Neighbor danger: Yellow fever virus epizootics in urban and urban-rural transition areas of Minas Gerais state, during 2017-2018 yellow fever outbreaks in Brazil

Background From the end of 2016 until the beginning of 2019, Brazil faced a massive sylvatic yellow fever (YF) outbreak. The 2016–2019 YF epidemics affected densely populated areas, especially the Southeast region, causing thousands of deaths of humans and non-human primates (NHP). Methodology/Principal findings We conducted a molecular investigation of yellow fever virus (YFV) RNA in 781 NHP carcasses collected in the urban, urban-rural interface, and rural areas of Minas Gerais state, from January 2017 to December 2018. Samples were analyzed according to the period of sampling, NHP genera, sampling areas, and sampling areas/NHP genera to compare the proportions of YFV-positive carcasses and the estimated YFV genomic loads. YFV infection was confirmed in 38.1% of NHP carcasses (including specimens of the genera Alouatta, Callicebus, Callithrix, and Sapajus), from the urban, urban-rural interface, and rural areas. YFV RNA detection was positively associated with epidemic periods (especially from December to March) and the rural environment. Higher median viral genomic loads (one million times) were estimated in carcasses collected in rural areas compared to urban ones. Conclusions/Significance The results showed the wide occurrence of YF in Minas Gerais in epidemic and non-epidemic periods. According to the sylvatic pattern of YF, a gradient of viral dissemination from rural towards urban areas was observed. A high YF positivity was observed for NHP carcasses collected in urban areas with a widespread occurrence in 67 municipalities of Minas Gerais, including large urban centers. Although there was no documented case of urban/Aedes YFV transmission to humans in Brazil during the 2016–2019 outbreaks, YFV-infected NHP in urban areas with high infestation by Aedes aegypti poses risks for YFV urban/Aedes transmission and urbanization.


Introduction
Yellow fever virus (YFV) (family Flaviviridae, genus Flavivirus) is endemic in tropical and subtropical regions of Africa, Central and South America. Yellow fever (YF) presents a broad spectrum of severity, with clinical manifestations ranging from self-limited febrile to fatal disease in humans [1,2]. There is no specific treatment for YF, and vaccination is recommended for people who live in or travel to at-risk areas [2]. Urban YF, transmitted by Aedes aegypti vector, was a severe threat to Brazil's human health until 1942 when the last case was registered in Acre state [1]. Nowadays, the virus is maintained in the sylvatic cycle involving Neotropical non-human primates (NHP) and sylvatic mosquitoes (Haemagogus spp. and Sabethes spp.) [2]. Although these mosquitoes mainly feed on NHP, they can incidentally transmit the virus to humans [2]. The reemergence of sylvatic YF outside the Brazilian Amazon basin has been reported, especially during the rainy season, called the epidemic period [3].
Since the eradication of urban YF, Brazil faced the largest sylvatic YF outbreak from the end of 2016 until the beginning of 2019 [4]. YFV was introduced into Southeast [4][5][6][7][8], from the Midwest region [8] spreading southward, causing thousands of NHP and human cases and deaths [4,[9][10][11]. From July 2016 until June 2019, 2,240 human cases and 760 deaths were confirmed, mostly in the Southeast region (99.2%), and over 2,590 YF epizootics were confirmed in Brazil [4,[10][11][12][13]. Minas Gerais, Southeast Brazil, was one of the most affected states during the YF outbreaks. Between July 2016 and June 2018, the Secretary of Health of Minas Gerais confirmed 1,002 human cases and 340 deaths, and more than 490 epizootics in NHP [14]. Minas Gerais is in the recommended area for YF vaccination since 2008, but until the end of 2016, the state presented low average vaccination coverage (57.26%), reaching 92.71% by the end of 2018 [15]. The rapid spread and the extent of the latest recent YF outbreaks in the Southeast region, including highly densely populated urban centers in the states of Minas Gerais, São Paulo, Espírito Santo, and Rio de Janeiro [10,11], have raised concerns regarding YFV reintroduction in epidemiologically receptive areas, with millions of naïve people for YFV. NHP is the primary host of the YF sylvatic cycle, and even though NHP are mostly observed in rural/sylvatic zones, these animals are commonly found in urban-rural transition and urban areas in Brazil. Given the high numbers of epizootics during YF outbreaks, this study aimed at the molecular investigation of YFV in NHP from Minas Gerais. Our results showed the widespread occurrence of YFV in Minas Gerais and, hastily, YFV in NHP carcasses collected inside urban matrixes and urban-rural transition areas, including large urban centers.

Study area and samples
During YF outbreaks in Brazil (January 2017 to December 2018), we analyzed 771 liver samples from carcasses of free-living NHP from different areas of Minas Gerais state. Data regarding ten NHP carcasses previously screened for the YFV RNA [7] were included in data analysis, totalizing 781 individuals analyzed here. The NHP carcasses consisted of a convenient sampling, collected in the context of the Brazilian Yellow Fever Surveillance Program [16], of epizootics that occurred in 2017 and 2018. Since biological samples were obtained from carcasses, blood sampling was not possible, and liver samples were selected for YFV investigation, due to the hepatotropic of YFV in humans and NHP [1,4,16,17]. Minas Gerais is in Southeast Brazil (Fig 1), with 586,521.121 km 2 and 21,040,662 inhabitants. The state has 853 municipalities grouped into 12 mesoregions. Three main biomes cover the state: Cerrado, Atlantic Forest, and Caatinga [18] (Fig 1).
The NHP carcasses were collected in the field by health surveillance agents (Brazilian Yellow Fever Surveillance Program [16], frozen (-20˚C), and transported to the Laboratory of Zoonosis of Belo Horizonte/Minas Gerais (LZOON-BH). At LZOON-BH, NHP carcasses were defrosted and visually inspected regarding the stage of decomposition, taking into account different tissue aspects: (i) brightness (bright or opaque); (ii) color (regular aspect, greenish, greyish or whitish); (iii) consistency (if the tissue was firm or softened); (iv) smell (whether the carcass was smelling rotten or not); and (v) presence or absence of parasites (larvae). Based on that, liver samples were categorized according to their preservation status: good (n = 465; 59.5%), intermediate (n = 211; 27.0%), or bad (n = 59; 7.5%). After defrosting, liver fragments (0.5 cm) were collected from the carcasses and placed into microtubes containing RNA later solution (Ambion, USA) by the LZOON team. The microtubes were kept at 4˚C for 16 hours and then stored at -20˚C until used for RNA extraction at Laboratório de Vírus/ Universidade Federal de Minas Gerais (UFMG).
Using the information on the address or geographic coordinates where carcasses were collected, each carcass was classified based on land use: urban (built-up) areas, urban-rural transition areas, and rural/sylvatic areas. Urban areas (built-up) corresponded to cities, towns, and villages or isolated urban areas, characterized by transformations resulting from urban development (e.g., paved streets, electric lighting, sewerage, buildings, and intense human occupation) [19]. A buffer zone (2.0 kilometers) was established on the outskirts of urban areas, delineating an urban-rural transition area (interface). The rural/sylvatic areas covered the entire area located outside these limits. Each localization of carcasses was checked using satellite imagery (Google Maps, available at https://www.google.com/maps) to determine the boundaries of the urban areas [19] urban-rural interface areas and rural/sylvatic areas as stated above. NHP carcasses analyzed here were collected in Cerrado and Atlantic Forest areas, but not in Caatinga regions.
NHP carcasses were identified at species or genus level, by the Veterinary team at LZOON or by the Zoology/Mastozoology teams at UFMG, using morphological criteria [20][21][22]. Out of 781 NHP analyzed in this study, 360 carcasses (collected in 2017) were forward to UFMG laboratories, and some of them were taxidermized and deposited in the Center of Taxonomic Collections of Universidade Federal de Minas Gerais (CCT-UFMG), Brazil [20][21][22]. The Committee on Ethics in the Use of Animals of UFMG approved the study (CEUA 98/2017).

YFV molecular investigation
Total RNA was extracted from approximately 30 mg of the liver preserved in RNA later solution (Ambion, USA), using RNeasy Minikit (Qiagen, USA). RNA extraction was performed in batches of 11 samples plus one negative control (nuclease-free water). RNA samples were submitted to a one-step real-time polymerase chain reaction (RT-qPCR), with primers targeting the gene coding for β-actin [23], using GoTaq 1-Step RT-qPCR System (Promega, USA). All samples were positive, showing the suitability of samples for RT-qPCR. RNA samples and the negative extraction controls were screened in duplicate for YFV RNA using RT-qPCR [24]. Non-template (nuclease-free water) and YFV 17DD RNA (provided by Bio-Manguinhos/FIO-CRUZ) were used as negative and positive controls, respectively. Samples were considered positive when they presented amplification in duplicate, considering the threshold for cycle quantification value (Cq) of 37. Since between the range of 37-40 Cq indicate minimal quantities of DNA, with low confidence for the results, Cqs>37 or undetermined were considered negative. The results of RT-qPCR runs were manually inspected for the correction of baseline and threshold parameters whenever necessary due to heterogeneity in the amount of input RNA among different samples.
The Cqs are inversely proportional to the input target nucleic acid in the sample, which could indirectly reflect the YFV genomic load in the samples. In that way, we performed a quantitative RT-qPCR (Bio Gene Research Febre Amarela PCR kit-Bioclin-Quibasa, Brazil) using RNA obtained from 28 YFV-positive samples to have an estimative of YFV genomic load in the liver. We determined the number of genomic copies per gram (gc/g) of the liver, and we estimated the genomic loads in all YFV-positive carcasses using linear regression and the observed Cqs.

Statistical analyses
For statistical analyses, samples were subdivided into different groups according to the period of sampling: (i) epidemic (December-May) and non-epidemic (June-November) periods; (ii) bimester within epidemic (December/January, February/March, and April/May) and within non-epidemic periods (June/July, August/September, and October/November); (iii) NHP genera: Callithrix, Callicebus, and Alouatta (Sapajus was not included due to the low number of specimens); (iv) environment: rural, urban-rural interface, and rural areas; and (v) environment together with NHP genera: urban Callithrix, urban-rural Callithrix, rural Callithrix, rural Callicebus, and rural Alouatta (given the low number of specimens, Alouatta and Callicebus from urban and urban-rural interface were not included in the analyses). The comparison of proportions of YFV-positive carcasses within groups was performed using chi-square and chisquare for trend. The Kruskal-Wallis test and multiple pairwise-comparisons between groups by Wilcoxon rank-sum test were used to compare the (median) Cq values observed for the positive samples within each group. Median Cq values were compared only using data from carcasses in good and intermediate conservation status (see Results). Statistical analyses were run in R v.3.6.0 (https://www.r-project.org/) and Epitools (http://epitools.ausvet.com.au/). Bonferroni correction for multiple comparisons was applied, with differences considered statistically significant at p< 0.0125.
Since Cqs reflect the YFV genomic load in liver samples, we analyzed the median Cq values for YFV-positive samples. There was no difference in median Cqs observed in carcasses in good versus intermediate conservation status (x 2 = 9.08, p< 0.010), carcasses in bad conservation status were discarded from this analysis, and only results obtained from carcasses in good and intermediate conservation status were considered. A negative linear regression was observed, and a significant equation was obtained [10E (-0.2986 x Cq + 13.5094)] (R 2 = 0.99 and p< 2.2E-16) to estimate the YFV genomic load using the Cqs in liver samples under the experimental conditions used here.
Partial nucleotide sequences of the envelope gene of 11 samples were obtained (Genbank accession numbers: MN517211-MN517221) (S2 Table), confirming wild-type YFV in NHP carcasses from urban and rural areas of different municipalities of Minas Gerais state. Phylogenetic analysis, involving 112 nucleotide (nt) sequences (621 nt) (S3 Table), indicated that YFV isolates obtained here formed a monophyletic cluster together with other YFV isolates from 2016 to 2018, within South America genotype I (S3 Fig).

Discussion
From late 2016 to 2019, Southeast Brazil faced large and extensive sylvatic YF outbreaks, with thousands of human cases and epizootics, especially in MG state [4,[10][11][12][13]. Based on demand for YFV investigation, we received liver samples belonging to NHP carcasses from January/ 2017 to December/2018. NHP are the primary hosts of the sylvatic YF cycle, and to better understand viral circulation in Minas Gerais state during recent outbreaks, we investigated YFV RNA in NHP carcasses. YFV was detected with high overall positivity of 38% in specimens of Alouatta, Callicebus, Sapajus, and Callithrix. We observed the widespread circulation of YFV in Minas Gerais, detecting the virus in 49 municipalities where YF was not previously reported during the recent YF outbreaks [14,25,26]. In Minas Gerais state, the mesoregions with the highest numbers of YFV-positive carcasses were the same ones with the highest records of human YF cases from 2016 to 2018 [4,14,23,24]. Our phylogenetic analyses confirmed the YFV in different urban areas of Minas Gerais, close to human populations. Besides, the YFV detected in urban and rural environments belonged to the same lineage, indicating the dissemination of YFV from rural/sylvatic towards the urban areas during the recent YF outbreaks.

Variable
Median Cq (range) A gradient of YFV dissemination was observed from rural towards urban-rural interface and urban areas, in conformity with the sylvatic pattern of YF [5], when one would expect a higher exposure to the virus in rural environments. According to the seasonal pattern, a higher positivity for YFV infection was observed during the epidemic periods, when a more extensive transmission of the virus and a higher number of cases would likely occur [9][10][11]. Our data indicate the period from December to March with the highest chances of YFV detection in NHP carcasses, what should be considered by the YF surveillance system, and strengthening the surveillance, especially at the beginning of the epidemic period. Nevertheless, the YFV circulation during the non-epidemic period of 2017 was confirmed here and by official bulletins [10,14] confirming the persistence of YFV in Minas Gerais, from 2017 to 2018 [7], followed by a decline by mid-2018 until 2019 [15].

Median YFV genomic copies (range)/ gram of liver
The NHP genera and species studied here are distributed along with the Atlantic Forest and Cerrado biomes [4,27], the biomes from which NHP carcasses were sampled. Alouatta spp. has been described as extremely sensitive to YFV infection, developing severe and fatal disease [1,4]. However, specimens of Alouatta presenting antibodies against YFV shows that howler monkeys can survive from YF [4,28,29]. Indeed, high rates of howler monkeys presenting protective or neutralizing antibodies against YFV were described in studies conducted in different regions of Panama (50.7% out of 203 individuals) [29] and Brazil (25.5% out of 192 specimens) [28]. There is little information regarding the susceptibility of Callicebus spp. to YFV infection. However, humoral immunity against YFV was demonstrated in 28.6% of 14 specimens investigated by Kumm and Laemmert (1950) [28], in Brazil. In the present study, Callicebus and Alouatta presented similar proportions of YFV-infected carcasses with similar low median Cqs in RT-qPCR. These results might indicate the high sensitivity to YFV of Callicebus, as C. nigrifrons and C. personatus. The YF surveillance and animal conservation programs should consider this information since some of those species are threatened (C. nigrifrons) and vulnerable (C. personatus) with decreasing population trends [30]. Similarly, species within Sapajus genus present population decreasing trends with some already considered critically endangered [30]. Here, only two specimens were sampled, and one carcass was YFV-positive. Although Sapajus specimens may develop severe disease, they have been considered less susceptible to YFV [4]. Our results and previous studies [5,7,31] showed YF outbreaks causing thousands of deaths of NHP, which lives in one of the most threatened ecosystems of the world, the Atlantic Forest [4,18].
Despite many marmoset carcasses received (mostly from urban areas), the overall positivity for YFV in Callithrix was lower than Alouatta and Callicebus. Due to their ability to adapt and survive in degraded and secondary habitats [32], free-living marmosets frequently inhabit urban green areas inside Brazilian cities, close to human populations. The proximity of Callithrix to the anthropic environment and its wide occurrence in urban areas could increase the chances of dead marmosets to be found in a good preservation status. However, we observed a strong negative correlation for detection of YFV in urban marmosets and higher median Cqs compared to Callicebus and Alouatta, suggesting lower genomic viral loads in those urban marmosets. Previous studies showed similar data when Callithrix presented higher Cqs (lower viral loads) than Alouatta during the recent YFV epizootics in Brazil [33]. On the other hand, we did not observe a difference among rural NHP (Callithrix, Alouatta, and Callicebus), which presented median estimates of genomic viral loads of 10E+6 times higher than Callithrix from urban and urban-rural transition areas. When only marmosets were evaluated, we detected a higher positivity in rural areas and a decreasing gradient towards urban areas, consistent with the sylvatic pattern of YF. In that way, the lower overall positivity for YFV could result from the lower exposure to YFV-infected sylvatic vectors in urban areas. Haemagogus spp. is highly competent and considered the primary vector of YFV in Brazil [1,4,34], including the recent outbreaks in the country [34,35]. These mosquitoes are primatophilic species, being found in the tree canopy [36]. However, some species as Haemagogus leucocelaenus and Haemagogus janthynomis can disseminate through great distances [34] and may adapt to modified environments, as urban-rural interface [34,35] and urban green areas [37].
In the present study, we observed YFV-positive carcasses collected in urban areas boundaries and different neighborhoods, including downtown areas or near to the central area of different cities. For example, 25-YFV positive marmosets were collected in urban areas of Belo Horizonte (2.5 million inhabitants), the state capital, and a total of 76 YFV-positive marmosets were sampled in the Metropolitan Region of Minas Gerais state (5.0 million inhabitants). Due to some carcasses location (for example, near the downtown area of Belo Horizonte), the land use, and marmosets territorial behavior [32], some of these NHP would have been likely infected in urban matrixes.
Although NHP have a crucial role in the epidemiology of YF and despite YFV-infected NHP in urban areas, there was no documented case of urban/Aedes transmission of YFV in Brazil since 1942 [1,4,[9][10][11]. The lower viral load (represented by the estimated viral genomic load) in urban NHP could lower the chances for further viral transmission to vectors. Besides, the closure of some parks in urban centers, the vaccination of humans, and the distinct behavior patterns and habitats observed for hosts (NHP and humans) and vectors in urban areas might have played a role in preventing the urban/Aedes transmission of YFV to humans. However, the high infestation of Brazilian urban areas by Aedes aegypti poses a risk for YFV urban/ Aedes transmission [4,34,37,38]. Concerning Aedes, previous studies in Brazil have confirmed the competence of Aedes aegypti and Aedes albopictus, although the latter has shown low transmission competency to YFV [33].
Here we demonstrated the wide occurrence of YFV-infected NHP in Minas Gerais state, including several urban areas, during the epidemic and non-epidemic periods. Although the number of YFV-infected NHP and humans dramatically decreased from the mid-2018, YFV persisted in the Southeast region [5,7,8], at least during three consecutive transmission seasons [39]. These data reinforce that YFV has suitable ecological and climate conditions for its maintenance in Southeast Brazil. Further studies to investigate the sensitivity of urban NHP and the competence of YFV vectors found in urban and urban-rural transition areas should be conducted to understand better the dynamics of YF and the risks for the occurrence of YF in urban centers.