Dispersion of Leishmania (Leishmania) infantum in central-southern Brazil: Evidence from an integrative approach

Leishmania (Leishmania) infantum is the zoonotic agent of visceral leishmaniasis (VL), a disease with a global distribution. The transmission scenario of VL has been undergoing changes worldwide, with the biologic cycle invading urbanized areas and dispersing the parasites into other previously free areas. The epidemiological cycle in Brazil has dispersed from the Northeast to other regions of the country. In this study, an integrative approach, including genotyping Brazilian strains of L. (L.) infantum for 14 microsatellite markers and reviewing historical records of the disease, was used to assess dispersion routes throughout central-southern Brazil. Our results support three L. (L.) infantum dispersion routes: A) dispersion from Bolivia to the states of Mato Grosso, Mato Grosso do Sul and São Paulo via the Bolivia-Brazil gas pipeline from 1998 to 2005; B) VL dispersion from Paraguay to the Brazilian side of the triple border (Foz do Iguaçu and Santa Terezinha de Itaipu) during after 2012; and C) emergence of a new L. (L.) infantum cluster in western Santa Catarina State and its dispersion to southern Paraná State (municipality of Pato Branco), after 2013. Hypotheses regarding possible entries of Leishmania (L.) infantum into the area of the triple border are presented and discussed. Understanding how VL has dispersed is vital to the development of control measures for this disease and to avoid future dispersion events.


Introduction
Human Visceral Leishmaniasis (hVL) is a widely distributed neglected disease caused by the protozoans Leishmania (Leishmania) infantum in Asia, Africa, Europe and Americas, and L. (L.) donovani in Asia and Africa [1]. These parasites use the domestic dog as a reservoir, in which it causes canine Visceral Leishmaniasis (cVL), and Phlebotominae sand fly species of Phlebotomus and Lutzomyia longipalpis as vectors in the Old and New World, respectively [2], although other phlebotominae species have been hypothesized as secondary vectors in the latter region (see Thomaz-Soccol et al. [3] for further discussion).
VL has recently experienced changes in its transmission profile in both the Old and New World [1,[4][5][6]. The disease has dispersed to places where it had not been previously described (e.g. United States, Uruguay, Madrid Spain), and has expanded its geographical distribution into previously free areas in endemic countries [7][8][9][10][11]. Thus, the number of cases of VL has increased in recent last years in both the Old (e.g. [6,12]) and New World [e.g. [13,14]. Currently, 1.69 billion people are estimated to be living in VL transmission areas worldwide, the disease presented 2.27 cases per 100,000 habitants in 2015, and 90% of global VL cases occurred in six countries, including Brazil [14][15][16].
Although known since 1913 [17] L. (L.) infantum is likely an invasive species in Brazil, arriving first in the Northeast Region carried by dogs transported with colonizers from Portugal and Spain [18][19][20][21]. Between 1920 and 1980, VL was restricted to rural areas in Northeast Brazil, where it has remained endemic [4,[22][23][24]. However, the disease subsequently began to invade urban and peri-urban areas in other regions of the country [25][26][27][28][29], with epidemics in the north region, especially in Teresina, state of Piaui, in 1981 and in São Luis, state of Maranhão, in 1982 [30,31]. In the subsequent decade, several epidemic outbreaks were reported, especially in the Southeast and Central-West regions, with high rates of cVL cases followed by clinical human cases in Belo Horizonte, state of Minas Gerais, Campo Grande, state of Mato Grosso do Sul, and Araçatuba, state of São Paulo. Now, L. (L.) infantum has spread throughout the states of Minas Gerais, Goiás, São Paulo, Mato Grosso, Mato Grosso do Sul, Rio de Janeiro and Espírito Santo [1,4,26,[32][33][34][35][36][37] Dispersion of VL in the Southern Region of Brazil has been more recent. The first records of cVL and hVL in this region were in the state of Rio Grande do Sul in 2006 and 2008, respectively [38], followed by the state of Santa Catarina in 2011 [39,40]. In the state of Paraná the first detection of vectors and dogs diagnosed with cVL was in 2012, while the first human case was recorded in 2016 [41][42][43][44][45]. In the South Region of Brazil, VL occurs primarily in cities bordering Paraguay, Argentina and Uruguay [3,[9][10][11]46]. Currently, the disease is classified as 'controlled' in Brazil according with to its epidemiological scenario [47].
Several hypotheses have been proposed for the spread of L. (L.) infantum throughout central-southern Brazil. For instance, the construction of the east-west route of the Bolivia-Brazil gas pipeline is thought to have allowed the dispersion of L. (L.) infantum into the centralsouthern Brazil through the migration of workers and infected dogs and deforestation in the 1990s [1,[48][49][50][51][52][53]. Moreover, the construction of railways and the immigration of infected dogs from other endemic areas seemed to have also facilitated the spread of the disease throughout central-southern Brazil [36,50,52,54]. Deforestation and climate and environmental changes have also been proposed as assisting the expansion of VL in different parts of Brazil [22,49]. Most of these studies used different data to test these hypotheses, including molecular markers (i.e. microsatellite markers, see Ferreira et al. [51]) and historical spatial data (e.g. [49,50,55]). However, no study has tested these hypotheses using an integrative approach that combines both methods.
As part of the IDRC #107577-002 research project (idrc.ca/en/project/ addressing-emergence-and-spread-leishmaniasis-bordersargentina-brazil-and-paraguay), the objective of the present study was to evaluate the dispersion of L. (L.) infantum in central-southern Brazil by integrating molecular markers and historical records for hVL and cVL in the region. Assessing which cluster of L. (L.) infantum is present in each city allows reconstructing potential dispersion routes, while integrating spatial-temporal analysis of the first descriptions of VL cases helps to determine the direction of dispersion. Knowing dispersion routes is essential for developing strategies to control this emergent disease and restrain its future dispersion.

Samples
One hundred and thirty-two isolates from dogs, humans and sand flies were genotyped for 14 loci of microsatellite markers to assess the dispersion of L. (L.) infantum in central-southern Brazil (Table 1). For this study, seventy samples were collected in four areas in the municipality of Foz do Iguaçu, Paraná (62 from dogs, four from sand flies and four from humans), and four samples from dogs collected in Santa Terezinha de Itaipu, Paraná, between 2013 and 2016 (see Thomaz Soccol et al. [45] for collection and parasite isolation details). Briefly, Leishmania strains from Foz do Iguaçu and Santa Terezinha de Itaipu were isolated from bone marrow, aspiration of lymph nodes and leukocyte layer of dogs, intestines of sand flies, and leukocyte layer of humans with clinical symptoms. These samples were inoculated in Neal, Novy and Nicole (NNN) culture medium with 0.9% saline solution for four weeks at 24ºC [56]. The promastigote cultures from the other samples were cultivated in Brain Heart Infusion (BHI) with 0.9% saline solution at 24ºC. After culture, parasites were centrifuged at 3,500 g at 4˚C and washed three times (0.9% saline solution, 0.3% saline solution and again 0.9% saline solution). The DNA of cultured promastigotes and biological samples was extracted using the phenol/ chloroform/isoamyl alcohol method [57].
The isolates from other regions of Brazil were acquired from the Molecular Biology Laboratory of the Graduate Program in Bioprocess Engineering and Biotechnology of Universidade Federal do Paraná (UFPR). Additionally, 10 samples from Asunción (Paraguay, PY) were provided by the Laboratorio de Medicina Tropical of Instituto de Investigaciones en Ciencias de la Salud of Universidad Nacional de Asuncion, and seven samples from the Old World

Ethics statement
The collection of human sampling was conducted in accordance with the International Ethical Guidelines for Biomedical Research in Humans. The samples were taken by the doctors. In addition, ethical approval was obtained from the Universidade Federal do Paraná Ethical Committee (number 684.244) and we complied with the minimum requirements of the Southern Common Market Treaty (Mercosur), Resolution No. 129/96. All individuals have signed the free consent clause indicating that they agree to use this sample. For dogs, all procedures were carried out in strict compliance with the rules defined by the National Council for the Control of Animal Experiments (CONCEA). Every effort was made to minimize suffering of the dogs. The work was approved by the Ethics Committee of the Federal University of Paraná (protocol number 044/2014). The owners have signed a consent form for the use of the samples.
The presence of null alleles, allele dropout and scoring errors was analyzed with Micro-Checker 2.2.3 [61]. The presence of loci under selection was assessed in the BayeScan v2.1 [62] only for populations with more than five individuals. Hardy-Weinberg disequilibrium, diversity (gene diversity, Ho and He) and genetic differentiation (FST and AMOVA, only for populations with more than five strains) analyses were performed using the software Arlequin [63]. Allelic richness was calculated in FSTAT 2.9.3.2 [64]. The critical p value was corrected using the B-Y method [65] in analyses with multiple comparisons. The probable number of genetic populations was assessed using the assign method implemented in STRUCTURE 2.3.3 [66] with three runs for each K (K between 1 and 8), composed of a burn-in period of 500,000 itinerations and 5,000,000 Markov Chain Monte Carlo (MCMC) iterations, and no-admixture model. The ad hoc method of Evanno et al. [67], implemented on the online tool Structure Harvester [68], was used to assess the most likely value of K. However, the main assumptions of Structure Analysis are that the population present Hardy-Weinberg and linkage equilibrium, while species of Leishmania species frequently deviate from these assumptions (see the Results section and [21,60,69]) regarding caution in interpreting Structure Analysis results for Leishmania spp.). Thus, we also assigned strains of L. (L.) infantum using Discriminant Analysis of Principal Components (DAPC), which is free from the assumptions of H-W and linkage equilibrium [70]. This analysis was performed using the package 'adegenet' [71] in R 3.5.0 software (R development core team [72]). The optimum number of retained PCs was assessed using both α-score and cross-validation, while the numbers of clusters was chosen based on the results of structure analysis. Due its more flexible assumptions, the results of the DAPC were preferably used to assess the dispersion of L. (L.) infantum in central-southern Brazil. Strains with posterior probability of belong to a cluster higher than 0.80 in the DACP analysis were assigned to that cluster, while strains with posterior probability of belong to a cluster lower than 0.80 remained undetermined. Subestructuration within the clusters was assessed in 3 runs of 5,000,000 MCMC (burn-in of 500,000 iterations) with K between 1 and 4 in STRUCTURE for each cluster. , Uruguay (UR)) were also added to the database to assess possible dispersion of the parasite from these countries. For this, a search was performed for publications available in the Scopus, PubMed, Google Scholar, and Scielo portals between 1913 and 2017 using the following keywords: "first case visceral leishmaniasis", "visceral leishmaniasis in dogs", "visceral leishmaniasis in human" or "Leishmania infantum". Descriptions of errant dogs were not considered due to uncertainty regarding origin (autochthonous or allochthonous). Records of autochtonous human VL cases in the SINAN database (Sistema de Informação de Agravos de Notificação, available in http://portalsinan.saude.gov.br/) were also considered. The SINAN is the database that presents the records of diseases with mandatory notification in Brazil, includinhg hVL. The hVL cases recorded in the SINAN ranges between 2001 and 2017. All cases were categorized into five ranges of years according the following events: 1. 1913 to 1980: population migration from the Northeast Region to central-southern Brazil (see [73][74][75]); 2. 1981 to 1997: beginning of rural exodus, with migration of people and their animals from rural to urban areas (see [74][75][76]

Population genetic analysis
Among the 14 microsatellite markers assessed, the loci List 7031, Li 41-56, Li 45-24 and TubCA exhibited recurrent evidence of null alleles for some populations and were thus removed from further analyses. No loci presented evidence of balancing or positive selection. Greater allelic diversity was observed in Brazilian populations from Campo Grande (MS 2 ) and Foz do Iguaçu (PR 5 ), while the populations from Foz do Iguaçu and Paraguay (PY) had greater intra-population allelic richness (see Table 1).
The Structure and DACP analysis with K = 3 divided Cluster 2 into two other clusters (named Clusters 2.1 and 2.2) (Fig 2).

First records of VL in cities of central-southern Brazil
The research for first records of VL cases in dogs and humans resulted in 52,029 articles published between 1913 and 2018, of which 350 were pre-selected due their epidemiological information or the report of the first case of VL in cities of central-southern South America. Among these articles, 55 were selected due to human or canine VL reports in Bolivia, Paraguay, Argentina, Uruguay or Brazil. Cases of VL were described for 672 cities (Table 3  Between 2006 and 2010 (Fig 3C), 173 cities in Argentina (3), Paraguay (5), Bolivia (1) and Brazil (164) had their first VL cases (141 hVL, 32 cVL). In Brazil, VL spread to cities northward and southward from the cities in the São Paulo state that is close to the Bolivia-Brazil gas pipeline and had their first records in previous period. Moreover, the first records of VL for the South Region of Brazil were during this period in three cities in the state of Rio Grande do Sul. New records of VL in Argentina occurred in cities close to gas pipelines and the frontiers with Paraguay and Brazil (i.e. Rio Grande do Sul state).

Table 2. Pairwise genetic differentiation (FST) and their significance (in parenthesis) between Leishmania infantum populations from central-southern Brazil and Paraguay (PY).
Only populations with more than five strains were used.

Discussion
The results of the microsatellite and historical report analyses support three dispersion routes for L. (L.) infantum in central-southern Brazil, as follow: A) Dispersion from west to east along the Bolivia-Brazil gas pipeline between 1998 and 2005 (A in the Fig 4); B) Dispersion from Paraguay to Brazilian cities at the triple border after 2012 (i.e. Foz do Iguaçu and Santa Terezinha de Itaipu) (B in the Fig 4); C) The emergence of a new cluster in western Santa Catarina State (i.e. the cities of São Miguel do Oeste and Descanso) in 2013, and its dispersion to southern Paraná State (Pato Branco) (C in the Fig 4). The evidence supporting each reconstructed route is provided below.
The first putative dispersion route of VL into central-southern Brazil probably followed the construction of the Bolivia-Brazil gas pipeline between 1998 and 2005 (Fig 4B), first in cities close to the gas pipeline, and then to the north and to the south of these cities. This dispersion route has previously been suggested by studies based on historical cases in Mato Grosso do Sul [49] and São Paulo [55], on historical and molecular data [79] and on microsatellite markers [51]. Our microsatellite data also supported this hypothesis. Our Cluster 2.1, which probably represents cluster POP-3 of Ferreira et al. [51], is present in populations from Mato Grosso do Sul (Campo Grande-MS 1 and Três Lagoas-MS 2 ) and São Paulo (Bauru-SP 1 and Andradina-SP 2 ). The construction of the Bolivia-Brazil gas pipeline started in 1998 in Bolivia and expanded into Mato Grosso do Sul (from Corumbá to Três Lagoas), São Paulo (from northwest, close to Andradina, to the southeast), Rio de Janeiro and Minas Gerais. During this construction, the displacement of employees and their pets from Corumbá (an endemic area) to other cities of Mato Grosso do Sul and neighboring states along the course of gas pipeline may have dispersed the parasite into the Southeast Region [49]. Moreover, the deforestation and ecological imbalance resulting from the construction itself may have favored the dispersion of the vector Lu. longipalpis into urban areas [85]. Alternatively, Oliveira et al. [36] found no correlation between the date of vector detection and confirmation of the autochthonous cVL and hVL, with the construction of the Bolivia-Brazil gas pipeline, and they found positive correlation between the presence of the Marechal Rondon highway and of transverse highways with the presence of the vector in the São Paulo State. Highways permit the transportation of people, host and vectors, including those infected with infectious diseases. Once the Bolivia-Brazil gas pipeline presets high collinearity with the BR-272 (in Mato Grosso do Sul) and the Marechal Rondon highway (in the São Paulo State), isolating the influence of these highways from those of the Bolivia-Brazil gas pipeline is difficult. In this way, the influence of the highways, or even a composed influence of highways and gas pipelines, should not be discarded.
Our results also support a possible dispersion of L. (L.) infantum from Bolivia to Rondonópolis (MT) though a secondary gas pipeline from Bolivia to Cuiabá (MT) (182 km from Rondonópolis), since the same cluster (Cluster 2.1) was found in this city and other cities of Mato Grosso do Sul and São Paulo. This result opposes that of Ferreira et al. [51], who found 86% of the 11 L. (L.) infantum strains from Cuiabá (MT) and 3 strains from Rondonópolis (MT) to belong to the POP-2 cluster (not found in our study), and 14% of belonging to the POP-1 cluster (our cluster 2.2). Thus, our samples are the first to record the POP 3 (cluster 2.1) in Mato Grosso state.
The seccond dispersion hypothesis represents the dispersion of L. (L.) infantum from Paraguay to Brazilian cities close to the triple border between Argentina-Brazil-Paraguay (Fig 4C). VL was first described in the Asunción Paraguay in 1998, and then was found to have dispersed to other cities in the country between 2006 and 2010, likely reaching the triple border around 2012. Supporting this, the Cluster 2.2 is the most prevalent cluster in Paraguay, Foz do Iguaçu and Santa Terezinha de Itaipu. On the Brazilian side of the triple border, Foz do Iguaçu borders Argentina and Paraguay and it is considered a potential place for VL dispersion due the intense flow of people and environmental conditions suitable for the emergence of Lu. longipalpis and VL (see Thomaz-Soccol et al. [3,45] for details). The first description of the vector (Lu. longipalpis) in this region was in 2012, the first cVL in 2013 and the first hVL in 2016, with increasing number of cases in subsequent years [41][42][43][44]. Currently, Thomaz-Soccol et al. [45] reported that about 24% of the dogs in Foz do Iguaçu possess cVL, and that the parasite and the vector are widely present in Foz do Iguaçu and the neighboring city Santa Terezinha de Iguaçu. Thus, Foz do Iguaçu is now considered an endemic city for L. (L.) infantum due to the generalized distribution of seropositive dogs, abundance of Lu. longipalpis and human cases in all areas of the city.    [82]. These L. (L.) infantum strains were assigned to a new cluster (here named cluster 1), and probably dispersed to southern Paraná State (i.e. Pato Branco) (Fig 4D). The allochthonous cVL case described for Curitiba in 2004 (a dog captured off the street, see Thomaz-Soccol et al. [85] for more details) was also assigned in this cluster. The origin of this cluster is still unclear, but it is genetically similar to MON-24 from Algeria, MON-108 from France and MON-198 from Spain. Therefore, this cluster may represent a new contemporary introduction of L. (L.) infantum from the Old World.
Molecular markers, including microsatellite markers, have been used for a wide range of groups to assess genetic structure, dispersion and to assign individuals to populations [86,87]. However, the usefulness of these markers is limited when the diversity is low, which is a common feature of invasive species, as is the case for L. (L.) infantum in the New World [18,19,29,48,49,58,59]. Moreover, most sample sizes of the present study, as well as that of other investigations (e.g. [21,48]), are low and may not represent the entire diversity of each population, and thus limit statistical power for testing more specific hypothesis about the genetic structure and dispersion of L. (L.) infantum in the Brazil. Collaboration among researchers with the exchange of samples is essential for increasing sample sizes and the distribution of sampling to better understand the dynamics of this and other parasites in Brazil. This information is vital for developing public policies for effective control and restraint of VL dispersion.
Integrating molecular markers with other forms of data, like historical reports, is important and could increase the reliability of reconstructions and hypotheses of dispersal of L. (L.) infantum. However, some inferences using these data must be done with care since limitations on the notification of the cases may bias analyses. First, most of the VL cases found in our research were reported in SINAN, and this database has not presented the hVL cases before 2001. Thus, some cities that SINAN reported the presence of VL in and after 2001 probably had already presented cases of VL before 2001. This factor may explain the increase in the number of hVL cases between 1998 and 2005, especially in Minas Gerais. Second, hVL and cVL may often be reported in cities that are not the site of the original infection (allochthonous cases). Third, several cVL cases may not be diagnosed and/or described. Cities without health professionals trained in hVL diagnosis may have several non-recorded cases. These issues could explain the higher number of cases described in Minas Gerais and São Paulo than in neighbor states that should also have a great number of cities with VL cases (i.e. Mato Grosso, Mato Grosso do Sul, Goiás e Rio de Janeiro). These two states are among the top five states regarding health systems in Brazil (4 th and 5 th , respectively), while their neighboring states rank much lower among the 27 Federal Units (26 states and the Federal District): Mato Grosso 11 th , Goiás 13 th , Mato Grosso do Sul 14 th , and Rio de Janeiro 23 th [88]. Moreover, the few cases reported for neighboring countries is probably due to the absence of unified databases and/or an obligation to describe VL cases. Thus, approaches that integrate different data allow the body of information to be increased, thus reducing potential bias in analyses.
The dispersion routes proposed herein are the result of complex scenarios that include population migrations and environmental changes. The recent urban dissemination of VL in medium and large cities and the spread to other regions of Brazilian suggest worsening scenarios in the future. The movement of people and their infected dogs [89], and the lack of joint policies with countries bordering Brazil (approximately 8 thousand km of border) are risks for VL dispersion to South Region of Brazil. The dispersion routes identified herein should be considered in developing plans to efficiently control VL and avoid its further dispersion in central-south Brazil.

Conclusion
In summary, our results highlight the need of the development of plans that efficiently avoid the dispersion of the visceral leishmaniasis in the central-southern Brazil that includes monitoring of this diseases and joint policies with countries bordering this Brazilian region.