study detected, across 3 years, a shift in
An. darlingi in LUP and CAH towards decreased exophagy. This shift occurred between distributions of LLINs in these villages; the most recently distributed LLINs in CAH were a year old and in LUP were over 2 years old by the start of this study, and LLINs in both villages were at least 3 years old (the expected lifetime of LLINs [
51]) by the end of the study. It is therefore possible that this shift represents a return to baseline biting behaviour in these villages following a previous shift towards increased exophagy driven by LLIN exposure.
IRS was conducted in CAH and LUP sporadically during this study. However,
An. darlingi is known to rest mainly outdoors following blood feeding [
47], so it is unlikely that IRS is effective against this vector. In this study, there was not a consistent effect of IRS on the
An. darlingi abundance or exophagic:endophagic ratio in either the month IRS was conducted or the following month in these two villages (Fig.
2).
This study further confirms the heterogeneous biting behaviour of
An. darlingi [
34,
43,
47,
77‐
79]. A range of local environmental or ecological changes can influence the ratio of exo/endophagic
An. darlingi. For example, in a gold mining area in Venezuela,
An. darlingi was found to be significantly more exo- than endophagic, attributed mainly to the location of villages within forested areas, and to houses with incomplete walls [
42]. A similar pattern of high exophagy/low endophagy in
An. darlingi was detected along a corridor of a highway deforested for power line construction in Porto Velho, Rondonia state, Brazil [
80].
Shifts in behaviour in vector anophelines towards increased exophagy following LLIN distribution are relatively common [
9‐
11,
14]. In these studies, such shifts were seen within 1 year after LLIN distribution, or during periods of high LLIN usage. Shifts in biting behaviour can result from physiological (insecticide-induced) or behavioural resistance, which can be difficult to distinguish [
12]. Because there is little documented physiological resistance in
An. darlingi, including populations sampled for this study [
81], the documented exophagic/endophagic shift may be evidence for behavioural resistance that emerged as a result of LLIN usage. However, some previously reported shifts may be the result of changes in species composition, or plasticity in feeding responses [
2,
13]. For example, using mark-release-recapture, individual
Anopheles farauti were found to feed both outdoors and indoors [
19].
In the current study, using 1021 genome-wide SNPs, there was no evidence of genetic differentiation between exophagic and endophagic An. darlingi, or among An. darlingi biting at different times during the night. These results were consistent across both model-based (STRUCTURE) and non-model-based analyses (PCA/DAPC). In addition, thorough exploration of the parameter space, by changing the exclusion criteria for low coverage samples and the STACKS settings for filtering of loci, consistently returned evidence that all samples belonged to a single homogeneous population. The lack of detectable population structural differences among the An. darlingi from this study suggests that the reported shifts in biting behaviour are due to behavioural plasticity resulting from reduced spatial repellence of aging LLINs, rather than genetically differentiated populations of exophagic and endophagic An. darlingi. However, the lack of population genetic structure does not preclude a genetic basis for changes in An. darlingi biting behaviour. It is possible that the methods used were unable to detect smaller-scale genetic differences between exophagic and endophagic An. darlingi that do not influence the overall genetic structure.
Previous studies have found microgeographic genetic differentiation between
An. darlingi by habitat [
52] and season [
77]. In addition, a recent study using whole-genome SNPs found genetic differentiation between
An. darlingi collected in two rural Brazilian villages ~ 60 km apart, which had experienced different levels of deforestation [
82]. Although these Brazilian villages were approximately the same distance apart as the villages in the current study, it is not surprising that there was no evidence of genetic structure between
An. darlingi collected in LUP and CAH, because both are riverine villages with similar ecological characteristics [
34,
52].
Across its broad distribution,
An. darlingi populations exhibit a wide range of peak biting times and patterns (unimodal, bimodal, trimodal, no peak) [
47,
83,
84]. Furthermore, in a study in Amapá state, northern Amazonian Brazil, Voorham [
85] found intra-population variation of biting activity in
An. darlingi to be as high as inter-population variation. Some variation is attributed to seasonality [
50,
77,
86], and some is assumed to be the result of interaction between local ecological and endogenous factors [
45]. The current study determined that in the peri-Iquitos area, more
An. darlingi were biting before midnight than afterwards, especially early in the evening, in agreement with observations in previous studies in Peru [
32‐
34] and some regions of Brazil [
43,
78]; although another Brazilian study found
An. darlingi biting throughout the night [
84]. A preponderance of early evening biting is likely related to the availability of humans as hosts while they are engaged in various activities prior to retiring under bed nets.
There was a second, smaller biting peak around 02.00 in all years and villages except 2013 in the exophagic individuals from LUP (Fig.
3). Although there were no genetic structural differences by biting time, it is possible that there are individual genes determining biting time in the population that do not influence the overall genetic structure (as investigated in [
24]). Alternatively, this additional peak may be the result of phenotypic plasticity in biting behaviour within the population.
This study shows similar overall patterns in abundance, HBR, and biting behaviour of
An. darlingi for LUP and CAH between 2013 and 2015. That the two communities are ~ 60 km apart and located on different rivers suggests that these populations may respond to some types of regional environmental or anthropogenic change as a single metapopulation. In 2012 [
34], peak monthly HBRs of exophagic populations of
An. darlingi were similar in LUP and CAH, and by 2015, they had declined similarly. In addition, the changing ratio of exophagic:endophagic
An. darlingi from year-to-year is quite congruent (Fig.
4). However, as the negative binomial regression results demonstrate, local context strongly influences the patterns of abundance in these populations of
An. darlingi. In summary, there is evidence for both metapopulation and local population behavioural patterns in
An. darlingi in Loreto, but the mechanisms involved have not yet been identified.
The aggregated monthly HBR of
An. darlingi (exophagic, endophagic and both together) was significantly correlated with monthly malaria case numbers in LUP and CAH. This is particularly interesting because cases of
P. falciparum have increased since 2015, especially in LUP, representative of a wider regional trend in Loreto [
29,
87], whereas
P. vivax cases peaked in LUP in 2014, and CAH experienced a major
P. vivax outbreak in 2012–2013 after which cases have fluctuated considerably (Additional file
1: Fig. S2). These data confirm that the HBR and malaria incidence are highly related, though it is clear that other parameters used in the calculation of vectorial capacity, such as vector survival rates [
88], are also valuable for predicting malaria risk.
Although in the present study the overall numbers of infected An. darlingi (n = 30) were insufficient for statistical analysis, there were endophagic and exophagic An. darlingi infected with both P. vivax and P. falciparum throughout the rainy season (January–June), before and after midnight. Thus, during the rainy season, villagers are at risk of malaria infection both inside their houses and in the peridomestic area throughout the night.
An explanation for the decline in the rainy season
An. darlingi population sizes in LUP and CAH over time is elusive. A massive flood in Loreto in April 2012 [
89], attributed mainly to an early La Niña event [
90], may have influenced survival or population dynamics, perhaps by destruction of breeding sites. Less massive flooding in Suriname, combined with several vector interventions and malaria case management, reduced malaria incidence to near zero, in a region where
An. darlingi was the principal vector [
91]. In the village of LUP, there was no discernible immediate effect of the flood on the peak HBR in April–May between 2011 (before the flood) and 2012 (immediately after the flood [
34]), although longer term effects cannot be ruled out.
Modelling has demonstrated that
P. falciparum is likely more sensitive than
P. vivax to changes in malaria vector survival rates due to longer sporogonic cycle duration [
88]. If
An. darlingi survival rates had been measured over the 3 years of this study, the relationship between survival and the increase in
P. falciparum cases could have been investigated. In addition, as this study analysed only
An. darlingi collected by HLC, it is possible that a sub-population of
An. darlingi, not sampled in this study because it feeds mainly on animals, also contributes to malaria transmission. Although a previous analysis of blood meal sources in resting
An. darlingi in LUP, CAH and SEM found that the majority of mosquitoes tested had fed on humans (human blood index (HBI): 0.58–0.87), with a similar infection rate to that found in the current study (0.42%) [
49], a substantial proportion of
An. darlingi feed on non-human hosts in this region. It is unknown whether
An. darlingi feeding on different hosts are genetically distinct, as has been demonstrated in
An. arabiensis in Tanzania [
25]. Another limitation of this study was the sporadic collections at SEM, which prohibited statistical and genetic comparisons with LUP and CAH. Finally, this study did not determine whether LLINs were used or IRS was conducted in the individual houses in which collections were conducted. However, collections rotated between different houses on different nights throughout the study in an effort to obtain an unbiased, representative sampling of
An. darlingi biting behaviour in the villages during the study period.