Ecology of Triatoma brasiliensis in northeastern Brazil: seasonal distribution, feeding resources, and Trypanosoma cruzi infection in a sylvatic population

Authors

  • Otilia Sarquis,

    1. Laboratório de Eco-Epidemiologia da Doença de Chagas, Instituto Oswaldo Cruz (IOC/Fiocruz), Av. Brasil 4365, 21045–900 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author
  • Filipe A. Carvalho-Costa,

    1. Laboratório de Sistemática Bioquímica (IOC/Fiocruz), Av. Brasil 4365, 21045–900 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author
  • Lívia Silva Oliveira,

    1. Laboratório de Eco-Epidemiologia da Doença de Chagas, Instituto Oswaldo Cruz (IOC/Fiocruz), Av. Brasil 4365, 21045–900 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author
  • Rosemere Duarte,

    1. Departamento de Ciências Biológicas, Escola Nacional de Saúde Pública (ENSP/Fiocruz), Av. Leopoldo Bulhões 1480, 21031–210 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author
  • Paulo Sergio D′Andrea,

    1. Laboratório de Biologia e Parasitologia de Mamíferos Silvestres Reservatórios (IOC/Fiocruz), Av. Brasil 4365, 21045–900 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author
  • Tiago Guedes de Oliveira,

    1. Laboratório de Eco-Epidemiologia da Doença de Chagas, Instituto Oswaldo Cruz (IOC/Fiocruz), Av. Brasil 4365, 21045–900 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author
  • Marli Maria Lima

    1. Laboratório de Eco-Epidemiologia da Doença de Chagas, Instituto Oswaldo Cruz (IOC/Fiocruz), Av. Brasil 4365, 21045–900 Rio de Janeiro, RJ, Brazil
    Search for more papers by this author

ABSTRACT:

We assessed some ecological parameters of Triatoma brasiliensis in rock piles in the state of Ceará during the rainy and dry seasons. The greatest density was in April (median = 12.5 triatomines/site). The greatest abundance was in December, when the insects were more dispersed and the density per site was lower (6 triatomines/site). The nutritional status of females and 5th instar nymphs was increased in July. The rate of T. cruzi infection reached its highest peak in July (10.9%). ELISA revealed that the principal food sources were birds (33.1%), followed by armadillos (18.8%). Food sources were more frequently identified during the rainy season. T. brasiliensis specimens collected in the drought tended to: i) present lower rates of T. cruzi infection and gut content reactivity to tested antisera, ii) have a poorer nutritional status, iii) exhibit lower fecundity, iv) be more dispersed among the studied collection sites, and v) be more abundant and easily collected in the surface of the rocks, possibly reflecting an increased searching for blood meals. Such findings underscore epidemiological concerns and allow inferences about the season when triatomines can more frequently invade the peridomestic environment in search of food and recolonize artificial structures.

INTRODUCTION

Hematophagous insects of the subfamily Triatominae (Hemiptera; Reduviidae) are vectors of Trypanosoma cruzi, the etiological agent of Chagas disease (American trypanosomiasis). These insects are widely distributed throughout the Americas and are the primary vectors of Chagas disease in endemic areas. Several triatomine species have great epidemiological importance because of their susceptibility to T. cruzi infection, as well as their capacity to invade and colonize domiciles, increasing the potential for transmitting the protozoa to humans (Silveira 2002).

Brazilian health authorities have reported that some autochthon species originally restricted to sylvatic ecotopes have increasingly been found invading and inhabiting domiciles (Alencar 1987, Silveira and Vinhaes 1998, Almeida et al. 2000, 2008, Fé et al. 2009). Triatoma brasiliensis is able to invade and colonize dwellings, enhancing parasite transmission to humans in the trypanosomiasis domestic cycle (Sarquis et al. 2004, Costa 1999, Costa et al. 2003).

Due to its biological and ecological characteristics, T. brasiliensis has been considered the most important vector in the Brazilian northeast region. Its presence has been observed in all states of this region (Silveira and Vinhaes 1998). This species also inhabits xerophytic ecosystems in Goiás and Tocantins, states bordering the northeast region of Brazil, situated in the west-central and northern regions of the country, respectively (Costa et al. 2003).

In sylvatic environments, T. brasiliensis lives in rock cracks and crevices, usually in close proximity to small mammals (Alencar 1987, Carcavallo et al. 1999). On the other hand, in an anthropic environment, this triatomine is able to colonize intradomiciles as well as a wide diversity of peridomestic structures such as corrals, pig styes, hen houses, and piles of bricks, tiles, and wood (Oliveira-Lima et al. 2000, Sarquis et al. 2006). This capacity to invade and adapt to different environments demonstrates the accentuated synanthropic behavior and epidemiological importance of this species. In this context, studies of the ecology and biology of sylvatic triatomines in their natural environments are very important to vector control management, especially regarding the species that invade and colonize domestic environments (Forattini et al. 1971, Lorenzo and Lazzari 1999, Lorenzo et al. 2000, Noireau et al. 2005, Abad-Franch et al. 2005, Cortez et al. 2007, Guhl et al. 2009).

We report some ecological and biological aspects of sylvatic populations of T. brasiliensis colonizing rocks in the Caatinga (scrublands) of northeastern Brazil, including the assessment of natural T. cruzi infection rates, in order to contribute to the knowledge of this species in its natural habitat.

MATERIALS AND METHODS

This study was carried out from September 2005 to July 2006 in the municipality of Jaguaruana, state of Ceará, Brazil (4º 50 ‘90“S; 37º 46 ‘48“W). This area is rich in small, sparse rock piles surrounded by xerophytic and arboreal vegetation, mainly different cactus species, which are a hallmark characteristic of the landscape of the semi-arid area northeast of Brazil (Figure 1).

Figure 1.

Caatinga landscapes at Ceará, Brazil, the environment of sylvatic T. brasiliensis, in rainy (A) and dry (B) seasons. Arrows indicate two of the investigated collection sites (rocks).

In this region, the hot/dry climate is dominated by a period of drought that lasts six months. Generally, the rainy season extends from January to June and the dry season from July to December, with peak rainfall occurring in April and May. Nevertheless, the dry season commonly extends beyond this six-month period (Leal et al. 2005). These periodic climatic changes greatly influence the vegetation and fauna, transforming the Caatinga landscape. The average annual temperature ranges from 23° C to 33° C, with an average rainfall of about 850 mm/year.

Within the sylvatic habitat of the triatomines, an area of ∼700m × 300 m was investigated, and 15 collection sites, each characterized by small assemblages of two to four rock piles as high as 3 m, were selected based on the presence of triatomine feces and/or exuviae.

Around and on top of the rocks, xerophytic vegetation proliferated, mainly Cereus squamosus, a typical and abundant cactus in that region (Figure 1), often strewn with bird nests. Rock cracks and crevices ∼50 cm in depth sheltered several animals such as birds, wild rodents (mainly Echimyidae (Thrichomys laurentius) and Caviidae (Galea spixii and Kerodon rupestris)), opossums (mainly Didelphis albiventris and Monodelphis domestica), armadillos, snakes, spiders and scorpions, in addition to a wide diversity of insects, thus comprising the fauna of the natural triatomine habitat. The collection sites were situated between two small rural communities about 2 km from the nearest house.

The survey was conducted during four periods of five consecutive days in July, September, and December, corresponding to the dry season, and in April, corresponding to the rainy season. Inspections took place at night from 18:00 to 20:00 and were performed by two technicians per collection site with the aid of 30 cm tweezers and a flashlight. No insect dislodging substances were used, and all triatomines were captured on the surfaces and in the crannies of the rocks.

Only adults and nymphs from 3rd to 5th instars were collected, since 1st and 2nd instars could not be detected with the flashlight. During the searches, the rock surface and crack temperatures were registered at 15-min intervals with a data-logger thermometer (Q-240, Quimis®).

Excluding females, all the captured triatomines were initially stored in labeled plastic containers identified according to the respective site of capture. Females were stored in the same way but in isolation, i.e., one female per container. All insects were individually weighed with a precision balance (GEHAKA®) and identified by stage for nymphs and sex for adults (Lent and Wygodzinsky 1979) in a temporary field laboratory.

After weighing, all triatomines were returned to their respective labeled plastic containers and forwarded to the laboratory in Rio de Janeiro, where the females remained isolated and were fed fortnightly. Oviposition was followed daily in the first month after capture, allowing an estimation of seasonal variations of fecundity. In order to assess the natural T. cruzi infection index, all specimens that arrived alive in the laboratory were examined. Feces were removed by abdominal compression, diluted in saline, and the numbers of epimastigote and trypomastigote forms per μl of feces were quantified using a Neubauer chamber and an optical microscope (400X).

The blood meal source in the gut contents of nymph and adult triatomines was identified in 27% of those captured in the dry season and 48% in the rainy, through the indirect enzyme-linked immunosorbent assay (ELISA) (Duarte and Marzochi 1997, Burkot et al. 1981). Briefly, samples were diluted in carbonate bicarbonate buffer and applied to 96 well polystyrene microplates. After a second wash, the conjugate (anti rabbit IgG horseradish peroxidase conjugated) was added, and a new incubation and wash was performed. This was carried out with the buffer application, and the plates were analyzed in a microplate reader with a 492 nm operational filter and a 600 nm reference filter (Burkot et al. 1981). The interpretation of the results was obtained after the cutoff point was established. The cutoff was defined as the mean value of the negative controls plus two standard deviations. Positive samples were those reading 10% over the cutoff. The following antisera were tested: armadillo (Dasypus novencinctus), chicken (Gallus gallus), cockroach (Periplaneta americana), dog (Canis familiaris), opossum (Didelphis marsupialis), horse (Eqqus caballus), human (Homo sapiens), lizard (Tupinambis meriane), rodent (Rattus norvegicus), goat (Capra hyrcus), and sheep (Ovis aries).

The Mann-Whitney non-parametric test was used to compare triatomine weights with respect to month of capture and parasite burden observed in insects collected in distinct seasons. The rate of infection in the different stages and months of capture was compared by the chi-square test. Statistical significance was accepted at p≤ 0.05.

RESULTS

A total of 475 specimens of T. brasiliensis was collected, including 123 (25.9%) adults and 352 (74.1%) nymphs (Table 1). Among adults, males were more abundant than females with a male:female ratio of 2.6:1. Among nymphs, the number of captured insects decreased with the instar; 5th instars were captured most, followed by 4th and 3rd instars. Table 1 shows that the greatest apparent density per collection site was in April (median=12.5, interquartile range [IQR]=6–23 triatomines/positive site). Although the greatest abundance was in December (178 insects), the density per site was lower (median=6, IQR=2–18 triatomines/positive site). Insect stage distribution included 89 (18.7%) adult males, 34 (7.2%) adult females, 73 (15.4%) 3rd instar, 117 (24.7%) 4th instar, and 162 (34%) 5th instar nymphs. In addition, 3rd instar nymphs were less frequently captured in July. With regard to nutritional status, adult females and 5th instar nymphs presented a significantly better nutritional status in July (Table 2, Figure 2).

Table 1.  Abundance and relative density of distinct developmental stages of T. brasiliensis captured at 15 collection sites in Ceará, northeastern Brazil. Thumbnail image of
Table 2.  Comparison of medians and interquartile intervals of weight (mg) of nymphs and adults of T. brasiliensis from a sylvatic environment in Ceará, Brazil, by month of capture.
 AdultsNymphs
  1. Statistically significant differences: a p-value = 0.002; b p-value = 0.010; c p-value = 0.030; d p-value = 0.037 (Mann-Whitney test).

 MalesFemalesN3N4N5
April     
Median weight (interquartile range)154 (136–179)150 (131–193)c12 (10–16)26 (21.5–38.5)66 (55–77)d
July     
Median weight (interquartile range)167 (122–201)157.5 (143.5–182.5)ab28 (28–28)28.5 (18.5–32)85.5 (59–121.5)d
September     
Median weight (interquartile range)135 (116–155)118 (101–132)ac12 (7–31)26 (20–40)78 (72–122)
December     
Median weight (interquartile range)146 (123–167)127 (89–147)b19.5 (10–25)32 (24–62.5)69 (59–88)
Figure 2.

Seasonal fluctuation of biotic and abiotic factors of Triatoma brasiliensis habitat.

As shown in Table 3, the T. cruzi-infection rate was significantly lower in September compared to July and April (p = 0.010 and p = 0.035, respectively), with no differences with respect to the developmental stages. Infected triatomines were captured exclusively at five collection sites, three of which were 5–10 m apart 200 m away from the second group of positive sites that were 5 m apart. The trypomastigote parasite burden was similar in insects collected during the rainy (median=107.5; interquartile range=52.5–120 trypomastigotes/μl of feces) and dry (median=102.5, IQR=67.5–252.5 trypomastigotes /μL of feces) seasons. Also, the epimastigote burden was similar during the two seasons: median=46.5, IQR=30–55 trypomastigotes/μL, during the rainy and median=47.5, IQR=42.5–187.5 trypomastigotes/μl, during the drought. Differences were not statistically significant (not shown). Table 4 shows that insect feces examined by ELISA revealed that birds (33.1%), followed by armadillos (18.8%), were the main food sources for the triatomines; these blood sources were more frequently identified in insects captured during the rainy season. Among the 17 T. cruzi-positive insects, nine could have its food source characterized: three had fed on armadillos and four had fed on birds. One insect was positive for hemolymph, and one to hemolymph plus bird. Among the three T. cruzi-positive triatomines which feed on armadillos, two were captured on the same day, on the same rock. The third one was collected four days later, on a rock 15 m away. Concerning the reproductive status, among 24 adult females examined, 19 (73.1%) were fertilized. Fertile females were collected throughout the entire study period. Females captured in July and December laid the higher (median=42, IQR=30–50) and lower (median=21, IQR=10–36) number of eggs, respectively (p=0,201; Mann-Whitney test).

Table 3.  Infection by T. cruzi in T. brasiliensis from a sylvatic environment in Ceará, Brazil, according to stage and capture period.
 AdultsNymphsTotal
  1. ap= 0.010; bp= 0.035.

 MalesFemalesN3N4N5 
April      
Positive/Examined0/160/41/74/442/617/132a
% positive (95% CI)0014.3 (0.4–57.9)9.1 (2.5–21.7)3.3 (0.4–11.3)5.3 (2.2–10.6)
July      
Positive/Examined1/100/30/11/83/245/46b
% positive (95% CI)10 (0.3–44.5)0012.5 (0.3–52.7)12.5 (2.7–32.4)10.9 (3.6–23.6)
September      
Positive/Examined0/250/110/120/200/210/89ab
% positive (95% CI)000000
December      
Positive/Examined3/342/80/390/380/505/169
% positive (95% CI)8.8 (1.9–23.7)25 (3.2–65.1)0003 (1–6.8)
Table 4.  Source of blood, as identified by indirect ELISA, in T. brasiliensis captured in rock piles.
 AprilJuly/September/December
 positive/tested% positive (95% CI)positive/tested% positive (95% CI)p
Blood source     
Reactive for any tested antisera54/7275 (63.4–84.5)37/8842 (31.6–53)<0.001
Bird32/7244.4 (32.7–56.6)21/8823.9 (15.4–34.1)0.005
Armadillo21/7229.2 (19–41.1)9/8810.2 (4.8–18.5)0.002
Haemolymph14/7219.4 (11.1–30.5)1/881.1 (0–6.2)<0.001
Opossum2/722.8 (0.3–9.7)5/885.7 (1.9–12.8)0.371
Rodent0/720 (0–5)3/883.4 (0.7–9.6)0.113
Equine3/724.2 (0.9–11.7)1/881.1 (0–6.2)0.221

DISCUSSION

The habitats of wild Triatominae are important to know because they provide the potential sources of household infestation or re-infestation (Lima and Sarquis 2008). In endemic areas, T. brasiliensis is able to adapt to distinct ecotopes and to act as a link between the natural enzootic and the artificial domestic cycle of Chagas disease (Borges et al. 2005).

In this study we examined some ecological characteristics of sylvatic T. brasiliensis. Our strategy of not dislodging substances during the collection of insects allowed the observation of some interesting findings regarding T. brasiliensis behavior in its natural habitat. Starting at 18:00, the insects initiated active movement upon the rock surfaces and continued for approximately 90 min, coinciding with dusk and temperature decrease. This behavior has also been described under laboratory conditions (Lazzari 1992). It has been proposed that triatomines leave their hiding places in search of food at sunset, since this period coincides with the activity of many mammals, one of their routine sources of blood.

Although it was not possible to estimate the actual number of bugs living in each collection site, we inferred that considering the number of insects that leave their hiding places to search for food during dusk, the T. brasiliensis sylvatic population size seems to vary along space and seasonality. Interestingly, although some collection sites displayed a reduction in the number of individuals at some times, the population never disappeared completely. This fluctuation is probably associated with changes in the availability of trophic resources following the dry season. Our data suggest that the greatest apparent density per collection site occurs in April, when the median number of insects per rock was higher, suggesting that the population of triatomines was concentrated in a fewer number of sites. However, the greatest abundance and dispersion was noticed in December, during the dry season. In fact, increasing numbers of insects could be collected throughout the dry season (51 in July, 97 in September, 178 in December). These data illustrate the adaptation of T. brasiliensis to the reduction of trophic resources; the reduction in the median number of insects per collection site in the dry months could reflect an active dispersal in search of food. Additionally, the large proportion of 3rd instar nymphs collected at the end of the dry period (December), combined with the absence of this stage in July, could indicate the eclosions of eggs after this month. This finding, associated with the better nutritional status and fecundity presented by adult females collected in July, suggests greater oviposition activity in this month in the sylvatic life cycle of T. brasiliensis.

Similar population fluctuations in R. neglectus, which also was able to maintain a minimum number of individuals even in months of nutritional privation, were observed (Diotaiuti 2007). Although it was not possible to access insects that did not leave the rocks to feed, sylvan T. brasiliensis colonies in our studied area appeared to be smaller than the large size peridomestic colonies reported in the same municipality (Sarquis et al. 2004, Sarquis et al. 2006). As previously reported, peridomestic T. brasiliensis colonies frequently have over 100 individuals per collection site (Sarquis et al. 2006). It has been observed that food availability is a key factor dictating triatomine population size in distinct ecotopes (Schofield 1980). In this context, artificial environments can provide a great availability of food, since peridomestic T. brasiliensis colonies are usually in close proximity to domestic animals. It has been suggested that the peridomestic environment plays a crucial role in the dynamics between the sylvatic and domestic T. brasiliensis populations, acting as a bridge between them and ensuring successful adaptation to both ecotopes (Borges et al. 2005).

It has been reported that the number of generations throughout the year may also determine colony size in sylvatic populations of T. infestans, T. sordida, and T. guasayana in the Bolivian Chaco (Noireau et al. 2000). T. brasiliensis produces two generations per year under laboratory conditions (Soares et al. 2000), so we should expect to find distinct development stages in the surveyed sites throughout the study period.

In previous studies, we observed that sylvatic and peridomestic T. brasiliensis populations from the same region display distinct weight profiles (unpublished data). In general, sylvatic triatomines are about 50% lighter than bugs collected in domestic and peridomestic habitats, a characteristic also observed in T. infestans populations (Schofield et al. 1980, Lopez et al. 1999). In this study we observed that despite relatively low weights, suggesting some degree of food deprivation, the great majority of the collected triatomines had some food residue in their guts, and starved bugs were rare. It has been reported that the worst nutritional status of triatomine specimens is mainly during dry seasons, as was observed in T. pseudomaculata populations studied in northeastern Brazil (Noireau et al. 2005). Our data points to a trend for a better nutritional status in specimens collected in July (beginning of the dry season). This difference was statistically significant for 5th instar nymphs and adult females. We suggest that triatomines could be improving their nutritional status during the rainy season and bugs collected at the beginning of the drought, a transitional period, are still of greater weight. Interestingly, in our study, triatomine males were slightly heavier than captured females, but this difference was not statistically significant. Field observations for peridomestic T. infestans suggested a better nutritional status for females (Ceballos et al. 2005). Also, T. sordida and T. guasayana females have a significantly higher weight when compared to males (Noireau and Dujardin 2001). Nevertheless, López et al. (1999) demonstrated that although T. infestans females have higher mean weight than males during periods of better food availability, this difference disappears during the summer.

In this investigation, the rate of T. cruzi infection in T. brasiliensis wild populations was relatively low and restricted to a few collection sites. This result contrasts with the higher infection rates not only in sylvatic T. infestans in the Bolivian Cochabamba Valleys (Cortez et al. 2006), but also in T. brasiliensis peridomicile populations of the same municipality (Sarquis et al. 2006). Differences in food sources should explain distinct T. cruzi-infection profiles between T. brasiliensis captured in different ecotopes and seasons of the year. In this context, we previously demonstrated that many T. brasiliensis captured in the peridomiciles of the studied municipality feed on rodents (7.5%), marsupials (8.3%), goats (13.3%), cats (5.5%), and sheep (12.8%)1, while in the present study T. brasiliensis collected in sylvatic ecotopes had a relatively poor pool of food resources, feeding almost exclusively on birds, members of the family Dasypodidae (armadillos), and probably insects. Birds are the most common vertebrates nesting among the rocks even after prolonged drought and are refractory to T. cruzi infection.

Food sources were better assessedin insects captured during the rainy season. During the dry season, despite the detection of some content in the insect gut, it was often not possible to identify the specific food source, as the residues were in advanced stages of digestion, reducing the sensitivity of the assay. The proportion of T.cruzi-infected triatomines was significantly higher in the rainy season (April) and at the beginning of the drought (July), probably reflecting a relatively better availability of blood sources, as discussed above. Armadillos were the very first wild mammals found to be infected with T. cruzi (Chagas 1909). T. cruzi-infected armadillos were observed in many countries on the American continent (Noireau et al. 2009). In addition, molecular genotyping studies have revealed that armadillos are mainly associated with T. cruzi-II (Yeo et al. 2005). In this study, analysis of the gut content of T. cruzi-infected triatomines identified armadillos as one of the main food sources and the only mammal reservoirs, suggesting that it plays a crucial role in the T. cruzi sylvatic cycle. Interestingly, T. cruzi-positive insects that feed on armadillos were captured in the same four-day interval and in close proximity to each other, suggesting that they could be infected by the same mammal reservoir. We argue that in the sylvatic environment, T. brasiliensis could prefer to feed on armadillos even when other blood sources are available in a given ecotope. Nonetheless, T. cruzi genotype Tc III, Tc II and Tc I have been found infecting T. brasiliensis in northeastern Brazil. This suggests that probably rodents and opossums, and also humans, have been utilized as food sources for these T. brasiliensis (Camara et al. 2010). Alencar (1987) observed that chickens and goats were the main food sources of T. brasiliensis. Interestingly, this author found a very low rate of reactivity to rodent antiserum, similar to our data in Jaguaruana. However the insect populations assessed by Alencar (1987) were not classified by collection site and it does not allow a direct comparison with our results. Examined together, these data emphasize the importance of sylvatic T. brasiliensis in the ecology of Chagas disease in the Caatinga ecosystem. Considering the dependence of T. cruzi transmission on the population dynamics of vectors and on the contact rate between vectors and reservoirs (Catalá et al. 1991), it is expected that, in the study area, the mobility of T. brasiliensis from one site to another can be low among nymphal instars. This might limit the transmission only to a few habitats harboring infected reservoirs. However, little is known about the dynamics of T. cruzi transmission in the wild xerophytic ecosystem of the Brazilian Caatinga.

In conclusion, we hypothesize that ecological differences concerning the rate of T. cruzi infection, gut content reactivity to tested antisera, dispersion, nutritional status, and fecundity of sylvatic T. brasiliensis follow a trend where two periods could be defined, from April (rainy) to July (transitional period at the beginning of the drought) and from September to December (the driest months). In this context, T. brasiliensis specimens collected in the drought tend to: i) present lower rates of T. cruzi infection and gut content reactivity to tested antisera, ii) have a poorer nutritional status, iii) exhibit lower fecundity, iv) be more dispersed among the studied collection sites, and v) be more abundant and easily collected on the surface of the rocks, possibly reflecting an increased activity in searching for blood meals. Such differences may underscore epidemiological concerns once they allow inferences to be made regarding the season when triatomines can more frequently invade the peridomestic environment in search of food and recolonize artificial structures. This knowledge should improve control measures directed against T. brasiliensis, the principal Chagas disease vector in northeastern Brazil.

Acknowledgments

Thanks to the Secretary of Health of the State of Ceará, the City Hall of Jaguaruana, Ceará, and Ana Luísa Barbosa, for technical assistance, transportation, and physical facilities; Marcos Eduardo Melo and Cleber Cesar Ramos, for assistance with the field work. We thank three anonymous reviewers who gave invaluable suggestions that improved substantially the paper. This study received financial support from Fiocruz and Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (Faperj).

Ancillary