Drivers of molecular and morphometric variation in Triatoma brasiliensis (Hemiptera: Reduviidae): the resolution of geometric morphometrics for populational structuring in micro-geographic scale

Background The protozoan Trypanosoma cruzi circulates in semiarid areas of the northeastern of Brazil in distinct ecotopes (sylvatic, peridomestic and domestic) where Triatoma brasiliensis is the most important Chagas disease vector. Methods We sampled populations of T. brasiliensis from distinct ecotypic and geographic sites in the Rio Grande do Norte (RN) and Paraíba (PB) States to compare the results of morphometric and genetic variations. The geometric morphometry was carried out with 13 landmarks on the right wings ( N =698) and the genetic structure was assessed by sequencing a region of cytochrome B mitochondrial gene ( N =221). Mahalanobis distance (DM) and molecular differentiation coefficient (Φ ST ) were calculated among all pairs of populations. Results The results of comparisons generated MD and Φ ST dendrograms, and graphics of canonical variate analysis (CVA). Little structure was observed for both markers for macro-geographic scales. Mantel tests comparing geographic, morphometric and genetic matrices showed low correlation (all R 2 <0.35). The factorial graphics built with the CVA evidenced population delimitation for the morphometric data at micro-geographic-scales. Conclusions We believe that T. brasiliensis carries in its genotype a source of information to allow the phenotypical plasticity across its whole distribution for shaping populations, which may have caused a lack of population delimitation for CVAs in morphometric analysis for macro-geographic scale analysis when populations are grouped. On the other hand, the pattern of morphometric results in micro-geographic scales highlights the potential of this tool to infer the source for infestation.


Background
Chagas disease is among the most important neglected diseases in Latin American countries. The transmission classically occurs via vector insects of the Triatominae subfamily (Hemiptera: Heteroptera: Reduviidae). These insect vectors may transmit the disease when they are infected by the parasite Trypanosoma cruzi -the etiological agent of the illness. Although the disease´s incidence has been declined in recent decades due to intense programs to combat the main domiciliated vector in this region -Triatoma infestans -it is estimated that six to eight million people are still infected worldwide [1].
More than sixty-five species of triatomines have been recorded in Brazil, but few exhibit synanthropic behaviors (T. infestans, Panstrongylus megistus, T. brasiliensis, T. pseudomaculata and T. sordida) [2]. The Brazilian Northeast, one of the poorest and underdeveloped regions of the country, raises concern in the context of Chagas disease transmission because it is an area of infestation by T. brasiliensis and T. pseudomaculatanative triatomines of difficult control and widespread in the Northeast semiarid. Whereas T. pseudomaculata is more associated with birds [3] -refractory to T. cruzi infection -T. brasiliensis is associated with mammals, which are potential parasite reservoirs. [4]-5] Additionally, T. brasiliensis is the most frequently Chagas disease vector collected in human domiciles and is considered the most important species under an eco-epidemiologic point-of-view [6].
A comprehensive set of studies to understand the relationships between T. brasiliensis at the time considered as"populations" of the semiarid region of Northeast Brazil has been conducted. The investigations approached biology, ecology, phylogeny, experimental crossings [7][8][9][10][11][12] among others. It aimed to address the differentiation observed in chromatic variations and geographic distribution of morphotypes of T. brasiliensis s.l.. The result was a detailed taxonomic review of the T. brasiliensis "populations". Therefore, some "populations" were raised to the specific taxonomic status, as T. juazeirensis [13] and T. melanica [14]. Later, T. sherlocki was included in this species complex by using the information on the variation in cytochrome B (Cyt B) and 16S mitochondrial genes [15]. Other species were included in this group or revalidated, as T. lenti and T. bahiensis [16]. Finally, T. petrocchiae was also considered a member of this monophyletic group in a study combining morphometric and molecular information (mitochondrial genes: 12S + 16S + COI + Cyt B. Recently, a key was released to identify each member of this species complex [17]. For the T. brasiliensis species complex, molecular studies have been mainly based on the variation of mitochondrial DNA genes [18][19] and, geometric morphometrics has already been applied for the species. [20] Molecular and morphometric markers have been combined, [8,[21][22] but not under the focus of gamma systematics (intra-specifically). In this study, we analyzed microevolutionary and demographic aspects of T. brasiliensis populations at the ecotypic, micro and macro-geographic scales by combining morphometrics and molecular approaches [23][24].

Materials And Methods
Insects. Triatomine captures were conducted in the State of Rio Grande do Norte (RN), in the municipalities of Caicó, Currais Novos and Marcelino Vieira, whereas in the State of Paraíba (PB), insects were collected in the municipalities of Condado, Cajazeiras, Santa Teresinha, São José de Espinharas, São Mamede and Emas (Fig. 1).
The sampling covered a range of 240 km (East-West) and 95 km (North-South) (-06 o 58'48',0" to -06 o 08'49,2" latitude and -38 o 35'27,6" to -36 o 29'09,6" longitude). The sampled spots were within the biogeographic zone known as Caatinga (Figs 7-9), a mosaic of xerophytic, deciduous, semiarid thorn scrub, and forest [25]. All captures were conducted manually with the aid of tweezers. We defined three distinct ecotopes (i) the sylvatic ecotopes (primary) were represented by rocky outcrops; (ii) the domestic ecotopes were considered as the indoor spaces where humans sleep or work; (iii) the peridomestic ecotopes were considered within a radius of 50 m of the domiciles, where domesticated animals are kept. In peridomiciles, most of the triatomines were captured in woodpiles, stone piles, chicken coops and goat/pig corrals. Captured specimens were identified according to taxonomic keys. [9,26] The insects were individually labeled and stored in 1.5 ml microtubes with 100% alcohol. The wings were kept dry in another tube. In the laboratory, the tubes were stored at <5°C.
Morphometric analysis. For this study, the anterior wing of 5-30 individuals per population was used. Considering the epidemiological importance, we did not exclude populations with lower numbers, as for the bug collected inside domiciles in Condado-PB. Photographic records of right wings were made using a stereomicroscope Zeiss™ Discovery V.12 with image capture system AxioCam 5.0™ and software ZEN™ version 2.0. Thirteen anatomical landmarks (Fig. 2) were used via CLIC package (Collection of Landmarks for Identification and Characterization. Jean-Pierre Dujardin, www.mome-clic.com/). To reduce the user effect, the same operator marked all landmarks. Additionally, the wing images of each gender were mixed and sorted before capturing the landmarks.
The shape variables were obtained by using the generalized Procrustes superposition algorithm and the subsequent projection of the Procrustes residues were defined in an Euclidean space. Both non-uniform and uniform components were used as shape variables [27]. These two components describe the differences in deviations from an average reference point configuration [28]. The centroid size of the isometric estimator (CS) derived from the analysis based on anatomical landmarks was used as a measure of the overall size [29][30].
For shape measurements, Mahalanobis distances (MD) between all pairs of populations were calculated and their significance was assessed using a non-parametric test based on permutations (10,000 simulations), which were illustrated in dendrograms. The percentage of phenotypic similarity between pairs of populations was calculated using the cross-validation test, which evidences the percentages of individuals that were correctly assigned to their populations. The relationship between centroid size (CS) and discrimination of shapes between groups (allometry) was estimated through a multivariate regression between the coordinates of Procrustes (dependent variables) and CS (independent variable). This analysis was performed for each module and among populations/population groups. These regressions show associations in all comparisons via random permutations. The P values were set as significant at <0.05. The degree of differentiation between individuals was illustrated through graphics of Principal Component Analyses (PCA) and Canonical Variate Analysis (CVA), which were assessed by the combination of variations in the 13 anatomical coordinates of each individual. All of these statistics were performed using the MorphoJ software [30]. Molecular analysis. The total DNA of 3 legs was carried out with a Blood and Tissue kit from Qiagen™ (Valencia, CA), following the manufacturer's specifications. The mitochondrial cytochrome b (Cyt b) gene was amplified by the polymerase chain reaction (PCR), using the primer pair CYTB7432F, and CYTBR and amplification conditions set by the authors [31][32] who designed the primers. Both strands (forward and reverse) were sequenced, which were then assembled into contigs for edition [33]. Edited sequences were compared to those from the GenBank database using the BLAST [34] program to check the specific status. Sequences were cropped into 447 bp to be with equal length. A phylogram was built with the resulting Cyt B sequences via Bayesian inference with unique haplotypes. The matrix was run in BEAST 1.8.0 [35] with 10 million generations and sampling every 1,000 simulations. A strict molecular clock model (with a default rate of 1.0) and the constant-size coalescent tree prior were applied for this analysis. Tracer 1.7 [36] was used to check parameter convergence with a stationary state, using a minimum threshold for ESS (Effective Sample Size) value of 200. TreeAnnotator (available on BEAST package) was used to summarize all the generated trees with the highest credibility, being visualized and edited using FigTree 1.4.4 (available at http://tree.bio.ed.ac.uk/software/figtree/). The evolutionary model was chosen according to the Bayesian Information Criterion (BIC) in MrModeltest [37]. Estimates of evolutionary divergence between sequences were based on Kimura 2-parameter model (K2P) [38] by using MEGA-X software [39].
For population analysis, summary statistics of genetic diversity of T. brasiliensis were calculated for insects collected in the same geographic spot. The null hypothesis of neutrality was tested using Tajima's D statistic [40] and Fu´s FS [41]. The following population genetic summary statistics were calculated: number of haplotypes (H), haplotype diversity (Hd) and nucleotide diversity (π). A hierarchical analysis of molecular variance (AMOVA) [42] was performed to access the population's genetic structure. The DnaSP v6 software [43] was used to generate the files to run the analyses described above in Arlequin 3.5 [44] software. The significance of molecular statistics (Tajima´s D, Fu´s FS, FCT, FSC, FST, ΦST) [40] was tested using 1,000 permutations, being considered significant those at P values < 0.05.

Morphometrics.
Overall, 733 wings were digitized. Forty-one samples were considered outliers and excluded. The 692 samples were distributed in 43 different populations. PCA explained 36% of the variation (PC1 = 21.6% and PC2 = 14.7%). The first comparison was made between the genders, with 298 females and 394 males. The graphic of PCA (Figures 3.A) shows a higher concentration of females on the negative pole of Component 2 with Mahalanobis Distance (MD=1.343) statistically significant (P<0.0001). The bloxspot (3.B) regarding the centroid size showed that the females were larger than the males (P <0.0001). The cross-validation test wrongly classified 27-28% of the individuals according to sex (Figure 3.C). Thus, it is assumed that sexual dimorphism may have influenced differentiation and we dealt with genders separately. We excluded populations with low numbers (N<5). Therefore, the resulting dataset included 349 samples divided into 24 populations for males  When samples are grouped according to the municipalities (independently of the collection site and ecotope), it is not possible to observe any grouping delimitation in the CVA graphics for each gender (Figures 5.A-5.B), even if they are sorted according to the ecotope inside each municipality (data not shown, because the pattern remains the same). Because all values were low (<0,01), it assured we were dealing with T. brasiliensis. The phylogenetic reconstruction by using unique haplotypes based on Bayesian Inference evidenced that posterior probabilities were in general low. For the cases it was higher than 80%, sequences had K2P lower than 0,009, confirming we were dealing with haplotypes with low variation. The

Molecular evolution model was HKY+G based on BIC (Supplementary File 3 and Supplementary File 1, Sheet 4).
The ΦST tree ( Figure 6.A and Supplementary File 1, Sheet 5) showed the geographic force of the distribution of genetic variation. For example, the majority of Currais Novos populations are associated in the same branch. This was observed for some populations of Cajazeiras and Emas (EM#). For both sexes, the similarity dendrogram built with morphometric information (MD distances) shows that there was a relative ecotypic grouping, which is evident in the red rectanglesn black rectangles, however, we can observe that there were both an ecotypic and geographical grouping at the municipal level, but with some exceptions. Most of the morphometric population comparisons were statistically significant (Supplementary File 1, Sheet 1 [for males] and 2 [for females]). The biggest discrepancy between the morphometric and molecular tree was that only the morphometrics showed an association between the populations of Marcelino Vieira (MV#) (Figures 6.B and 6.C).
When we run the CVA within each municipality almost all groups were delimited. The clear exception is for both sexes for populations from Currais Novos, and for this reason we sent it to the Supplementary File 2. In Caicó-RN, Females the three spots of capture had morphometric delimitation, whereas the males did not have structuring. The opposite was observed for the populations from Marcelino Vieira. However, the number of bug per males from this city was low (<6).
For Emas-PB, all populations were delimited. However, for females, we did not have enough samples to compose the third population (EM94P). Populations from São José dos Espinhais-PB were also well delimited for both genders, despite the fact that there was some sobreposition for the sylvatic female population SJ93S and the peridomestic SJ35P.
For Cajazeiras, clear delimitation was observed for males and a reasonable delimitation was observed for some female populations, as the peridomestic CZ18P and CZ19P. The overlapping of the female CZ27S with some peridomestic (CZ21P and CZ16P) and a domestic population (CZ15D) may indicate similarities of populations from distinct ecotopes, as a signal of the source of infestations. For Condado municipality we only have robust number for morphometric analisis for females. But it can be seen that this peridomestic population (CD46D) was clearly delimited in a group separated from other samples. We summarized observations of morphometric structuring/grouping delimitation in Table 1.
Analysis of Molecular Variance (AMOVA) (Table 2) demonstrated that when populations are grouped according to the municipality, it is observed that 18% of the variation is among municipalities and 36% among populations within municipalities. Following the hypothesis that the genetic structure may be driven by ecotypic force (predominant migration between the same ecotopes), a second test was performed by nesting groups of populations defined by their ecotopes within each municipality. In this kind of nesting, the percentage of variation between groups was even lower (10%) and the variation within groups was higher (41%). In both kinds of nesting populations, most of the variation was observed within populations (46-49%).
Correlation between matrices. All Mantel tests comparing the matrices showed a significant correlation (P <0.05) but in general, it was low. The highest correlation occurred between the MD between males and females (R 2 = 0.74), as expected. The second highest correlation was observed between the ΦST distances and the geographical distance (R 2 = 0.36). For both sexes, the correlations between MD and ΦST were low (R 2 = 0.13-0.15).

Discussion
The differences between the distributional patterns of morphometric and genetic variations have been extensively explored for insect vectors [45][46]. For the T. brasiliensis complex, interspecific molecular and morphometric variations have also been studied [22,[47][48]. Population variations based on morphology can be influenced by ecotypic and environmental variations, but it is also influenced by the individual's genotypic information [49]. Therefore, geometric morphometrics approaches are also used to recognize microevolutionary processes [50]. The first study in this sense for T. brasiliensis used classical morphometry [51]. The second was conducted at a micro-geographic scale perspective and had a restricted sample number (N = 97), but it was a pioneer study by using geometric morphometrics for looking for the variation among ecotypic populations [20]. Our study brings up the first effort by using a large sampling (initial N of 733 insects) on a wider geographic scale -ranging 95 km (North-South) to 240 km (East-West) -for T. brasiliensis. Besides, morphometric information is here associated with genotypic variations. Some authors [23,52] concluded that morphometrics combined with genetic information can improve the understanding of the re-infestation process and contribute to vector control strategies when the approach is used in the micro-geographic scale. In this study, molecular variation was used as a tool for monitoring and control, as the chosen marker (variation in the mitochondrial gene Cyt B) does not suffer the influence of the environmental variation [53]. Thus, this work is also a pioneer in confronting the ecotypic effect on morphometric variation for Brazilian Chagas disease vectors.
Considering all the sampling grouped according to the ecotopes and for each sex, it can be observed in the PCA that there was a relative ecotypic grouping. Similarly, in Ceará, Batista and cols [20] found significant differences between T. brasiliensis collected in different ecotopes. The peridomestic individuals were significantly larger than the wild ones. Costa and cols [12] mentioned that wild populations of T. brasiliensis are usually found hungry in the field, which may explain the difference in size. We suggest that domestic insects did not have numbers enough for inferring the size. Geometric morphometrics effectively minimizes the size factor, but it does not eliminate it totally [50]. The ecotopes seem to influence the morphometric variation, which may represent the bias of geometric morphometry for inferences on genetic structuring. However, as discussed below, some peridomestic populations were more morphometrically associated with some wild bugs than other peridomestic ones, when it is observed under a micro-geographic scale. It indicates that this tool can be used to assess population structure, microevolutionary processes and to look for foci of reinfestation.
In the dendrograms derived from the Mahalanobis distances (MD), it was possible to observe that some groups were formed by nearby populations, indicating geographical structure. The District of Patos is composed of the municipalities of Emas, São José de Espinharas, Santa Teresinha, Condado, all in the State of Paraíba. Within this district, many population groups did not show significant differentiation among them. Low MD (and not significant) were observed between the groups of Emas versus São José de Espinharas and between São José de Espinharas and Santa Teresinha (< 1.54), which makes sense from a geographical point of view [54]. However, the Mantel test showed a weak correlation between the morphometric and geographic distance matrices, which may indicate that forces, other than geographic, maybe also driving the morphometric variation. For T. infestans in Argentina, the resolution of geometric morphometry to detect differences in macro-geographic scales was more evident in comparisons of individuals separated by municipality [23]. However, the authors did not apply any matrices correlation tests (Mantel tests). Additionally, these authors did not include insects from wild populations.
Principal Component Analysis (PCA) is the most popular method to characterize shape variation [30]. However, when the data contains subgroups, the PCA should be used as a method for ordering large groups, with the analysis of canonical variables (CVA) being the most appropriate for the subgroups. The factorial maps constructed with the CVAs in populations defined by collection point within each municipality (in microgeographic scales) were the only ones to delimit the populations. Pilot tests with PCAs could not delimit populations even for micro-geographic scales.
For both genders, the structure was not defined for the populations of Currais Novos; and for Caicó there was definition only for females in CVA graphics. The lack of correlation between the CVA graphics between the genders for Caicó may indicate a different migration pattern. However, for overall results, the Mantel test showed 73% correlation between males and females for MDs. For T. dimdiata [55], authors found no significant differences in gene flow between the sexes of insects by using microsatellite markers. However, for T. juazeirensis -a member of the T. brasiliensis complex -Carbajal de la Fuente and cols [56] found more males than females in light traps, which capture flying insects (dispersants). Lilioso and cols [57] also captured T. brasiliensis via active search, and they found more than twice males than females. This relationship was inverse for peridomestic insects. It is important to recognize the difference in methods of capture in peridomestic and wild environments. These latter authors reported the nocturnal capture of insects select the dispersing in the wild environment, whereas for the peridomestic environment the captures are during the daytime, selecting the inactive insects (sessile). As all Caicó populations were sylvatic, it may indicate distinct patterns of migration between genders.
Low variation was found among sequences (all K2P were <0.01). This value is low if compared to the ones reported between the subspecies T. b. brasiliensis and T. b. macromelasoma (K2P = 0.03) [58], assuring we were dealing with close-related entities. For mosquitoes, Lorenz and cols [59] presented interesting clues about selective patterns observed by molecular markers, which were corroborated by morphometric markers. However, the neutrality tests here applied pointed out to a random ("neutral") evolution for our sampling. In other words, we did not evidence directional or balancing selection, as well as expansion or demographic contraction. Therefore, we could not compare molecular and morphometric results in this sense. Indeed, the neutral evolution for the molecular marker chosen is in agreement with the results of Almeida and cols [18,19] which used the same gene fragment.
The hierarchical molecular variance analysis (AMOVA) was conducted to evaluate the role of geographic and ecotypic forces in the genetic structure of T. brasiliensis. In this case, it is known that the ecotope does not influence the genotype, but for some triatomine vectors, gene flow occurs predominantly between ecotopes, as observed for T. infestans in the Andean valleys [60]. Here, neither the geographical force nor the ecotypic force explained the distribution of molecular variation -based on the observation that the greater percentage of variation was noticed within groups and within geographic and ecotypic populations. These results corroborate Almeida and cols [18,19]; although these authors have noticed some degree of influence of ecotypic force. However, at this point, the morphometric and molecular results were in agreement. Remarkably, molecular and morphometric markers were in agreement to indicate an association between a wild population and a peridomestic population (MV18P) of Marcelino Vieira. It is important to note that the latter was collected at the site where the Chagas disease outbreak occurred [61].
Both molecular and morphometric markers can be applied to understand macroevolutionary and microevolutionary processes in vector insects [62]. Concerning microevolutionary processes, in general, the dendrogram constructed with the ΦST matrices revealed that the geographic force plays a role in the distribution of genetic variation. However, most populations of Currais Novos were in the same cluster. This was observed for some populations of Cajazeiras and Emas-although AMOVA has not shown statistical significance for geographic structuring. On the other hand, it corroborates the results of morphometry, in which Mantel tests did not evidence correlation between matrices of geographic and Mahalanobis distances.
The comparison of the results of the molecular and morphometric tools showed low conformity by Mantel tests. But in both dendrograms, the populations of Cajazeiras and Currais Novos were grouped in the same branch. It evidences that in some cases the markers are in agreement. This conformity has already been tested by other authors and for other species, such as T. dimidiata in the Yucatan peninsula in Mexico [63] and T. infestans in Argentina [23]. For T. dimidiata the authors did not find evident conformity between markers [64], which may be explained by the distinct resolution of molecular and morphometric markers.
The strategy of observing variations within a smaller geographic scale aimed to seek clues about the processes of (peri) household infestation. This was the first attempt in the literature to test geometric morphometrics for this purpose focusing on vectors of Chagas disease in Brazil. Schachter-Broide and Cols [24] applied the approach to this inference for T. infestans in Argentina, concluding that geometric morphometrics represents a useful tool for this purpose. However, the populations analyzed by these authors came all from artificial ecotopes. We must emphasize that our results evidenced that the ecotope seems to have an important effect on phenotypic variation. Therefore, the morphometric associations between populations of different ecotopes in the micro-geographic scale can present important clues to detect infestation processes -when analyzed within the same municipality. In this sense, geometric morphometrics was more sensitive to delimit populations on a microgeographic scale than on a macro-geographic scale, which was evidenced in the factorial maps. For example, in the maps that illustrate the analysis of canonical variables -for both males and females -the peridomestic populations of Cajazeiras and São José de Espinharas presented associated with sylvatic, which questions the role of the ecotope in the distribution of morphometric variation. These findings corroborate the idea that this tool can be used to detect infestation foci [50].
The genome of the species T. brasiliensis may carry information that allows equal morphological plasticity throughout the entire collection range which prevents structuring in the macro-geographic scale. However, demographic events, such as bottlenecks or founding effects, might be acting to maintain homogeneity in groups of insects collected at the same point. Retention of the ancestral character in all sampled populations can also help to explain the results of morphological plasticity. This retention was verified in this work also by molecular tools: the ancestral haplotype of Cyt B is the most frequent in the entire collection range, and this phenomenon is well known for T. brasilensis by other authors [5,18,58].The power demonstrated by geometric morphometrics to delimit populations in micro-geographic scales indicates that this tool has more resolution to structure populations than the molecular marker used since Almeida and cols [18] could not delimit populations using the Cyt b gene in micro-geographic scales. The ecologic niche model was used for the first evaluation of the distributional potential of members of T. brasiliensis in the present and onto futures climate scenarios -as global warming. Model projections indicated the potential for expansions of T. brasiliensis for 2050, highlighting the need for constant monitoring T. brasiliensis distribution [65].
The most remarkable result here obtained was the resolution in the population delimitation in micro-geographic scale for geometric morphometrics -which was sometimes in agreement with the results of molecular variation. The combination of these markers has the potential to be used in decision-making for more rational, targeted and resource-saving control measures (as for spraying domiciles with insecticides). Additionally, the geometric morphometrics based on the venation of the wings of triatomines is a low-cost and quick-to-perform technique. Therefore, the results presented here indicate that there is potential applicability of this technique in vector control measures in the near future.

Supplementary File 3:
Estimates of evolutionary divergence between fragments (510bp) of Cyt B gene of Triatoma brasiliensis conducted using the Kimura 2-parameter model [38].

Supplementary file 4.
Bayesian phylogenetic tree based on Cyt B gene (447 bp) for Triatoma brasiliensis using the molecular evolution model TN93+I. Posterior probabilities are indicated in the nodes. The outgroup is represented by T. petrocchiae (GenBank accession MN177967).

Declarations
Ethics approval and consent to participate. SISBIO IBAMA 58373, SISGEN AF53F3B. We obtained permission from house owners/residents to search for insects in all properties and we visited four houses in each spot (permission "Informed Consent Form" "TCLE 2.665.746").

Consent for publication was uploaded with files as letter properly signed.
Competing interests. The authors declare the no competing intersts exist. . The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Table 1. Summary of the observation in graphics of Canonical Variate Analysis (CVA) regarding the population delimitation in gra municipality: D = delimited (shaded), N=Non delimited and I = insufficient number for robust analysis (see       N 393 B) shows the canonical variable using the discriminating function of the variation between the genders C Boxplots representing variations in the size of the wing centroid    The locations where Triatoma brasiliensis populations were collected in Cajazeiras and Condado in the State of Paraíba are on the left side. On the right side it is shown the Canonical Variate Analysis (CVA) based on the CVA 1 and CVA 2 for both genders. We did not have enough samples from males from Condado