This study was conducted in Awash Sebat Kilo, Ethiopia, an area of perennial malaria transmission, during April–September 2019. We examined aquatic habitats for immature-stage Anopheles mosquitoes by standard dipping (10×/site) for 5 consecutive days (4). We assessed mosquito resting, feeding, and host-seeking behavior by 5 methods: CDC miniature light traps model 512 (John W. Hock Company, https://www.johnwhock.com), human landing collection, pyrethrum spray sheet collection, aspiration from animal shelters, and cattle-baited traps (5). We identified adult mosquitoes by using standard keys and confirmed identification by targeted sequencing of nuclear internal transcribed spacer 2 (ITS2) and mitochondrial cytochrome oxidase subunit 1 gene (COI) (6). To generate clade topologies, we compared An. stephensi mosquito DNA sequences with those in publicly available libraries (7). We determined mosquito blood meal sources by using multiplex PCR targeting cytochrome b (8) and infection status by using 18S rRNA nested PCR (9).

Adult An. stephensi mosquitoes reared from immature mosquitoes from local water sources and a colony of An. arabiensis mosquitoes (≈120 each) were fed in the dark for 30 min on membrane feeders containing fresh blood from Adama malaria clinic patients with microscopy-confirmed mono- and mixed-species infections with P. vivax and P. falciparum (10). Unfed and partially fed mosquitoes were removed; fully engorged mosquitoes were maintained on sugar solution. At 7 or 12 days after feeding, mosquitoes were dissected, their midguts were examined for oocysts, and their salivary glands were examined for sporozoites. To compare infection status between An. arabiensis and An. stephensi mosquitoes, we performed logistic regression. We used individual mosquito data and a fixed effect for each patient to account for correlations between mosquito observations from the same donor. Bland-Altmann plots were generated for differences in infectivity between mosquito sources by using the Pitman test of difference in variance. For analyses, we used STATA version 13 (StataCorp., https://www.stata.com/company) and GraphPad Prism 5.3 (GraphPad Software Inc., https://www.graphpad.com). Raw data have been deposited in the DRYAD data depository (https://datadryad.org/stash/dataset/doi:10.5061/dryad.gf1vhhmnt).

An. stephensi larvae were detected in 75.3% (64/85) of the 85 artificial water sources surveyed (Table 1). A total of 49,393 immature Anopheles larvae and pupae were collected during 20 weekly collections in April–September 2019, of which 45,316 (91.7%) emerged as adult mosquitoes in the laboratory. Morphologic identification of adults confirmed that all were An. stephensi. During monthly rounds of entomologic surveillance in August and September (6 days each), we collected 89 adult female Anopheles mosquitoes (72 [80.9%] An. stephensi, 16 An. gambiae, and 1 An. pharoensis). We detected P. vivax in 2.8% (2/72) and P. falciparum in 1.4% (1/72) of wild-caught An. stephensi mosquitoes. Blood meal source was identified for 35.0% (28/80) blood-fed wild-caught An. stephensi mosquitoes; exclusive human blood meal was identified for 17.2% (5/29). The remainder fed (multiple blood meals) either on humans and animals (n = 9) or animals only (n = 14) such as goats (n = 21), cows (n = 4), and dogs (n = 5). Successful sequencing of ITS2 for 76 and COI for 45 Anopheles mosquitoes confirmed that all were An. stephensi. According to ITS2 sequences, An. stephensi mosquitoes from Ethiopia formed a well-supported monophyletic clade with isolates from the Arabian Peninsula and Southeast Asia (Appendix). The COI tree was more resolutive, suggesting that An. stephensi mosquitoes from Ethiopia were most closely related to mosquitoes from Djibouti (64%) and Pakistan (54%).

Figure 1

Figure 1. Comparison of feeding efficiency and infection rates for Anopheles stephensi and An. arabiensis mosquitoes in paired feeding experiments in study of An. stephensimosquitoes as...

Figure 2

Figure 2. Comparison of relative oocyst numbers and infection rate for Anopheles stephensi and An. arabiensis mosquitoes in paired feeding experiments in study of An. stephensimosquitoes...

We conducted 47 paired-membrane feeding experiments by using blood from patients with microscopy-confirmed P. vivax or P. falciparum infection (Table 2). The proportion of blood-fed mosquitoes was generally higher for An. arabiensis (median 80.5%; interquartile range [IQR] 72.5–85.0) than An. stephensi mosquitoes (median 53.5%, IQR 44.0–68.0; p<0.001; Figure 1, panel A). The proportions of the 2 mosquito species infected with P. vivax were strongly associated (ρ = 0.82, p<0.001; Figure 1, panel B); a significantly higher proportion of An. stephensi (median 75.1%, IQR 60.0–85.9) than An. arabiensis mosquitoes were infected (median 58.4%, IQR 40.0–85.6; p<0.042). Allowing for the number of dissected mosquitoes for each set of paired feeding experiments, the odds of an individual mosquito becoming infected was higher for An. stephensi mosquitoes (odds ratio [OR] 1.99, 95% CI 1.52–2.59; p<0.001) (Figure 1, panel C). The number of oocysts per infected midgut was also higher for An. stephensi (median 17, IQR 6–33) than An. arabiensis mosquitoes (median 13, IQR 4–30); p<0.001 (Figure 2, panel A). The number of oocysts was positively associated with the proportion of infected mosquitoes for An. stephensi (ρ = 0.553, p<0.001) and An. arabiensis mosquitoes (ρ = 0.576, p<0.001; Figure 2, panel B). Among paired feedings, sporozoites were detected in 52.2% (47/90) An. arabiensis and 75.0% (84/112) An. stephensi mosquitoes. A much higher proportion of An. stephensi (51.8%, 58/112) than An. arabiensis mosquitoes (31.1%, 28/90) had high sporozoite load (+3 and +4); p = 0.011. After accounting for the number of examined salivary glands, the odds of detecting high sporozoite intensity were substantially higher for An. stephensi than An. arabiensis mosquitoes (OR 4.6, 95% CI 2.2–9.9; p<0.001).

An. stephensi mosquitoes have spread from Asia throughout the Horn of Africa, detected in Djibouti in 2012 (11), Ethiopia in 2016 (12), and Sudan in 2019 (3). The widescale presence of An. stephensi mosquitoes in developmental stages in artificial water bodies demonstrates that these mosquitoes are firmly established in an urban setting in Ethiopia, located on the main transportation corridor from Djibouti to Addis Ababa. Detection of 4 haplotypes suggests independent arrival of different populations or heterogeneity arising after importation of the mosquito species. Our mosquito feeding experiments predominantly included highly infective patients with clinical P. vivax infection (10,13). Although feeding rates for the membrane-adapted colony of An. arabiensis mosquitoes were high, mosquito infection rates were significantly higher for An. stephensi mosquitoes. Our detection of salivary gland sporozoites establishes that sporogonic development of local P. vivax can be completed by An. stephensi mosquitoes. We recruited fewer patients with clinical P. falciparum infection, who were less likely than P. vivax patients to infect mosquitoes (10). Despite a modest number of observations, our findings demonstrate that local P. falciparum isolates are also capable of infecting An. stephensi mosquitoes and are further supported by detection of P. falciparum– and P. vivax–infected wild-caught adult mosquitoes.

Spread of An. stephensi mosquitoes poses risk for increased P. falciparum and P. vivax receptivity and local transmission in urban Africa. Given mosquito preference for human-made containers (14), our findings support integrated vector management recommended by the World Health Organization under the Global Vector Control Response (15). Management may include integrated surveillance and control of other vectors such as Aedes aegypti mosquitoes for larval source management.