Predicted distribution of sand fly (Diptera: Psychodidae) species involved in the transmission of Leishmaniasis in São Paulo state, Brazil, utilizing maximum entropy ecological niche modeling

ABSTRACT

Leishmaniasis is a public health problem worldwide. We aimed to predict ecological niche models (ENMs) for visceral (VL) and cutaneous (CL) leishmaniasis and the sand flies involved in the transmission of leishmaniasis in São Paulo, Brazil. Phlebotomine sand flies were collected between 1985 and 2015. ENMs were created for each sand fly species using Maximum Entropy Species Distribution Modeling software, and 20 climatic variables were determined. Nyssomyia intermedia (Lutz & Neiva, 1912) and Lutzomyia longipalpis (Lutz & Neiva, 1912), the primary vectors involved in CL and VL, displayed the highest suitability across the various regions, climates, and topographies. L. longipalpis was found in the border of Paraná an area currently free of VL. The variables with the greatest impact were temperature seasonality, precipitation, and altitude. Co-presence of multiple sand fly species was observed in the cuestas and coastal areas along the border of Paraná and in the western basalt areas along the border of Mato Grosso do Sul. Human CL and VL were found in 475 of 546 (86.7%) and 106 of 645 (16.4%) of municipalities, respectively. Niche overlap between N. intermedia and L. longipalpis was found with 9208 human cases of CL and 2952 cases of VL. ENMs demonstrated that each phlebotomine sand fly species has a unique geographic distribution pattern, and the occurrence of the primary vectors of CL and VL overlapped. These data can be used by public authorities to monitor the dispersion and expansion of CL and VL vectors in São Paulo state.

Introduction

Leishmaniasis, a protozoan parasitic disease transmitted via the sand fly, is expanding into previously unaffected areas throughout the state of São Paulo. This expansion is associated with environmental-land use change and an increase in the human population in new urban and peri-urban areas [1,2]. There is a need to determine whether this expansion is also related to changes in local sand fly populations.

Phlebotomine sand flies (Diptera: Psychodidae, subfamily phlebotomine) are vectors of leishmaniasis, a group of diseases caused by intracellular protozoa of the Leishmania genus, that manifest as cutaneous/mucocutaneous (CL/MCL) or visceral (VL) disease forms [3]. In Brazil, the main vectors of CL are Nyssomyia whitmani, Lutzomyia flaviscutellata, Nyssomyia umbratilis, Nyssomyia (N.) intermedia, Psychodopygus wellcomei, Migonemyia migonei. Lutzomyia neivai, and Pintomyia fischeri have not been confirmed as vectors of CL but have been found frequently in the peridomicile areas of disease transmission [4]. Lutzomyia longipalpis and Lutzomyia cruzi are the only known vectors of VL in Brazil; the latter has been confirmed as a vector only in Mato Grosso state [5]. However, permissive vectors are involved in the transmission of VL in some regions. In a study conducted in the state of Pernambuco, in the northeast region of Brazil, females of M. migonei were naturally infected by Leishmania infantum, suggesting that the phlebotomine may be the permissive vector of VL in areas where L. longipalpis is not found [6].

Sand flies have a wide ecological distribution throughout Brazil, but there is a dearth of knowledge on how environmental relationships affect the geographic range and drive vector expansion from established areas into new areas. Differences between sand fly species and geographic distributions are associated with the adaptation of each species to different ecological niches [7]. Furthermore, in some settings, the presence of vectors of VL does not correspond with the presence of infected individuals; conversely, in some places, the presence of infected individuals does not correspond with the presence of vectors [1–5]. Between 1990 and 2019, 737,224 cases of CL were notified; the annual average was 24,574 cases and the average incidence rate was approximately 14.2 cases per 100,000 inhabitants. Data on MCL are scarce and are included with the total cases of CL by the Brazilian Ministry of Health (BMH). Between 1990 and 2019, 93,614 cases of VL were registered. In São Paulo state, CL, zoonotic cutaneous leishmaniasis and mucocutaneous leishmaniasis have been found in several species of wild animals (rodents and marsupials) and domestic animals (dogs, cats, and horses). However, in the urban area, dogs are the main source of VL. In the wild, the reservoirs are foxes which act as maintainers of the disease cycle, and marsupials [4,5]. The average annual number of cases was 3120, and the average incidence rate was approximately 1.74 cases per 100,000 inhabitants [8]. According to the BMH, epidemiological surveillance actions within the scope of CL comprise (1) health care and surveillance of human cases and deaths and adverse drug events; (2) entomological surveillance and vector control; (3) surveillance of reservoirs and hosts, including wild reservoirs, domestic animals, and clinical and laboratory diagnosis in dogs; (4) health education activities involving multi-professional and multi-institutional teams. Epidemiological surveillance actions within the scope of VL comprise (1) early diagnosis and proper treatment of human cases and adverse drug events; (2) vector control depending on the epidemiological and entomological characteristics of each location, including chemical control. Minor measures recommended by the BMH for vector control include the following: the use of repellents when exposed to environments where vectors usually can be found; avoid exposure during vector activity hours; use of fine-mesh mosquito nets and screens on doors and windows; correct environmental management; proper disposal of organic waste; periodic cleaning of domestic animal shelters, and a distance of 400 to 500 m between the residences and the forest is recommended in potential transmission areas; (3) control of the canine reservoir, including euthanasia and the correct disposal of corpses; (4) health education activities requiring the effective involvement of multi-professional and multi-institutional institutions [4,5].

Geospatial analytic techniques are useful tools in epidemiological studies focused on infectious diseases because they enable the development of surveillance and response system strategies for control and prevention based on vector occurrence points and environmental data [9]. The principle of maximum entropy was first proposed by E.T. Jaynes in two papers in 1957. He propounds that the entropy of statistical mechanics and the information entropy of information theory are basically the same thing. Thus, statistical mechanics should be seen as just a particular application of a general tool of logical inference and information theory. Ecological niche modeling is a general-purpose method for making predictions or inferences from incomplete information [10]. For leishmaniasis vectors, ecological niche modeling enables the tracking of their dispersion and helps elucidate their relationship with the expansion of CL and VL. In a study conducted in Bahia state, ecological niche modeling was able to identify climatic variables that contributed most to the distribution of VL vectors in the state [11]. In this study, ecological niche models (ENMs) were developed for VL and CL and the sand flies that may be involved in the transmission of leishmaniasis in São Paulo state.

Materials and methods

Study design

São Paulo state is located in the southeastern region of Brazil (23.5432° S, 46.6292° W) and borders the states of Mato Grosso do Sul (MS), Paraná (PR), Minas Gerais (MG), and Rio de Janeiro (RJ) (Figure 1). In December 2019, São Paulo state had an estimated population of 45,919,049 inhabitants; it occupies an area of 248,216.63 km² with 645 municipalities. The stepwise framework displayed in Figure 2 illustrates the research approach used to create ENMs for phlebotomine sand flies.

Figure 1.

Geospatial location of South America, Brazil, São Paulo state and bordering states of Mato Grosso do Sul (MS); Paraná (PR) and Minas Gerais (MG)

Figure 2.

Procedure framework designed for the study. Baseline, climatic and environmental variables, sand fly collection; regional aggregation datasets, we chose the 645 municipalities of São Paulo state to represent suitability for vector VL; statistical and geospatial analysis: (a) mapped the 16 prediction models for sandflies that occurred in at least 10% of the municipalities of São Paulo; (b) selected the most and the least significant climate variables for suitability prediction; results, suitability of sand flies, AUC, and the jackknife test outcome

The base maps used in the Geographic Information System (GIS) were road maps and municipality shapefiles provided by the Brazilian Institute for Geography and Statistics (Instituto Brasileiro de Geografia e Estatística [IBGE]) [12]. Sand fly vector records were aggregated by municipality. Latitude and longitude coordinate points were extracted from the centroid of each municipality. A spreadsheet detailing vector information for each municipality was created for presence-only ENM creation using the Maximum Entropy Species Distribution Modeling (MaxEnt) software version 3.3.3 (https://www.gbif.org/tool/81279/maxent) (Figure 3). Current global climate data were obtained from WorldClim [13], clipped to the municipalities, and converted from raster Georeferenced Tagged Image File Format (GeoTIFF) to American Standard Code for Information Interchange (ASCII) files. To determine which sand fly species would be included in the MaxEnt analysis, a selection criterion that included only sand fly species that occurred in at least 8% of the state municipalities during the period from 1985 to 2015 was used. This criterion was chosen because previous research attempting to examine the association between vector presence and environmental variables demonstrated that a minimum of 40 samples from São Paulo municipalities were required to infer the geography of the sand flies species N. whitmani, M. migonei, Lutzomyia pessoai, and N. intermedia [14].

Figure 3.

Ecological niche models of (a) Lutzomyia longipalpis; (b) Nyssomyia intermedia; (c) Nyssomyia whitmani; (d) Migonemyia migonei; (e) Nyssomyia neivai; (f) Psathyromyia shannoni; (g) Brumptomyia brumpti; (h) Expapillata firmatoi. Source: SUCEN, IBGE (base maps). The points of presence shown on the map are (a) Nyssomyia whitmani, 27.75%; (b) Expapillata cortelezzii, 8.83%; (c) Migonemyia migonei, 30.38%; (d) Lutzomyia ubiquita, 22.01%; (e) Lutzomyia longipalpis, 19.37%; (F) Nyssomyia intermedia, 37.36%; (g) Nyssomyia neivai, 22%; (h) Pintomyia monticola, 9.92%; (i) Pintomyia fischeri, 30.69%; VL, 14.10%; CL, 73.64%. The coordinates are available as supplementary files

Entomological and epidemiological data

The sand fly vectors analyzed in this study were L. longipalpis, N. intermedia, N. whitmani, M. migonei, N. neivai, Psathyromyia shannoni, Evandromyia lenti, Expapillata firmatoi, Pintomyia monticola, Evandromyia cortelezzii, Evandromyia edwardsi, Psychodopygus arthuri, Psathyromyia lanei, Psychodopygus geniculatus, Brumptomyia brumpti, and Pintomyia fischeri. Data on phlebotomine sand flies identified in entomological collections collected during epidemiologic surveillance activities by Endemic Control Supervision (Supervisão de Controle de Endemias [SUCEN]) from 1985 to 2015 were used. The collected sand flies were processed and were identified using the taxonomic key of Galati [15]. São Paulo state vector control utilized entomological surveys, entomological foci research, and fixed stations with CDC light traps to provide epidemiological surveillance. The SUCEN identification survey protocol discontinues surveillance within an area after at least one sand fly is collected [5]. The sand flies collected for this study were captured using CDC light traps and Shannon traps. Representative sites associated with peridomicile and extra-domicile habitats were selected for sampling as per the SUCEN surveillance protocol [5]. The classification of species utilized was created by Shimabukuro et al. [16]. Representative sites associated with peridomicile (the external area of a residence within a radius not exceeding 100 m; or in the balconies of houses) and extra-domicile habitats (about 15 m from the house) were selected for sampling as per SUCEN surveillance protocol [5].

Environmental data

Environmental data were collected from the Global Climate Data website [13], corresponding to climate data with a spatial resolution of 1 km². BIOCLIM variables are derived from WorldClim monthly temperature and rainfall averages spanning 1970–2000. The BIOCLIM variables are annual mean temperature (BIO1), mean diurnal range (BIO2), isothermality (BIO3), temperature seasonality (BIO4), the maximum temperature of the warmest month (BIO5), the minimum temperature of the coldest month (BIO6), temperature annual range (BIO7), mean temperature of the wettest quarter (BIO8), mean temperature of the driest quarter (BIO9), mean temperature of the warmest quarter (BIO10), mean temperature of the coldest quarter (BIO11), annual precipitation (BIO12), precipitation of the wettest month (BIO13), precipitation of the driest month (BIO14), precipitation seasonality (BIO15), precipitation of the wettest quarter (BIO16), precipitation of the driest quarter (BIO17), precipitation of the warmest quarter (BIO18), precipitation of the coldest quarter (BIO19), and one topographic variable, elevation (ELEV).

Ecological niche modeling

Ecological niche modeling was used to relate the geographic distribution of species with their potential ecological niches [7,10,11]. The geographic variations in the occurrence of species are often profoundly influenced by specific climatic and environmental constraints [17,18]. In ecological niche modeling, the presence data observed (and pseudo-absence data), together with ecological variables at the sample region, are used to provide a reasonable likelihood of the presence of the species at other locations. Such projections assume that the distribution of a species is mainly determined by its environmental requirements and not by other factors [7,10,11]. MaxEnt is a presence-only niche modeling software tool used to identify the potential distribution of species. This method estimates the distribution of species by finding the most uniform distribution of sampling points compared with background locations given the constraints derived from the data [18]. MaxEnt has emerged as a reliable technique for predicting the distribution of species comparable with genetic algorithms (GARP) and regression trees [19,20]. The advantage of maximum entropy methods such as MaxEnt is their ability to predict relative probabilities of occurrence rather than the simple presence or absence of species [19,20].

Evaluating model performance using the area under the curve

The area under the curve (AUC) of the receiver operating characteristic (ROC) is a common tool used to evaluate ENM performance and accuracy. The ROC has the advantage of being threshold independent and therefore does not require decisions regarding thresholds of what constitutes a prediction of presence versus a prediction of absence. The AUC value of the ROC provides a measure of model performance, or how accurately the model predicts the presence and absence of species. Previous literature [21] has set AUC value thresholds as follows: <0.7, non-informative; >0.7–0.9, informative; >0.9, highly informative. In this study, the AUC value of each ENM’s ROC curve was used to investigate the overall predictive power and performance of the model.

To determine the variable importance for each ENM, MaxEnt’s jackknife test outputs were used. The jackknife test evaluates the contribution of each variable to the ENM, demonstrating which environmental variables are important for each sand fly species niche prediction.

MaxEnt was run using the logistic output to estimate the suitability index (predicted probability of presence), which gives values between 0 and 1, indicating impossible conditions (0) to highly suitable conditions (1) [21,22].

Comparing ecological niche model overlap of cutaneous and visceral leishmaniasis with sand fly vectors

ENMTools software was utilized to mathematically assess the degree to which the ecological niches produced for VL and CL overlapped with the various sand fly vectors, as well as assess the niche overlap between the vectors themselves. ENMTools measured the similarity between MaxEnt niches produced using both Shoener’s D and the I metric [20–22]. These statistics measure niche overlap by taking the difference between suitability score for each species for each grid cell after the suitabilities have been standardized so that they sum to 1 for the entire measured geographic space. These metrics range from 0 (species have completely discordant ENMs) to 1 (species have identical ENMs). The ENMs analyzed by ENMTools were created with MaxEnt 3.4.0 using the clog log output format.

Human cutaneous and visceral leishmaniasis in municipalities of São Paulo state

Data on human leishmaniasis were obtained from public agencies, including the Epidemiological Surveillance Center of São Paulo, which organizes and publishes data on infectious diseases recorded in São Paulo state annually [23,24]. The databases for the maps were collected from the IBGE website. The maps showing the spatial distribution were made using the ArcGIS 10.2 platform (ESRI 2014).

Results

Ecological niche model: phlebotomine distribution in São Paulo state

From 2005 to 2016, 9208 cases of CL were reported and from 1999 to 2018, 2952 cases of VL were reported. Since the vector L. longipalpis was first discovered in Araçatuba in 1996, 193 of the 645 municipalities (29.9%) have reported its presence.

The sand fly vectors found in the highest number of municipalities were as follows: N. intermedia (199), N. whitmani (169), L. longipalpis (164), Pintomyia fischeri (156), M. migonei (154), N. neivai (141), L. pessoai (112), B. brumpti (92), P. monticola (62), P. shannoni (59), E. cortelezzii (56), and E. firmatoi (53). N. intermedia, N. whitmani, and L. longipalpis were found in 30.9%, 26.2%, and 25.4% of the 645 municipalities, respectively.

Eight sand fly species, the main vectors of leishmaniasis reported in São Paulo state, were selected for construction of an ENM: L. longipalpis, N. intermedia, M. migonei, N. whitmani, N. neivai, P.shannoni, B. brumpti and E. firmatoi (Figure 3).

Suitability for L. longipalpis it was found in an extensive area covered by two main topographies: the basalt coastal plateau in the western region of São Paulo state on the border of Mato Grosso do Sul and the coastal area along the main roads between the cities of Rio de Janeiro and São Paulo (Figure 3(a)).

Conversely, higher suitability areas for N. intermedia were found in a large area between the central plateau region and the eastern region of cuestas, including the coastal region. The highest areas of suitability were found on the border of Minas Gerais and Rio de Janeiro, the south of São Paulo along the Paraná state border, and in areas of the Basalt Coast Plateau in the western region (Figure 3(b)).

N. whitmani suitability had a broad distribution and was located predominantly along the borders between São Paulo state and Mato Grosso do Sul, Minas Gerais, and Paraná states. Higher suitability was also demonstrated in the cuestas and coastal areas along the Paraná state border (Figure 3(c)).

Areas of higher suitability for M. migonei were in the cuestas and coastal areas within the southeastern region of the state. The borders of Mato Grosso do Sul and Paraná states were found to be less suitable areas (Figure 3(d)).

N. neivai demonstrated suitability in different regions of São Paulo state, predominantly along the border of Mato Grosso do Sul and Paraná states in the southwest and the border of Minas Gerais state in the northeast. Small areas of higher suitability were also found in the cuestas region (Figure 3(e)).

The distribution of P. shannoni was clearly delineated, occupying 30–40% of São Paulo state. Beginning in the western region plateau areas along the Mato Grosso do Sul border, P. shannoni it was present along 80% of the Paraná border (Figure 3(f)).

The suitability distribution of B. brumpti (Figure 3(g)) was similar to that of M. migonei (Figure 3(d)).

The predominant area of suitability for E. firmatoi was in the eastern region of São Paulo state, in areas of cuestas and along the coasts, the Rio de Janeiro border in the north region, and the Paraná state border in the south (Figure 3(h)).

Co-presence of the sand flies was demonstrated at the regional level. N. intermedia, N. whitmani, M. migonei, N. neivai, P. shannoni, and E. firmatoi were found in the southern region, in cuestas and along the coast, and along the northeast area of Paraná state border, a region covered by preserved Atlantic forest. Co-presence of L. longipalpis, N. neivai, P. shannoni, and B. brumpti was found in the basalt and coastal areas in the west, along the border of Mato Grosso do Sul and Paraná states. This area is an environmentally modified region.

Environmental analysis and the contribution of the variables

The importance of the 19 BIOCLIM variables and altitude for ENM creation was evaluated for 16 different species of sand flies recorded in São Paulo state from 1985 to 2015. Frequency tables were created to describe the most and the least influential variables for each species and the most frequent variables found when analyzing all vectors together. The most frequent variable found in the ENMs of 11 of the 16 species (68.8%) was BIO2 (mean diurnal range), followed by BIO15 (precipitation) in 10 of 16 (62.5%), and altitude in 8 of 16 (50.0%).

The variables with the highest percentage contribution to the potential distribution model for all species were mean diurnal range and minimum temperature of the coldest month. Conversely, the variables with the lowest percentage contribution were precipitation of the wettest month and the mean temperature of the warmest quarter (Table 1).

Table 1.

_{Variables that produces the largest and smallest AUC when included.}

Species	Variable that produce the largest AUC when included	(%) contribution	AUC value	Variable that produces the smallest AUC	(%) contribution	AUC value
Lutzomyia longipalpis	BIO14	21.3	0.835	BIO13	0.1	0.653
Nyssomyia intermedia	BIO15	30.9	0.642	BIO3 (x100)	8.5	0.633
Nyssomyia whitmani	BIO4	33.8	0.697	BIO6	10.2	0.667
Migonemyia migonei	BIO4	29.3	0.665	BIO5	8.3	0.581
Nyssomyia neivai	BIO2	39.9	0.666	BIO9	7	0.609
Psathyromyia shannoni	BIO15	55.6	0.591	Altitude	2.4	0.688
Evandromyia lenti	BIO2/BIO7 (x100)	37.1	0.648	BIO15	2.8	0.482
Expapillata firmatoi	BIO14	52.1	0.678	BIO16	2.8	0.491
Pintomyia monticola	BIO9	56.4	0.754	BIO7	6.4	0.622
Evandromyia cortelezzii	BIO16	29.8	0.701	BIO9	8.9	0.722
Evandromyia edwardsi	BIO2	53.6	0.762	BIO13	6.6	0.511
Psychodopygus arthuri	BIO6	67.2	0.891	BIO10	0.8	0.699
Psathyromyia lanei	BIO2	70.4	0.935	BIO18	5.4	0.542
Psychodopygus geniculatus	Altitude	56.6	0.631	BIO2	8.1	0.572
Brumptomyia brumpti	BIO14	31.7	0.581	BIO5	8.5	0.547
Pintomyia fischeri	BIO2	23.1	0.724	BIO11	8.2	0.581

The BIOCLIM variables are as follows: mean diurnal range (BIO2), isothermality (BIO3), temperature seasonality (BIO4), maximum temperature of the warmest month (BIO5), minimum temperature of the coldest month (BIO6), temperature annual range (BIO7), mean temperature of the driest quarter (BIO9), mean temperature of the warmest quarter (BIO10), mean temperature of the coldest quarter (BIO11), precipitation of the wettest month (BIO13), precipitation of the driest month (BIO14), precipitation seasonality (BIO15), precipitation of the wettest quarter (BIO16), precipitation of the warmest quarter (BIO18).

Figure 4 displays the AUC values for the different sand fly species. P. lanei and L. longipalpis produced the largest AUC values (0.935 and 0.835, respectively) and E. lenti produced the smallest AUC (0.482) for the test data. The AUC values were >0.80 for the P. lanei (0.935), P. arthuri (0.891), and L. longipalpis (0.835) models. ENMs were also created and produced high AUC values for P. monticola (0.754), P. fischeri (0.724), and E. edwardsi (0.762) (Table 1). The average AUC value overall was 0.71 (95% confidence interval, 0.65–0.76).

Figure 4.

Receiver operating characteristic (ROC) curve for the different species of sand fly vectors. Nyssomyia whitmani: training data, 0.761; test data, 0.697. Evandromyia cortelezzii: training data, 0.803; test data, 0.701. Migonemyia migonei: training data, 0.750; test data, 0.665 (0.5). Lutzomyia ubiquita: training data, 0.711; test data, 0.669. Lutzomyia longipalpis: training data, 0.894; test data, 0.835. Nyssomyia intermedia: training data, 0.692; test data, 0.642. Nyssomyia neivai: training data, 0.716; test data, 0.666. Pintomyia monticola: training data, 0.793; test data, 0.754. Pintomyia fischeri: training data, 0.763; test data, 0.724. All data: specificity versus sensitivity is measured according to the AUC, considering the training data, test data, and random prediction. The maximum achievable AUC is less than 1. The random prediction is 0.5

The jackknife test of variables for Lutzomyia longipalpis

L. longipalpis is the primary vector involved in the dissemination of VL throughout the state of São Paulo and was found in different regions, climates, and topographies (Figure 3(a)). Climate variables with the highest contribution to the ENM, according to the jackknife tests performed by MaxEnt, were altitude, precipitation of the driest month (BIO14), precipitation seasonality (BIO15), precipitation of the wettest quarter (BIO16), and precipitation of the driest quarter (BIO17). As displayed in Figures 5, Figure 6, and Figure 7, the jackknife test of variable importance revealed that both the test gain and test AUC were high for altitude (0.40 and 0.76), precipitation of the driest quarter (BIO 17, 0.35 and 0.72), and precipitation of the coldest quarter (BIO 19, 0.45 and 0.73), and may therefore highly influence vector occurrence (Figure 5).

Figure 5.

Jackknife test for regularized training gain for L. longipalpis.

Figure 6.

Jackknife of AUC for L. longipalpis.

Figure 7.

Jackknife of test gain for L. longipalpis.

Ecological niche overlap analysis

The ecological niche overlap analysis was performed using the ENMs for the sand fly vectors N. whitmani, B. brumpti, E. firmatoi, N. intermedia, L. longipalpis, M. migonei, N. neivai, P. shannoni, and ENMs for both VL and CL. Niche overlap was evaluated using both Shoener’s D (Table 2) and the I metric (Table 3). For both metrics, the highest vector niche overlap for CL was with N. intermedia (D = 0.8826, I = 0.9858) and the highest vector niche overlap for VL was with L. longipalpis (D = 0.6947, I = 0.9112).

Table 2.

_{Ecological niche overlap analysis of sand fly vectors, cutaneous leishmaniasis (CL), and visceral leishmaniasis (VL)using Schoener’s D metric.}

Species	Nyssomyia whitmani	Brumptomyia brumpti	CL	Expapillata firmatoi	Nyssomyia intermedia	Lutzomyia longipalpis	Migonemyia migonei	Nyssomyia neivai	Psathyromyia shannoni	VL
Nyssomyia whitmani	1.0000	0.8392	0.8499	0.7368	0.8167	0.5925	0.8525	0.7852	0.7557	0.4500
Brumptomyia brumpti	x	1.0000	0.8383	0.7665	0.8238	0.5811	0.8490	0.8228	0.8172	0.4707
CL	x	x	1.0000	0.7774	0.8826	0.5548	0.8730	0.8521	0.7717	0.4014
Expapillata firmatoi	x	x	x	1.0000	0.8383	0.4533	0.8155	0.7490	0.7185	0.3438
Nyssomyia intermedia	x	x	x	x	1.0000	0.5247	0.8960	0.8047	0.7766	0.3781
Lutzomyia longipalpis	x	x	x	x	x	1.0000	0.5451	0.5433	0.6102	0.6947
Migonemyia migonei	x	x	x	x	x	x	1.0000	0.8292	0.7878	0.4056
Nyssomyia neivai	x	x	x	x	x	x	x	1.0000	0.7296	0.3991
Psathyromyia shannoni	x	x	x	x	x	x	x	x	1.0000	0.5253
VL	x	x	x	x	x	x	x	x	x	1.0000

Table 3.

_{Ecological niche overlap analysis of sand fly vectors, cutaneous leishmaniasis (CL), and visceral leishmaniasis (VL) using the I metric.}

Species	Nyssomyia whitmani	Brumptomyia brumpti	CL	Expapillata firmatoi	Nyssomyia intermedia	Lutzomyia longipalpis	Migonemyia migonei	Nyssomyia neivai	Psathyromyia shannoni	VL
Nyssomyia whitmani	1.0000	0.9698	0.9700	0.9180	0.9499	0.8696	0.9549	0.9473	0.9361	0.7524
Brumptomyia brumpti	x	1.0000	0.9758	0.9565	0.9717	0.8517	0.9763	0.9695	0.9682	0.7630
CL	x	x	1.0000	0.9550	0.9858	0.8309	0.9837	0.9783	0.9603	0.7086
Expapillata firmatoi	x	x	x	1.0000	0.9736	0.7389	0.9706	0.9432	0.9401	0.6522
Nyssomyia intermedia	x	x	x	x	1.0000	0.8051	0.9868	0.9629	0.9646	0.6888
Lutzomyia longipalpis	x	x	x	x	x	1.0000	0.8172	0.8216	0.8553	0.9112
Migonemyia migonei	x	x	x	x	x	x	1.0000	0.9717	0.9643	0.7075
Nyssomyia neivai	x	x	x	x	x	x	x	1.0000	0.9387	0.6995
Psathyromyia shannoni	x	x	x	x	x	x	x	x	1.0000	0.7981
VL	x	x	x	x	x	x	x	x	x	1.0000

Cumulative incidence and spatial distribution of cases of cutaneous leishmaniasis in São Paulo state:1998–2019

Higher incidences (58–3793) of individuals infected with CL were distributed in all areas of São Paulo state, but they were more concentrated on the Atlantic coast: in the south, on the border of Paraná state; in the north, they were found in the region of the Paraíba Valley, on the border of Rio de Janeiro. However, high numbers of individuals infected with CL were found on the other side of São Paulo state, in the western region, on the border of Mato Grosso do Sul, with some municipalities bordering the Raposo Tavares highway (Figure 8).

Figure 8.

Cumulative incidence and spatial distribution of cutaneous leishmaniasis in municipalities of São Paulo state from 1998 to 2019. Source: Centro de Vigilância Epidemiológica Alexandre Vranjac (2019). Base map: Malhas digitais do IBGE 2010

Cumulative incidence and spatial distribution of cases of human visceral leishmaniasis in São Paulo state: 1999–2019

The mean incidence of cases was higher in the western region, on the border of Mato Grosso do Sul and in the central region of the state, following the Marechal Candido Rondon highway, in the direction of the metropolitan area of São Paulo, the capital. Some municipalities showed higher cumulative incidences (>14 per 100,000 inhabitants) parallel to municipalities with lower incidence rates (<2.9). Transmission by epidemic leaps has been verified, i.e., infected individuals were identified in distant municipalities in which the disease has not been established previously (Figure 9).

Figure 9.

Cumulative incidence and spatial distribution of human visceral leishmaniasis in municipalities of São Paulo state from 2000 to 2019. Source: Centro de Vigilância Epidemiológica Alexandre Vranjac (2019). Base map: Malhas digitais do IBGE 2010

Discussion

The geographic distribution of infectious diseases in tropical and subtropical regions is poorly characterized. The lack of regional level maps displaying an overview of disease characteristics, particularly for leishmaniasis, has been reported previously [2,11]. Globally, around 9.6 million deaths occur annually due to parasites, bacteria, viruses, and fungi, and leishmaniasis ranks third among them [25,26]. Brazil currently harbors 96% of individuals infected with VL in Latin America and the burden of the infection is mediated by vectors [27].

In the time period studied, the most important vectors of CL and VL, N. intermedia and L. longipalpis, were found in 30.9% and 25.4% of the 645 municipalities, respectively, highlighting the importance of these vectors in the burden of leishmaniasis in São Paulo state. Suitability for L. longipalpis, the main vector of VL, was found countrywide in extensive areas with different topographies [11,28,29]. In São Paulo, with increased levels in the Basalt Coast Plateau along the Mato Gross do Sul border in the western region, the sand fly distribution is spreading toward Paraná state, an area currently free of VL (Figure 3(a)). The reasons for increase in suitability for L. longipalpis throughout this area are not well established. The area is considered the poorest region of the state, with dry soil in the winter and rainy/wet soil in the summer, and temperatures have been increasing in recent decades. Between 1990 and 2010, the rainforest was replaced by pasture and sugar monoculture, facilitated by the influx of a large number of migrant rural workers engaged in sugarcane plantation work. The modified environment may have had an impact on the temperature, humidity, and precipitation, well-known risk factors favorable for the survival, adaptation, and dissemination of sand flies [30–32].

Following the Marechal Candido Rondon highway, L. longipalpis sand flies are disseminating in the direction of the capital, São Paulo, the cuestas, and coastal areas (Figure 2(a)). In the ROC curve analysis P. lanei and L. longipalpis had high AUC values and consequently the most paired training and test data (0.894), demonstrating excellent model power and performance. Altitude, precipitation of the driest quarter, and precipitation of the driest month were the climate variables that most contributed to L. longipalpis suitability and may highly influence the occurrence of the vectors (Figure 5).

It is well known that N. neivai, P. shannoni, B. brumpti, and M. migonei may be permissive vectors of VL in areas where L. longipalpis it has not been recorded, but human cases do occur [33,34]. In São Vicente Ferrer, Pernambuco, north-eastern Brazil, cases of human VL and canine VL have been reported, but in entomological studies, L. longipalpis they were not found despite extensive searches [6]. In Fortaleza, another capital in the north-eastern region and endemic for human VL, a total of 14,237 specimens of M. migonei were captured and suggested as a potential vector of VL in different areas of the city [35]. Studies from La Banda, Argentina, have correlated the absence of L. longipalpis and the presence of M. migonei with cases of autochthonous VL in humans and dogs, indicating their role as a permissive vector of VL. In all these settings, the vector species was associated with environmental degradation, social vulnerability, periurban–rural transition habitats, and domestic animals [36].

In Minas Gerais, on the border of São Paulo state, nine species of Lutzomyia sand flies, including L. neivai, were found to be naturally infected with Leishmania infantum [33]. Permissive sand flies can adapt to different Leishmania species, thereby introducing leishmaniasis transmission in new areas. This adaptability has epidemiological consequences as globalization increases. Presently, changes in climate, deforestation, expansion of human settlements, and the displacement of large numbers of people and animals to different regions and continents have led to an increased risk of vector-borne disease dissemination, including leishmaniases [25,26].

Many authors have stressed the potential value of mapping vector-borne diseases [9,16,37,38]. Our data suggest that there are important gaps in knowledge of the distribution of L. longipalpis in São Paulo state due to data collection difficulties. Research often refers only to recent discovery of the vector, but there is a chance that it was already present in previous times and instead was just not found due to a lack of vector survey data or a scarce number of cases of VL recorded in each municipality [8]. The highest suitability of N. intermedia, the main vector involved in CL in São Paulo state, was found in the eastern region of cuestas, including the coastal region on the border of Minas Gerais and the northern border of Paraná state (Figure 2(b)). However, due to additional competent vectors, CL has a broad distribution throughout the state (Figure 2(c,d,e)). In a timeline analysis spanning 2007 to 2018, 4107 patients with CL were reported. The municipalities of Registro (617 cases), Sorocaba (235 cases), and Itapeva (157 cases), located in the region of the Ribeira Valley, which consists of large areas of preserved Atlantic rain forest, and is close to Paraná state, reported the highest number of cases (Figure 2(b,c,d)). Presidente Venceslau, in the western region, also reported an increased number of cases of CL, and this is an area where L. neivai has increased suitability (Figure 2(e)) [23]. In the same way that the distribution of L. longipalpis is linked to cases of VL, the presence of CL vectors is linked to the areas where patients infected with CL were found [24] (Figure 6 and Figure 7). To quantify the niche overlap between the vectors and CL in our study, the same approach utilized to assess VL was used and demonstrated that N. intermedia it had the highest vector niche overlap for cutaneous leishmaniasis (D = 0.8826, I = 0.9858). This again demonstrates the important role that N. intermedia plays in disseminating CL throughout the state [39].

As ENMs are consolidated models used to relate the geographic distribution of species and their potential ecological niches, the influence of climatic variables on the presence of sand flies and on leishmaniasis dissemination has been previously demonstrated. Depending on the region, weather, and species analyzed, temperature seasonality, humidity, and altitude were strongly linked to the population of phlebotomine sand flies [18,40–43]. In this study, the variables that most influenced the presence of species of sand flies were the mean diurnal range (68.8%), precipitation (62.5%), and altitude (50.0%). Our results are similar to previous research, which found temperature seasonality and annual mean precipitation were the variables that most influenced the suitability of phlebotomine sand flies in the urban area of Timóteo, MG [44]. Additional research has demonstrated that temperature seasonality, annual precipitation, and seasonality of precipitation were some of the climatic variables associated with suitability and expansion of N. whitmani and L. intermedia in CL in Brazil [38], and L. flaviscutellata in South America [7]. N. whitmani has also been considered a secondary vector of CL, with high density in the endemic areas of the states of São Paulo, Minas Gerais, and Espirito Santo. The vector has been implicated in the link between the wild environment and areas surrounding homes. In the southeast, N. whitmani tends to expand toward areas occupied currently by N. intermedia [38]. Temperature was also associated with VL vector suitability in Bangladesh [40] and with CL vector suitability in Tunisia [45].

The limitations of our research should be noted: (1) only the AUC was used to evaluate the models; (ii) vector collection was not uniform and was biased toward areas where disease was found previously; (3) possible collinearity of BIOCLIM variables; (4) entomologic studies were performed only on randomly selected areas; (5) due to financial limitations, the presence of vectors of CL has not been recorded since 2015.

Conclusion

This research demonstrated that using ENMs model, each species of phlebotomine sand flies has unique geographic distribution patterns. They were found in different regions, climates, and topographies, and their co-presence was observed at regional levels. The occurrence of primary vectors, N. intermedia and L. longipalpis, correspond with endemic regions of human CL and VL. The broad distribution and expansion of other vectors, such as N. whitmani and M. migonei, especially in areas where L. longipalpis and N. intermedia were not recorded but human cases do occur, highlight that they are reliable sources of permissive vectors in VL and CL dispersion. Furthermore, the capacity of these vectors to adapt and expand to new environments was demonstrated. These data can be used by public authorities to better monitor the dispersion and expansion of CL and VL vectors in São Paulo state.

Funding Statement

FAPESP, Grant Number 2014/12494-0.

Disclosure statement

No potential conflict of interest was reported by the authors.