Species composition and blood feeding patterns of understudied Anopheles mosquitoes in southern Zambia, an area of low malaria transmission

Abstract

Background

Understanding the composition, host feeding patterns, and infection status of all potential vectors is essential to understand changing disease transmission dynamics and inform future vector control strategies. This study assessed anopheline species composition, abundance, distribution, and host feeding preferences in Southern Province, Zambia to identify potential contributors to residual malaria transmission beyond primary vectors.

Methods

Adult mosquitoes were collected along a 100 km transect between the districts of Pemba, Monze and Gwembe in Southern Province, Zambia, an area of low malaria transmission. Mosquitoes were sampled from 12 randomly selected households at 5 km intervals along the transect. Centers for Disease Control and Prevention light traps were used to collect adult mosquitoes indoors and outdoors. Mosquitoes were identified using morphological keys and molecular approaches. A multiplexed PCR assay was utilized to identify mammalian hosts and Plasmodium falciparum (Pf) sporozoite infection was determined using an enzyme linked immunosorbent assay.

Results

A total of 5697 female anopheline mosquitoes were collected from both indoor and outdoor traps, revealing a diverse range of species. Nine species of Anopheles mosquitoes were identified, dominated by Anopheles rufipes (n = 1527, 26.8%), and An. pretoriensis (n = 1427, 25.1%). Host blood meal analysis of 401 visibly blood-fed mosquitoes revealed that Anopheles species in the study area fed primarily on goats (n = 176; 62.5%), humans (n = 55; 19.4%) and cows (n = 29; 10.2%). The human blood index across all captured anophelines was estimated at 0.22. Human blood meals were detected in secondary malaria vectors, with notably high HBI values of An. pretoriensis and An. rufipes mosquitoes captured outdoors near goat pens. Out of the 5697 female anophelines tested for sporozoite infectivity, one An. rufipes specimen was positive for P. falciparum.

Conclusion

This study demonstrates that malaria transmission risk in the surveyed areas may extend beyond well-recognized primary vectors. Species that are not well studied, often exophagic and assumed to be zoophagic, were found to have fed on humans and the study showed some preliminary evidence of P. falciparum infectivity. These findings indicate that peri-domestic human activity may sustain residual transmission despite high Insecticidal-Treated Nets and indoor residual spraying coverage.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12936-026-05799-y.

Background

Zambia has made steady progress in reducing the burden of malaria since the early 2000s. The combined approach of vector control interventions using insecticide treated nets (ITNs) and indoor residual spraying (IRS), improved case management with rapid diagnostic tests (RDTs), artemisinin-based combination therapy (ACTs), reactive test-and-treat strategies and mass drug administration (MDA), have resulted in a 74% reduction in malaria-related deaths between the years 2000 and 2018 in Southern Province [1–5]. Since 2018, malaria incidence rates have remained low in Southern Province [5]. In 2021, the prevalence by RDT for children under the age of five years was approximated to be 3%, the lowest in Zambia and the closest to potentially achieving elimination nationally [6].

Understanding local malaria vector composition is crucial for targeted vector control. This makes routine vector surveillance critical to track and respond to changes in the composition and behavior of vector species. Longitudinal shifts in vector species composition or behavior are inevitable as malaria vectors respond to the selective pressure from the use of insecticides [7–10]. These changes are also influenced by the development of insecticide resistance [11], climatic change [12], and human behavior including those of mobile populations [8, 13, 14]. For example, historically, the recognized malaria vectors in the southern region of Zambia were Anopheles arabiensis [6, 15] and An. funestus sensu stricto [9, 10]. These species were documented to be highly anthropophilic and endophagic [15, 16] making them efficient malaria vectors and, at the same time, susceptible to indoor-based vector-control interventions [8, 17]. However, following the sustained use of ITNs and IRS, combined with reduced rainfall and droughts, declining densities of An. arabiensis and near-complete decimation of An. funestus populations have been reported in Southern Province [6, 15]. The decrease in the abundance of these primary vectors has changed the malaria vector landscape, creating ideal ecologies for potential secondary vectors to potentially contribute to residual malaria transmission [18].

Over the years, several species that have received limited attention such as An. squamosus s.l., An. coustani s.l. and An. rufipes, have been implicated as potential vectors of malaria in Southern Province [19, 20]. Further, since many of these anophelines are zoophagic, exophagic and/or exophilic, they may evade indoor-directed interventions such as IRS or ITNs and potentially contribute to sustaining residual malaria in the province [19, 22].

This study was conducted in three administrative districts of Southern Province characterized by low sustained malaria transmission. The objectives were to assess anopheline species composition, relative abundance, spatial distribution, and host feeding preferences and evaluate sporozoite infection status of collected mosquitoes. By characterizing the ecology and behaviour of understudied Anopheles species, this study aims to improve understanding of residual malaria transmission and inform vector surveillance and control strategies appropriate for settings approaching elimination.

Materials and methods

Study area

The study was conducted in Pemba, Monze and Gwembe districts, in Southern Province, Zambia (Fig. 1). All districts experience three distinctive seasons: a warm and wet season from November to March, a cool and dry season from April to August, and a hot and dry season from late September to November. Temperatures typically range from 14 to 28 °C with the warmest temperatures occurring from October to December and the lowest between May and June. Malaria transmission peaks between March and May annually.

Fig. 1.

Map of Zambia showing the locations of Southern Province (top left), Choma, Pemba, Monze and Gwembe districts (bottom left) and households that were used for entomological collection along the transect (in red, right)

Study design

A cross-sectional sampling design was employed and was conducted between May and September 2018. Anopheles funestus had historically been reported in the districts of Pemba, Monze, and Gwembe [9], but not in Choma District. To identify potential residual populations of An. funestus, a transect was established extending from Choma District—where extensive entomological studies had been conducted—to the edge of Lake Kariba, crossing three districts. The first sampling point was at the Mutama River along the Choma-Pemba boundary, and the last sampling point was at of the shoreline of Lake Kariba (Fig. 1).

Sampling was conducted at 5 km intervals along this transect. At each point, 12 households accessible from the roadway were randomly selected, all within 5 km radius of the most central house. Consenting household heads were interviewed for risk factors for malaria including house construction, presence of animals, and number and age of household residents. Households with no identifiable sleeping structure and households with no adults over the age of 16, the minimum legal age for obtaining written consent in Zambia, were not considered for inclusion in the study. Houses sampled were those along or close to roadways to facilitate access and so may not have captured the full diversity of house types in the districts.

Mosquito collections

Miniature Centers for Disease Control and Prevention light traps (CDC-LTs) (Model 512, John W Hock, Florida, USA) were used to collect adult mosquitoes in consenting households. Collections were conducted for a single night, during which two CDC-LTs were deployed at each household. The first trap was set indoors, suspended 1.5 m from the ground. The trap was set adjacent to beds with mosquito nets and positioned at the foot end. The second trap was set 1.5 m from the ground outdoors near livestock pens (goat, pig, and/or cattle) where these were available within the premises of the householder. When no animal pen was present, only one indoor trap was used. Traps were switched on in the evening at 1800. At 0600 in the morning the household owners tied the trap collection cups. Collection cups and traps were transported to the nearest health facility, where the cups were placed in a freezer at− 20 °C for approximately 30 min to kill the collected mosquitoes. Mosquitoes were then placed in labelled Petri dishes and transported to the Macha Research Trust (MRT) laboratory for processing.

Morphological identification

At MRT, all Anopheles mosquitoes were morphologically identified using taxonomic keys [22] and the physiological status of each female was noted as either unfed, fed or gravid. All morphologically identified Anopheles mosquitoes were then individually placed in 0.6 mL microcentrifuge tubes containing silica gel desiccant (Fisher Scientific, Waltham, Massachusetts, USA) and cotton wool and stored at room temperature (averaging between 15 and 26 °C) for further molecular analysis. Mosquitoes were stored for less than five days before molecular analysis to facilitate weekly processing due to the large number of mosquitoes collected at some sites which could not be processed entirely at once. All culicine mosquitoes were counted and discarded.

DNA extraction and PCR amplification for species identification

Mosquitoes were dissected into two parts: the head and thorax combined, and the abdomen. DNA was extracted from abdominal tissue using a modified salt extraction method [23]. Members of the An. funestus group and An. gambiae complex were further identified to sibling species level by PCR [24, 25]. For the molecular analyses, a subsample of the total catch of Anopheles gambiae complex mosquitoes were used, by randomly selecting up to 10 An. gambiae s.l. specimens from those caught at every sampling point and trap location. Where fewer than 10 An. gambiae s.l. mosquitoes were collected, all specimens were processed. For An. funestus group, all mosquitoes were processed molecularly to maximize the possibility of detecting any residual An. funestus populations, following the sizeable reduction of this vector in Southern Province.

Specimens that did not produce an amplicon with either the Gambiae or Funestus group assays and those initially not identified morphologically (e.g. due to missing wings, legs, maxillary palps, abdomen, or overall damage [26]) were identified molecularly using the internal transcribed spacer-2 (ITS2) ribosomal-DNA PCR (ITS2-PCR), which produces amplicons of varying band size depending on the mosquito species [19, 27]. The ITS2-PCR assay of An. maculipalpis, An. pretoriensis and An. rufipes all resulted in a 500 base pair amplicon [20]. Thus, any specimen that was not morphologically identifiable and resulted in a 500 bp ITS-2-PCR amplicon was considered as An. maculipalpis/pretoriensis/rufipes.

Blood meal source identification and Plasmodium falciparum detection

Blood meal sources were determined by extracting the DNA from the abdomens of visibly engorged mosquitoes and hosts identified using a previously described multiplexed PCR assay [23]. All collected female Anopheles mosquito head/thoraces were tested for Plasmodium falciparum circumsporozoite proteins (CSPs) using a standard enzyme-linked immunosorbent assay (ELISA) [28]. ELISA kits were obtained from BEIR Resources through the Malaria Research and Reference Reagents Resource (MR4) Centre. Sporozoite infection rates were determined separately for mosquitoes caught indoors and outdoors, and by species.

Statistical analyses

All data were entered and stored in an Excel spreadsheet and exported to R version 3.5.1 for statistical analysis [29]. Descriptive statistics were used to summarize the mean catches per trap per night and proportions of mosquitoes caught indoors, and outdoors. The human blood index (HBI) and sporozoite infection rate (SIR) were calculated as an estimate of malaria transmission risk. HBI was calculated as the proportion of mosquitoes identified to have a human blood meal from the total mosquitoes that were confirmed to have blood fed i.e. those specimens for which a host amplicon was produced following the multiplexed PCR [30–32]. Mixed animal and human blood meals were included in these calculations. Sporozoite infection rates were calculated as the proportion of Anopheles mosquito specimens testing positive for sporozoites of the total number of mosquitoes examined for sporozoites [30].

To quantify the effect of housing structure on the density of indoor Anopheles generalized linear models were employed. First, bivariate tests were conducted to explore the variables that were significantly associated with indoor anopheline counts. The non-parametric Kruskal Wallis, Mann–Whitney U and Spearman Rank correlation tests were used to select which categorical variables to include in the models. The level of significance was set at 0.05. Factors included in the bivariate tests were roof type, eave type, whether IRS had been conducted in the past six months, the number of ITNs hanging at the time of collection, presence or absence of fire burning in the house and the number of people sleeping in the house, due to the reported importance of these factors in affecting indoor mosquito count [33]. Wall type was not included in the tests as only one wall type (brick) was found during collections. The model was fitted using the glm function in R [29]. For the mapping of female anophelines both indoors and outdoors count data for all anopheline species were pooled at each sampling point. All maps were generated using ArcGIS (Environmental Systems Research Institute, Redlands, California, USA).

Results

Mosquitoes were caught across 26 sampling points along the transect, which comprised 314 indoor collections, and 272 outdoor collections. Of these outdoor collections, 204 were set in goat pens, 66 in outdoor cattle kraals and two in pig pens.

Species composition

A total of 5697 female anopheline mosquitoes were collected across all locations and traps. Overall, nine species of Anopheles mosquitoes were morphologically and molecularly identified; An. rufipes (n = 1527, 26.8%), An. pretoriensis (n = 1427, 25.0%), An. gambiae s.l. (n = 421, 7.4%), An. longipalpis (n = 366, 6.4%), An. funestus group (n = 356, 6.2%), An. coustani (n = 277, 4.9%), An. squamosus (n = 276, 4.8%), An. pharoensis (n = 31; 0.5%) and An. maculipalpis (n = 15; 0.3%). A total of 1001 (17.6%) specimens could not be identified morphologically due to missing wings, legs, maxillary palps or abdomen.

Collectively, there were significantly more mosquitoes collected in goat pens (n = 5171/5697; 90.8%) compared to indoors (n = 367/5697; 6.4%) (indoors (χ² = 14.8, df = 1, P > 0.001) or in cattle kraals (n = 159/5697; 2.8%) (χ² = 12.3, df = 1, P > 0.001). Figure 2 shows the proportions of the different species within the trapping locations.

Fig. 2.

Proportions of Anopheles species associated with trap location from across the entire study. The number in parenthesis indicates the total number collected

As described in the methods, due to the large sample size and resource limitation, species-group PCRs were performed on a random subsample female anophelines morphologically identified as those from the An. gambiae complex; 229 of 421 (54.4%). Of the 229 An. gambiae s.l. speciated using the Gambiae PCR, 81% (n = 187) successfully produced amplicons, revealing two species within the complex; An. arabiensis (n = 111/187, 59.4%) and An. quadriannulatus (n = 76/187, 40.6%).

Species-specific PCRs run on mosquitoes morphologically identified morphologically as of the An. funestus group resulted in the successful amplification of 164 of the 356 specimens (46.1%) revealing four sub-species; An. parensis (n = 92/164; 56.1%), An. leesoni (n = 53/164; 32.3%), An. rivulorum (n = 13/164; 7.9%) and An. rivulorum-like (n = 6/164, 3.7%). Indoor and outdoor collection locations of specimens within the An. gambiae complex and An. funestus group are shown in supplementary Figure S1 and S2, respectively.

Of the 1001 mosquitoes that could not be identified morphologically, a random subsample of 577 (57.6%) mosquitoes was processed for molecular identification using the ITS2-PCR. Of these, 292 (50.6%) produced a 500 bp amplicon, corresponding to An. maculipalpis/pretoriensis/rufipes. The remaining samples produced non-specific amplicons or failed to produce a PCR product (n = 153). Forty-two samples that failed to produce amplicons on the Gambiae-specific and 192 on the An. funestus specific PCRs and were re-analyzed on the ITS2-PCR. No samples previously ran on the Gambiae-specific PCR amplified, whilst 81 (42.2%) that failed the An. funestus specific PCR were identified as An. longipalpis. The remaining 111 did not produce any amplicons.

Blood meal host analysis

A total of 401 anopheline mosquitoes were visually assessed to be blooded. Blood meal hosts were successfully identified by PCR from 289 of these (72.1%) specimens. Of the 289, 273 (94.5%) identified blood meals came from traps set in goat pens, eight from inside the house and eight from cattle kraals. Five different vertebrate hosts were identified, human (Homo sapiens), goat (Capra hircus), cattle (Bos taurus), pig (Sus scrofa domesticus), dog (Canis lupus familiaris) and chicken (Gallus gallus domesticus). Tables 1, 2 and 3 show blood meal sources from mosquitoes caught at goat pens, cattle kraals and indoors respectively. Of the identified blood meals, goat was most common (n = 176; 62.5%), with humans as the second most frequently detected vertebrate blood meal (n = 55; 9.4%), followed by cattle (n = 29; 10.2%).

Table 1.

Blood meal sources from Anopheles mosquitoes caught at goat pens

Blood meal sources	An. pretoriensis	An. rufipes	An. coustani	An. longipalpis	An. gambiae sl	An. funestus group	An. squamosus	ITS2 500 bp	Total	Proportion (%)
Single host blood meal
Human	30	21	0	1	0	0	0	0	52	19.0%
Goat	85	58	18	10	1	2	1	0	175	64.1%
Cattle	7	5	5	1	2	0	0	2	22	8.1%
Pig	5	0	0	0	0	0	0	0	5	1.8%
Chicken	0	2	0	0	0	0	0	0	2	0.7%
Dog	0	0	0	1	0	0	0	0	1	0.4%
Mixed blood meal
Human/animal1	4	0	1	0	0	0	0	2	7	2.6%
Cattle/goat	5	3	0	1	0	0	0	0	9	3.3%
	136	89	24	14	3	2	1	4	273

Samples producing amplicons of 500 bp on the ITS2 PCR could be An. rufipes, An. maculipalpis or An. pretoriensis

¹“Mixed blood meals refer to mosquitoes that fed on more than one host species, which may include humans and animal hosts (cattle and/or goats)”

Table 2.

Blood meal sources from Anopheles mosquitoes caught at cattle kraals

Blood meal sources	An. pretoriensis	An. gambiae s.l	An. coustani	An. longipalpis	Total	Proportion
Cattle	1	1	4	1	7	87.5%
Human/Cattle	1	0	0	0	1	12.5%
Total	2	1	4	1	8

Table 3.

Blood meal sources from Anopheles mosquitoes caught indoors

Blood meal sources	An. pretoriensis	An. rufipes	An. coustani	An. longipalpis	Total	Proportion
Human	0	2	1	0	3	37.5%
Goat	0	1	0	0	1	12.5%
Cattle	1	1	0	1	3	37.5%
Dog	0	0	0	1	1	12.5%
Total	1	4	1	2	8

Human blood index

The overall HBI for mosquitoes from all collections in the three districts was estimated at 0.22. Species-specific HBIs are shown in Table 4. Anopheles rufipes and An. pretoriensis had identical human blood indices (0.25). Of the 63 blood meals containing human DNA, 93.7% (59/63) were from mosquitoes captured in CDC-LTs placed at goat pens (Table 1). In contrast, only 4.8% (3/63) of mosquitoes collected inside homes contained human blood meals (Table 3).

Table 4.

Human blood index for Anopheles mosquitoes¹

Species	Number analyzed for blood meal	Number positive for human blood meal	Human blood index
An. pretoriensis	139	35	0.25
An. rufipes	93	23	0.25
An. coustani	29	2	0.07
An. longipalpis	17	1	0.06
ITS2 500 bp (An. rufipes, maculipalpis or. pretoriensis)	4	2	0.5
Total	282	63	0.22

¹Mixed blood meals with human blood included in calculation

Sporozoite infectivity rate

Circumsporozoite protein (CSP) ELISA was performed on the head/thorax of all female anopheline mosquitoes collected, to detect the presence of P. falciparum (Pf) sporozoites. Of the 5697 female anopheline processed, only a single specimen of An. rufipes, collected outdoors at a goat pen, was positive for Pf sporozoites, to give an overall sporozoite rate of 0.018% for all specimens tested and a sporozoite rate of 0.07% for An. rufipes.

Spatial distribution of Anopheles mosquitoes

Anopheles mosquitoes were collected at all but one of the 26 sampling points along the 100 km transect. Across all sites, An. rufipes and An. pretoriensis represented the dominant species (Fig. 3). The highest number of mosquitoes were collected at Nabuyuni (n = 1129, 19.8%) in Pemba. The sampling point with no mosquitoes collected either indoors or outdoors was Chawama (within Monze District, shown by a single yellow dot on the map). More mosquitoes were found in Gwembe District starting from Musisa through to Masua. Figure 3 illustrates the anopheline species composition and numbers collected at each sampling point along the transect.

Fig. 3.

Map of Anopheles species composition at each sampling point across collection sites in Pemba, Monze and Gwembe Districts. “Other” in the map legend refers to a combination of An. pharoensis and An. maculipalpis which were found in smaller numbers. “Unidentified” were mosquitoes that could not be identified morphologically. Background was obtained freely from Google Maps

Effects of housing characteristics on mosquito counts

Results of the six housing characteristic variables included in the bivariate analysis (roof type, open/closed eaves, whether IRS was conducted in the past six months, number of ITNs hanging at the time of collection, whether a fire was burning in the house and the number of people sleeping in the house) are shown in a supplementary file (Table S1). Following backward stepwise selection, three variables-“roof type”, “fire burning indoors during the night of collection”, and “IRS conducted within the previous six months”- were retained in the final negative binomial regression model. Results are shown in Table 5.

Table 5.

Negative binomial regression results for household factors associated with indoor adult Anopheles counts

Variable	n	IRR*	95% CI	p-value
Roof type: thatch vs iron	219	2.23	1.17–4.27	0.010
Fire burning indoors (yes vs no)	149	2.17	0.70–6.77	0.182
IRS conducted (yes vs no)	149	0.75	0.40–1.41	0.365

^*IRR = Incidence Rate Ratio (IRR)

IRR > 1 indicates higher mosquito counts; IRR < 1 indicates lower counts relative to the reference category. Model adjusted for relevant covariates. Reference categories: roof type = iron sheet, fire burning = no, IRS = no

Among the variables analyzed, only “roof type” was significantly associated with mosquito counts. There was a 2.23 times higher incidence rate ratio (IRR) associated with those houses with thatched roofs compared to those with iron roofs (95% CI 1.17–4.27, P = 0.01)”. Fire burning indoors showed a potential association with higher counts of mosquitoes (IRR = 2.17), but this was not statistically significant (95% CI 0.70–6.77, P = 0.182) likely reflecting the limited sample size and therefore wide confidence intervals. IRS treatment was associated with a non-significant 25% reduction in anopheline counts (IRR = 0.75, 95% CI 0.40–1.41, P = 0.365).

Discussion

This study provides evidence that malaria transmission in the surveyed Zambian districts may involve mosquito species traditionally considered secondary or of limited epidemiological importance. By combining species composition, host blood meal analysis, sporozoite detection and spatial analyses, the study revealed some of the ecological and behavioral patterns that have direct implications for surveillance and vector control policy.

Secondary vectors as overlooked contributors to transmission

Collections conducted for this study demonstrated An. rufipes and An. pretoriensis as the two most abundant species across the 26 collection points. Anopheles rufipes and An. pretoriensis are both widely distributed across Africa including Zambia [20, 34–37]. Blood meal analysis showed that An. rufipes and An. pretoriensis obtained most blood meals from goats, confirming the zoophagic behavior of these species [20, 31, 35, 38, 39]. However, 25% of blood meals of these two understudied anopheline species came solely from human hosts. Our results are similar to those reported by Tedrow et al. [35] in Madagascar, where their analysis demonstrated that the secondary vectors An. rufipes and An. pretoriensis took 22% of their blood meals from humans. This provides further evidence that An. rufipes and An. pretoriensis are opportunistic feeders and readily feed on humans when available [31]. It also underscores their potential role in sustaining residual transmission outside of sleeping hours and away from the protection of ITNs or IRS [18].

Outdoor and peri-domestic exposure risks

Analysis of the trap locations of blood-fed mosquitoes containing human DNA showed that only 4.8% were captured indoors. Most human blood-fed mosquitoes, 93.7%, were captured in traps placed at goat pens. This may indicate that in this area human biting can occur outdoors in peri-domestic spaces. It could also be argued that these captured mosquitoes may have been those that fed indoors then exited the house, in search of another host (if they did not complete a blood meal) or to find a place to rest [40]. In either case, these mosquitoes may have avoided contact with insecticides sprayed on indoor walls (IRS) and/or ITNs and thus, potentially compromising current vector control interventions [40]. Detection of multiple hosts blood meals further demonstrates feeding plasticity in these mosquitoes [35].

Other understudied species, including An. coustani, An. longipalpis and An. squamosus, demonstrated strong exophagic and zoophilic tendencies, feeding mainly on goats, consistent with previous findings in the study area [16, 19, 20, 38] and elsewhere [34, 35, 41, 42]. Such peri-domestic exposure routes are not addressed by indoor-focused interventions, and they highlight the need for integrated approaches that target exophagic and zoophilic vectors [43, 44].

Low sporozoite infection rates and vector contributions to residual malaria transmission

Low sporozoite infection rates were recorded amongst Anopheles collected in the study. Only a single specimen of An. rufipes was found with P. falciparum sporozoites. These findings are similar to those reported previously supporting the potential role of this species in residual malaria transmission [20, 34]. This study provides further evidence that in the pre-elimination settings of Southern Province, Zambia, outdoor foraging mosquitoes may be contributing to malaria transmission [15, 18, 20, 21, 45].

No P. falciparum sporozoites were observed in any other species. These findings contrast studies in Kenya [41, 46], Madagascar [35, 42], Cameroon [39] and earlier studies in eastern and southern Zambia [19, 20, 47] which implicated An. pretoriensis, An. coustani, An. pharoensis, An. squamosus and An. maculipalpis as vectors of P. falciparum. However, collections in the current study were spatially dispersed and conducted in five separate months and therefore may not have captured transmission hotspots. It is also possible that the sampling may have been conducted during a time when mosquito densities and malaria transmission potential were naturally lower. Nationally, malaria transmission is low during the sampling period [32, 48]. Furthermore, high-coverage of vector-control interventions such as ITNs and IRS that reduce mosquito survival and human-vector contact may have contributed to the low infection rates observed [1, 6]. Additionally, this region previously benefitted from a MDA campaign using dihydroartemisinin and piperaquine [1, 49], combined with on-going test and treat campaigns using RDTs and ACTs [2, 5, 48]. These can reduce the reservoir of Plasmodium in the human population [2, 49] resulting in fewer opportunities for mosquitoes to acquire and subsequently transmit the parasite [40, 50]. Thus, collectively, widely distributed and high infection rates in this very low transmission setting would be unexpected.

This study confirms An. arabiensis as the dominant species within the An. gambiae complex in the study area, similar to previous reports [20, 32]. Although a primary vector of malaria in many parts of sub-Saharan Africa [17, 50], in this study An. arabiensis was not found infected with P. falciparum sporozoites nor was it found with human blood. Anopheles quadriannulatus, another member of the An. gambiae complex was also captured during this study but is not a recognized vector [17]. Combined morphological and molecular identifications by PCR revealed that the An. funestus group mosquitoes collected for this study were predominantly made up of An. parensis, a largely exophagic and zoophilic species which has been implicated as vector of human malaria elsewhere[51, 52]. However, in this study, this species was not found with P. falciparum sporozoites or human blood meals.

Roofing types affect indoor mosquito counts

Grass thatched roofs were associated with a higher mosquito count compared to those with metal roofs. These findings are consistent with other studies that have shown that houses with metal roofs tend to have lower Anopheles densities [53, 54]. Metal-roofed houses generally have higher interior temperatures than houses with grass thatched roofs. The higher temperatures reduce vector survival and thereby may have reduced the number of mosquitoes captured [53, 54]. These results, however, should be interpreted with caution. As stated previously, this was a cross-sectional study carried out over a short period of five months. Whilst the geographical scope was extensive, the temporal scope dedicated to each collection point was extremely limited. The surveys also transitioned from the cool and dry season (May and June) to warmer temperatures (July and August/Sept). This rendered making true comparisons of anopheline counts and related factors of house structure effect limited. Further, mosquito counts indoors were low for comprehensive analysis. Future studies should consider a longitudinal study covering different seasons of households selected upon the transect to better study the effect of household structure on mosquito counts and Pf infection.

Spatial analysis of mosquito counts showed higher numbers in Gwembe District starting from the cluster of Musisa to Masua. Approximately 38% of all mosquitoes came from this location. Another area with high mosquito numbers was Lweendo and Munyama (together contributing ~ 33% to the total catch), likely due to their proximity to the Mutama River. Several flat-bottomed valleys were observed as the descent of the Zambezi escarpment begins after Musisa area through to Masua area. These valleys are characterized as having clay soil with evidence of seasonal swamps [55]. Informal conversations with residents revealed the rivers flow to some degree even during the dry season. This may explain the higher anopheline counts observed in Gwembe. If a longitudinal study was undertaken within each district, it could also determine the contribution of larval habitats to mosquito densities. This could guide future targeted interventions.

Operational considerations

Findings of this study reinforce calls for vector control tools that protect against outdoor exposure to mosquitoes, such as spatial repellents, targeted livestock-based interventions, animal or synthetic human odor-baited traps and toxic sugar baits [43, 44, 50]. The association between these understudied anopheline species, animal shelters, and human blood meals suggests that interventions placed strategically in peri-domestic spaces may have measurable public health benefits. Vector control product profiles should consider these ecological realities, particularly in areas nearing elimination or with persistent low-level transmission [50].

Study limitations

Several factors may have limited the study. Human biting rates were not estimated nor was any information on peri-domestic human activity collected. Combining mosquito biting with human behaviors by time and space is critical for accurately assessing human exposure and may identify key activities that may expose people to bites [13]. Whilst CDC-LTs can sample those times and localities when both vector and host may interact, they under-sample blood-fed mosquitoes and do not establish whether biting itself is taking place [18]. It is also possible that light may play a greater role in the number of mosquitoes caught than the human cues [56]. Further, daylight biting was not investigated in this study. This is a growing concern as demonstrated in a recent study in Western province Zambia [57]. This factor should be an important consideration for future studies. Another limitation was that blood meal analysis was restricted to five host types only, excluding several potentially important animal groups such as sheep (Ovis aries), donkeys (Equus asinus), and ducks (Anas platyrhynchos) which were observed in the study area. Broader analysis is recommended for assessing foraging behavior and the potential for zoonotic disease transmission [18, 58]. Plasmodium falciparum infection in mosquitoes was confirmed using CSP ELISA, without molecular confirmation by PCR. While sporozoite detection using CSP ELISA has been used extensively and is relatively inexpensive, it may be relatively insensitive at very low levels of infection [59, 60]. Further, there are findings of both false positives and negatives when using CSP ELISA if dissections are not done well and also if non-specific proteins and proteases are present in mosquito homogenates [61]. Boiling of mosquito homogenates prior to CSP ELISA is recommended to improve assay specificity and sensitivity by denaturing these proteins and proteases [28, 61] and should be considered for further analyses. CSP ELISA can be complemented with molecular techniques such as PCR, which have higher sensitivity and specificity, particularly for detecting low-density infections. Combining these methods would provide a more accurate assessment of P. falciparum infection rates in mosquito populations and improve the reliability of transmission studies.

Conclusion

This study demonstrated the exophagic, zoophagic and opportunistic foraging patterns of malaria vectors that are largely understudied. The detection of human blood in these Anopheles mosquitoes suggests human biting is taking place. Whilst human blood meals were few in the study and low sporozoite infectivity was reported, outdoor foraging mosquitoes may be responsible for residual malaria transmission in Southern Province, a region that is advancing towards elimination. This finding calls for an expansion of surveillance to include more species with collections inside houses and outdoor or peri-domestic spaces. Such an approach would enable more targeted interventions and support efforts to assess and address residual transmission in rural African settings. Future work should focus on more detailed studies of potentially incriminating these lesser studied species as malaria vectors and estimate the epidemiological contribution of these species through longitudinal sampling, human behavioral studies, and modelling.

Abbreviations

CDC LTs

Centers for Disease Control and Prevention light traps

CSP

Circumsporozoite protein

ELISA

Enzyme-linked immunosorbent assays

HBI

Human biting index

IRS

Indoor residual spraying

ITN

Insecticidal treated nets

PCR

Polymerase chain reaction

SIR

Sporozoite infectivity rate

Author contributions

JCS and DEN conceived the study and wrote the main protocol. JSC, WJM and DEN contributed to resources and funding acquisition. LS and GM supervised the field work. LS, HH, PC, SC and GM implemented the mosquito surveys. TM, LS, GM performed all the laboratory analyses, assisted by MM and BK. ML performed all spatial analysis using geographical information system (GIS) and designed the maps. KS and MEG analyzed the data. KS, LS and TM, drafted the manuscript with contributions from MEG. MEG, JCS and DEN supervised this work. The manuscript was revised by ES, PET, WJM, MEG, DEN and JCS. All authors read and approved the final manuscript.

Funding

Funding for this study was provided in part by the National Institutes of Health funded Southern and Central Africa International Centers of Excellence for Malaria Research (2U19AI089680), as well as support from Bloomberg Philanthropies and the Johns Hopkins Malaria Research Institute, and NIH T32 Training Grants (T32 AI007417 & AI38953).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol and informed consent forms were reviewed and approved by the Macha Research Trust Institutional Review Board (IRB approval number: E.2017.02), Johns Hopkins School of Public Health Institutional Review Board and the Zambia National Research Health Authority. Written permission to undertake the study was obtained from the Ministry of Health through the Southern province health office (PHO). Participation in the study was voluntary, and informed consent was obtained from household heads and every participant above the age of 18 years. Written consent was obtained from household heads before any mosquito collections.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.