**3. Discussion**

Our analyses indicate that additional data are required to better identify routes of WNV importation into and within Brazil and to more generally understand the local transmission dynamics of the virus. Interestingly, our data sugges<sup>t</sup> that the circulation of the virus may have resulted from multiple independent introductions, since the new isolates did not group with the previously sequenced genome in 2019 from the Espirito Santo state. This suggests that intra-continental introduction events due to the mobility of infected birds or mosquitoes might be a more plausible mechanism for the multiple introductions of WNV in South American countries, including Brazil. This scenario is consistent with previous studies that showed that multiple independent introductions into Latin America occurred during the initial outbreak in US in 1999; detailed revision is provided in [28]. While migrating birds are a convenient explanation of WNV dispersal, other possible ways of dispersion exist, such as infected mosquitoes that are accidentally transported via airplane or by road transport [29]. Another likely scenario is commercial legal or ilegal human transportation of birds and/or mosquitoes, which could be transported on airplanes [29].

The current data scarceness prevents definite conclusions on key aspects of WNV epidemiology. For example, given the unconfirmed nature of the reported cases by SINAN for Piauí and Espírito Santo, it is unclear what the proportion of cases truly reflect WNV occurrence and seasonality, hampering our ability to ascertain how representative our theoretical projections are. For Piauí, we would speculate that reported cases may indeed reflect some aspects of WNV seasonality, given that this state had the largest number of cases reported while also being the region of Brazil for which we estimated higher transmission potential and that our estimated transmission potential was well correlated with reported cases (albeit with a possible lag of one month typical of mosquito-borne viruses). At the same time, while inferred trees including the new genome sequences sugges<sup>t</sup> that inter-continental introduction events might be frequent in Brazil, the lack of higher spatio-temporal sampling restricts our ability for definite conclusions on viral movement and persistence.

The phylogenetic and epidemiologic perspectives presented in this study, based on both existing and novel data as well as theoretical projections, sugges<sup>t</sup> that both scenarios of sporadic and endemic local transmission are possible [30]. Similarly to sudden changes in WNV epidemiology and transmission as recently observed in other countries, the occurrence of a WNV outbreak affecting humans in Brazil may simply be a matter of time. Shifting from passive to active WNV screening and sequencing in animal reservoirs (e.g., equines, birds, vectors) in Brazil must be implemented to better understand the virus' local epidemiology and to be able to act accordingly in preventing and controlling any future epidemics with spill-over to humans.

### **4. Materials and Methods**

### *4.1. Sample Collection, Viral RNA Isolation and PCR Screening*

Samples (red blood cells, RBCs) from three horses with suspected WNV infection obtained from southern (Minas Gerais and São Paulo) and northeastern (Piauí) Brazilian states were sent for molecular diagnosis at the Laboratório de Patologia Molecular at the Federal University of Minas Gerais (UFMG).

Sample 1 from 11 July 2018 was collected from a 9-month-old female horse in a farm in the state of Minas Gerais, Mangueiras neighbourhood (Sabará), 15 km from the capital Belo Horizonte. Clinical findings were consistent with bilateral blindness. Neurological examination revealed no other abnormalities. The ophthalmological exams (direct and indirect pupillary light reflex (PLR), fluorescein eye stain test, fundus examination, and intraocular pressure) were consistent with retinal disease, mainly with chorioretinitis.

Sample 2 from 30 July 2019 was collected from a 13-year-old male horse that presented seizure episodes, muscle stiffness, tremor retinal, and flaccid paralysis in a farm located in São Bernardo do Campo countryside of the São Paulo state. Twenty-four days after the onset of neurological signs, the animal had severe pain in the forelimbs from laminitis, and it was euthanized due to hoof decumulation.

Sample 3 from 21 August 2020 was collected from a male horse, 5 years old, which died 72 h after presenting neurological signs, in a farm located in the municipality of Parnaíba, Piauí state. The animal presented motor incoordination, paddling movements, loss of sensitivity over the spine column, and behavioral changes. In this municipality, the tenth human case in Brazil was also detected, presenting neuroinvasive disease compatible with WNV infection, confirmed by serological assay (IgM) in both serum and cerebrospinal fluid (CSF) samples during acute and convalescent phases.

Whole blood samples obtained from the three horses were centrifuged at 1260× *g* for 20 min, and the plasma and buffy coat fractions were collected and stored at 4 ◦C. Red blood cells (RBC) were washed by centrifugation three times in phosphate-buffered saline (PBS) at 1260× *g* for 10 min and stored also at 4 ◦C [15]. RNA from each unit (washed RBC, plasma and buffy coat) were extracted using the QIAmp Viral RNA Mini kit (Qiagen, Hilden, Germany), following manufacturer's recommendations.

Diagnostic investigation of arboviruses was performed by a generic RT-PCR targeting the flavivirus non-structural protein 5 (NS5) gene [31] and alphavirus non-structural protein 1 gene (nsP1) [32]. West Nile virus-specific degenerated primers: forward primers (+) AACCKCCAGAAGGAGTSAAR and reverse primers ( −) AGCYTCRAACTCCAGRAAGC were used in second reaction of nested PCR targeting the NS5 gene after a genus specific flavivirus RT-PCR amplification [22]. A synthetic gene fragment of partial NS5 gene (gblocks gene fragment, Integrated DNA Technologies) was used as a positive control. The 25 μL PCR "master-mix" comprised 2.5 μL of 10× PCR buffer, 1.5 mM MgCl2, 0.4 μM of each primer (forward and reverse), 0.8 μM dNTP mixture (Phoneutria, Sao Paulo, Brazil), 1 U Taq DNA polymerase (Platinum Taq DNA polymerase; Invitrogen, Carlsbad, CA, USA), 2 μL of template DNA (sample or gBlock), and DNA/RNAse-free water. The thermocycling conditions involved 40 cycles, and reaction conditions were previously reported in [18]. As an internal control for amplification efficiency, primers for the beta actin gene were used. As a negative control for the reactions, we used RNA extracted from equine washed RBC, plasma, and buffy coat that previously tested negative for arboviruses, equine herpesvirus 1 and 4, and borna disease. The amplicons were analyzed by 1% (*w/v*) agarose gel electrophoresis, stained with ethidium bromide, and visualized under UV light. Nested PCR were performed for equine herpesvirus 1 (EHV-1) [33] for borna disease [34,35], both with negative results in the 3 horses.

### *4.2. cDNA Synthesis and Multiplex Tiling PCR*

Then, WNV-positive (in nested RT-PCR) RNA samples from washed RBCs were submitted to a cDNA synthesis protocol [36] using a Superscript IV cDNA Synthesis Kit. Then, a multiplex PCR primer scheme was designed (Table S1) to generate complete genomes sequences by means of portable nanopore sequencing, using Primal Scheme (Supplementary Table S1) (http://primal.zibraproject.org accessed on 31 December 2019) [37]. The published WNV genome from Brazil (MH643887) was used to generate a mean 98.4% consensus sequences that formed the target for primer design. The thermocycling conditions involved 40 cycles, and reaction conditions were previously reported in [37].

### *4.3. Library Preparation and Nanopore Sequencing*

Amplicons were purified using 1× AMPure XP Beads, and cleaned-up PCR products concentrations were measured using Qubit ™ dsDNA HS Assay Kit on a Qubit 3.0 fluorimeter (Thermo Fisher Scientific, Waltham, MA, USA). DNA library preparation was carried out using the Ligation Sequencing Kit and the Native Barcoding Kit (NBD104, Oxford Nanopore Technologies, Oxford, UK) [37]. Purified PCR products pools were pooled together before barcoding reactions (taking in consideration each amplicon pool DNA concentrations), and one barcode was used per sample in order to maximize the number of samples per flow cell. Sequencing library was loaded onto a R9.4 flow cell, and data were collected for up to 6 h, but generally less.

### *4.4. Generation of Consensus Sequences*

Raw files were basecalled using Guppy and barcode demultiplexing was performed using qcat. Consensus sequences were generated by de novo assembling using Genome Detective (https://www.genomedetective.com/app/ accessed on 31 December 2019) [38]. New genomes were deposited in the GenBank with accession numbers MW420987, MW420988, and MW420989 (Table 1).

### *4.5. West Nile Virus Typing Tool: Classification Method and Implementation*

The classification pipeline we present comprises two components. One for species and sub-species assignment that enables assignment at these levels by BLASTing the query sequences against a set reference sequences [39]. An assignment is made when BLAST reports a result that exceeds the present threshold.

The other component constructs a Neighbor Joining (NJ) phylogenetic tree that is used to make assignments at the lineages and sublineages level. For this component, the query sequence is aligned against a set of reference sequences using the profile alignment option in the ClustalW software [40], such that the query sequence is added to the existing alignment of reference sequences. Following the alignment, a NJ phylogenetic tree with 100 bootstrap replicates is inferred. The tree is constructed using the HKY distance metric with gamma among-site rate variation, as implemented in the PAUP\* software (https: //paup.phylosolutions.com/ accessed on 31 December 2019) [41]. The query sequence is assigned to a particular genotype if it clusters monophyletically with that genotype clade with bootstrap support >70%. If the bootstrap support is <70%, the genotype is reported to be unassigned (Supplementary Figure S1).

For each of these steps, the earlier discussed reference strains were used with respect to the appropriate typing level (i.e., virus species, lineages, and sublineages). Testing revealed that a BLAST cut-off value of 200 allowed accurate identification of the virus species and WNV using sequence segments >200 base pairs. Note that the species classification procedure is implemented as separate BLAST steps. This enables the tool to efficiently perform large throughput species classification, such as for the classification of shorts sequencing reads. An instance of the web application is publically available on a dedicated server (https://www.genomedetective.com/app/typingtool/wnv/ accessed on 31 December 2019). The web interface on this server accepts up to 2000 whole-genome or partial genome sequences at a time.

### *4.6. Phylogenetic Analysis*

The 3 new sequences reported in this study were initially submitted to a genotyping analysis using the new phylogenetic West Nile virus subtyping tool, which is available at https://www.genomedetective.com/app/typingtool/wnv (accessed on 31 December 2019). To put the newly WNV sequences in a global context, we constructed phylogenetic trees to explore the relationship of the sequenced genomes to those of other isolates.

We retrieved 2321 WNV genome sequences with associated lineage date and country of collection from GenBank (Supplementary Figure S2). From this dataset, we generated a subset that included the highly supported (>0.9) clade containing the newly WNV strains obtained in this study plus 29 globally sequences (randomly sampled) from all lineages 1A, 1B, 2, 3, 4, 5, 7, and 8 (Supplementary Table S3). Sequences were aligned using MAFFT [42] and edited using AliView [43]. Those datasets were assessed for the presence of phylogenetic signal by applying the likelihood mapping analysis implemented in the IQ-TREE 1.6.8 software [44]. A maximum likelihood phylogeny was reconstructed using IQ-TREE 1.6.8 software under the HKY+G4 substitution model [44]. We inferred time-scaled trees by using TreeTime [45].

### *4.7. WNV Epidemiological Data*

Human reported cases presenting neurological disease compatible with WNV infection collected between November 2015 and early 2020 were obtained from SINAN. We reinforce the nature of the reports as suspected (not confirmed), being officially defined as cases presenting neurological syndromes compatible with WNV infection, registered as suspected occurences of West Nile virus infection (code A923). As such, the spatiotemporal series of suspected cases should only be interpreted as a proxy for the possible spatio-temporal dynamics of WNV infections [46].

### *4.8. Modeling Transmission Potential*

To estimate the transmission potential of WNV, we employed the computational approach from Lourenço et al. recently applied in Israel [25] and Portugal [26]. This approach estimates the suitability index P using climatic variables only. The index measures the transmission potential of single adult female mosquitoes (spp. Culex) in the animal reservoir and is thus interpreted as a summary measure of the risk for spill-over into human populations. The theory and practice of estimating the index P for mosquito-borne viruses has been previously described in full by Obolski et al. [27]. The epidemiological priors used were the same as in the original study by Lourenço et al. in Israel, which relate to spp. Culex, WNV, and an average bird species. Climatic data were obtained from Copernicus.eu (https://www.copernicus.eu (accessed on 31 December 2019)); in particular, we used the dataset "essential climate variables for assessment of climate variability from 1979 to present" [47]. This dataset offers climatic variables at a time resolution of 1 month and gridded spatial resolution of 0.25 × 0.25.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/pathogens10070896/s1, Figure S1: WNV typing tool, Figure S2: Maximum likelihood phylogenetic tree of 2321 WNV complete genomes. Colors indicates different lineages. Highlighted red clade include the WNV viral strain obtained in this study, Table S1: Primer scheme, Table S2: WNV suspected cases reported between 2014–2020 in each Brazilian state, according to SINAN, Table S3: Globally reference WNV sequences from the subset *n* = 29 used in this study.

**Author Contributions:** Conception and design: É.A.C., M.G., J.L. and L.C.J.A.; Data collection: É.A.C., M.G., L.S.C., V.F., M.A.d.C.e.S.V., D.F.H., D.B.d.A.M., K.R.L.J.C., N.F.O.d.M., A.P.M.R. and L.C.J.A.; Investigations: F.F.A., F.L.L.C., A.M.B.d.F., R.V.d.C., É.A.C., M.G., J.X., V.F., M.I.M.C.G., B.S. Á.d.S.S., A.S.G.S., R.d.P.A.M., N.R.d.C.F., R.F.d.S., R.B.C.T. and J.L.; Data Analysis: M.G., V.F., F.F.A. and J.L.; Writing—Original: É.A.C., M.G., L.S.C., V.F., M.A.d.C.e.S.V., J.L. and L.C.J.A.; Draft Preparation: É.A.C., M.G., L.S.C., V.F., M.A.d.C.e.S.V., J.L. and L.C.J.A.; Revision: É.A.C., M.G., L.S.C., V.F., F.F.A., F.C.d.M.I., M.A.d.C.e.S.V., D.F.H., D.B.d.A.M., M.I.M.C.G., B.S. Á.d.S.S., A.S.G.S., T.d.O., K.R.L.J.C., N.F.O.d.M., A.P.M.R., C.F.C.d.A., L.C.S.F., J.J.M.B., R.B.C.T., O.L.L., S.d.C.S., R.d.P.A.M., R.F.d.S., J.L. and L.C.J.A. Methodology: F.F.A., F.L.L.C. Writing—review & editing: F.F.A., F.L.L.C., A.M.B.d.F., R.V.d.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was founded by CNPq (440685/2016-8, 421598/2018-2), by CAPES (88887.130716/ 2016-00), by the Pan American Health Organization (IOC-007-FEX-19-2-2-30), by the Fundacão Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ, gran<sup>t</sup> number E-26/2002.930/2016 by the European Union's Horizon 2020 Research and Innovation Programme under ZIKAlliance Grant Agreement no. 734548, by the Horizon 2020 through ZikaPlan and ZikAction (grant agreemen<sup>t</sup> numbers 734584 and 734857) and by the National Institutes of Health USA gran<sup>t</sup> U01 AI151698 for the United World Antiviral Research Network (UWARN). MG and LCJA is

supported by Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ). JL is supported by a lectureship from the Department of Zoology, University of Oxford.

**Institutional Review Board Statement:** This project was reviewed and approved by the Comissão Nacional de Ética em Pesquisa (CONEP) [National Research Ethics Committee] from the Brazilian Ministry of Health (BrMoH), as part of the arboviral genomic surveillance efforts within the terms of Resolution 510/2016 of CONEP, by the Pan American Health Organization Ethics Review Committee (PAHOERC) (Ref. No. PAHO-2016-08-0029), by the Animal Welfare Committee of Universidade Federal do Piauí, under n◦065/19 and by the Oswaldo Cruz Foundation Ethics Committee (CAAE: 90249218.6.1001.5248). All experiments were performed in accordance with relevant guidelines and regulations.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Newly generated WNV sequences have been deposited in GenBank under accession numbers MW420987, MW420988 and MW420989.

**Acknowledgments:** The authors thank the important contributions of the Municipal and Piaui State Health Department (SESAPI, FMS), Municipal and Piaui State Animal Health Department (ADAPI), Laboratório de Saúde Pública do Piauí (LACEN-PI), and the colleague Thiago dos Santos Silva. We also thank the sponsoring institutions: Saint Louis Zoo WildCare Institute and Institute for Conservation Medicine (USA), Universidade Federal do Piauí (UFPI), Fundação de Amparo a Pesquisa do Estado do Piauí (FAPEPI). The authors also thank the Municipal and State Health Department of São Paulo and Minas Gerais state.

**Conflicts of Interest:** The authors declare no conflict of interest.
