Characterization of Triatoma infestans (Klug, 1834) (Hemiptera: Heteroptera, Reduviidae) from Residual Foci in the States of Bahia and Rio Grande do Sul, Brazil, Using Geometric Morphology

Simple Summary Foci of Triatoma infestans were found in the states of Rio Grande do Sul and Bahia, Brazil. The objective of the present work is to evaluate the morphometric differences between populations of residual and wild focus of T. infestans through head geometric morphometry. It is possible to show significant morphometric differences between the studied populations. Furthermore, it is possible to establish a morphometric relationship between these populations and to list hypotheses about the distribution and maintenance of residual foci of T. infestans in Brazil. Abstract Triatoma infestans is one of the main vectors of Chagas disease in Latin America. Although the species is under control in most Latin countries, it is still necessary to maintain epidemiological surveillance. The present study aims to characterize T. infestans populations from residual foci in Bahia and Rio Grande do Sul, Brazil, comparing them with natural populations in Argentina and Bolivia. For this, we adopt the geometric morphometry of the heads. It is possible to report the morphometric variety of the studied populations. In addition, we show that the size of the heads contributes to the differentiation between populations, while the shape has less power to discriminate groups. Furthermore, we show that some natural populations have morphometric proximity to residual populations, suggesting a relationship between these triatomines. Our data do not support the origin of residual populations but demonstrate the importance of new studies with other techniques to understand the dynamics of distribution and reintroduction of these vectors in Brazilian territory.


Introduction
The subfamily Triatominae (Hemiptera, Reduviidae) is made up of hematophagous insects, and all the described species are potential vectors of the protozoan Trypanosoma cruzi (Chagas, 1909), the etiological agent of Chagas disease (CD) or American trypanosomiasis (WHO) [1].
Triatoma infestans (Klug, 1834), started to occupy the domestic niche in the Bolivian Inter-Andean valleys as long ago as the Precambrian period [2]. In the 20th century, T. infestans occupied three distinct areas of dispersal, one isolated by the Andes mountain range (Chile), and the other two separated by regions where climatic conditions were  * RS-Rio Grande do Sul; ** BA-Bahia; *** residual foci.

Obtention of Images and Landmarks
Geometric morphometry (GM) was used to evaluate variations in the shape and size of the heads of adult males from T. infestans populations. The methodological procedure

Obtention of Images and Landmarks
Geometric morphometry (GM) was used to evaluate variations in the shape and size of the heads of adult males from T. infestans populations. The methodological procedure was the same for all analyses presented in this study. The heads were photographed and digitized using a Leica DMC2900 microscope (Leica Microsystems, Buffalo Grove, IL, USA) with a 3.1 megapixel digital camera (Leica Microsystems) and LAS software (Leica Application Suite United States). The images followed the same settings, aiming to standardize the capture [15].
Eight reference points were selected for the generated matrix ( Figure 2). The landmarks were selected as in Oliveira et al. (2017) [16], collected and processed with the TPS package utilities: TPSdig 2.3.2 and TPSutil 1.81 [17,18]. The reference landmarks of all populations were collected three times, seeking to minimize collector effects [17,18]. The order of introduction of the anatomical landmarks were the same for all specimens, as this is a requirement to establish homology of structures. Raw coordinates are used for generalized Procrustes analysis (GPA) in MorphoJ [19]). GPA is a statistical analysis method that is performed to delete all information related to size, position, and orientation [17,18]. Subsequently, the generated matrix is projected onto a Euclidean space to generate a set of scores: partial warps [20]. An average setting, known as a "consensus", is calculated and allows the average variation between the raw data to be determined [20]. All additional statistics were performed using Procrustes residues to analyze differences in size and shape. was the same for all analyses presented in this study. The heads were photographed and digitized using a Leica DMC2900 microscope (Leica Microsystems, Buffalo Grove, IL, USA) with a 3.1 megapixel digital camera (Leica Microsystems) and LAS software (Leica Application Suite United States). The images followed the same settings, aiming to standardize the capture [15]. Eight reference points were selected for the generated matrix ( Figure 2). The landmarks were selected as in Oliveira et al. (2017) [16], collected and processed with the TPS package utilities: TPSdig 2.3.2 and TPSutil 1.81 [17,18]. The reference landmarks of all populations were collected three times, seeking to minimize collector effects [17,18]. The order of introduction of the anatomical landmarks were the same for all specimens, as this is a requirement to establish homology of structures. Raw coordinates are used for generalized Procrustes analysis (GPA) in MorphoJ [19]). GPA is a statistical analysis method that is performed to delete all information related to size, position, and orientation [17,18]. Subsequently, the generated matrix is projected onto a Euclidean space to generate a set of scores: partial warps [20]. An average setting, known as a "consensus", is calculated and allows the average variation between the raw data to be determined [20]. All additional statistics were performed using Procrustes residues to analyze differences in size and shape.

Size and Shape Variables
A factorial map was generated using MorphoJ 1.07a [18] by the general patterns of morphological variation in the multidimensional data obtained with principal component analysis (PCA) using the Procrustes covariance matrix. Coordinates were obtained after the GPA of the original reference coordinates. Procrustes ANOVA [18] was adopted to assess shape variations using MorphoJ 1.07a [18] and to infer differences between groups [20]. To determine the size variables, the isometric estimator defined as centroid size (CS) was used [21]. The CS is derived from raw coordinate data [22] generated in MorphoJ 1.07a [18]. Mahalanobis distances between pairs of populations were calculated for shape measurements and their significance was assessed using a nonparametric permutationbased test (10,000 replicates, Table 2) in MorphoJ 1.07a [18]. Furthermore, a dendrogram was constructed with Mahalanobis distance data, using PAST v.3.25 [23].

Size and Shape Variables
A factorial map was generated using MorphoJ 1.07a [18] by the general patterns of morphological variation in the multidimensional data obtained with principal component analysis (PCA) using the Procrustes covariance matrix. Coordinates were obtained after the GPA of the original reference coordinates. Procrustes ANOVA [18] was adopted to assess shape variations using MorphoJ 1.07a [18] and to infer differences between groups [20]. To determine the size variables, the isometric estimator defined as centroid size (CS) was used [21]. The CS is derived from raw coordinate data [22] generated in MorphoJ 1.07a [18]. Mahalanobis distances between pairs of populations were calculated for shape measurements and their significance was assessed using a nonparametric permutationbased test (10,000 replicates, Table 2) in MorphoJ 1.07a [18]. Furthermore, a dendrogram was constructed with Mahalanobis distance data, using PAST v.3.25 [23].

Evaluation of Variations between Groups
To determine the relationships between populations (nine regions and three countries), canonical variable analysis (CVA) was performed, using MorphoJ 1.07a [18]. Multivariate statistics were performed using Procrustes coordinates [24]. The CVA was performed associated with a resampling method (10,000 repetitions). A factor map of the first and second canonical factors was generated in MorphoJ 1.07a [18].

Results
From the collection of nine morphometric landmarks from 245 images and the heads of adult males of the T. infestans species, it was possible to demonstrate the morphological relationships of the residual foci in Brazil with the populations of Argentina and Bolivia.

Analysis of Size and Shape Variables between Populations of Nine Localities
The size variables estimated through the CS show different size means between the studied populations ( Figure 3). All recovered averages were significant (p < 0.001). The Mahalanobis distance is a useful distance estimator for determining similarity between samples. Our results allow us to estimate the distances between populations evaluated pairwise ( Table 2). The Argentine populations TI25 and TI26 present relatively approximate values, showing proximity between these populations. The population of Rio Grande do Sul, TI24, in relation to the other populations, is close to the Argentine  The populations of T. infestans from Rio Grande do Sul, Brazil, TI124, TI27, TI28, and TI131, are larger than TI156. Likewise, the two populations of Argentina, TI25 and TI26, show differences in CS, with TI25 being higher. The population of Santa Rosa, Rio Grande do Sul (TI156) and Argentina (TI26) show approximate values. The population from Cordoba, Bolivia has a different metric from the other populations, as well as TI169 from Bahia, Brazil. The size relationship can be described as: TI24 > TI27 > TI25 > TI131 > TI165 > TI26 > TI168 > TI28 > TI156 ( Figure 3).
The Mahalanobis distance is a useful distance estimator for determining similarity between samples. Our results allow us to estimate the distances between populations evaluated pairwise ( Table 2). The Argentine populations TI25 and TI26 present relatively approximate values, showing proximity between these populations. The population of Rio Grande do Sul, TI24, in relation to the other populations, is close to the Argentine populations (p < 0.001). Unlike other Brazilian populations, Bahia and Rio Grande Sul differ significantly from the Argentine specimens and show a relative proximity to TI65 from Argentina. In summary, the populations that present the most discrepant Mahalanobis distance metric are the two Argentinean populations.
The dendrogram retrieved from the Mahalanobis distance data and the Procrustes distance data clearly illustrates the results ( Figure 4). The Procrustes distance is a useful method for comparing shape, and, in this study, we built a similarity dendrogram through the retrieved data. Both dendrograms clearly show the close relationship between populations. Populations TI24, TI25, and TI126 form a clade ( Figure 4). Populations from Bahia (TI169), Rio Grande do Sul (TI28), and TI165 from Bolivia form a second clade. Populations TI27, TI131, and TI156 from Rio Grande do Sul do not constitute a clade; however, they are closer to the second clade (TI169, TI128, and TI165).

Principal Component Analysis between Populations of Nine Localities
Principal component analysis (PCA), which is defined as an exploratory analysis based on the averages of each population, was able to identify morphological patterns among populations of T. infestans ( Figure 5).

Principal Component Analysis between Populations of Nine Localities
Principal component analysis (PCA), which is defined as an exploratory analysis based on the averages of each population, was able to identify morphological patterns among populations of T. infestans ( Figure 5).

Principal Component Analysis between Populations of Nine Localities
Principal component analysis (PCA), which is defined as an exploratory an based on the averages of each population, was able to identify morphological pa among populations of T. infestans ( Figure 5).  Principal components explain 100% of the shape variables. The PCA of T. infestans populations from different localities showed the following variables: the first principal component (PC1) accounted for 42.39% of the variations, the second component explained 24.14%, and together they accounted for 66.53% of T. infestans form variables. The disposition of the ellipses in the space of the Cartesian plane shows an overlapping of the ellipses indicating few differences between the studied forms (heads). Even with similar shapes, head length has greater power of discrimination among the studied specimens.

Canonical Variance Analysis between Populations of 9 Localities
The CVA was used to discriminate groups across datasets. The first component (CV1) was responsible for 55.47% of the shape variations and the second component explained 23.04% of the variation; together they accounted for 78.51% of shape variables in T. infestans populations from different localities ( Figure 6). The space of the Cartesian plane groups the populations TI24 and TI25. The Bolivian population (TI165) is separated from the other study species that appear overlapping, showing little potential for discrimination ( Figure 6). festans populations from different localities ( Figure 6). The space of t groups the populations TI24 and TI25. The Bolivian population (TI165 the other study species that appear overlapping, showing little poten tion ( Figure 6).

Analysis of Size and Shape Variables between Populations of Dif
To better understand the morphometric diversity of T. infestans populations by grouping them into four groups distributed in Argentin two populations from Brazil, Bahia and Rio Grande do Sul, respecti was to characterize the morphometric patterns of geographically close

Centroid Size between Populations of Different Regions
The CS averages obtained from the GPA show the size variables be from different regions. The highest averages of size were T. infestans f lations of T. infestans from Argentina and Rio Grande do Sul show ap while T. infestans from Bahia shows the lowest average. The size relat scribed as: T. infestans Bolivia > T. infestans Argentina (Córdoba/Santa F Grande do Sul > T. infestans Bahia. The size variation is illustrated by t

Analysis of Size and Shape Variables between Populations of Different Regions
To better understand the morphometric diversity of T. infestans, we evaluated the populations by grouping them into four groups distributed in Argentina and Bolivia, and two populations from Brazil, Bahia and Rio Grande do Sul, respectively. The objective was to characterize the morphometric patterns of geographically close populations.

Centroid Size between Populations of Different Regions
The CS averages obtained from the GPA show the size variables between populations from different regions. The highest averages of size were T. infestans from Bolivia. Populations of T. infestans from Argentina and Rio Grande do Sul show approximate values, while T. infestans from Bahia shows the lowest average. The size relationship can be described as: T. infestans Bolivia > T. infestans Argentina (Córdoba/Santa Fe) > T. infestans Rio Grande do Sul > T. infestans Bahia. The size variation is illustrated by the box plot (Figure 7). In summary, differences in centroid size (CS) between the different populations of Argentina, Bolivia, and Brazil were possible (Figure 7).  Mahalanobis distance values retrieved from show morphometric dissimilarity in size between male heads ( Table 3). All values are statistically supported (<0.001). The values show that the populations of the Argentine group are more distant, with less difference with the Rio Grande do Sul group, Brazil. Bolivia's population is far from all populations. The population of Bahia, Brazil, according to the sampled data, is different from the other Brazilian population (Table 3). Using data from Mahanalobis distances and Procrustes distances, we recovered two similarity dendrograms (Figure 8). The dendrograms retrieved with different data (Mahalanobis and Procrustes) show a similar topology. The dendrograms enable clear visualization of the distances between the study populations. Again, as in the previous results, the population of Argentina is relatively close to the populations of southern Brazil. In contrast, the populations of Bahia, Brazil, and Bolivia are also close. Mahalanobis distance values retrieved from show morphometric dissimilarity in size between male heads (  Using data from Mahanalobis distances and Procrustes distances, we recovered two similarity dendrograms (Figure 8). The dendrograms retrieved with different data (Mahalanobis and Procrustes) show a similar topology. The dendrograms enable clear visualization of the distances between the study populations. Again, as in the previous results, the population of Argentina is relatively close to the populations of southern Brazil. In contrast, the populations of Bahia, Brazil, and Bolivia are also close.

Principal Component Analysis between Populations of Different Regions
Principal component analysis (PCA) describes shape and size variables in a Cartesian plane (Figure 9). The first principal component (PC1) was responsible for 63.44% of the variations, and the second component explains 23.58%; together they accounted for 87.02% of shape variables in T. infestans populations from different regions (Figure 9). The disposition of the ellipses in the space of the Cartesian plane shows relative proximity of form in the heads of the populations of the studied regions; however, the population of

Principal Component Analysis between Populations of Different Regions
Principal component analysis (PCA) describes shape and size variables in a Cartesian plane (Figure 9). The first principal component (PC1) was responsible for 63.44% of the variations, and the second component explains 23.58%; together they accounted for 87.02% of shape variables in T. infestans populations from different regions (Figure 9). The disposition of the ellipses in the space of the Cartesian plane shows relative proximity of form in the heads of the populations of the studied regions; however, the population of Bolivia and Bahia presented greater capacity of differentiation in the x-axis. The populations of Argentina and Brazil (Rio Grande do Sul) maintained the approximate midpoint overlap (Figure 9). tina (Santa Fé) (TI26); BA/Bahia/Brazil/Residual Foci (TI169); Bolivia/Cochabamba (T Grande do Sul/Brazil (TI24, TI27, TI28, TI131, TI156 residual foci).

Principal Component Analysis between Populations of Different Regions
Principal component analysis (PCA) describes shape and size variables in plane (Figure 9). The first principal component (PC1) was responsible for 6 variations, and the second component explains 23.58%; together they ac 87.02% of shape variables in T. infestans populations from different regions (Fi disposition of the ellipses in the space of the Cartesian plane shows relative p form in the heads of the populations of the studied regions; however, the p Bolivia and Bahia presented greater capacity of differentiation in the x-axis. tions of Argentina and Brazil (Rio Grande do Sul) maintained the approxima overlap ( Figure 9).

Canonical Variance Analysis between Populations of Different Regions
The CVA was able to group the populations through the estimated multivariate correlation of the head of T. infestans populations. The analysis of canonical variables explained 100% of the shape variables. The first component (CV1) was responsible for 48.82% of the variations, and the second component (CV2) explained 33.6%; together they were responsible for 82.41% of the shape variations in T. infestans populations of different regions. The disposition of the ellipses in the space of the Cartesian plane shows a similarity of shape in the heads of the populations. The populations of Bolivia and Bahia (Brazil) maintain approximate midpoint overlap ( Figure 10). In contrast, Argentina and Rio Grande do Sul also overlap in the space of the Cartesian axes ( Figure 10). plained 100% of the shape variables. The first component (CV1) was responsible for 48.82% of the variations, and the second component (CV2) explained 33.6%; together they were responsible for 82.41% of the shape variations in T. infestans populations of different regions. The disposition of the ellipses in the space of the Cartesian plane shows a similarity of shape in the heads of the populations. The populations of Bolivia and Bahia (Brazil) maintain approximate midpoint overlap ( Figure 10). In contrast, Argentina and Rio Grande do Sul also overlap in the space of the Cartesian axes ( Figure 10).

Discussion
Despite the positive impact obtained from the reduction in T. infestans populations through vector control, this species still has epidemiological importance because of its ability to adapt to domestic and peridomiciliary environments [6,25]. However, for unknown reasons, T. infestans is still found in residual foci in the states of Bahia and Rio Grande do Sul, Brazil [26]. These episodic findings raise questions about the existence and persistence of residual foci in these regions, so our study sought to evaluate the morphometric relationship of T. infestans populations from residual foci and natural environments by geometric head morphometry.
Our results show that different ecotypic populations of T. infestans, and from different geographical sites, present significant morphometric differences, demonstrated through head geometric morphometry. We also show that macro-geographic patterns are better discriminated than micro-geographic patterns by geometric morphometry. The multivariate method has been shown to be useful in phylogenetic, systematic, or biogeographical studies with Triatominae [18,[27][28][29][30][31].
Passive transport by humans is one of the main hypotheses to explain the distribution of T. infestans in South America [5]. In this way, it is possible that maintenance of residual foci of species such as T. infestans or R. prolixus in Brazil can be sustained by the passive

Discussion
Despite the positive impact obtained from the reduction in T. infestans populations through vector control, this species still has epidemiological importance because of its ability to adapt to domestic and peridomiciliary environments [6,25]. However, for unknown reasons, T. infestans is still found in residual foci in the states of Bahia and Rio Grande do Sul, Brazil [26]. These episodic findings raise questions about the existence and persistence of residual foci in these regions, so our study sought to evaluate the morphometric relationship of T. infestans populations from residual foci and natural environments by geometric head morphometry.
Our results show that different ecotypic populations of T. infestans, and from different geographical sites, present significant morphometric differences, demonstrated through head geometric morphometry. We also show that macro-geographic patterns are better discriminated than micro-geographic patterns by geometric morphometry. The multivariate method has been shown to be useful in phylogenetic, systematic, or biogeographical studies with Triatominae [18,[27][28][29][30][31].
Passive transport by humans is one of the main hypotheses to explain the distribution of T. infestans in South America [5]. In this way, it is possible that maintenance of residual foci of species such as T. infestans or R. prolixus in Brazil can be sustained by the passive transport of specimens from focal regions, where the wild environment or the lack of control keeps populations close to or in contact with human dwellings [5,6]. Our data demonstrate that Brazilian populations from Bahia and Cochabamba, Bolivia show few morphological differences. Similarly, populations from southern Brazil show greater morphometric proximity to populations from Argentina. The proximity of the border between the southern region of Brazil and Argentina is related to dispersion through the Argentinean Gran Chaco, as suggested by Piccinali et al. (2011) [10,11] and Waleckx et al. (2011) [9]. Even if our data do not support the hypothesis, the proximity between populations is demonstrated in this study.
In Brazil, so far, passive transport of T. infestans specimens has not been observed, as described in southern Chile [32], only the presence of residual foci in Bahia and Rio Grande do Sul. However, Brazil is Bolivia's main economic market, as Bolivia exports 42% of its products to Brazil and, not unlike Argentina, has an active economic relationship with Brazil [33,34]. The results presented are not enough to trace the origins of the Brazilian residual populations, but all dynamics of the distribution of triatomines in the Americas support the need for investigations with other methods, such as phylogeographic studies, as in Campos-Soto [33], where the origin of insular species of Mepraia was described through phylogeographic studies that used molecular phylogenies, a molecular clock, and biogeography.
The Brazilian triatomine fauna is distributed by natural and artificial environments, the latter mainly associated with the transmission of Chagas to humans [35]. According to Forattini [5], maintenance of triatomines in artificial environments is mainly the result of anthropic action in natural environments. In addition to direct human actions on natural environments, secondary factors such as climate change can also significantly interfere with the radiation of insect vectors to new environments [36,37].
Triatomines are insects whose morphology is easily modeled by the environment. Rapid morphological changes in response to ecological factors can make specific identification difficult, as similar genetics but with marked morphological differences are common. [16] The T. brasiliensis complex is an interesting model to illustrate how morphology can obscure knowledge of the group's species [16]. The T. brasiliensis species has a morphology shaped by ecotypic and environmental variations, as described in Kamimura [29]. Our studies support the phenotypic plasticity of T. infestans and show that geometric morphometry is useful to characterize morphological variation.
In this study, we used the PCA to characterize the variation in shape. The results show little variation in the shape of the heads, and length is the factor with the greatest potential for discrimination between populations of T. infestans. The CVA grouping method is consistent with the result of the PCA, showing little potential for discrimination between the studied populations. These methods are suitable for intra and interspecific studies [18,29]. However, the analyses of the CS and the Mahalanobis and Procrustes distances were useful to discriminate between the specimens, showing phenotypic variability of T. infestans.
Triatomines are excellent models to study phenotypic plasticity, as they are capable of rapidly modifying morphology in response to new habitats. Phenotypic plasticity is evident in triatomines. Our study shows, through head geometric morphometry, the morphometric variations in different populations of T. infestans. In addition to showing the phenotypic relationship between natural and residual populations of T. infestans, we point out the importance of new studies to confirm the relationships and origins between morphometrically close populations, as these data are important to elucidate the distribution dynamics of triatomines, as well as for collaborating in entomological control and surveillance.
Author Contributions: S.P., wrote the article, conceptualization, methodology, editing; T.B., general review of the article, methodology, software; C.G., general review of the article and writing; D.R., general review; F.F., contribution to the making of maps and review; N.F., contribution to the making of maps and review. All authors have read and agreed to the published version of the manuscript.
Funding: This research received support from CAPES Coordination for the Improvement of Higher Education Personnel Program.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The database used and/or analyzed during the current study is available from the corresponding author on reasonable request.