Rhodnius (Stål, 1859) (Hemiptera, Triatominae) genus in Bolivian Amazonia: a risk for human populations?

Background Chagas disease, one of the most important neglected tropical diseases in the countries of Latin America, is considered to be a particularly important public health concern in the Amazon region due to increases in the number of outbreaks of acute Chagas disease and increased local transmission in the last 20 years. However, relative to other countries, in Bolivia there is little information available on its transmission in the Amazon region. The aim of this study was to investigate the infestation of palm trees, the main habitat of Triatominae in the region, in several localities, to evaluate the danger they represent to inhabitants. Methods Triatominae were collected using live bait traps left overnight in six localities in Pando and Beni Departments, Bolivia. DNA extraction and sequencing were used to establish the Triatominae species (Cytb, 16S and 28S-D2 gene fragments), and the blood meal sources (Cytb fragment). Trypanosoma sp. infection was analyzed by sequencing gene fragments (GPX, GPI, HMCOAR, LAP, PDH and COII) or by mini-exon multiplex PCR. Results A total of 325 Rhodnius were captured (97.3% of nymphs) from the 1200 traps placed in 238 palm trees and 32 burrows/ground holes. Sequence analyses on DNA extracted from 114 insects and phylogeny analysis identified two triatomine species: Rhodnius stali (17%) and Rhodnius montenegrensis (equated to Rhodnius robustus II, 83%). These were found in palm trees of the genera Attalea (69%), Astrocaryum (13%), Copernicia (12%), Euterpe (2%) and Acrocomia (1%). The infection rate was around 30% (165 analyzed insects), with 90% of analyzed insects infected by Trypanosoma cruzi (only the TcI discrete typing unit was detected), 3% infected by Trypanosoma rangeli (first time found in Bolivian Triatominae) and 7% infected by mixed T. cruzi (TcI)-T. rangeli. Rhodnius specimens fed on Didelphidae, rodents, gecko and humans. Conclusions The results of this study highlight the epidemiological importance of Rhodnius in the Bolivian Amazon region. The huge geographical distribution of Rhodnius and their proximity to the human dwellings, high infection rate and frequent meals on the human population highlight a risk of transmission of Chagas disease in the region. Graphical Abstract Supplementary Information The online version contains supplementary material available at 10.1186/s13071-022-05423-3.

Background Chagas disease remains one of the most important neglected diseases in the countries of Latin America. It is estimated that 6-7 million people are infected worldwide [1]. The disease is caused by a parasite, Trypanosoma cruzi (Chagas, 1909), which is mainly transmitted by hematophagous insects of the subfamily Triatominae (Hemiptera: Reduviidae). Other less frequent modes of transmission include congenital transmission, transfusion of contaminated blood products, transplantation of infected organs and oral transmission through ingestion of food or drinks contaminated with infected triatomine feces [1].
The Amazon region has long been considered a nonendemic region, despite the general belief that T. cruzi has circulated in the sylvatic environment for millions of years [2]. The first human case of Chagas disease in this region was reported in French Guiana in 1941 [2]. Since then, acute cases of Chagas disease have been diagnosed in almost all Pan-Amazonian countries [2]. In 2004, the Amazon Countries Initiative (AMCHA) was launched to monitor and prevent the disease in the nine countries sharing the Amazon rainforest: Bolivia, Brazil, Colombia, Ecuador, Guyana, French Guiana, Peru, Surinam and Venezuela [3]. An international workshop held in 2002 concluded that although disease occurrence remains sporadic, there is a high potential for spread due to the wide distribution of infected vectors in the region and intensive human migration [4]. In Brazil, national surveys and other studies have reported various outbreaks of acute Chagas disease in the Amazon region since 1975, mostly due to oral transmission [2]. The risks of endemic Chagas disease in the Amazon region are: (i) presence of a huge reservoir of wild mammals infected with T. cruzi; (ii) the presence of Triatominae found naturally infected with T. cruzi; (iii) the occurrence of acute oral Chagas disease transmission; (iv) uncontrolled deforestation; (v) the increasing migration of people and their domestic animals from other Chagas disease endemic regions; (vi) the lack of knowledge of Chagas disease transmission; and (vii) the poor surveillance campaigns in the region [2]. Despite the efforts and support of the Pan American Health Organization, very little has been done to fulfill the goals of the AMCHA [3].
In Bolivia, 60% of the territory is endemic to Chagas disease due to household infestation by the main vector Triatoma infestans (Klug, 1834). Since 1999, the authorities have prioritized the control of Chagas disease by focusing on this vector, using such strategies as insecticide spraying, screening blood donations and providing treatment for all acute cases and infected children under the age of 15 years [5]. The Amazon region of Bolivia represents 36-40% of the country's territory and extends completely across the two departments of Beni and Pando and partly across the departments of La Paz, Cochabamba and Santa Cruz [5][6][7]. The Amazon region has been considered to date to be a non-endemic area for Chagas disease [5]. However, at least seven species of Triatominae have been reported to live in this region, namely Rhodnius pictipes (Stål, 1872), Rhodnius robustus (Larousse, 1927), Rhodnius stali (Lent, Jurberg & Galvão, 1993), Panstrongylus geniculatus (Latreille, 1811), Microtriatoma trinidadensis (Lent, 1951) and Eratyrus mucronatus (Stål, 1859), with some of these found in peridomestic and domestic environments [8][9][10][11][12][13][14][15]. In these areas, palm trees are known to be epidemiologically important in Chagas disease; in addition to providing a refuge for many mammals and birds, they also have the potential to contain large numbers of Triatominae, mostly of genus Rhodnius [16][17][18]. Palms are often of great importance to the local human population who used them mainly for house construction, food and handicrafts. These palms can be found throughout wild forest and near dwellings.
Very little is known about Chagas disease in the Bolivian Amazon region in terms of the presence and abundance of triatomine species or the infection rate of the human population. In the Amazonian part of the department of La Paz, a process of domiciliation of R. stali has been described in the peridomicile, and local transmission of Chagas disease to the human population has been detected [10,13,19]; this species has also been captured in palms in the same region [11]. An outbreak of acute Chagas disease by oral transmission has been reported in Beni Department, municipality of Guayaramerín in 2010, as a result of the ingestion of a palm fruit juice (Oenocarpus bataua; Majo in Bolivian) [19,20]. A study in a small town at the foot of the Andes (Beni Department, on the border with La Paz Department) showed the presence of R. robustus and probably R. stali in palms [21]. Finally, in the tropics region of Cochabamba (Municipality of Villa Tunari), P. geniculatus (16 specimens) and R. robustus (5 specimens, species determined only by morphological examination) were found mainly in intra-and peridomiciles; 57% of these were infected by T. cruzi [22].
The identification of Triatominae, and in particular of Rhodnius species, is currently the subject of intense research, as many species belong to cryptic species complexes [23]. Three groups of Rhodnius are generally described: pictipes (7 species), prolixus (11 species) and pallescens (3 species) [24,25]. Within the prolixus group, R. prolixus (Stål, 1859), R. robustus, R. montenegrensis (da Rosa et al., 2012) and R. marabaensis (Souza et al., 2016) are very close morphologically (sibling species, or pseudo-cryptic species, see [26]) and are generally grouped under the compound term of "R. prolixus-R. robustus cryptic-species complex" [27]. This complex is currently under study, and a recent publication has characterized all five nymphal instars and established the differences between R. prolixus, R. robustus and R. marabaensis in each of their instars [28]. Studies have shown that R. prolixus is a monophyletic group whereas R. robustus sensu lato appears to be paraphyletic with five subclades noted I to V [23,29,30]. Recently, the paraphyletic lineage R. robustus II was described as R. montenegrensis [27].  (Zhao, Galvão & Cai, 2021), the latter being a newly described species from Bolivia [24].
The aim of the present study was to increase basic entomological knowledge of Chagas disease in the Bolivian departments of Beni and Pando and to obtain a preliminary assessment of the risk that wild Triatominae can represent in this region. These two departments cover a vast area of approximately 278,000 km 2 . Six cities were selected as study sites, and the exploration for Triatominae was done by passive searching of palms using traps. In parallel to this study, a survey on the incidence of Chagas disease in the population of these six cities was carried out in collaboration with the representative of the Amazonian Indigenous People, with the results showing that between 2% and 12% of the population had positive serological test results for Chagas disease [32].

Area description
The department of Pando, located in the north of Bolivia, on the border with Brazil and Peru, in the Amazon rainforest, receives between 1500 and 2100 mm of rain per year (https:// www. world wildl ife. org/ ecore gions/ nt0166). The department has a surface area of 63,827 km 2 , and a population of 110,000 inhabitants. The department of Beni is located in the country's frequently flooded lowlands area and is characterized by the presence of savannas. The countryside shows a succession of forest fragments and pastures, with scattered ranches. Beni has a surface area of 213,564 km 2 , and a population of 420,000 inhabitants. In both departments, poverty levels are high, and access is difficult, especially during the rainy season. According to Census 2012 (https:// www. ine. gob.  bo/ index. php/ censos-y-banco-de-datos/ censos/), the houses are typically made of bricks (46 ± 17%) or wood (34 ± 14%) walls, with dirt (46 ± 22%) or cement/tiled (42 ± 8%) floors, and with roofs of sheet metal (42 ± 29%), tiles (37 ± 28%) or palms (15 ± 14%).

Selection of capture locations
The presence of triatomines was investigated in six different locations during two different trips. The cities of Cobija, Riberalta and Guayaramerín were surveyed in August 2016, and those of Trinidad, San Borja and San Joaquin were surveyed in August 2017 (Fig. 1). All six cities are situated at an altitude falling within the range of 140-280 m a.s.l. Trinidad (capital of Beni department) has the highest number of inhabitants (125,000), followed by Riberalta (99,000), Cobija (capital of Pando department, 56,000), Guayaramerín (39,000), San Borja (25,000) and San Joaquin (4500).
For each location, three capture points (only 2 in Guayaramerín) were selected within a 30-km radius of the city. These were named using the initial of the city (SB: San Borja, SJ: San Joaquin, T: Trinidad, C: Cobija, G: Guayaramerín, R: Riberalta) and a number ranging from 1 to 3. These points were characterized by the presence of palms and accessibility to our team (no flooding and permission of local health authorities and/or owners). The environment of each capture point was described based on Google Earth photos using a circle of 500 m in radius around the capture point. The categories of land use were crops, forest, water, pasture/herbaceous, pasture/herbaceous with isolated palms, road, construction, open building plot and others. This last category comprised vacant land, land without vegetation and peridomiciles. The area of each category for each capture point was calculated (see details of surfaces and illustration of each capture point in Additional file 1: Figure S1). A principal component analysis (PCA) was realized with all of the capture points and land use categories using the psych [33], FactoMineR [34] and ggrepel [35] packages in R software [36]. The number of principal components (PCs) to be kept was chosen based on the Kaiser criterion (SS loadings > 1). Each component was explained using the correlation between axes and variables, and the square cosine value (Cos2) of the variables. The V-test was used to detect whether supplementary qualitative variables were significantly different from 0 [37].
Data on triatomine detection/non-detection in palms were stratified by ecoregion (rain forest/flooded lowlands-savannas), landscape, locality and palm species, with the percentage of palms with at least one captured triatomine and its 95% confidence interval (CI; infestation index) given for each stratum. For comparison within a stratum, conditional maximum-likelihood odds ratios (ORs) with mid-P exact 95% CI were calculated, using OpenEpi 3.01. Finally, following the methodology presented in [16], single-season siteoccupancy models were used to estimate the prevalence of palm infestation and to investigate ecological correlates of infestation taking into account detection failure [38,39]. These models were run in the Presence v.2.13.14 software package [40]. The result of each sampling event, i.e. a trap in one palm, was coded 1 if at least one triatomine was captured, and 0 otherwise. As around five traps were placed overnight at the same time for the same palm, this coding provided a history of detection for each palm to build the models (see [16] for more details). Ecological covariates included ecoregion [Amazonian Forest vs flooded lowlands (savanna)], landscape and palm genus.

Triatomine collection
Triatominae were captured using adhesive bait-traps with mice mostly settled in palms [41], only during one night without repetition. In each capture point, at least five palms were explored for the presence of Triatominae by positioning four to five traps inside the crown of each palm using a 10-m ladder in the late afternoon. Some additional traps were placed, when possible, in holes in the ground. Each trapping site was described, photographed and georeferenced. On the next morning, Triatominae stuck on each trap were collected and the genus and stages of the insect recorded. Adults were transported alive to the laboratory for species determination using dichotomic keys [42,43] prior to molecular characterization, while the nymphal instars were preserved in 70% alcohol (1 specimen per 1.5-ml tube). The project (field work and use of mice for triatomine capture) was approved by the Bolivian National Bioethics Committee, Department of Research Ethics (CEI-CNB).

DNA extraction
DNA was extracted to determine the species of triatomine and to identify the blood meal source. For young nymphal-instars (N1-N3), DNA extraction was performed using the total insect. For the oldest nymphal instars (N4, N5) and adults, specimens were first dissected, separating the legs, head and thorax into one tube and the abdomen into another. Equipment used for handling insects was decontaminated with a 10% chlorine solution and rinsed with bi-distilled water before and between each insect dissection. The universal rapid saltextraction method was applied to all samples [44]. The method is described in detail in Additional file 1: Text S1a.
Due to financial limitations, DNA extraction and analysis were not possible for all specimens. Thus, the processed samples were considered to provide genetic data that were representative coverage of all capture points and positive palms. When fewer than 15 insects were collected at a capture point, DNA was extracted on all the insects to cover the entire available genetic pool. In contrast, when the number of Triatominae collected at a capture point was higher (> 70), only one to two insects per positive palm were processed to cover the greatest genetic variation at that particular capture point. When possible, analysis of the oldest nymphal instars was preferred to that of the youngest.

Molecular characterization of Triatominae and blood meal source
Characterization of Triatominae was based on the amplification of two fragments of mitochondrial genes, the large ribosomal subunit 16S (LSU-rRNA) and cytochrome B (Cytb), and one fragment of a nuclear gene, the D2 variable region of the 28S RNA gene (28S-D2). The primer sequences used are described in Additional file 1: Table S1 [30,45,46], and the methodology is described in Additional file 1: Text S1b. The characterization of the blood meal source was achieved using the set of primers for the Cytb fragment (Additional file 1: Table S1) [47], using the methodology described in Additional file 1: Text S1c. Negative control tubes were filled with water and handled in a manner similar to the other tubes containing the insect samples; these were used in each PCR batch.
Both methods (PCR sequencing and MMPCR) lead to the identification of T. cruzi DTUs, but the former gives results about haplotypes while the latter does not. As PCR sequencing is very expensive, MMPCR was used instead of this method for the large collection of bugs in year 2 (2017). Negative control tubes were filled with water and handled in a manner similar to the other tubes containing insect samples; these were used in each batch of PCR.

Sequence analysis
Direct sequencing of both strands of the amplified PCR products was performed at Macrogen Inc. (Seoul, Korea). The sequences were edited, corrected and aligned using MEGA7 software [51]. Haplotypes were calculated using DnaSP5 software [52]. The BLAST algorithm was used to compare the sequences with those deposited in GenBank (http:// www. ncbi. nlm. nih. gov/ BLAST). Vector species (Rhodnius) were characterized by constructing phylogenetic trees (maximum likelihood) from the obtained sequences and the sequences were deposited in GenBank using IQ-TREE [53]. The best model was determined by ModelFinder [54], and the phylogeny was tested using 1000 ultrafast bootstrap [55]. As the UFBoot values correspond roughly to the probability that a clade is true, a minimum of 95% is considered necessary for a clade to be supported. For each gene fragment, two trees were built: (i) using Rhodnius GenBank sequences showing 100% coverage with our haplotypes; (ii) using Rhodnius GenBank sequences with less coverage. After alignment of the GenBank sequences, only one randomly selected sequence was retained per haplotype. The second tree (< 100% cover) allowed us to include more species of interest (especially those of R. pictipes and R. prolixus complexes) and observe how our haplotypes position themselves among the species. A sequence of T. infestans was used as an outgroup. Trees were drawn using the R software packages ggtree [56] and treeio [57]. Haplotype median-joining networks per locus were constructed for Cytb, 16S and 28S-D2 haplotypes with POPART [58,59].

Insect captures
A total of 1200 traps were positioned during the field study, of which 99.4% were located in 238 palms and 0.6% in 32 burrows or ground holes (Table 1). A total of 325 Triatominae were captured, all in palms and all belonging to genus Rhodnius (see Additional file 2: Table S2 for the details of each specimen). Across the whole collection of Triatominae, 12% of the traps and 28% of the palms were positive, resulting in 1.6 ± 0.7 insects per positive trap and 3.1 ± 2.7 Triatominae per positive palm. Insects were found at all six locations, with some disparities among them. Most triatomines were found near Trinidad (34% of the insects), San Joaquin (31%) and San Borja (25%). Among the study locations, the positivity of the traps and of the palms ranged from 4% to 19% and from 15% to 43%, respectively ( Table 1).
All nymphal stages were found in the six regions, except in Guayaramerín, where only N1 and N5 were captured (Additional file 1: Table S3). Adults represented only 2.7% of the captures, and were found in only one location in Cobija and in two locations in San Borja. Based on the morphological keys of genus Rhodnius [24,42,43], in Cobija (location C1; see Table 1 for site identification code ) the one captured specimen, a male, was determined to be R. robustus; in San Borja (SB2), the San Joaquin three captured specimens, one male and two females, were determined to be R. robustus (they came from three different palms); in San Borja (SB3) one female and one male from the same trap were identified as R. stali, and one male and two females from three different palms were identified as R. robustus.
Analysis of the infestation index by ecoregions (Fig. 5) revealed that palms are most likely infested in flooded lowlands (savannas) rather than in the Amazonian rainforest (OR: 2.56, 95% CI: 1.37-4.89). The relationship between palm infestation and landscape was also studied (Fig. 5). Of the four landscape categories, namely forest, pasture, peri-forest and peri-pasture (described based on the environment of each capture sites; see Additional file 1: Table S4; Additional file 1: Figure S1), and taking the forest landscape category as the reference category, this analysis showed that palms in the pasture landscape were 2.38-fold more likely to be infested (95% CI: 1.05-5.46), palms in peri-forest were 4.09-fold more likely to be infested (95% CI: 2.00-8.67) and palms in the peri-pasture category were less likely to be infested (OR: 0, 95% CI: 0.00-0.83). In terms of localities (Fig. 5 The palm infestation index calculated over all sites was 0.121 (145 positive traps/1200 surveyed traps). A null single-season site-occupancy model that considered a constant occupancy and a constant rate of detection, estimated an overall palm occupancy rate of 0.306 (95% CI: 0.246-0.467), which is 39.5% higher than the observed index. Estimates of palm occupancy by ecoregion showed that the odds of occupancy are 2.79-fold higher at a savanna site than at a forest site (95% CI: 1.43-5.41). The palm occupancy estimate for savanna sites was 0.395 (95% CI: 0.309-0.468), which is 44% higher than the observed index (0.174); in comparison, it was only 26% higher for the Amazonian forest (estimate palm occupancy: 0.189, 95% CI: 0.120-0.285). The models on the influence of landscape showed that compared to forest sites, the odds of occupancy at sites in peri-forest (border between human settlement and forest) were 2.91-fold higher (95% CI: 1.34-6.29) and the odds of occupancy at sites in pasture were 2.53-fold higher (95% CI: 1.03-6.24). Finally, the model on palm genus, and using Copernicia as reference, showed only one significant result where the odds of occupancy of Attalea were 6.49-fold higher (95% CI: 1.84-22.85).

Molecular characterization of triatomines
DNA was extracted from 114 insects (35% of the total number of captured insects; see DNA extraction section in Methods section) for which at least one gene fragment (Cytb, 16S, or 28S-D2) was sequenced. From the 110 Cytb samples sent for sequencing, 100 sequences were obtained, of which 99 were based on both forward and reverse directions, and one was obtained from the forward direction only. The details of each sequence, and the BLAST result for each, are presented in Additional file 2: Table S2. Two sequences were removed from the haplotype analyses, one being too short compared to the others (only 465 bp), and one having two polymorphic loci (C/T). Hence, based on the 98 aligned sequences (607 bp), 17 haplotypes were detected: 14 corresponding to R. robustus/R. prolixus (100% cover, 99.34-100% identity) and three belonging to R. stali (96-100% cover, 98.68-100% identity) as putative species, with 81 and 17 individuals, respectively (Additional file 1: Table S5). For the 16S fragment, 87 insects were analyzed and 72 sequences were obtained using both strands, except for one sequence based only on the reverse sequence. Eight haplotypes of 431 bp were detected, four corresponding to R. prolixus (99-100% cover, 98.83-99.70% identity) or R. robustus (69-78% cover, 98.82-100% identity) and four corresponding to R. stali (100% cover, 98.60-100% identity); the number of involved specimens was 59 and 13, respectively (Additional file 1: Table S5). For the 28S-D2 fragment gene, 29 individuals were analyzed, and 24 sequences were obtained (578 bp long after alignment), all from both strands except for one based on the forward sequence. Two sequences showing two polymorphic loci (T/C) were deleted before haplotype analysis. Four haplotypes were detected, all corresponding to the putative species R. robustus (100% cover, 99.48-100% identity, Additional file 1:  Table S2). All sequences (Cytb, 16S, 28S-D2) described in the present study have been deposited in GenBank (accession numbers are given in Additional file 2: Table S2).
According to the above results, the putative current species belong to two complexes: the R. pictipes complex, and the R. prolixus-R. robustus cryptic species complex. A first phylogenic analysis of Cytb sequences with those    Table S4; Additional file 1: Figure S1): forest (R1, R2, R3, G1, SJ1, SB2), pasture (T1, T2, SJ2), peri-forest (G2, C1, C2, C3 SJ3, SB3) and peri-pasture (T3, SB1) UFboot, apart from R. pictipes (Fig. 6). These results confirm the putative species identified by the BLAST analysis. Using shorter sequences (369 bp) it was possible to include R. brethesi and R. montenegrensis species in the Cytb analysis (Additional file 1: Figure S2). In this case, haplotypes 1-14 are still grouped in the R. robustus II clade with 98% UFBoot support, along with the R. montenegrensis sequence. It is worth noting that due to the use of a shorter sequence length (369 bp), the species of the R. pictipes complex are no longer well discriminated, i.e.   Figure S2). Phylogenic analyses of the 16S gene fragment using 100% coverage resulted in the inclusion of nine GenBank sequences belonging to eight Rhodnius species (Additional file 1: Figure S3). Hap1 to hap4 were included in a clade comprising R. domesticus, R. neglectus, R. neivai and R. prolixus, with 99% UFBoot support, but without separation between species, well separated from hap5 to hap8 that clustered with R. stali, R. brethesi and R. pictipes. The tree using the sequences with < 100% coverage allowed the inclusion of more species, especially R. robustus (Additional file 1: Figure S3). Hap1 to hap4 were included in a clade with R. robustus, R. neglectus, R. prolixus, R. domesticus and R. neivai with 96% UFboot support. Hap5 to hap8 were separate from the R. pictipes clade (96% UFboot), but they were not distinguishable between R. stali and R. brethesi.
For the 28S-D2 fragment, only six sequences belonging to three species of genus Rhodnius with 100% coverage with our haplotypes (578 bp) were available in GenBank: R. robustus, R. prolixus and R. nasutus; no species conclusion can be extracted from the phylogenetic tree (Additional file 1: Figure S4). Using shorter sequences, 21 GenBank sequences belonging to nine species were included in the phylogenic tree. The four haplotypes were classified in a clade (96% UFBoot) comprising R. colombiensis, R. ecuadoriensis, R. nasutus, R. neglectus, R. pallescens, R. prolixus and R. robustus, separated from R. stali and R. pictipes.
Our results appear to show some geographic structuring of haplotypes after construction of the network,  Table 2). For R. robustus, this involves hap1 and hap2, which were found only in Trinidad; hap4, hap5 and hap6, in San Joaquin; hap7, hap8, hap9 and hap10 in San Borja; hap13 in Cobija; and hap14 in Riberalta. The remaining haplotypes were shared in different regions: hap3 was shared between San Borja and Trinidad; hap11, between Cobija and Guayaramerín; and hap12, between Cobija and Riberalta. San Borja and Cobija showed the greatest haplotype diversity for Cytb (Table 2). For R. stali, each of the three haplotypes was found in only one region: hap15 in Trinidad, hap16 in San Borja and hap17 in Riberalta. For the 16S characterizing R. robustus, hap1 was found only in San Borja, hap2 only in San Joaquin and hap3 and hap4 in the three regions of San Borja, San Joaquin and Trinidad. For the 16S characterizing

Blood meal source
A total of 63 blood meal sources could be determined from 68 DNA extractions, with all sequences based on the two strands (Additional file 2: Table S2). These were determined from DNA extracted from 61 nymphal instars (2 N2, 32 N3, 19 N4 and 8 N5) and two adults (1 male and 1 female), from collection sites at San Borja, San Joaquin and Trinidad. Seven blood meals were attributed to Mus musculus (Linnaeus, 1758) and were likely due to insects that fed on the mouse-bait used for trapping. Four other blood meals were detected from two species of Didelphidae (Didelphinae subfamily), all in San Borja (SB3): two blood meals on Philander canus (Osgood, 1913), from two nymphs captured in two different palms, and two blood meals on Metachirus nudicaudatus (Geoffroy, 1803), from one N3 and one male captured with the same trap. Both species of Didelphidae are reported in Beni Department, Bolivia. Three blood meals were attributed to rodents, all from nymphs from Trinidad (T1): two on Oecomys mamorae (Thomas, 1906) (Cricetidae family, Sigmodontinae subfamily), a widespread species in the country, from two insects captured in two different palms; and one blood meal on Coendou prehensilis (Linnaeus, 1758) or Coendou bicolor (Tschudi, 1844) (Erethizontidae family), both reported from the region. One blood meal was attributed to Reptilia species: Hemidactylus mabouia (Moreau de Jonnès, 1818) (Gekkonidae family), found in a nymph from San Joaquin (SJ3). This species, native to Africa, has been introduced by humans and is widely reported in South America. Finally, 48 blood meals had a human origin; these were found in San Borja [in SB2 (1 female) and in SB3 (11 nymphs in 7 palms], in San Joaquin (SJ3; 19 nymphs in 5 palms) and in Trinidad (T1; 17 nymphs in 7 palms). According to the description of the capture points (Additional file 1: Figure S1; Additional file 1: Table S4), SB2 is located in a forest area, with the nearest house 41 m away, SB3 is located in an indigenous village, with the shortest distance to a house being 9 m, SJ3 is a brickyard, with the smallest  distance to a house being 9 m, and T1 was in a forest fragment at the border of a pasture, in front of a ranch (58 m). All individuals that had fed on human blood were found in A. phalerata (Additional file 2: Table S2). It is worth noting that the negative controls used in each batch of PCR never amplified any DNA.

Infection
A total of 165 insects were analyzed for detection and identification of Trypanosoma sp., of which 136 were analyzed using MMPCR and 29 were analyzed by sequencing of gene fragments. From these, 48 samples were found to be positive (Additional file 2: Table S2), infected by T. cruzi (42 insects, 87.5%), T. rangeli (2 insects, 4.2%) and a mixed infection T. cruzi-T. rangeli (4 insects, 8.3%), as shown after agarose gel electrophoresis (Additional file 1: Figure S5). All T. cruzi belonged to the TcI strain (Additional file 2: Table S2). The infection rate by locality was 44.4% (20 infected/45 insects) in San Borja, 24.5% (12/49) in Trinidad and 16.7% (7/42) in San Joaquin, with a difference observed between capture points within a locality (Fig. 8). Trypanosoma rangeli was found in Trinidad and in San Borja. According to the putative species, 35% (22/62) of the R. robustus and 21% (3/14) of the R. stali were found infected by Trypanosoma sp. Trypanosoma rangeli was found in one R. robustus (mixed infection).
Of the 32 insects captured in Cobija, Guayaramerín and Riberalta, 29 were studied for Trypanosoma infection by sequencing several gene fragments from positive and negative PCR products: six fragments in four insects and two fragments (GPI and GPX) in 25 insects were sequenced. Readable sequences were obtained for only nine insects, with all readable sequences corresponding to the TcI DTU (Additional file 2: Table S2). The number of infected insects per locality was seven in Cobija (2 N2, 3 N4, 2 N5), one N5 in Guayaramerín and one N5 in Riberalta (Fig. 8). Seven infected insects were determined to be R. robustus; the other two insects were unidentified.

Triatomine abundance
Knowledge about the vectors of Chagas disease in the Bolivian Amazon region is scarce. This study, with its wide geographical range, highlights the ubiquity of Rhodnius in the region, which is an important novelty. Indeed, these insects were captured in each of the six study localities. Insect abundance (> 80 vs < 17 insects) and trap positivity (> 15% vs < 6%) were higher in the three flooded lowland (savanna) capture points than in the three Amazonian forest ones surveyed 1 year before. Palm occupancy models showed that palms in flooded lowlands are 2.8-fold more likely to be infested than palms in the Amazonian forest. Our estimates are in the range of previous estimates found for central (24%) and southwestern Amazonia (37%) [16]. The landscape is also important. According to our results, palms at the border between human settlement and forest are 2.9-fold more likely to be infested than palms in forest, and that palms in a pasture environment are 2.5-fold more likely to be infested than palms in the forest. Although all collections were conducted during the dry season (August) and with the same methodology, local variations (climate, wildlife, etc.) between years may explain these differences between savanna and Amazonian forest. More specifically, these regions are under the influence of global climate variations, such as El Niño and La Niña, which can result in drought or flood years. Precipitation data collected from the SENAMHI (National Service of Meteorology and Hydrology, Bolivia; http:// senam hi. gob. bo) at each collection point except Guayaramerín (for which no data were available) showed that the total amount of precipitation during the year preceding the field missions was higher for San Borja (4265 mm), similar for San Joaquin (3308 mm) and Riberalta (3300 mm) and lower for Cobija (2986 mm) and Trinidad (2975 mm). Thus, the total amount of precipitation during this previous year does not seem to explain the differences observed between the flooded lowlands and the Amazonian forest ecoregions, although the importance of monthly variation cannot be discarded. Further studies with a specific design should be carried out to truly assess the abundance and stability of Rhodnius populations over time, study the possible factors related to annual variation and confirm the present results. It is important to note that very few adults (2.7%) were found, while all the nymphal instars were well represented. The density (no. of Rhodnius/palm: 0.13-8.16) and crowding (no. of Rhodnius/positive palm: 1.00-9.80) indices observed in this study are similar to those obtained in a previous study in Bolivia conducted in Yucumo, 50 km southwest of San Borja (density: 0.06-1.46; crowding: 1.33-6.33 [21]). They are also similar to those obtained in a study of R. ecuadoriensis in Ecuador (density: 0.06-2.55, crowding: 1.00-10.21 [17]).

Triatomine species and geographic distribution
In Bolivia, four species of Rhodnius have been described to date: R. robustus, R. stali, R. pictipes and newly described R. micki [24]. Our results using the Cytb fragment demonstrate that hap1 to hapl14 are clustered with clade II of R. robustus, now named R. montenegrensis [27]. Cytb hap15 to hap17 were identified as R. stali species, with a high Uboot support value of 98%. Rhodnius micki, a newly described species from the R. pictipes complex, was not included in this analysis due to the lack of sequences available in GenBank. Nevertheless, the adults captured in this study were morphologically consistent with R. stali and not with the description of R. micki [24]. In conclusion, all analyzed insects belong to two species: R. montenegrensis and R. stali. This makes the existence and true distribution of R. robustus sensu stricto in Bolivia doubtful, and more sampling and genetic analysis are essential to resolve this important issue.
Rhodnius montenegrensis was found in all six geographic localities, whereas R. stali was only captured in San Borja, Trinidad and Riberalta. The number of specimens was also higher for R. montenegrensis. Interestingly, R. stali was predominant in a previous study of Rhodnius in palms in Yucumo, 50 km southwest of San Borja (Andean foothills, October 2012), where the authors identified as putative species 23 R. stali and one R. robustus [21]. In the Alto Ben region of La Paz Department, 65 km south of Yucumo, R. stali was the only Rhodnius species found in palms (Attalea phalerata) [11]; this species was very abundant and in a peridomicile domiciliation process in the same region in 2006 [10]. This panorama allows us to speculate that R. stali, although widely distributed in the lowlands of the Amazon region of Bolivia, could be more abundant at the foot of the Andes. Rhodnius robustus (identified only by morphology, so potentially R. montenegrensis) was also captured in the municipality of Villa Tunari of the department of Cochabamba [22]. Therefore, it would appear that R. montenegrensis has an extensive home range in Bolivia. For both species, with the majority of haplotypes observed in only one location, a geographic structuring of haplotypes seems to exist, which was especially highlighted with Cytb. Haplotype diversity tended to be highest in San Borja, which is the closest point to the Andean foothills and an entry point into the Amazonian region for the human population. Interestingly, the four sequences collected in Riberalta corresponded to three haplotypes, also suggesting high diversity in the Amazonian forest. Further research on insect collection and haplotype diversity should be conducted in the future to investigate diversity in these regions.

Infection of Triatominae by Trypanosoma
Both species were detected naturally infected with Trypanosoma sp., with an infection rate of 35% for R. montenegrensis and 21% for R. stali. Most infections were due to T. cruzi, specifically for the DTU TcI. A variation in infection rate was observed between localities, being higher in San Borja (44.4%) and lower in San Joaquin (16.7%), for a similar number of Triatominae tested (45 vs 42, respectively). Six specimens were found to be infected with T. rangeli (4 individuals with mixed infection T. rangeli-TcI). This is interesting because T. rangeli is known to occur in primates (Squirrel Monkey) in Bolivia [60], but this is the first time this species has been shown in Triatominae. Trypanosoma rangeli is a well-known parasite that is restricted to Central and South America, of which Bolivia is the southern border. It can infect a wide range of mammals, including humans (not pathogenic in humans), and is mainly transmitted by Rhodnius species. In species such as R. ecuadoriensis [61] or R. pallescens [62], infection with T. rangeli and mixed T. rangeli/T. cruzi infection are common occurrences. Our results allowed us to identify one of the species infected by this parasite, namely, R. montenegrensis. Infection of this species with T. rangeli has been previously observed in Brazil [63]. Rhodnius stali has also recently been found to be infected with T. rangeli in Brazil [64]. This is a very important discovery because there is a tendency to believe that T. rangeli is not present in Bolivia. The two species are not easily distinguishable by microscopy, and cross-reactions occur with serological tests, causing false-positive results [65]. This is a problem that the Ministry of Health and the Chagas Program in Bolivia should take into account.

Palm species and their infestation
Palms are an emblematic group with > 2500 species distributed throughout the world [66]. In the Amazon rainforest, palms represent 60% of the most common species [67]. They are key species for human populations, which use their leaves for the construction of roofs, their fruits for food and handicraft, among others [68]. It is customary for local people to keep some palms near their dwellings to provide shade and fruits for domestic animals. These palms are also important for wildlife, providing shelter and food [66]. Palms have been shown to provide Triatominae with a refuge characterized by stable abiotic conditions [69]. Local wildlife in association with palms is the food source that Triatominae needs to thrive [69]. Rhodnius species have been found in close association with palms, with some species being captured exclusively in one palm species, such as R. brethesi and Leopoldinia piassaba (Wallace), while others live in various palm species [69]. In our study, R. stali was found in two palm species, i.e. Copernicia alba and Attalea phalerata, while R. montenegrensis was captured in A. phalerata, A. speciosa and A. princeps. Other non-identified Rhodnius specimens were also found in Astrocaryum sp. and Euterpe oleracea. Copernicia alba is a synonym of C. australis, where R. pictipes has been found previously [70]. Our study is the first to report the presence of R. stali in a palm species other than A. phalerata [69], although it is possible that this reported R. pictipes in R. alba was in fact R. stali. Rhodnius montenegrensis has been found previously in A. speciosa [63,71], and R. robustus has previously been found in various palms, including Astrocaryum aculeatum (G.Mey.), Astrocaryum murumuru (Mart.), Attalea phalerata and A. speciosa. Here, we report two new species for R. montenegrensis: Attalea princeps and A. phalerata; these two palms are widely distributed in Bolivia and are known locally as Motacú. Finally, Acai (E. oleracea) is known to be infested by Triatominae [Alberprosenia malheiroi (Serra, Atzingen & Serra, 1987)] [16]. Unfortunately, we were unable to determine which species inhabited these palms in Bolivia. It is interesting to note that both Rhodnius species were found twice sharing the same palm (A. phalerata), in San Borja (SB3): two N1 nymphs in one case (same trap) and one N2 (R. montenegresensis) and one N4 (R. stali) in two different traps in the other case. The crown structure and abundance of vegetable matter and epiphytes were unfortunately not described in our study. They are factors known to increase the probability of infestation and the size of the triatomine colonies in palm [17,72]. We have captured Rhodnius specimens in species such as Astrocaryum sp. and C. alba that have less developed crowns than Attalea sp. Therefore, it would be interesting to investigate this point in future work.
Triatominae were captured regardless of the environment type. Based on occupancy models, palms surrounded by pastures were 2.5-fold more likely to be infested than palms in the forest, and palms at the border between human settlements and forest were 2.9-fold more likely to be infested than palms inside the forest. A total of 48 human blood meals were detected in Triatominae in the three localities where the analysis was conducted: San Borja (SB2 and SB3), San Joaquin (SJ3) and Trinidad (T1). These blood meals were found in 47 nymphal instars, and one female, in both Rhodnius species. Looking in detail at the land use of these capture points, SB3 can be defined as a peridomicile-sylvatic (forest) area inside a Tsimané village (indigenous people of the lowlands of Bolivia), and SJ3 as peridomicile in a small brickyard neighborhood surrounded by forest. At these two capture points, the number of captured insects was high, and the smallest palm-to-building distance was 7 and 9 m, respectively. On the contrary, SB2 and T1 were in a wilder environment: a forest fragment (SB2) and a small forest fragment among pastures (T1). In both cases, a ranch was not very far away, at 41 and 58 m, respectively. This is not the first time that human blood meals have been discovered in wild nymphs of Rhodnius: for example, 7% of the blood meal detected in R. pallescens captured in palms in Panama had a human origin [76]; in Ecuador, 41.5-57.3% of R. ecuadoriensis found in the sylvatic environment had fed on human blood, half of them being nymphs [77]. These authors also showed multiple sources of blood meal in younger nymphal instars and hypothesized that cleptohematophagy (bug feeding on the blood meal ingested by another bug) had a role to play. This behavior may explain the human blood meals in nymphal instars while they can only move by walking. Adults, on the other hand, can fly several hundred meters and sometimes kilometers [78]; hence a distance of < 60 m is not that far to go and return to the forest fragment. Another explanation is that humans rest at the base of the palms so that the nymphs can reach them and return to the palm crown; although plausible, this explanation is probably less likely.

Conclusions
To what extent does Rhodnius represent a danger to the human population in the Amazon region of Bolivia? First, it is well-known that palms near dwellings represent a risk because they can be inhabited by Rhodnius populations that can invade human dwellings [16]. One of the most common palms in peridomicile area in Bolivia is Motacú (A. princeps, A. phalerata), a frequently infested palm. In this study, the odds of infestation of A. phalerata are particularly high (OR = 71.6 compared to A. speciosa and 27.8 compared to A. princeps); the OR was only 1.55 compared to A. speciosa in [16]. It should be kept in mind that in our study the number of palms was limited. Furthermore, the non-independence of samples within a capture site can lead to an overestimation of the prevalence of infestation and to an underestimation of covariate effects [16]. In the present study, Rhodnius was found in every locality, showing a wide distribution. With a T. cruzi infection rate of around 30%, these Rhodnius populations definitively represent a risk of being a T. cruzi vector, as their contact with the human population was clearly indicated by the identification of human blood meals. In addition, R. stali has been described as a vector in the Alto Beni region, being in a process of domiciliation in peridomicile [10], and R. robustus was recently found invading dwellings in the tropics of Cochabamba [22]. At the time this study was conducted, a serological diagnosis of Chagas disease was organized in these six localities [32]. It was noted at that time that the local populations had little knowledge about Chagas disease. The results of that survey also showed that the seropositivity of Chagas disease was 4.1% in Cobija (49 analyzed persons), 6.7% in Riberalta (60 persons analyzed), 6.0% in Guayaramerín (83 persons tested), 5.5% in Trinidad (55 persons analyzed), 2.1% in San Joaquin (48 persons tested) and 11.6% in San Borja (43 people tested). These results clearly show that cases of Chagas disease exist in the Bolivian Amazon. They may be due to local vector or oral transmissions, the latter being frequent in this region [19,20,79]. They may also be imported cases, as the Amazon region is a region of intense migration of people who previously lived in areas endemic for Chagas disease [2]. There is an urgent need to develop research to improve the assessment of the local transmission risks and to understand both the modes of transmission and the biological and anthropological factors related to these transmissions in order to generate surveillance and control policies if necessary.
Additional file 1: Text S1. Methodology used for the different molecular characterization of Triatominae, blood source meal and Trypanosoma. Table S1. Description of primers used in molecular characterization of triatomines, blood source meals and Trypanosoma. Table S3. Distribution of the insects by capture points and developmental stages. Figure S1. a Environmental description of capture points in Trinidad, San Joaquin and San Borja. b Environmental description of capture points in Riberalta, Guayaramerín and Cobija. Table S4. General description of the environment around each capture point inside a circle of 500 m in radius. Table S5. Description of the haplotypes found for the Cytb, 16S, and 28S-D2 gene fragments. Figure S2. Maximum likelihood tree constructed with the 17 Cytb haplotypes jointly with < 100% coverage sequences deposited in GenBank. Figure S3. Maximum likelihood tree constructed with the eight 16S haplotypes jointly with sequences deposited in GenBank. Figure S4. Maximum likelihood tree constructed with the four 28S-D2 haplotypes jointly with sequences deposited in GenBank. Figure S5. Agarose gel electrophoresis (3% agarose) of PCR-amplified products using mini-exon multiplex PCR (MMPCR). Table S2. Details on analyses and results obtained for each captured insect: capture location, species determination, blood meal origin, and infection by Trypanosoma.