We trapped sandflies using CDC light traps (John W. Hock, https://johnwhock.com) in villages of the Marigat district, Baringo County, Kenya, in February 2014 (Figure 1). We homogenized pools of 5–50 female specimens in minimum essential medium (MEM), inoculated an aliquot of the clarified supernatant (50 μL) into Vero cells, and incubated it for up to 14 days with daily monitoring for cytopathic effect (CPE). We passaged CPE-positive supernatant onto fresh Vero E6/7 cells; virus stock solution was generated from the first passage and used for all further experiments. We determined the infectious titer by 50% tissue culture infectious dose (TCID₅₀) assay using 10-fold serial dilutions from^– 10^–1 to 10^–12 of the virus stock inoculated in 5 wells each of subconfluent Vero cells seeded in a 96-well plate. We calculated the virus titer according to Reed and Muench (20). We estimated the minimum infection rate (MIR) in sandflies using the formula [number of positive pools/total specimens tested] × 1,000.

We extracted RNA from cytopathic cell culture supernatant (200 μL aliquot) using the Viral RNA Mini Kit (QIAGEN, https://www.qiagen.com) and eluted it in 50 μL of buffer. We performed cDNA synthesis using SuperScript III reverse transcription (Invitrogen, https://www.thermofisher.com) and random hexamer primers (TIB Molbiol, https://www.tib-molbiol.com), followed by reverse transcription PCR (RT-PCR) for orthobunyaviruses, alphaviruses, and flaviviruses (21–24). We screened for sand fly–borne phleboviruses by targeting the RNA-dependent RNA polymerase (RdRp) gene using degenerate primers (25). We compared the nucleotide sequences to GenBank using blastn (https://blast.ncbi.nlm.nih.gov/Blast.cgi) and subjected them to initial phylogenetic analysis in MEGA6 (26). We further extracted RNA from pooled sand flies and tested it by real-time PCR using NPV-specific primers (forward, 5′-GCAAGAAAGCACTGTGGTGG; reverse, 5′-CGTATGATGATCGGCCACCA; probe, 5′-6-FAM-ACAGCCACCTCTGATGATGC-IBFQ).

We amplified the barcode region of the cytochrome c oxidase subunit I (COI) gene using published primers (27). We extracted genomic DNA from individual bloodfed sand fly specimens using the QIAGEN DNeasy Blood and Tissue Kit (QIAGEN). We amplified a 500-bp fragment of the 12S mitochondrial rRNA gene as described (28), sequenced the PCR products, and compared them to GenBank database data. We inferred species-level identification on the basis of ≥98% identity spanning >300 bp, as described by Valinsky et al. (29).

We purified and concentrated virions from the supernatant of infected Vero cells by ultracentrifugation through a 36% sucrose cushion. We extracted viral RNA using the QIAGEN RNeasy Kit according to the manufacturer’s instructions. We generated cDNA using the Maxima H Minus Double-Stranded cDNA Synthesis Kit and random hexamer primers (Thermo Fisher Scientific, https://www.thermofisher.com). We prepared DNA libraries using the Nextera XT DNA Sample Preparation Kit and analyzed them on an Illumina MiSeq instrument with the MiSeq Reagent Kit v3 (Illumina, https://www.illumina.com). We identified viral reads by reference mapping to phleboviruses as well as by BLAST comparisons against a local amino acid sequence library containing translations of open reading frames (ORFs) of phleboviruses. We closed sequence gaps by conventional RT-PCR followed by Sanger sequencing. We performed genome assembly using Geneious (http://www.geneious.com) and confirmed genome terminal sequences by rapid amplification of cDNA ends (RACE-PCR; Life Technologies, https://www.thermofisher.com). We identified ORFs using Geneious, compared nucleotide and amino acid sequences with other sequences by blastn and blastx searches against the GenBank database, and identified protein motifs by web-based comparison to the Pfam database (http://www.pfam.janelia.org). We identified putative transmembrane regions by prediction of the hydropathy profile using TMHMM (http://www.cbs.dtu.dk/services/TMHMM-2.0) and predicted N-linked glycosylation sites using the NetNGlyc 1.0 server (http://www.cbs.dtu.dk/services/NetNGlyc).

We aligned nucleotide and amino acid sequences of the ORFs of the respective genome segments with related viral sequences in Geneious using MAFFT (30). Phylogenetic trees were inferred by the maximum-likelihood (ML) method using the best suitable substitution matrix (LG) identified by Modeltest, as implemented in MEGA. We performed confidence testing based on 1,000 bootstrap iterations (31).

We infected cell lines from insects (LL-5, sand fly; C6/36, mosquito), humans (HEK293-T), small mammals (BHK-21, hamster; VeroE6/7, primate; MEF, mouse; EidNi, bat) and livestock (PK-15, swine; ZN-R, goat; DF-1, chicken; KN-R, cattle) in doublets, at a multiplicity of infection of 0.1. We harvested aliquots of infectious cell culture supernatants every 24 h for periods of 7 d and quantified viral genome copies by real-time RT-PCR with plasmid-based quantification standards.

We intracerebrally inoculated 100 μL of the viral stock of the first passage, as well as 3 consecutive 2-fold dilutions, into 3–4-day-old Swiss Albino suckling mice. The doses used in the experimental infection were quantified by plaque assays in Vero cells as described previously (32) and corresponded to viral titers of 4 × 10⁶, 2 × 10⁶, 1 × 10⁶, 5 × 10⁵, and 2.5 × 10⁵ PFU/mL. We included noninfectious MEM as a negative control. We observed all mice 2 times/day for up to 14 days for signs of disease. We homogenized brains from recently dead mice in 1 mL of cell culture media and plaque-titrated them on Vero cells.

Archived serum samples from the Marigat district hospital, taken during 2010–2011, and from Sangailu Health Centre in the Hulugho subcounty in northeastern Kenya, collected during 2010–2012, were available (Figure 1). We performed a virus neutralization test using 2-fold serial dilutions of serum samples (1:20 to 1:640). We mixed 50 μL of the serial serum dilutions with 70 TCID₅₀ of NPV. Mixtures were incubated at 37°C in the presence of 5% CO₂ for 1 h, then used for infection of a confluent Vero E6/7 cell monolayer seeded in 96-well culture plates with 2 wells/dilution. After 7 days of incubation, we recorded the highest serum dilution at which no CPE was observed in at least 50% of the wells as the neutralization titer.

We tested NPV-reactive human serum samples with RVFV, Gabek Forest virus (GFV), and Karimabad virus (KARV) for serologic cross-reactivity, as described earlier in this article. In addition, we tested GFV- and KARV-positive serum samples with NPV in 2-fold serial dilutions from 1:5 to 1:20 (Table).

Approval for the study was granted by the Scientific and Ethical Review Unit and Animal Care and Use Committee of the Kenya Medical Research Institute (SSC Protocol nos. 1560 and KEMRI/SERU/CVR/003/3312). All animal experiments were carried out in accordance with the regulations and guidelines of the Kenya Medical Research Institute.

The NPV genome was deposited in GenBank under accession nos. MF695810–MF695812. The COI sequence obtained from the virus-positive sand fly pool was deposited in GenBank under accession no. MG913288.

In total, 6,434 sand flies were trapped (Figure 1). A subset of 5,481 sandflies was pooled and the resulting 111 pools individually inoculated in VeroE6/7 cells. One pool consisting of 8 females induced CPE 4–5 days postinfection. Sequence analysis of the COI gene of the sand flies of this CPE-positive pool suggested that sand flies were of the genus Sergentomyia. We identified blood-meal hosts for 62 blood-fed specimens sampled at the same place and time as the pooled specimens. Results revealed that 56 (90.3%) had fed on humans, 2 (3.2%) on snakes, and 1 (1.6%) each on a frog, lizard, cow, and ostrich. The infectious cell culture supernatant tested negative for RVFV, orthobunyaviruses, alphaviruses, and members of genus Flavivirus. We amplified a 0.5-kb fragment of the RdRp gene of sand fly–borne phleboviruses using degenerated primers (25). The sequence showed the highest pairwise identity of 79% to GFV and 75% to KARV. We sampled a subset of 953 individual sand fly samples 2 years after the initial study and tested it in pools of 10 by specific RT-PCR for the cultured virus; results were negative.

Analysis of the complete genome by next-generation sequencing confirmed isolation of a novel phlebovirus. The virus was tentatively termed Ntepes virus, after the location where the sand flies were collected. The virus exhibits the characteristic tripartite-segmented genome organization of phleboviruses, comprising the large (L) segment, which encodes the RdRp protein; the medium (M) segment, encoding a glycoprotein precursor protein (GPC) that is posttranslationally cleaved into 2 viral surface glycoproteins (Gn and Gc) and a nonstructural protein (NSm); and the small (S) segment, encoding the nucleocapsid (N) protein and a nonstructural protein (NS) in an ambisense manner (Figure 2). Highest sequence similarities to GFV were 93% to RdRp, 88% to GPC, 79% to Nsm, 85% to N, and 90% to NS. NPV has the typical conserved genome termini shared among phleboviruses (5′-ACACAAAG and CUUUGUGU-3′) (8).

Phylogenetic analyses of NPV RdRp, Gn, Gc, and N proteins and all available sand fly–borne phlebovirus sequences indicate that NPV forms a strongly supported clade with GFV and KARV. NPV branches as a sister taxon to GFV in all genes, suggesting NPV to be a member of the Karimabad species complex (Figure 3). However, the designation of the Karimabad species complex is not yet officially approved by the ICTV. For a provisional genetic classification, we analyzed the intragenetic distances among established phlebovirus species and unclassified isolates based on the RdRp gene. Pairwise nucleotide and amino acid distances between established species ranged from 38% to 62% for nucleotide distances and 39% to 68% for amino acid distances (Appendix Figure). For example, amino acid distance between Punta Toro virus and Candiru virus was 39% and between SFTS virus and sandfly fever Naples virus was 68%. Pairwise nucleotide distances ranged from 20% to 59% and amino acid distances from 6% to 69% when unclassified tentative species and variants pertaining to established species were included. For example, amino acid distance between Ponticelli virus and Adana virus was 6% and between Naples virus and SFTS virus was 69% (Appendix Figure). NPV showed 7% amino acid distance to GFV and 19% amino acid distance to KARV.

Classical criteria for species demarcation in phleboviruses are based on serology, with established species showing at least 4-fold differences in 2-way neutralization tests (11). We confirmed that NPV did not react with antiserum against its next closest relatives, GFV and KARV, in neutralization tests (Figure 4). NPV Gn protein was 13% different and Gc 4% different from GFV. The Gn protein of phleboviruses is the key component for neutralization and is recognized by specific neutralizing antibodies (33).

Although sequence-based species demarcation criteria have not been determined for phlebovirus species, such criteria exist for the related orthobunyaviruses. Species demarcation criteria are now based on the RdRp gene, which shows >6% difference to the closest related virus. Previously unique orthobunyavirus species were defined on >10% difference in N protein sequences (11). The N proteins of NPV and GFV differ by 15% (GFV itself is not formally classified as a species, and any of the formally classified phlebovirus species are markedly more distant from NPV in this and other genes; Appendix Figure). We conclude, upon cumulative evidence, that NPV constitutes a putative novel species within the phlebovirus genus.

To obtain initial data on permissiveness, we performed in vitro growth analyses in a broad range of cell lines derived from different insect species (sand fly and mosquito), peridomestic wildlife (rodent, nonhuman primate, and bat), and livestock (swine, goat, chicken, and cattle) species, as well as from humans. Results revealed a broad susceptibility to NPV, with peak genome copy numbers in cells derived from swine and rodents (Figure 5, panel A). Cells derived from sand flies but not from mosquitoes were permissive, despite using C6/36 mosquito cells that are normally broadly susceptible to arboviruses because of a defect in their antiviral RNA interference response (34). These findings suggest a host range for NPV similar to those of KARV and GFV, which are transmitted by sand flies and infect rodents (35). It is not known whether rodents are amplificatory or dead-end hosts.

Because GFV is known to induce fatal disease in laboratory mice (36,37), we explored similarities in pathogenicity with NPV. We intracranially inoculated 3–4-day-old Swiss Albino suckling mice, causing tremors, hind-limb paralysis, prostration, and death 5–8 days postinfection (Figure 5, panel B). Time to death was clearly correlated with virus dose. All animals had high infectious virus concentrations in the brain (mean 2.9 × 10⁶ PFU/mL). Taken together, the in vitro and in vivo pathogenicity studies of NPV, including the pathogenicity in suckling mice, may suggest that rodents and sand flies may be involved in the maintenance cycle of NPV.

To test whether NPV infects humans, we analyzed 187 archived serum samples: 59 samples from the Marigat district hospital in the area where NPV-infected sand flies were trapped, and 128 samples collected in northeastern Kenya at Sangailu Health Centre (Figure 1). All patients from Marigat, as well as 98 patients from Sangailu, had symptoms compatible with acute infectious diseases. The remaining 30 samples from Sangailu came from healthy controls.

Twenty-six (13.9%) serum samples neutralized NPV, with titers ranging from 1:20 to 1:320 (Figure 4). Women and men were infected at equal rates. Positive samples originated from Marigat (10.2%) and Sangailu (15.6%), without statistical differences in rates (Fisher exact test odds ratio [OR] 0.6, 95% CI 0.19–1.70; p = 0.37). Detection rates in Sangailu did not differ between healthy and febrile patients (Fisher exact test OR 1.4, 95% CI 0.40–4.3; p = 0.58) (Table; Figure 4). No NPV nucleic acids were detected in serum samples by NPV-specific RT-PCR, suggesting no causative link to the present symptoms with NPV. The detection of NPV neutralizing antibodies in geographically unlinked regions of Kenya suggests widespread previous human exposure and infection.

Because NPV is genetically most closely related to GFV and KARV, we tested all NPV-positive serum samples for ability to cross-neutralize GFV or KARV. All tests yielded negative results, providing further support for the classification of NPV as a separate serotype (and species). Because RVFV frequently causes outbreaks in East Africa, we also tested against RVFV, which, according to its phylogenetic relationship with NPV, is not expected to cross-react with NPV. Seven of 26 NPV-neutralizing serum samples were also reactive with RVFV, showing titers that did not correlate in height with titers against NPV (Table; Figure 4). Absence of correlation of titers suggests previous RVFV infection rather than cross-reactivity between RVFV and NPV.