Antibodies against arthropod saliva have shown to be a good marker of bite exposure. Because Anopheles albimanus Wiedemann (Diptera: Culicidae) is the principal malaria vector in Haiti, we evaluated the immune response against salivary gland extract (SGE) of this species in malaria-positive and malaria-negative subjects from this country. The results showed that the level of anti-SGE immunoglobulin (Ig)G antibodies was higher in patients with clinical malaria than those in malaria uninfected people living in the same region. In addition, a significant positive correlation between the level of anti-An. albimanus IgG and IgM antibody levels was observed. These results suggest that antibodies against An. albimanus saliva, especially IgG, are useful markers of mosquito bite exposure in Haiti.
Hispaniola is one of the few islands in the Caribbean where Plasmodium falciparum malaria is endemic and chloroquine can be still used effectively (White 1996). The main malaria vector in Haiti, Anopheles albimanus Wiedemann (Diptera: Culicidae), is primarily exophilic, preferring to rest outdoors (Zimmerman 1992, Molez et al. 1998). However, there is some evidence that An. albimanus prefers to feed indoors, especially between 7:00 and 9:00 p.m., after which it heads outdoors to rest (Bown et al. 1993). Although An. albimanus has been reported to be zoophilic (Hobbs et al. 1986), no studies to date have confirmed this in Haiti. Anopheles pseudopunctipennis Theobald, which was recently introduced in the south of Haiti, has been incriminated as a malaria vector in other parts of the Americas (Manguin et al. 1995). This species has yet to be implicated as a malaria vector in Haiti. Peaks in mosquito density occur in November and June, just after the start of the two rainy seasons in Haiti (Hobbs et al. 1986). Peaks in vector density are followed by peaks in malaria incidence (Vanderwal and Paulton 2000).
To our knowledge, only Hobbs et al. (1986) have attempted to measure malaria transmission intensity using the entomological inoculation rate in Haiti (Hobbs et al. 1986). Their efforts took place over a one-month period in March 1984 in four low-lying villages of northern Haiti. Of the 1,219 midguts and 102 salivary glands examined with dissection, none were found to be infected with P. falciparum or other malaria species, resulting in no detectable transmission. These data suggest that traditional entomological methods of directly measuring malaria transmission may be insufficient in areas of low seasonal transmission, especially when resources are limited.
Malaria parasites are transmitted during the mosquito blood-feeding process, when mosquitoes inject infective sporozoites into the skin of their hosts (Vanderberg and Frevert 2004). Saliva, which contains numerous pharmacologically active compounds that have antihemostatic, anti-inflammatory, and immunosuppressive properties that facilitate the mosquito blood-feeding process (Titus and Ribeiro 1990, Titus et al. 2006), is also injected into the host skin during the bloodmeal uptake by the female mosquito (Ribeiro 1987). Many of these salivary components include highly immunogenic enzymes and inhibitory proteins that elicit strong immune responses (Ribeiro and Francischetti 2003). Therefore, salivary proteins can modulate the immune response and affect antibody production, host cell regulation, or both (Titus et al. 2006). In fact, vector saliva has been shown to enhance pathogen infectivity, including not only vector-transmitted viruses (arboviruses) (Schneider and Higgs 2008) but also bacteria (Nuttall and Labuda 2004) and parasites (Theodos and Titus 1993).
Previous studies found that people living in a malaria endemic region in western Thailand presented higher immunoglobulin (Ig)G and IgM antibodies against the saliva of Anopheles dirus (Peyton & Harrison) than people living in Bangkok, a nonendemic area for malaria also in Thailand (Waitayakul et al. 2006). In addition, people with clinical malaria presented higher IgG antibody level than malaria uninfected people living in the same region (Remoue et al. 2006, Waitayakul et al. 2006). Previous information suggests that the presence and level of human antisaliva antibodies are highly dependent on the area-specific mosquito population and potentially represent a proxy measure of the level of mosquito bite exposure in humans (Waitayakul et al. 2006). In addition, previous studies suggest that specific IgG responses are closely related to exposure to mosquito bites, with the decline in the antibody concentration over time, indicating that anti-saliva IgG antibodies have a short half-life (Remoue et al. 2006, Orlandi-Pradines et al. 2007). Using anti-mosquito saliva antibody levels as proxy indicators of mosquito bite exposure may aid in helping to better define the risk potential of acquiring malaria in an endemic area over time (Remoue et al. 2006).
As a pilot study, our hypotheses stated that the presence and level of anti-saliva antibodies represent the level of mosquito bite exposure to the main mosquito vector in the area. This study quantifies the level of anti-An. albimanus salivary protein antibodies in individuals living in the Artibonite River valley in Haiti as baseline data for determination of the utility of An. albimanus salivary gland extract to appraise mosquito bite exposure. Blood samples from individuals from Colombia, Guinea, and the United States are used as controls. The association between antibody type, level and malaria status or parasite exposure also is explored.
Materials and Methods
The protocols and research methods for these studies were reviewed and approved by the Tulane University Institutional Review Board; Hospital San Francisco de Asis, Universidad de Pamplona; and the Hôpital Albert Schweitzer (HAS). Written informed consent was obtained from each participant before collecting data.
Study Site and Population.
In the Artibonite Valley, as is typical of most low-lying areas in Haiti, malaria transmission is highly seasonal with the primary peak occurring from November to January (Bonnlander et al. 1994, Eisele et al. 2007). In Haiti, in addition to irrigation associated production, An. albimanus is found in a variety of other habitats, including road side ditches and hoof- and footprints. Rice, Oryza sativa L., fields are very important in maintaining the population during the dry season (Caillouët et al. 2008). P. falciparum was the only malaria parasite species found in the current study; the last reported case of Plasmodium vivax was in 1983 (PAHO 2000). Health care needs of this population primarily are served by HAS located in Deschapelles.
Samples also were collected in three other countries: Colombia, Guinea, and the United States. An. albimanus is the primary vector of P. falciparum and P. vivax (Gutiérrez et al., 2008) in rural and periurban areas of both the Caribbean and Pacific coasts of Colombia. This mosquito is reported to be present in at least 60% of all adult and larval mosquito collections in these coastal zones (Quiñones et al. 1987). In Guinea, where malaria is also seasonal, the main malaria vectors are Anopheles gambiae (Patton) and Anopheles funestus (Guiles) (Dia et al. 2003, Coetzee and Fontenille 2004). The United States is malaria free, although several Anopheles mosquito species, including An. albimanus, are endemic to the country (Darsie and Ward 2005). Individuals from the United States included in this study were not living in an area where An. albimanus occurs; thus, the purpose of including these individuals was to compare the level of anti-mosquito saliva antibodies in people exposed versus not exposed to An. albimanus saliva.
As has been described previously (Londono et al. 2009), 21 serum samples were collected from laboratory confirmed malaria-positive subjects in Haiti who presented at HAS between November and December 2007. In addition, a subset of 128 blood spots on filter paper (22 polymerase chain reaction [PCR]-confirmed malaria-positive and 106 PCR-confirmed malaria-negative subjects) were included in the study; samples were collected by a combination of active and passive case detection during the high malaria transmission season in 2006 and in 2007 (Londono et al. 2009).
Samples from Colombia (n = 30) and Guinea (n = 30) collected from malaria uninfected individuals and individuals with nonfebrile illnesses were included in the study. Malaria in these individuals was diagnosed by microscopy.
Nine serum samples from individuals living in United States, including banked neonate sera from a West Nile Virus study (Tulane University), also were tested for the presence of antibodies against An. albimanus salivary gland extract (SGE).
Filter Paper Sample Preparation for Detection of Anti-An. albimanus Antibodies.
Six-millimeter-diameter disks were cut from filter paper cards and soaked in 250 μl of phosphate-buffered saline (PBS) overnight at 4°C. After adding 250 μl of 5% dry milk in PBS (blocking buffer), 100 μl of the blood lysate was added to each well and incubated overnight at 4°C. Samples from filter paper were tested for IgG and IgM antibodies against An. albimanus only. Each sample was tested in duplicate.
P. falciparum Malaria Diagnosis by PCR.
The detection of P. falciparum from blood samples and from filter paper samples was performed by PCR for conserved sequences in 18S small subunit RNA, with a single reverse primer for all Plasmodium species and a P. falciparum-specific forward primer using the methods described previously (Eisele et al. 2007, Londono et al. 2009). The reaction was carried out on a MyCycler thermal cycler (Bio-Rad Laboratories, Hercules, CA). An initial 5-min incubation at 95°C was followed by 43 cycles, each of 45 s at 95°C, followed by 90 s at 60°C, and then a final step at 72°C for 5 min. The amplified products were visualized by electrophoresis on a 1% agarose gel stained with ethidium bromide. The length of the expected products for P. falciparum was 276 bp.
To ensure reproducibility of our results, mosquitoes used in the current study were kept under controlled conditions. An. albimanus (STECLA strain from MR4, donated by Benedict MQ and Seawright JA) and Aedes aegypti (L.) (Rockefeller strain) mosquitoes were reared at 25–28°C, 70–80% RH, and a photoperiod of 16:8 (L:D) h and maintained on a 10% sucrose solution during adult stages. In addition, 5–8-day old adult female Anopheles stephensi (Liston) mosquitoes were donated by the Tulane National Primate Research Center.
Mosquito Dissections and Preparation of Salivary Gland Extract.
Female mosquitoes from 5 to 10 day old were cold anesthetized, washed in 70% ethanol, and placed in PBS, pH 7.2, for salivary gland dissection. Salivary glands were allowed to freeze at −80°C and thaw at 4°C four times to induce cell rupture and release of proteins. The resulting SGE was kept in PBS at −80°C until use (Wanasen et al. 2004, Waitayakul et al. 2006). Protein concentration was determined by the method of Bradford (1976) (Bio-Rad protein assay).
Detection of Antibodies to Mosquito Salivary Proteins by Enzyme-Linked Immunosorbent Assay (ELISA).
To optimize working conditions of the ELISA test, a checkerboard titration was performed to establish salivary proteins and serum conditions. Anopheles and Aedes SGE at 0.5, 0.8, and 1 μg/ml; and human serum at 1/50, 1/100, and 1/200 (vol:vol) were tested. Based on the results from the titration, 96-well ELISA plates (Nunc–Maxisorp, Nalge Nunc International, Rochester, NY) were coated with 100 μl/well of 0.5 μg/ml of either Anopheles or Aedes SGE prepared in coating solution (Kirkegaard and Perry Laboratories, Gaithersburg, MD) and incubated overnight at 4°C. Plates were blocked for 1.5 h with 5% dry milk in PBS (Invitrogen, Carlsbad, CA) at 37°C and incubated with 100 μl/well of 1/100 serum dilution at 37°C for 2.5 h (filter paper blood lysate was incubated overnight at 4°C). Plates were washed three times with wash solution (PBS and 0.1% Tween) and incubated with 100 μl/well of either goat anti-human IgG (1/1000) or IgM (1/10,000) horseradish peroxidase (HRP)-conjugated antibodies (Caltag Laboratories, Burlingame, CA)) at 37°C for 1.5 h. Colorimetric development was obtained using 100 μl/well tetra-methyl-benzidine (TMB, one-solution microwell, GeneScript, Piscataway, NJ) incubated for 15 min at room temperature. The reaction was stopped with 100 μl/well of stop solution (1 M phosphoric acid), and absorbance was measured at 450 nm. Each sample was tested in duplicate (Remoue et al. 2006, Waitayakul et al. 2006). Two controls were included in each plate: 1) control blank: two wells without SGE to control for nonspecific induction of color for any of the reagents used in the test; and 2) negative control: two wells with SGE but without human serum to control for any nonspecific color induction of the coating antigen.
Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis (SDS-PAGE) Electrophoresis and Immunoblot.
Serum samples from each country were pooled (one pool of each country) with 30 μl of each individual sample in a 1.5-ml tube and kept at −20°C until use. Immunoblotting was performed according to the protocol published by Waitayakul et al. (2006) and, Cornelie et al. (2007). In brief, 1.4 μg/well SGE was separated in 10% polyacrylamide gel (NuPAGE, Invitrogen) at 90 V during 2.5 h and transferred to polyvinylidene difluoride (PVDF) membranes (Invitrolon PVDF, 0.45-μm pore size, 8.5 by 13.5 cm, Invitrogen) at 30 V during 1 h. Membranes were blocked for 1 h with blocking buffer (WesternBreeze, Invitrogen) and incubated with pooled human sera (1/100) at 4°C overnight. Each membrane was washed five times with wash solution (WesternBreeze, Invitrogen) and incubated with HRP-conjugate (IgG, 1/1000 or IgM, 1/10,000) during 1 h at 37°C. Color development was obtained with HRP chromogenic substrate TMB (Novex, Invitrogen). Prestained broad-range molecular weight marker 16.5–210 kDa (Thermo Fisher Scientific, Waltham, MA) was used for the estimation of the size of the proteins.
Antibody level was expressed as adjusted optical density (OD) calculated for each sample subtracting the mean OD value of the controls from the mean OD value of the duplicates for each sample. After verifying that our values did not meet the normal distribution (P > 0.05; skewness and kurtosis normality test), the difference in the antibody levels between two groups (malaria-positive and malaria-negative individuals) was tested using the nonparametric Mann–Whitney U test. The difference between more than two groups (i.e., Haitian, Colombian, and Guinean samples) was assessed using Kruskal–Wallis and Dunn's multiple comparison tests. All differences were considered significant with a probability of committing a type 1 error set at P < 0.05 (two-tailed). Spearman's correlation coefficient was used to evaluate correlation between the level of each type of antibody (IgM and IgG). All statistical tests were computed using Prism version 5.02 (GraphPad Software Inc. 2007).
Anti-Anopheles Antibody Levels and Presence of the Mosquito in the Area.
In total, 218 subjects were included in the study (Table 1). An. stephensi and Ae. aegypti SGEs were included to determine the specificity of the anti-SGE protein antibodies. IgG anti-SGE antibodies were more variable than IgM antibodies against An. albimanus and Ae. aegypti in Haitian subjects (Fig. 1). When comparing the antibody levels to both mosquito genera, we found that IgG anti-An. albimanus SGE antibodies were significantly higher than IgG antibodies against Ae. aegypti SGE (P < 0.001; two-tailed Mann–Whitney U test). In contrast, there was no significant difference in the level of IgM antibodies (P = 0.8602; two-tailed Mann—Whitney U test). When comparing IgG versus IgM levels within each country, IgM was significantly higher than IgG in Colombia (P = 0.0076; two-tailed Mann–Whitney U test), whereas the level of both antibodies was similar in Haiti (P = 0.9199; two-tailed Mann–Whitney U test).
Results showed a significant positive correlation between the level of anti-An. albimanus IgG and the level of anti-IgM antibody levels (Spearman's correlation coefficient, R = 0.2810, P = 0.0070). In addition, the antibody level was highly associated with the presence of An. albimanus in the area; thus, antibody levels were significantly higher in subjects residing in Colombia and Haiti than those in subjects residing in Guinea and the United States where An. albimanus does not occur (P < 0.001; two-tailed Mann–Whitney U test) (Fig. 2). In contrast, there was no significant difference in the levels of IgG antibodies against An. stephensi SGE between the two groups (P = 0.302; two-tailed Mann–Whitney U test).
Malaria Status and Anti-Mosquito Antibody Levels.
In addition to serum samples, 128 blood spots in total on filter paper were collected from Haitian subjects: there were seven malaria cases from 2006 and 15 cases from 2007; there were 54 uninfected individuals from 2006 and 52 from 2007. Levels of anti-An. albimanus antibodies in malaria patients and malaria uninfected controls are shown in Fig. 3. The IgG antibody level was significantly higher in malaria patients than in malaria-negative controls from the same region (P = 0.025; two-tailed Mann–Whitney U test). However, anti-An. albimanus IgM antibody levels were not significantly different between these two groups (P > 0.050; two-tailed Mann–Whitney U test). In addition, the level of IgG antibodies against An. albimanus SGE was significantly higher in 2006 in both malaria-negative (P < 0.0001; two-tailed Mann–Whitney U test) and malaria-positive individuals (P = 0.016; two-tailed Mann–Whitney U test) than in 2007, but the level of IgM antibodies was not statistically different between the 2 yr (P = 1.000 in 2006 and P = 0.7640 for 2007; two-tailed Mann–Whitney U test). These results were consistent with the antibody levels found in serum samples where IgG antibody levels were also higher in individuals from Haiti (malaria patients) than in individuals from the other countries (malaria-uninfected individuals) (P = 0.0163; two-tailed Mann–Whitney U test).
Level of antibodies against Ae. aegypti SGE was evaluated to determine the specificity of IgG anti-An. albimanus antibody levels in malaria patients because Aedes mosquito bite exposure should not affect the malaria outcome. We found no significant differences in the level of IgG anti-Ae. aegypti SGE antibodies between the two groups (P = 0.3238; two-tailed Mann–Whitney U test).
Identification of Anopheles SGE Proteins Recognized by Sera from the Study Subjects.
Due to constrains in sample quantity, the immunoblots were performed using pooled sera from each country (n = 4 pools). The SDS-PAGE protein analysis revealed several proteins for An. albimanus and An. stephensi mosquitoes. These proteins ranged from 32 to >110 kDa (Fig. 4). To identify the SGE proteins from these two mosquito species that were recognized by human serum, we pooled the sera from Haiti, Guinea, and Colombia and tested each pool by immunoblot (Fig. 4). Interestingly, only one band of ≈90 kDa was highly recognized by the pooled sera from Haiti and Colombia but not visible in the Guinea pooled sera.
Chloroquine treatment of malaria cases and indoor-residual spraying helped to reduce the malaria indices in Haiti, but eradication attempted in the 1960s–1980s was unsuccessful mainly because of limited knowledge of malaria ecology that could guide the effective use of the scarce resources toward the most affected areas (Bonnlander et al. 1994, Kachur et al. 1998, Keating et al. 2008). Detection of antibodies against mosquito saliva can be a practical cost-effective method to evaluate population exposure to mosquito bites before and after malaria control activities (Billingsley et al. 2006, Remoue et al. 2006).
Knowledge of the local mosquito population's feeding preference might be useful in implementing this method as an evaluation tool for mosquito bite exposure, taking into account that it is the number of bites that will determine the level of antibodies. It follows that the antibody levels will be a reflection of human–mosquito contact and consequently an indication of human risk of contacting vector-transmitted diseases such as malaria. Knowledge of antibody level could be a useful tool to evaluate the efficacy of antivector devices such as the insecticide-treated bed-net; it also can guide local disease control authorities to pinpoint the places where people are being fed on more frequently. This information will help in deciding what control method should be used and how to educate the community to reduce mosquito bite exposure.
According to the present results, antibody levels were highly dependent on the presence of the mosquito species in the area because the level of IgG antibodies against An. albimanus SGE were significantly higher than IgG antibody levels against An. stephensi, which is not present in America. In addition, the level of antibodies against Ae. aegypti mosquitoes in Haitian individuals was significantly lower than An. albimanus antibodies. Caillouët et al. (2008) found that the percentage of water bodies in the Artibonite valley of Haiti positive for An. albimanus mosquitoes was higher than for Ae. aegypti mosquitoes (25 versus 5.4%). They also found that the abundance of An. albimanus was associated with elevation and distance from the Artibonite River, whereas the presence of Ae. aegypti was more related to the availability of human maintained water containers (Caillouët et al. 2008). This also could explain some of the differences in antibody levels, especially IgM, because the distribution of these two mosquito species can be quite different even in a small geographical area shared by the two species.
This also reinforces the findings of previous studies in which IgG antibodies were very specific in recognizing the salivary proteins of a particular mosquito species (Remoue et al. 2006, Waitayakul et al. 2006). Although highly specific, these antibodies also can cross-react with saliva from other arthropods (Peng and Simons 1997, Jeon et al. 2001). We found that subjects living in An. albimanus nonendemic areas presented antibody titers sufficient to detect An. albimanus SGE proteins, although the level of such antibodies were significantly lower. These results suggest the presence of an important cross-reactivity among anopheline species saliva capable of inducing antibodies that may interact with SGE proteins from other species in the genus. More work is needed to determine whether this cross-reactivity lessens with increased phylogenetic distance between species.
High levels of anti-SGE antibodies in malaria infected individuals may suggest the following. First, people with malaria present high antibody levels due to factors such as increased skin temperature (Gilles 1980) and CO2 excretion (Costantini et al. 1996), and even body odor and breath (Knols et al. 1995; Mukabana et al. 2004) that increase their attractiveness to mosquitoes, which in turn increases their chance of being bitten by infected mosquitoes and getting malaria. Second, it could be that individuals with malaria are at more risk of being bitten by infected, and more importantly, by noninfected mosquitoes because of their malaria infection (Billingsley et al. 2006), gametocytemia, or both (Lacroix et al. 2005). Regardless, higher antibody levels would indicate not only a higher risk of people of having malaria but also of more mosquitoes acquiring and transmitting the parasite. Several studies have shown that host odors are important in determining mosquito host preferences (Dekker et al. 2001, Pates et al. 2001).
Based on passive case detection of confirmed malaria cases identified by HAS in the Artibonite Valley from 2004 to 2006 (Eisele et al. 2007) transmission in 2006 seems to have been substantially lower than in previous years. If this trend of decrease in transmission continued in the area, the phenomena could explain the significant decrease of IgG antibody levels in 2007 in comparison with 2006. Interestingly, a decrease in malaria cases does not a necessarily mean a decrease in mosquito bites in 2007 in comparison with 2006, because the level of IgM in both years was practically the same. Even though current malaria control in Haiti relies on chloroquine treatment only, the new funding for malaria control from the Global Fund to Fight AIDS, Tuberculosis, and Malaria, is expected to include free and subsidized distribution of insecticide-treated bed-nets, enhanced surveillance, and case management (Keating et al. 2008). The evaluation of anti-mosquito salivary protein antibodies after these malaria control interventions, especially IgG antibody levels, will give us an approximation of the reduction in both mosquito–human contact and malaria parasite exposure in the Haitian community.
The relationship between age and antibody level is still controversial. Waitayakul et al. (2006) found no correlation between IgG antibody levels against An. dirus and age, whereas Remoue et al. (2007) found no significant differences in the level of antibodies against Aedes saliva with age. Conversely, we found a relative decrease in the level of antibodies against mosquito salivary proteins with age (data not shown), but our small sample size did not allow us to measure these findings significantly. However, several studies have suggested an impact of the seasonality of mosquito density on the age–antibody level and age–antibody type relationship (Brummer-Korvenkontio et al. 1990, Remoue at al. 2006). We are working on increasing our sample number to test in depth whether the level of antibodies is influenced by age.
This study has several limitations. The main limitation is that sample collections were restricted to the Artibonite Valley of Haiti; as such, our results may not be generalizable to other areas in Haiti. In addition, the lack of sufficient serum samples did not allow us to perform individual immunoblot tests that are very important in identifying immunogenic proteins conserved through the individuals and those that are variable.
In conclusion, IgG antibody levels against An. albimanus SGE salivary proteins are a potentially useful indicator of mosquito bite exposure and malaria infection status in areas of low and seasonal transmission such as Haiti. In addition, salivary proteins specific to An. albimanus can be further determined through two-dimensional electrophoresis and mass spectrometry and used as specific antigen for An. albimanus bite exposure. The use of genus-, species-specific, or a combination of salivary gland protein to evaluate the level of anti-mosquito saliva antibodies may be useful, for example, for populations living in malaria endemic areas with highly seasonal transmission where the level of antibodies will indicate their increasing or decreasing risk of malaria during high or low transmission periods. In addition, in places where malaria control methods, such as insecticide-treated mosquito nets, are implemented, the measurement of antibody levels before and after the implementation of these tools would indicate the level of mosquito–host contact, and success (or failure) of the control method. Level of antibodies may also be useful in monitoring the risk potential for malaria transmission in areas that have achieved elimination, allowing researchers and malaria control programmers to assess how much exposure the population has to the vector, even if they are not infective (Orlandi-Pradines et al. 2007).
We thank Sylvie Cornelie and Papa Makhtar for support and guidance; Matt Ward, Camille Dieugrand, Adam Bennett, and the HAS data collectors, who were essential for data collection in 2006 and 2007; Don Krogstad for advice and use of laboratory space at the Department of Tropical Medicine; Lina Moses and Ken Swan for Guinea and US serum samples, respectively; and Chad Massey for providing and taking care of the An. stephensi colony. We also thank IDEA (Caracas–Venezuela) for the training in proteomics; PECET (Universidad de Antioquia), Diva Palacios (DASALUD, Choco), and Daisy Carvajal (Universidad de Pamplona) for collaboration with samples and advice; the Haitian, Colombian, and Guinean communities for participating in this study. These studies were supported in part by a grant for doctoral studies to B.L.L. from COLCIENCIAS (Bogota, Colombia); by Tulane University Research Enhancement Awards to D.J.K., J.K., and T.P.E.; and by a grant from the U.S. Agency for International Development to T.P.E.