Human, Nonhuman Primate, and Bat Cells Are Broadly Susceptible to Tibrovirus Particle Cell Entry

Introduction RNA interference (RNAi) is a process by which intracellular long double stranded RNA (dsRNA) is cleaved into small RNA (sRNA) effector molecules that direct the silencing of complementary RNA sequences. Multiple pathways, including exo- and endo-small interfering RNA (siRNA), microRNA (miRNA) and PIWI-interacting RNA (piRNA), contribute to this process. The exo-siRNA pathway, in which silencing is triggered by exogenously derived dsRNA molecules, is thought to comprise the main antiviral response in mosquitoes [1]–[7]. This pathway is initiated when Dicer 2 (Dcr2) binds to and cleaves long dsRNA molecules that exist within cells as viral replicative intermediates and/or RNA secondary structures into 20–22 nucleotide siRNAs [8]–[11]. The resulting siRNAs are loaded by Dcr2 and R2D2 into the multi-protein RNA-induced silencing complex (RISC), which includes Argonaut 2 (Ago2), and unwound, after which the 3′ terminus of the retained guide strand is 2′-O-methylated [12], [13]. The siRNA-loaded RISC identifies single stranded RNAs complementary to the guide strand, which are cleaved by the endoribonucleolytic activity of Ago2 [14], [15]. The advent of high-throughput sequencing technologies allows researchers to characterize viral-derived siRNAs (viRNAs) and to quantitatively map the areas of the viral genome most often targeted [1], [5], [6], [16], [17]. Arthropod-borne viruses (arboviruses) are a diverse group of viruses maintained in nature by horizontal transmission between hematophagous, arthropod vectors and vertebrates. Arbovirus infections can cause an acute, pathogenic outcome in the vertebrate host, but establish a persistent, relatively non-pathogenic infection in the invertebrate vector, with some noted exceptions [18]–[20]. The reasons for this difference are not fully understood, but may be related to the highly inflammatory innate antiviral immune response in vertebrates (type-I interferon-mediated) compared to invertebrates, which rely on antiviral mechanisms such as RNAi and the Toll, JAK/STAT, and Imd/Jnk signaling pathways [7], [18], [21]–[23]. Of these, RNAi appears to be the primary means of limiting viral infections in the vector [2]–[4], [7], [16], [24]–[27]. High-throughput sequencing has identified viRNAs from Aedes aegypti and Culex pipiens quinquefasciatus mosquitoes infected with Sindbis virus (SINV, Togaviridae: Alphavirus) and West Nile virus (WNV, Flaviviridae: Flavivirus), respectively, and Drosophila infected with Flock House virus (FHV, Nodaviridae: Alphanodavirus) [1], [6], [16]. In each case, the viRNAs were predominantly 21 nt in length, asymmetrically distributed across the length of the genome and derived from both the positive and negative RNA strands. Aedes albopictus C6/36 and Drosophila melanogaster S2 cells are commonly used, immortalized, invertebrate cell lines that have become powerful and convenient tools for studying many host/virus interactions at the molecular level. Originally established from mosquito larvae homogenates, C6/36 cells are easy to maintain and highly permissive to numerous arboviruses. Likewise, S2 cells are easy to maintain and manipulate and critical specific reagents, such as antibodies and specific knockouts, are available commercially. However, immortalized cells may not accurately model the natural environment encountered by viruses in the whole organism. For example, Vero (African green monkey kidney) cells, which are used to study many human viruses because of their inherent permissiveness, lack a functional type-I interferon response [28]. A recent report on WNV infection in insect cell culture models raised the possibility that C6/36 cells may be similarly deficient in key components of their antiviral defense mechanisms, leading to their permissiveness for arbovirus infection [29]. Therefore, we investigated the RNAi response in C6/36 cells compared to S2 cells. Specifically, C6/36 and S2 cells were infected with representatives of three diverse arbovirus families and small RNA populations of infected cells were characterized by deep sequencing. Materials and Methods Viruses The viruses used in these studies are representative of each of three major arbovirus families that include many human pathogens; Flaviviridae, Togaviridae, and Bunyaviridae. WNV (Flaviviridae; Flavivirus) was generated from an infectious cDNA clone derived from the NY99 strain [30]. SINV (Togaviridae, Alphavirus) was generated from the SINV TE3′2J infectious clone [31]. The LACV (Bunyaviridae; Orthobunyavirus) used in these studies was the LACV/Human/1960 strain (GenBank accession nos. EF485032.1, EF485031.1, and EF485030.1). Originally, isolated from the brain of a LaCrosse encephalitis patient in 1965 the virus was subsequently passaged in suckling mice three times and baby hamster kidney cells (BHK-21) an additional six times. Stock virus was prepared in BHK-21 cells [32]. Cell Culture, Virus Infections and RNA Extractions The C6/36 Aedes albopictus cells were grown in Dulbecco's modified essential medium (DMEM) with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 100 µg/ml streptomycin, L-glutamine, and sodium bicarbonate at 28°C with CO2. The S2 Drosophila melanogaster cells were grown in Schneider's Drosophila medium with 10% FBS, penicillin/streptomycin, and L-glutamine at 28°C without CO2. C6/36 and S2 cell cultures were infected in triplicate with each of the three viruses at a multiplicity of infection (MOI) of 0.1 and 1, respectively. Considering that Drosophila is not a natural host for any of these viruses, higher MOIs were required to ensure infection. Virus stocks were diluted in maintenance medium with a FBS concentration of 2%, inoculated onto confluent cell monolayers, and allowed to adsorb for one hour at room temperature. The virus inocula were removed and replaced with maintenance medium. Five days post infection, WNV and SINV infected C6/36 cells were harvested. LACV infection of C6/36 cells and all infections in S2 cells were maintained an additional two days before harvesting. Cells were pelleted by centrifugation and resuspended in mirVana lysis buffer. RNA from each sample was extracted using the mirVana miRNA Isolation Kit (Ambion, Austin, TX) according to the manufacturer's instructions. RNA quantity and integrity was determined on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Preparation of Small RNA Libraries and High-Throughput Sequencing Equal amounts of RNA from the three replicates were pooled and ethanol precipitated. Approximately 10 µg of total RNA from each experimental group was size fractionated on a TBE/urea 15% polyacrylamide gel and small RNA (sRNA) populations (17–30 nt) recovered. 5′ and 3′ sequencing adapters (5′ adapter: 5′-GUUCAGAGUUCUACAGUCCGACGAUC-3′, 3′ adapter: 5′-P-UCGUAUGCCGUCUUCUGCUUGU-3′) (Oligonucleotide sequences © 2007-2009 Illumina, Inc. All rights reserved) were then ligated to sRNAs using T4 RNA ligase, of which the 3′ adapter was not pre-adenylated. The sRNAs were reverse transcribed and PCR amplified according to the manufacturer's instructions (Illumina, San Diego, CA). The resulting libraries were sequenced at the National Center for Genome Resources (Santa Fe, NM) using an Illumina Cluster Station and Genome Analyzer II or IIx. Assembly and Analysis of sRNA Libraries Reads from sRNA libraries were trimmed of adapter sequences and aligned to genome sequences of either the WNV NY99 infectious clone, the SINV TE3′2J infectious clone or the LACV/Human/1960 strain using the Short Oligonucleotide Alignment Package v.1 (SOAP) (http://soap.genomics.org.cn/). A seed size of eight and a maximum of two mismatches were permitted. Gaps and further trimming were not allowed. Quality scores represent an average of the confidence in each sequenced nucleotide in each sRNA. These are based on the Illumina scoring system where 1 is a minimum and 40 is the maximum. Additional analyses were performed using Microsoft Excel and GraphPad. Results Characterization of Small RNA Profiles To assess the antiviral RNAi response in two model invertebrate cell culture systems, we sequenced small RNAs (sRNAs) from C6/36 and S2 cells infected with three arboviruses, WNV, SINV and LACV (Bunyaviridae: Orthobunyavirus). These viruses were chosen as representative members of three major arbovirus families with both positive- (WNV, SINV) and negative-polarity (LACV) RNA genomes. The total number of sRNA reads obtained varied greatly between samples and most likely reflects differences in cell density at the experimental endpoint, since the majority of the sRNAs originate from the host in the form of miRNAs. In WNV infected cells, there were 3.4 and 9.9 million reads from C6/36 and S2 cells, respectively (Table 1). Of these, only 12,539 reads (0.37%; 6,151 unique) in C6/36 cells and 4,431 (0.045%; 3,127 unique) in S2 cells aligned to the WNV genome, where unique reads represent individual viRNAs that code for a specific nucleotide sequence. In both cell lines, greater than 87% of the reads perfectly aligned with the WNV genome. Analysis of the viRNA sizes revealed that over 76% of the viRNAs isolated from S2 cells were 20–22-mers (mean length 21.6 nt) with 54% constituting 21-mers. In contrast, 57% of the viRNAs from C6/36 cells were 20–22-mers (mean length 21.3 nt) and 17.2% were 21-mers (Figure 1A) with a high proportion of the WNV viRNAs being 17–18 mers (data not shown). 10.1371/journal.pntd.0000856.g001 Figure 1 Size and abundance of small RNA reads Mmapping to the viral genomes. The abundance of 19–30-mer sRNA reads mapping to the WNV (A), SINV (B) and LACV (C) genomes based on size. Abundance is represented as a percentage of the total viRNAs from each sample. The black bars correspond with samples collected from S2 cells and white bars from C6/36 cells. 10.1371/journal.pntd.0000856.t001 Table 1 Small RNA profiles from C6/36 and S2 infected cells. Reads Aligning To Viral Genome Reads With Mismatches n (%) Small RNA Library Total # of Reads (×106) Total # (unique #) Average Length Average Quality Score 0 1 2 WNV C6/36 3.4 12,539 (6,151) 21.3 39.8 11,399 (91) 964 (8) 176 (1) WNV S2 9.9 4,431 (3,127) 21.6 29.4 3,912 (88) 363 (8) 156 (4) SINV C6/36 14.4 1.6×106 (1.2×105) 25.2 30.8 1.5×106 (92) 1.1×105 (7) 19,411 (1) SINV S2 7.3 3.5×105 (46,103) 21.0 31.4 3.1×105 (91) 27,070 (8) 4,202 (1) LACV C6/36 11.2 9.4×105 (73,098) 25.2 31.2 8.7×105 (92) 65,265 (7) 9,672 (1) LACV S2 8.6 6,777 (2,201) 20.4 29.2 5,899 (87) 683 (10) 195 (3) Analysis of the sRNA populations in SINV infected cells revealed that there were 14.4 and 7.3 million reads in C6/36 and S2 cells, respectively (Table 1). Of these, 1.6×106 reads (11.11%; 1.2×105 unique) in C6/36 and 3.5×105 reads (4.79%; 46,103 unique) in S2 cells aligned to the SINV genome, of which greater than 90% perfectly aligned. As with the WNV infected cells, greater than 93% of the viRNAs isolated from S2 were 20–22-mers (mean length 21 nt) and 76% were 21-mers (Figure 1B). In C6/36 cells, only 9% of SINV viRNAs were 20-22 nt in length. Mean length was 25.2 nt, with the majority of the viRNAs (76%) being between 24–27 nt. Finally, there were 11.2 and 8.6 million reads in LACV infected C6/36 and S2 cells, respectively (Table 1). In LACV infected C6/36 cells, 9.4×105 reads (8.39%; 73,098 unique) aligned with the LACV genome, with greater than 90% of the reads having zero mismatches. In contrast to the C6/36 infected cells, only 6,777 reads (0.078%; 2,201 unique) from S2 cells aligned with the LACV genome; however, 87% of these reads matched perfectly to LACV RNA. Consistent with the results from WNV and SINV, the majority of the LACV viRNAs in S2 cells were 20–22-mers (80%), with 43% of these reads being 21 nt in length (mean length 20.4 nt) (Figure 1C). In C6/36 cells, only 9% of the total viRNAs were 20–22-mers. Mean length of LACV viRNAs from C6/36 cells was 25.2 nt, with the majority (79%) being between 24–28 nt in length. Distribution and Abundance of viRNAs To more closely examine the viRNA populations, viRNAs from each experimental sample were aligned to the input viral genome and viRNA coverage intensity determined per nucleotide across the length of the genome. Included in these analyses are all viRNA reads 19–30 nucleotides in length. Over 90% of the WNV genome was targeted by at least one viRNA in both C6/36 and S2 cells. Inspection of the intensity of viRNA coverage of the WNV genome revealed significant positional and regional differences between C6/36 and S2 cells (Figure 2). viRNAs isolated from infected C6/36 cells were asymmetrically distributed across the length of the genome and were derived almost exclusively from the positive sense viral genomic RNA strand (vRNA) (99.9%) (Figure 2A). The most highly targeted site was genome position 206 within the capsid coding sequence with 776 reads covering this position. Expansion of the data set to include 17–18-mers revealed that the first 17 nucleotides of the complementary, negative RNA strand were the most highly targeted (1,019 reads mapping to these sites) within the genome (data not shown). Alignment of viRNAs obtained from WNV infected S2 cells revealed that they were asymmetrically distributed and were derived from both the positive sense vRNA (84%) and complementary cRNA (16%) (Figure 2B). Nucleotide 125 located in the capsid coding sequence was the most targeted site within the genome (99 reads mapping to this site).These results are strikingly similar to those generated from WNV infected Cx. p. quinquefasciatus midguts [1]. 10.1371/journal.pntd.0000856.g002 Figure 2 viRNA coverage of the WNV genome in C6/36 and S2 cells. Complete genome of WNV (11,029 nt.) showing intensity at each nucleotide of the genome in C6/36 (A) and S2 (B) cells. Plotted are the 19–30-mer viRNA reads. Reads originating from the genomic, positive strand are represented in blue above the x-axis and those originating from the negative strand are represented in red below the x-axis. The viral genome coverage, viRNA frequency and distribution for the SINV infected C6/36 and S2 cells were very similar to those observed for WNV infected cells. In both cell lines 100% of the viral genome was targeted by at least one viRNA. However, the percent coverage in the C6/36 cells is deceiving as the majority of viRNAs were directed at only a few “hotspots”. The viRNAs from C6/36 cells were unevenly distributed across the genome with both positive sense vRNA (71%) and negative sense cRNA (29%) targeting (Figure 3A). The region just 3′ of the SINV subgenomic promoter was the most highly targeted with 201,104 reads mapping to nucleotide 8,013, located in the viral capsid coding region. The increased targeting of the subgenomic transcripts may reflect the overall abundance of these transcripts in comparison to the full length genome. This predilection for targeting the subgenomic transcript was not observed in the SINV infected S2 cells (Figure 3B) or SINV infected Ae. aegypti [6]. Further similarities between the S2 cell viRNA profile and that from Ae. aegypti included the asymmetry of distribution across the genome and the proportion of viRNAs derived from the positive sense vRNA (52% in S2 cells and 54% in Ae. aegypti) [6]. The most highly targeted site within the viral genome was nucleotide 4,601 (4,803 reads) in the viral nsp3 gene. 10.1371/journal.pntd.0000856.g003 Figure 3 viRNA coverage of the TE3′2J SINV genome in C6/36 and S2 cells. Complete genome of TE3′2J SINV (11,385 nt.) showing intensity at each nucleotide of the genome in C6/36 (A) and S2 (B) cells. Plotted are the 19–30-mer viRNA reads. Reads originating from the genomic, positive strand are represented in blue above the x-axis and those originating from the negative strand are represented in red below the x-axis. The green vertical line represents the location of the subgenomic promoter. Analysis of sRNAs revealed 100% viRNA coverage of the LACV genome in both C6/36 and S2 cells. viRNAs from both cell lines were asymmetrically distributed across the genome and were derived from both the vRNA (negative-sense) and cRNA (positive-sense) strands (Figure 4). The predominance of positive-polarity strand targeting observed for WNV and SINV in C6/36 cells was also observed in both cell types infected with the negative-sense RNA LACV. However, a difference in the proportions of negative- and positive-sense targeting was noted. In S2 cells, 84% of the viRNAs were derived from the cRNA (positive-sense) strand compared to 72.5% in C6/36 cells (Figure 4B, D, F and 4A, C, E, respectively). The intensity of viRNA targeting for each of the three segments of the tri-partite LACV genome (L segment 6,980 nt, M segment 4,526 nt and S segment 984 nt) was determined. Values are presented as the total number of viRNAs per kb per segment. The S segment was the most frequently targeted in both cell types with 4.3×105 and 3,829 viRNA reads per kb in C6/36 and S2 cells, respectively. As expected, the most frequently targeted sites within the genome were located in the S segment. The predominant S segment region targeted in LACV infected S2 cells was located between nt 904-923 at the 3′ end (∼3,500 hits), while nt 482 (89,871 hits) was the most targeted site in C6/36 cells (Figures 4E and 4F). In each cell line there was an approximately 10-fold reduction in the targeting of the M and L segments as compared to the S segment (M = 3.0×104 and L = 5.5×104 in C6/36 cells and M = 263 and L = 261 in S2 cells) (Figure 4A–D). The observed S segment targeting bias most likely reflects the abundance of S segment mRNA in comparison to M and L segment mRNA in bunyavirus-infected cells [33]. 10.1371/journal.pntd.0000856.g004 Figure 4 viRNA coverage of the LACV/Human/1960 strain genome in C6/36 and S2 cells. Complete genome of LACV/Human/1960 strain showing intensity at each nucleotide of the genome in C6/36 (A,C,E) and S2 (B,D,F) cells. A and B correspond with the L gene segment (6,980 nt), C and D the M gene segment (4,526 nt), and E and F to the S gene segment (984 nt). Plotted are the 19–30-mer viRNA reads across the length of each segment represented by the x-axis. Reads originating from the genomic, negative strand are represented in red below the x-axis and those originating from the positive strand are represented in blue above the x-axis. Discussion Cell culture systems have become invaluable tools in the study of host-virus interactions. However, they may not faithfully model certain molecular features of the host organism-virus interaction. As a result, interpreting data generated in these systems can have limitations and clearly defining the limitations of such systems is crucial to glean as much accurate information as possible from these studies. C6/36 Aedes albopictus cells are a mosquito cell line commonly used to study arbovirus-vector interactions [34]–[36]. Recently, it was demonstrated that WNV-specific siRNAs could not be detected by northern blot hybridization following infection of C6/36 cells, although siRNAs were found in Drosophila S2 cells [29]. The authors concluded that WNV actively evaded the antiviral RNAi response either through the activity of a, as yet unidentified, WNV encoded viral suppressor of RNAi or by sequestration of viral replicative complexes within protective membranous vesicles [37], [38]. However, it remains unclear why these observations were limited to C6/36, but not S2 cells. Therefore, we tested the hypothesis that C6/36 cells lack a fully functional antiviral RNAi response. To this end, C6/36 and S2 cells were infected with members of three taxonomically diverse arbovirus families and the RNAi response characterized by high-throughput sequencing of viRNAs. Examination of three unrelated arboviruses allowed us to determine if the observed results were specific to a particular virus and its associated strategies for evading host defense, or a general defect in the cells themselves. We prepared six sRNA libraries from WNV, SINV and LACV infected C6/36 and S2 cells. In general, the total number of sequence reads and the average base call quality scores were comparable, with the exception of the WNV infected C6/36 cell library, which had fewer total reads and higher quality scores (Table 1). This observed difference may be attributable to the library preparation or sequencing efficiency, as the WNV infected C6/36 cell library was prepared and sequenced independently of the other samples. Nevertheless, the results (average read lengths and sRNA reads perfectly aligning to the viral genomes) were consistent among samples and with previous studies and therefore provide confidence that informative comparisons can be made between libraries. Inspection of sRNA sequencing results from infected C6/36 and S2 cells revealed obvious differences between the proportions of total reads mapping to the viral genomes. Of the sRNAs from WNV-infected C6/36 cells, 0.37% of all reads matched the WNV genome, whereas only 0.045% from infected S2 cells were WNV-specific. Likewise, the LACV infected S2 cells (0.078% virus-specific) had a proportion ≥100-fold lower than observed in infected C6/36 cells (8.39%). In contrast to the WNV-infected samples, the proportions of sRNAs matching viral genome RNA were much higher in SINV-infected cells and the difference between cell types was considerably less (11% in C6/36 cells and 4.8% in S2 cells). These findings are consistent with the 13.9% of sRNAs matching the SINV genome observed in Ae. aegypti four days post SINV inoculation [6]. These results may reflect differences in the replication kinetics of the viruses in each of the two cell lines. For instance, WNV infectious titers in S2 cells seven days post infection are usually 2–3 logs lower than titers in C6/36 cells five days post infection (data not shown) [29]. A relatively small proportion of small RNAs in flavivirus-infected cultured mosquito cells and mosquitoes are virus-specific as compared to mosquito infections by members of other arbovirus families. Although this has not previously been shown for bunyaviruses, it was independently determined in previous studies of flaviviruses [1], [7] (Scott et. al., submitted) and alphaviruses [5], [6] and is confirmed by our current results. This could reflect a more effective mechanism of evasion of innate immunity by flaviviruses, such as sequestration of the viral replication complex in membrane-enclosed vesicles in mosquito cells as well as mammalian cells [38], [39]. Analysis of sRNA reads aligning to the various viral genomes revealed obvious differences that may be related to their biogenesis. The average length of viRNAs mapping to WNV, SINV and LACV genomes from infected S2 cells was approximately 21 nt with the majority being 20-22-mers (Figure 1). These observations are indicative of Dcr2 processing of viral RNA and are consistent with previous analyses of viRNAs from WNV infected Cx. p. quinquefasciatus midguts, SINV infected Ae. aegypti and O'nyong-nyong infected Anopheles gambiae [1], [5], [6]. viRNA populations generated in C6/36 cells were markedly different. The average length of sRNA reads mapping to the SINV and LACV genomes was 25.2 nt with a comparatively small proportion composed of 20–22-mers. These results are consistent with those from two Flaviviruses, dengue virus (DENV) and cell fusing agent virus (CFAV) [40]. The abundant of 24-28 nt long viRNAs may represent products of the piRNA pathway. piRNAs are typically derived from a positive-polarity strand in a Dcr1 and Dcr2 independent manner and are thought to control the development of reproductive tissues and the transcription of transposons [41]-[43]. Recently, what appear to be virus-derived piRNAs have been identified in Drosophila ovary somatic sheet cells [44]. While SINV and LACV sRNAs were not limited to a single strand, the size distributions suggest their biogenesis may have occurred through the piRNA pathway. Interestingly, the size distribution of viRNAs from WNV infected C6/36 cells was quite different from the other viruses (SINV and LACV) (Figure 1) and from DENV and CFAV [40], lacking the characteristic peak at ∼27 nt and casting doubt on the role of the piRNA pathway in their biogenesis. However, the WNV viRNAs were almost exclusively derived from the positive strand, a characteristic of piRNA biogenesis (Figure 2A). The reasons for this apparent paradox are not clear, but may be due to either technical problems or biological mechanisms. A technical explanation seems unlikely because all libraries were processed under the same conditions using identical protocols. Future experiments are required to fully examine this observation. Pre-adenylated 3′ adapters, which would have minimized our sampling of small RNA degradation products, were not utilized in these experiments [17]. Nevertheless, the paucity of 20–22 nt Dcr2-like sRNAs we observed strongly suggests that the majority of the sRNA reads mapping to the viral genomes from C6/36 cells were derived either from the piRNA pathway or cellular degradation pathways. Examination of the polarity (positive sense vs. negative sense) of the sRNA reads mapping to the viral genomes further highlighted the differences between C6/36 and S2 cells. Whereas 16% of the likely Dcr2 generated viRNAs in WNV infected S2 cells were derived from negative-sense strand, almost no viRNAs from C6/36 cells originated from the negative-sense strand (0.1%). This may explain why WNV-derived sRNAs were not detected in C6/36 cells by Chotkowski et. al. as the northern blot probes used in that study were homologous to the NS1 positive sense-strand [29], and would have been unlikely to detect the extremely small proportion of negative sense sRNAs from C6/36 cells mapping to this region. Similarly, there was a 20% excess of positive-sense (vRNA) targeting in SINV infected C6/36 cells compared to S2 cells. The ratio of vRNA- to cRNA-derived viRNAs in infected S2 cells more closely resembled the ratio found in SINV infected Ae. aegypti [6]. Likewise, notable differences were observed in strand polarity ratios between S2 and C6/36 cells infected with LACV. In both cell types, the positive-sense (cRNA) strand was targeted: 72.5% and 84% of the viRNAs in C6/36 and S2 cells, respectively. Although this propensity for positive-sense (cRNA) LACV targeting differed from the observed positive-sense (vRNA) targeting of WNV and SINV, this discrepancy might be due to the marked differences in viral replication and gene expression mechanisms between positive- and negative-sense RNA viruses. The genomes of positive-sense RNA viruses serve as both mRNA and templates for asymmetrical replication through a negative sense RNA intermediate; however, the negative-sense RNA genomes of bunyaviruses serve as templates for both full length complementary RNA replicative intermediates and transcription of highly abundant subgenomic mRNA [45]. Furthermore, since dsRNA is undetectable by staining with a specific antibody in either LACV-infected mammalian cells [46] or mosquito cells (K. Poole-Smith, personal communication) the trigger for initiation of RNAi is unknown. Together, our data for all three viruses suggest that both replicative intermediates containing both genome-sense and anti-sense RNAs, and intra-strand secondary structures within mRNA are targeted by RNAi or other cellular nucleases. The mRNA strand bias observed in all the samples most likely reflects the proportionate abundance of mRNAs as well as use of an alternative small RNA processing pathway in C6/36 cells [40]. For both WNV and SINV in S2 cells, our findings are consistent with small RNA processing by the exogenous siRNA pathway as seen in infected mosquitoes [1], [6]. On the other hand, the increased positive-sense RNA targeting in C6/36 cells suggests that an alternate mechanism may be acting upon viral RNAs. A deep sequencing sRNA dataset was generated from WNV infected DF-1 chicken cells. Upon comparison of the WNV derived sRNAs to the C6/36 cell WNV viRNAs, it was determined that the intensity of viRNA targeting of each nucleotide of the genome was significantly correlated (Spearman r = 0.8882; p<0.0001) (data not shown). However, no correlation was observed between the S2 samples and the C6/36 (p = 0.3126) or DF-1 (p = 0.8467) samples. Since the role of RNAi in vertebrate cellular innate immunity is currently unclear and mRNA turnover pathways are conserved among metazoans, including C6/36 cells, we propose that the WNV derived sRNA populations from C6/36 cells are most likely degradation products or virus-derived piRNAs [17], [44], [47]–[49]. Further, in SINV infected C6/36 cells, the region 3′ to the subgenomic promoter on the positive-sense vRNA was intensely targeted. Were this region highly susceptible to RNAi targeting, then a similar topography would have been observed in S2 cells and Ae. aegypti, but this was not the case (Figure 3) [6]. A more likely explanation is that the highly abundant subgenomic transcripts were not targeted by RNAi, but rather by RNA degradation pathways or the piRNA pathway [44]. Together these results suggest that arboviruses are targeted by the antiviral exogenous siRNA pathway in Drosophila S2 cells, but not C6/36 mosquito cells. The results presented in this study demonstrate that in C6/36 cells, the absence of typical siRNAs, the hallmark of RNAi mediated antiviral immunity, is not limited to WNV and is evident in infections by other diverse arboviruses, such as SINV and LACV. There are multiple steps within the antiviral RNAi response that may be responsible for the observed dysfunction. However, a recent study suggests that it may be related to lack of Dcr2 activity. Studies with cell-free lysates of C6/36 cells revealed that they are unable to process 500 bp dsRNA into 21-mers; however, complementation of C6/36 cell lysates with recombinant human Dcr restored normal dsRNA processing and the presence of detectable 21-mers [40]. Further, when C6/36 cells were co-transfected with an EGFP expression plasmid and either EGFP siRNA or EGFP long dsRNA only the siRNA was able to suppress EGFP expression (J Scott et. al., submitted). These combined with our results suggest that the observed dysfunction is indeed related to lack of dicing activity and more precisely Dcr2 itself. These findings suggest that C6/36 cells may fail to accurately model important aspects of mosquito-arbovirus interactions.

Introduction Viral hemorrhagic fever (VHF) encompasses a group of diseases characterized by fever, malaise, bleeding abnormalities, and circulatory shock [1], [2], [3]. Quality research on these infections is hindered by the fact that they are sporadic and often occur in geographically remote and politically unstable regions of the developing world. Most VHF diseases are associated with a short incubation period (2–21 days), abrupt onset, rapid clinical course, and high mortality, placing VHF agents amongst the most virulent human pathogens [4]. All known VHFs are zoonoses, and to date have been attributed to only four families of enveloped, single-stranded RNA viruses – Arenaviridae, Bunyaviridae, Filoviridae and Flaviviridae. Viruses from these families have caused major deadly outbreaks on the African continent (Fig. 1). Lassa fever virus (Arenaviridae) causes an estimated 500,000 cases each year in West Africa [5]. Crimean-Congo hemorrhagic fever (CCHF) and Rift Valley Fever viruses (Bunyaviridae) are associated with outbreaks in West, South and East Africa [6]. Ebola and Marburg viruses (Filoviridae) have caused several sporadic human outbreaks with high mortality (50–90%) in Central Africa, where they have also decimated local great ape populations [7]. Yellow fever and dengue viruses (Flaviviridae) are widely distributed throughout Sub-Saharan Africa where they cause both endemic and sporadic epidemic diseases in human populations [8]. 10.1371/journal.ppat.1002924.g001 Figure 1 Map of Africa showing countries that are affected by viral hemorrhagic fever (VHF) outbreaks. Ebola VHF is pictured in orange, Marburg VHF in green, Crimean-Congo HF in violet, Lujo VHF in pink, and Lassa VHF in blue. Yellow fever and dengue VHF, which exhibit a wide geographic distribution throughout Sub-Saharan Africa, are not shown. Mangala village, located in the Bas-Congo province in DRC, is represented by a red star. Rhabdoviruses are members of the family Rhabdoviridae and order Mononegavirales and are enveloped viruses with single-stranded, negative-sense RNA genomes [9]. Their genomes encode at least five core proteins in the following order: 3′-nucleoprotein (N), phosphoprotein (P), matrix protein (M), glycoprotein (G) and large protein, or RNA-dependent RNA polymerase (L)-5′ (N-P-M-G-L). Rhabdoviruses are currently divided into six genera, with the two genera Ephemerovirus and Vesiculovirus, together with about 130 unclassified viruses, forming the dimarhabdovirus supergroup (“dipteran mammal-associated rhabdovirus”) [10]. Notably, although rhabdoviruses span all continents and exhibit a wide host range, infecting plants, invertebrates, vertebrate animals, and humans, relatively few are known to cause human infections. Rabies virus (RABV) and related viruses from the Lyssavirus genus and Chandipura virus (CHPV) from the Vesiculovirus genus are known to cause acute encephalitis syndromes [11], [12]. Other viruses from the genus Vesiculovirus cause vesicular stomatitis (mucosal ulcers in the mouth) and “flu-like” syndromes in both cattle and humans [13]. Unbiased next-generation or “deep” DNA sequencing is an emerging method for the surveillance and discovery of pathogens in clinical samples [14]. Unlike polymerase chain reaction (PCR), deep sequencing does not rely on the use of target-specific primers. Thus, the technique is particularly useful for the identification of novel pathogens with high sequence divergence that would elude detection by conventional PCR assays. Deep sequencing has been used previously to discover a new hemorrhagic fever-associated arenavirus from southern Africa, Lujo virus [15], as well as a new polyomavirus in human Merkel cell carcinoma [16]. With the depth of sequence data now routinely extending to >100 million reads, de novo genome assembly of novel viruses directly from primary clinical samples is feasible, as demonstrated by assembly of the 2009 pandemic influenza H1N1 virus genome from a single patient's nasal swab without the use of a reference sequence [17]. Here we report the critical role of deep sequencing in the discovery of a novel rhabdovirus associated with a small outbreak of fulminant hemorrhagic fever in the remote village of Mangala, Bas-Congo province, Democratic Republic of Congo (DRC), between May 25 and June 14, 2009. Results Case Reports from an Acute Hemorrhagic Fever Outbreak Patient 1 The first case was a 15-year-old boy who presented to the health center in Mangala village (Boma Bungu Health Zone) on May 25, 2009 with malaise, epistaxis (nose bleeding), conjunctival injection, gingival bleeding, hematemesis (vomiting with blood), and watery diarrhea with blood (Table 1). No fever or respiratory symptoms were noted. Hemorrhagic symptoms initially appeared on May 24, and the patient died 2 days later from sudden circulatory collapse. The patient lived in the Tshela neighborhood of Mangala village and attended the local public school. All close contacts were monitored for 21 days, and none developed any signs of illness. 10.1371/journal.ppat.1002924.t001 Table 1 Demographics of and clinical symptoms developed in the three patients suspected to be infected by Bas-Congo virus (BASV). Patient 1 Patient 2 Patient 3 Sex Male Female Male Age 15 13 32 Village Mangala Mangala Mangala Neighborhood Tshela Tshela Tshela Occupation Schoolboy Schoolgirl Nurse Disease onset May 24 June 4 June 13 Time until death 2 days 3 days survived Fever (T>39°C) No Yes Yes Weakness No No Yes Malaise Yes No No Headache No Yes Yes Abdominal pain No Yes Yes Epistaxis (nose bleeding) Yes Yes Yes Ocular hemorrhage/conjunctival injection (eye bleeding) Yes Yes Yes Oral hemorrhage (mouth bleeding) Yes Yes Yes Hemorrhagic vomiting Yes Yes Yes Hemorrhagic diarrhea Yes Yes Yes Patient 2 The second case was a 13-year-old girl. She attended the same public school as Patient 1 but was in a different class. She also lived in the Tshela neighborhood of Mangala village, about 50 meters from Patient 1′s house. They knew each other but had no known face-to-face contact during the previous weeks. This patient presented to the health center on June 5, 2009 with headache, fever, abdominal pain, epistaxis, conjunctival injection, mouth bleeding, hematemesis, and diarrhea with blood. She was examined by a nurse and received acetaminophen and dipyrone for fever and quinine for possible malaria. Symptoms appeared on June 4, and the patient died suddenly on June 7, three days after onset. None of her close contacts developed symptoms during the 21 days of monitoring after her death. Patient 3 The third case was a male nurse aged 32 years working in the health center visited by Patients 1 and 2. His disease appeared suddenly on June 13, 2009 with epistaxis, ocular and oral hemorrhage, hematemesis, and diarrhea with blood. Two days after the onset of hemorrhagic symptoms, he developed fever, anorexia, headache, fatigue, and abdominal pain. He was transferred to the regional general hospital of Boma (Fig. 1), a city of about 200,000 inhabitants, where a serum sample was obtained on June 15, just prior to treatment with fluid resuscitation, blood transfusion, and empiric antibiotics. Laboratory tests for malaria, tuberculosis, dengue, and bacterial sepsis were negative, and the patient recovered spontaneously a few days later. All persons in Mangala and Boma who had contact with Patient 3 were monitored for 21 days, and none became ill. Patient 3, like the two other patients, lived in the Tshela neighborhood of Mangala village, about 50 meters from Patients 1 and 2. Importantly, patient 3 was directly involved in the care of Patients 1 and 2 when they presented to the health center with hemorrhagic symptoms. No disease outbreaks had been reported in the past in Boma Bungu Health Zone with the exception of a cholera diarrheal outbreak in 2006, and, notably, no cases of hemorrhagic disease had previously been reported. In addition, although DRC is a country endemic for filovirus infection (Fig. 1), no outbreaks of Ebola or Marburg fever have ever been described in Bas-Congo province. No animal die-offs or other unusual events in association with these cases were noted. Initial Sample Collection and Diagnostic Testing A cluster of three human cases of typical acute hemorrhagic fever occurred between May 25 and June 13, 2009 in Mangala village, located in a remote tropical forest region in Central Africa. Cases were characterized by abrupt disease onset, high fever of >39°C when present, overt hemorrhagic symptoms with epistaxis, conjunctival injection, mouth and gastrointestinal bleeding, followed by death within 3 days of symptom onset in two patients (Table 1). The first patient, who died 140 million paired-end Illumina reads. The horizontal lines (red) depict regions of the genome successfully assembled at the end of each cycle. PCR and Sanger sequencing were performed to confirm the assembly and genomic organization of BASV (green lines). The coverage of BASV achieved by deep sequencing was at least 10-fold across nearly the entire genome and included 29,894 reads out of ∼140 million (0.021%) (Fig. 2). The viral load in the patient's serum was 1.09×106 RNA copies/mL by quantitative RT-PCR. The only moderately high titer is consistent with the fact that the sampled patient was a survivor of BASV infection and would thus be anticipated to have relatively lower viral titers in the blood, as also seen for survivors of Ebola virus infection [18]. Cultivation of the patient's serum in Vero, BHK, LLC-MK2 (rhesus monkey kidney), CCL-106 (rabbit kidney) and C6/36 (Aedes albopictus mosquito) cell cultures failed to show cytopathic effect, and serial quantitative BASV RT-PCR assays on primary and passaged cell culture supernatants turned negative. Subsequent electron microscopy of inoculated cell cultures was negative for viral particles. In addition, no illnesses or deaths occurred in suckling mice inoculated intracerebrally with the BASV-positive serum and observed over 14 days. Phylogenetic Analysis of BASV and Comparison with other Rhabdoviruses Phylogenetic trees reveal that BASV belongs to the dimarhabdoviridae supergroup and is distantly related to members of the Tibrogargan group and the Ephemerovirus genus, although it clusters separately from other rhabdoviruses in an independent deeply rooted branch (Figs. 3 and 4; Fig. S1). Comparative analysis of the concatenated BASV proteins with representative dimarhabdoviruses reveals very low overall amino acid pairwise identity of 25.0 to 33.7%, depending on the virus (Fig. 5). Notably, BASV diverges significantly from either of the two main recognized human pathogens among rhabdoviruses, rabies virus or Chandipura virus. 10.1371/journal.ppat.1002924.g003 Figure 3 Phylogenetic analysis of the L proteins of BASV and other viruses in the order Mononegavirales. The host from which each virus was isolated is represented by a specific color. To generate the Mononegavirales (Rhabdoviridae, Filoviridae and Paramyxoviridae) phylogeny trees, all complete sequences of the large (L) protein, or RNA-dependent RNA polymerase (2000–2300 amino acids in length) were downloaded from GenBank. Abbreviations and accession numbers used for the phylogenetic analysis are provided in Methods. 10.1371/journal.ppat.1002924.g004 Figure 4 Phylogenetic analysis of the L proteins of BASV and other rhabdoviruses. The geographic distribution for each virus or group of viruses is indicated with a specific icon, while diseases associated with infection by certain rhabdoviruses are indicated by specific colors. Abbreviations and accession numbers used for the phylogenetic analysis are provided in Methods. 10.1371/journal.ppat.1002924.g005 Figure 5 Schematic representation of the genome organization of BASV and its protein similarity plot compared to representative rhabdoviruses. The similarity plots are generated by aligning the concatenated rhabdovirus proteins and calculating scanning amino acid pairwise identities using a window size of 50 bp. The horizontal bar under each similarity plot shows the percent identity of the rhabdovirus protein relative to its corresponding protein in BASV. Genes coding for the 5 core rhabdovirus proteins are shown in green, while the accessory U1, U2, or U3 genes are shown in blue. Black bars correspond to accessory proteins which are not present in the genome. Note that BEFV contains 3 genes between G and L; only the alignment between the alpha-1 protein of BEFV and the U3 protein of BASV is shown (asterisk). The x-axis refers to the nucleotide position along the ∼12 kb genome of BASV. The sequence divergence of BASV relative to other rhabdoviruses is also correlated with differences in genome structure (Fig. 5). The prototype genome organization of rhabdoviruses, found in lyssaviruses, is N-P-M-G-L. However, molecular analysis of novel rhabdoviruses has often revealed more complex genomes, with up to 10 additional open reading frames (ORF) located within an existing gene or interposed between the five core genes [19], [20], [21]. Rhabdoviruses from the Tibrogargan group (TIBV and CPV) share a distinctive genome structure with three additional genes, two between M and G (U1 and U2) and one between G and L (U3) [22]. Interestingly, BASV also has these three additional genes (U1–U3), confirming the phylogenetic relationship and overall structural similarity to the Tibrogargan group viruses. Based on their size, the U3 proteins of TIBV, CPV, and presumably BASV are candidate viroporins [22]. BASV is more distant structurally and phylogenetically from the Ephemero and Hart Park Group rhabdoviruses (Figs. 3 and 4), which do not contain U1 or U2 genes, but rather an additional two or three genes between G and L (including a putative U3 viroporin in BEFV referred to as the alpha-1 protein) (Fig. 5, asterisk). Moussa virus (MOUV), another rhabdovirus recently discovered in Africa (Fig. 4), does not contain any accessory genes but instead, shares the prototype N-P-M-G-L rhabdovirus structure [23]. BASV Serological Testing of the Case Patient and Close Contacts To confirm that BASV is infectious to humans, convalescent sera were collected in early 2012 from surviving Patient 3 as well as five additional health care workers from Mangala identified as close contacts and tested in a blinded fashion for the presence of neutralizing antibodies to BASV (Fig. 6). Two of the six sera tested strongly positive with 50% protective doses between 1∶1,000 and 1∶5,000 (Figs. 6A and 6F). Moreover, the observed neutralization was highly specific for BASV-G, since no neutralization was observed with pseudoviruses harboring the vesicular stomatitis virus glycoprotein (VSV-G). One of the neutralizing sera had been collected from surviving Patient 3 (Fig. 6A, “Patient 3”), whereas the other serum sample, containing even higher titers, corresponded to an asymptomatic nurse directly caring for Patient 3 during his period of acute hemorrhagic illness (Fig. 6F, “Contact 5”). Specifically, Contact 5 was the primary health care provider to Patient 3 at the health center and during his transfer to the general hospital at Boma. All 6 individuals, including Patient 3, tested negative for BASV viremia by specific RT-PCR (data not shown). 10.1371/journal.ppat.1002924.g006 Figure 6 Detection of antibodies to BASV by serum neutralization of VSVΔG-GFP pseudotypes. Infectivities of VSVΔGFP pseudotypes bearing the glycoproteins of VSV or BASV, respectively, after incubation with 5-fold serial dilutions (1∶10, 1∶50, 1∶250, 1∶1,250, 1∶6,250, 1∶31,250) of sera from six individuals are depicted as percent of infectivity in the absence of serum. The six individuals tested include a patient with hemorrhagic fever (panel A, “Patient 3”), the nurse directly caring for him (panel F, “Contact 5”), and other health care workers in Mangala village (panels B–E). All data points represent the average of triplicate assays; error bars indicate standard deviations. Similar results were obtained in an independent experiment using murine leukemia virus (MLV)-based pseudotypes (data not shown). Epidemiological Screening for BASV in the DRC BASV was not detected by PCR in 43 serum samples from other unknown cases or outbreaks of hemorrhagic fever reported in the DRC from 2008–2010 (Fig. 7A, pink). Five of these 43 samples originated from the Bas-Congo outside of Mangala village and the Boma Bungu Health Zone. In total, the unknown hemorrhagic cases/outbreaks spanned 9 of the 11 provinces in the DRC, and all 43 samples also tested negative by PCR for the known hemorrhagic fever viruses circulating in Africa (data not shown). Fifty plasma samples collected from randomly selected blood donors in the Kasai-Oriental province of DRC (Fig. 7A, star; Table S2) were also screened and found to be negative for BASV-neutralizing antibodies (Fig. 7B). 10.1371/journal.ppat.1002924.g007 Figure 7 BASV Screening in DRC, Africa. (A) All 43 serum samples corresponding to unknown hemorrhagic fever cases or outbreaks in 2008–2010 from 9 provinces in DRC (pink) tested negative for BASV by PCR. (B) Sera from 50 donors in Kasai-Oriental province, DRC (Panel A, star) were tested for BASV-neutralizing antibodies. Sera at 1∶50 (dark blue) or 1∶500 dilution (light blue) were tested. Serum from the surviving Patient 3 was included as a positive control (grey shaded area). Data points represent an average of duplicate assays. Discussion Among more than 160 species of rhabdoviruses identified to date, fewer than 10 have been isolated from humans [24]. In addition, while human infection by rhabdoviruses has previously been associated with encephalitis, vesicular stomatitis, or “flu-like” illness, the discovery of BASV is the first time that a member of the Rhabdovirus family has been associated with hemorrhagic fever in humans with a fulminant disease course and high fatality rate. To our knowledge, this is also the first successful demonstration of de novo assembly of a novel, highly divergent viral genome in the absence of a reference sequence and directly from a primary clinical sample by unbiased deep sequencing. Several lines of evidence implicate BASV in the hemorrhagic fever outbreak among the 3 patients in Mangala. First, this virus was the only credible viral pathogen detected in the blood of the lone survivor during his acute hemorrhagic illness by exhaustive deep sequencing of over 140 million reads. Analysis of the Illumina deep sequencing reads for the presence of other viral pathogens yielded only endogenous flora or confirmed laboratory contaminants (Table S1 and Fig. S2). Some enteric pathogens, such as E. coli O157:H7, Campylobacter, Shigella, and Salmonella, are diagnosed through fecal laboratory testing and not blood, and have been associated with hemorrhagic diarrhea [25]. However, these outbreaks are typically foodborne and associated with larger clusters and much greater numbers of clinical cases than reported here [26], [27], [28]. Furthermore, enteric diarrheal cases rarely present with systemic symptoms such as fever or generalized mucosal hemorrhage, with bleeding most often limited to the gastrointestinal tract, and overall mortality rates are generally low [26]. Thus, the clinical syndrome observed in 3 patients with hemorrhagic fever in the DRC, a region endemic for viral hemorrhagic fevers, is much more consistent with infection by a VHF disease agent. BASV is a plausible hemorrhagic fever candidate because it is a novel, highly divergent infectious virus, thus of unknown pathogenicity, and was detected at a titer of >1 million copies/mL in blood from an acutely ill individual. In addition, there is ample precedent for hemorrhagic disease from rhabdoviruses, as members of the genus Novirhabdovirus cause severe hemorrhagic septicemia in fresh and saltwater fish worldwide [29] (Fig. 4). The detection of BASV seropositivity in an asymptomatic close contact (Fig. 6) is not surprising given that up to 80% of patients infected with Lassa virus do not exhibit any hemorrhagic fever symptoms [30], [31]. Prior to the BASV outbreak, no hemorrhagic disease cases had been reported in Boma Bungu Health Zone. BASV was also not detected in 43 serum samples from unknown, filovirus-negative cases or outbreaks of hemorrhagic fever from 2008–2010 spanning 9 of the 11 provinces in the DRC (Fig. 7A). In addition, a serosurvey of 50 random blood donors from Kasai-Oriental province in central DRC was negative for prior exposure to BASV (Fig. 7B). Taken together, these data suggest that the virus may have emerged recently and locally from Boma Bungu in Bas-Congo, DRC. We were unable to isolate BASV despite culturing the RNA-positive serum in a number of cell cultures and inoculation into suckling mice. One explanation for these negative findings may be that the virus inoculation titers of 1 million copies per mL of BASV in blood from the lone survivor. A more likely explanation is viral inactivation resulting from the lack of adequate cold chain facilities in remote Boma Bungu. Viral RNA can often still be detected by RT-PCR in sera that is culture-negative [32]. In support of this premise, we have observed that the BASV-G/VSVΔG-GFP pseudotyped virus efficiently infects and replicates in a variety of insect and mammalian (including human) cell lines (Steffen, et al., manuscript in preparation). In the absence of a positive culture, a “reverse genetics” approach to produce recombinant BASV particles, if successful, would greatly facilitate further study of the virus, as established previously for other rhabdoviruses such as VSV [33]. Based on our findings, some speculations on the origin of and routes of transmission for BASV can be made. All 3 patients became ill with acute hemorrhagic fever over a 3-week period within the same 2500-m2 area of Mangala village, suggesting that all 3 cases were infected with the same pathogen. Waterborne or airborne transmission would be expected to result in more numerous cases than the 3 reported. There were no reports of animal die-offs that would suggest potential exposures to infected wild animals or livestock. Taken together, these observations suggest that an unknown arthropod vector could be a plausible source of infection by BASV. This hypothesis is consistent with the phylogenetic and structural relationship of BASV to rhabdoviruses in the Tibrogargan group and Ephemerovirus genus, which are transmitted to cattle and buffalo by Culicoides biting midges [9]. In addition, the recent discovery of Moussa virus (MOUV), isolated from Culex mosquitoes in Cote d'Ivoire, Africa [23], implies the presence of hitherto unknown arthropod vectors for rhabdoviruses on the continent. Nevertheless, at present, we cannot exclude the possibility of other zoonotic sources for the virus or even nosocomial bloodborne transmission (as Patients 1 and 2 have not clearly been established to be BASV cases by serology or direct detection), and the natural reservoir and precise mode of transmission for BASV remain unknown. A community-based serosurvey in Boma Bungu and an investigation to track down potential arthropod or mammalian (e.g. rodents and bats) sources for BASV are currently underway. Although we cannot exclude the possibility of independent arthropod-borne transmission events, our epidemiologic and serologic data do suggest the potential for limited human-to-human transmission of BASV. Patient 3, a nurse, had directly taken care of Patients 1 and 2 at the health center, and another nurse (Contact 5), who had taken care of Patient 3 (but not Patients 1 or 2) had serologic evidence of asymptomatic BASV infection. We present a hypothetical model for BASV transmission during the hemorrhagic fever outbreak in which the initial infection of two children in Mangala (Patients 1 and 2) was followed by successive human-to-human transmission events involving two healthcare workers (Patient 3 and Contact 5) (Fig. 8). This pattern of transmission from the community to health care workers is also commonly seen in association with outbreaks of Ebola and Crimean-Congo hemorrhagic fever [6], [34]. 10.1371/journal.ppat.1002924.g008 Figure 8 Proposed model for BASV transmission during the hemorrhagic fever outbreak in Mangala. Patients presenting with symptoms of acute hemorrhagic fever are depicted in red. Dashed red lines represent potential routes of BASV transmission. Contacts 1 through 5 are health care workers at the local health center in Mangala village. Abbreviations: HCW, health care worker; y/o, year-old; Ab, antibody. While rhabdoviruses are distributed worldwide, some authors have suggested that the Rhabdoviridae family probably originated from tropical regions of the Old or New World [9]. The discovery of BASV in Central Africa suggests that additional rhabdoviruses of clinical and public health importance likely await identification, especially in these poorly investigated geographic regions. Active epidemiological investigation and disease surveillance will be needed to fully ascertain the clinical and public health significance of BASV infection in humans, as well as to prepare for potentially larger human outbreaks from this newly discovered pathogen. Methods Ethics Statement Written informed consent for publication of their case reports was obtained from the sole survivor of the hemorrhagic fever outbreak and the parents of the two deceased children. Written informed consent was obtained from the surviving patient and 5 of his close contacts for analysis of the serum samples reported in this study. Samples were analyzed under protocols approved by the institutional review boards of University of California, San Francisco, the University of Texas Medical Branch, and the National Institute of Biomedical Research (INRB) and CIRMF in Gabon, and the Institutional Animal Care and Use Committee (IACUC) of the University of Texas Medical Branch. Diagnostic Samples No diagnostic samples were available from Patient 1 or Patient 2. Blood was collected in a red top serum tube from Patient 3 on June 16, during the acute phase, three days after hemorrhagic onset. The sample was transported at 4°C to the BSL-4 facility at CIRMF. Serum was obtained by centrifugation at 2300 rpm for 10 min. No other acute samples from Patient 3 were available. In January of 2012 (∼2.5 years after the outbreak), convalescent sera were collected from Patient 3 and close contacts (other workers at the health center) for BASV neutralization testing. Forty-three serum samples from other unknown hemorrhagic fever cases or outbreaks representing 9 of 11 provinces in the DRC were available for BASV PCR testing (Fig. 7A). Fifty available plasma samples from random blood donors (median age 27.5 years; age range 1–76 years) in Kasai Oriental province, DRC, were also tested for antibodies to BASV (Fig. 7A and B; Table S2). Nucleic Acid Extraction and Viral PCR Testing RNA was extracted from 140 µl of serum using the QIAamp viral RNA mini kit (Qiagen). Taqman real-time reverse-transcription-PCR (RT-PCR) testing for known hemorrhagic fever viruses was performed using primers and probes specific for Marburg virus (MARV), all four species of Ebola virus (Zaire, ZEBOV; Sudan, SEBOV; Côte d'Ivoire, CIEBOV, and Bundibugyo, BEBOV), Crimean-Congo hemorrhagic fever virus (CCHFV), Yellow fever virus (YFV), Dengue virus (DENV), Rift Valley fever virus (RVFV) and Chikungunya virus (CHIKV) (available upon request). Discovery of the BASV Rhabdovirus by 454 Pyrosequencing 200 µL of serum sample were inactivated in 1 mL of TRIzol (Invitrogen), and nucleic acid extraction and purification were performed according to the manufacturer's instructions. Roche 454 pyrosequencing using randomly amplified cDNA libraries was performed as described previously [35]. Viral sequences were identified using BLASTn or BLASTx by comparison to the GenBank nonredundant nucleotide or protein database, respectively (E-score cutoff = 10−5). De novo Assembly of the BASV Genome by Illumina Sequencing To recover additional BASV sequence, two sets of cDNA libraries were prepared from DNase-treated extracted RNA using a random PCR amplification method as described previously [36], or random hexamer priming according to the manufacturer's protocol (Illumina). The libraries were then pooled and sequenced on two lanes of an Illumina HiSeq 2000. Raw Illumina sequences consisting of 100 base pair (bp) paired-end reads were filtered to exclude low-complexity, homopolymeric, and low-quality sequences, and directly compared using BLASTn or BLASTx alignments to a library consisting of all rhabdovirus sequences in GenBank. The initial read obtained by 454 pyrosequencing as well as other reads aligning to rhabdoviruses were then inputted as “seeds” into the PRICE de novo assembler [37] (Fig. 2), with a criterion of at least 85% identity over 25-bp to merge two fragments. De novo assembly of the BASV genome was performed iteratively using PRICE and the Geneious software package (Biomatters) [38]. The near-complete whole genome sequence of the novel rhabdovirus (∼98.2% based on protein homology to other rhabdoviruses) was determined to at least 3× redundancy by de novo assembly as well as PCR and Sanger sequencing of low-coverage regions. Sanger sequencing was also performed to verify the accuracy of the assembly and confirm the genomic organization of BASV (Fig. 2). Deep Sequencing Analysis of the BASV Serum Sample for Other Pathogens Rapid classification of the ∼140 million 100-bp paired-end Illumina reads was performed using a modified cloud computing-based computational analysis pipeline [17] (Veeraraghavan, Sittler, and Chiu, manuscript in preparation). Briefly, reads corresponding to human sequences were taxonomically classified using SOAP and BLAT software [39], [40]. Other reads were then identified using BLASTn or BLASTx by comparison to GenBank-derived reference databases (E-score cutoff = 10−5). PCR Quantitation of BASV Burden To estimate the viral load in the patient's serum, we first designed a set of specific PCR primers for detection of BASV targeting the L protein, BASV-F (5′- CGCTGATGGTTTTTGACATGGAAGTCC-3′)/BASV-R (5′-TAAACTTCCTCTCTCCTCTAG-3′), for use in a SYBR-Green real-time quantitative RT-PCR assay. A standard curve for the assay was constructed as described previously [36]. The viral load in the patient's serum was determined by comparison to the standard curve. Structural Features and Phylogenetic Analysis Predicted open reading frames (ORFs) in the BASV genome were identified with Geneious [38]. Multiple sequence (Figs. 3 and 4; Fig. S1) and pairwise (Fig. 5) alignments of BASV proteins relative to corresponding proteins from other rhabdoviruses were calculated using MAFFT (v6.0) with the E-INS-i option and at default settings [41]. To generate the phylogeny trees, all rhabdoviruses in GenBank were included as well as representative members of other families within the order Mononegavirales. Bayesian tree topologies were assessed with MrBayes V.32 software (20,000 sampled trees; 5,000 trees discarded as burn-in) [42]. Convergence was confirmed by the PSRF statistic in MrBayes, as well as by visual inspection of individual traces using TRACER from the BEAST software package [43]. Trees were visualized after midpoint rooting with FigTree V1.31 [43]. Virus Cultivation in Cell Cultures or Suckling Mice Initial attempts were made to culture the virus using a total of 200 µL of BASV-positive serum inoculated onto confluent monolayers of Vero E6 and C6/36 (Aedes albopictus mosquito) cells in 6-well plastic tissue culture plates at 37°C and 28°C, respectively, in a 5% CO2 environment as previously described [44]. From 20–50 µL of serum were used to inoculate the cells, which were examined daily for cytopathic effect (CPE) at days 5, 7, and 14. Supernatants were harvested and two additional blind passages were performed, each passage followed by 14 days of observation for CPE. Cell culture supernatants were also monitored for evidence of viral replication by quantitative RT-PCR. Using the remaining 100 uL of BASV-positive serum, further attempts were made to culture the virus in 5 cell lines and in suckling mice. The serum sample was split in half and diluted 1∶20 or 1∶10 in phosphate-buffered saline with 20% fetal bovine serum (FBS) to allow sufficient volume to inoculate cell cultures or mice, respectively. The first diluted sample was inoculated intracerebrally into a litter (n = 12) of 1 day old mice. Pups were observed daily for 14 days for lethality or signs of clinical illness. The second diluted sample was inoculated into 12.5 cm2 tissue culture flasks of Vero, BHK, LLC-MK2 (rhesus monkey kidney), CCL-106 (rabbit kidney) and C6/36 cells. Vertebrate cells were held at 37°C for 14 days and observed for evidence of CPE. Mosquito cells were maintained at 28°C for 10 days. Since no CPE was observed in any of the cultures, cells were subsequently fixed for transmission electron microscopy to see if viral particles could be visualized [45]. Construction of VSVΔG-GFP Pseudotypes and BASV Serum Neutralization Testing A pseudotype system based on a vesicular stomatitis virus (VSV) construct carrying a reporter gene for green fluorescent protein (VSVΔG-GFP) and bearing the predicted synthesized BASV glycoprotein (BASV-G) was used to generate a serum neutralization assay for BASV. Briefly, the predicted BASV glycoprotein (BASV-G) was synthesized (Genscript) and subcloned into the pCAGGS expression plasmid. Human embryonic kidney 293T cells were seeded (DMEM + 10% FBS + penicillin/streptomycin + Glutamax (Gibco) + non-essential amino acids (Gibco)) in 10 cm culture dishes 24 hours prior to transfection. Cells were transfected with 20 µg BASV-G, VSV-G, or empty pCAGGS DNA per dish following a calcium phosphate transfection protocol [46]. The culture medium was replaced 15 hours post-transfection and cells were stimulated with 6.2 mM valproic acid for 4 hours before the medium was replaced again. At 36 hours post-transfection the transfected cells were infected with VSVΔG-GFP/VSV-G pseudotypes at a multiplicity of 0.1–0.3. The inoculum was removed after 4 hours and replaced by fresh culture medium. At 24 hours post-infection, infectious supernatants were harvested, filtered through 0.45 µm filters, and concentrated 10-fold by centrifugation through a 100-kDA filter (Millipore). Concentrated viruses were aliquoted and stored at −80°C. For serum neutralization testing, human hepatoma Huh-7 cells were seeded (DMEM +10% FBS + penicillin/streptomycin + Glutamax (Gibco) + non-essential amino acids (Gibco)) in 48-well plates 24 hours prior to infection. Per well 10 µl of pseudovirus harboring either BASV-G or VSV-G (adjusted to obtain 25–50% infection of target cells) was mixed with 10 µl of the respective serum dilution and incubated for 45 minutes at 37°C. Subsequently, the mix was added to the target cells (performed in triplicate) and cells were incubated for 24 hours at 37°C. The infected cells were detached with trypsin and washed with PBS before fixing with 2% paraformaldehyde for 1 hour at room temperature. GFP expression in infected cells was quantified by flow cytometry using a LSR II (BD Biosciences) and the collected data was analyzed with FlowJo software (TreeStar). Abbreviations and Nucleotide Sequence Accession Numbers The annotated, nearly complete sequence of BASV has been submitted to GenBank (accession number JX297815). Deep sequencing reads have been submitted to the NCBI Sequence Read Archive (accession number SRA056894). Accession numbers used for the phylogenetic analyses in Figs 3, 4, and S1 are listed as follows, in alphabetical order: ABLV, Australian bat lyssavirus (NP_478343); ARAV, Aravan virus (ABV03822), BEFV, Bovine ephemeral fever virus (NP_065409); BYSMV, Barley yellow striate mosaic virus (BYSMV); CDV, Canine distemper virus (AAR32274); CHPV, Chandipura virus (ADO63669); CPV, Coastal Plains virus (ADG86364); COCV, Cocal virus (ACB47438); DURV, Durham virus (ADB88761); DUVV, Duvenhage virus (ABZ81216); EBLV1, European bat lyssavirus 1 (ABZ81166), EBLV2, European bat lyssavirus 2 (ABZ81191); EBOV, Ebola virus (AAG40171, AAA79970, BAB69010); EVEX, Eel virus European X virus (CBH20130); FDLV, Fer-de-lance virus (NP_899661); FLAV, Flanders virus (AAN73288); HeV, Hendra virus (NP_047113); HIRRV, Hirame rhabdovirus (ACO87999); HMPV, Human metapneumovirus (L_HMPVC); HPIV-1, Human parainfluenza virus type 1 (AAA69579); HPIV-2, Human parainfluenza virus type 2 (CAA40788); HPIV-3, Human parainfluenza virus type 3 (AAA46854); HPIV-4, Human parainfluenza virus type 4 (BAJ11747); INHV, Infectious hematopoietic necrosis virus (NP_042681); IRKV, Irkut virus (ABV03823); ISFV, Isfahan virus (CAH17548); KHUV, Khujand virus (ABV03824); LBV, Lagos bat virus (ABZ81171); LNYV, Lettuce necrotic yellows virus (YP_425092); MARAV, Maraba virus (AEI52253); MARV, Marburg virus (YP_001531159); MeV, Measles virus (AF266288); MMV, Maize mosaic virus (YP_052855); MOKV, Mokala virus (ABZ81206); MOUV, Moussa virus (ACZ81407); MUV, Mumps virus (AF201473); NCMV, Northern cereal mosaic virus (NP_597914); NDV, Newcastle disease virus (ADH10207); NGAV, Ngaingan virus (YP_003518294); NiV, Nipah virus (AAY43917); OVRV, Oak Vale rhabdovirus (AEJ07657); PFRV, Pike fry rhabdovirus (ACP28002); RABV, Rabies virus (NP_056797); RSV, Respiratory syncytial virus (NP_056866); RYSV, Rice yellow stunt rhabdovirus (NP_620502); SIGMAV, Sigma virus (ACU65444); SCRV, Siniperca chuatsi rhabdovirus (YP_802942); SHRV, Snakehead virus (AAD56771); SMRV, Scophthalmus maximus rhadovirus (ADU05406); SVCV, Spring viremia of carp virus (NP_116748); SYNV, Sonchus yellow net virus (NP_044286); TIBV, Tibrogargan virus (ADG86355); TUPV, Tupaia virus (AAX47602); TVCV, Tomato vein clearing virus (YP_224083); VHSV, Viral hemorrhagic septicemia virus (NP_049550); VSIV, Vesicular stomatitis virus, Indiana (NP_041716); VSNJV, Vesicular stomatitis virus, New Jersey (P16379); WCBV, West Caucasian bat virus (ABV03821); WONGV, Wongabel virus (YP_002333280). Supporting Information Figure S1 Phylogenetic analysis of the N, P, M, and G proteins of BASV and other rhabdoviruses. Each phylogenetic tree is rooted by using the corresponding protein from human parainfluenza virus type 1 (HPIV-1), a paramyxovirus, as an outgroup. Abbreviations and accession numbers used for the phylogenetic analysis are provided in Methods. (TIF) Click here for additional data file. Figure S2 Confirmation of laboratory contamination by rotavirus and absence of rotavirus in BASV serum by specific PCR. An RT-PCR assay for detection of Group A rotaviruses was performed using primers NSP3F (5′-ACCATCTWCACRTRACCCTCTATGAG-3′) and NSP3R (5′- GGTCACATAACGCCCCTATAGC-3′), which generate an 87-bp amplicon (Freeman, et al., (2008) J Med Virol 80: 1489–1496). PCR conditions for the assay were 30 min at 50°C, 15 min at 95°C for the reverse transcription step followed by 40 cycles of 95°C, 30 s/55°C, 30 s/72°C, 30 s and 72°C/7 min for the final extension. PCR products are visualized by gel electrophoresis, using a 2% agarose gel and 1 kB ladder. Rotavirus is readily detected in extracted RNA from a stool sample taken from an ongoing study of viral diarrhea in the laboratory (lane 1), but not in two separate aliquots of extracted nucleic acid from the BASV serum sample (lanes 2 and 3). (TIF) Click here for additional data file. Table S1 Viral reads in the deep sequencing data corresponding to the BASV-positive serum sample. (DOCX) Click here for additional data file. Table S2 Demographics of 50 blood donors from Kasai-Oriental province, DRC, randomly selected for BASV antibody screening. (DOCX) Click here for additional data file.

There is convincing in vitro evidence that the muscular form of the nicotinic acetylcholine receptor (nAChR), the neuronal cell adhesion molecule (NCAM), and the p75 neurotrophin receptor (p75NTR) bind rabies virus and/or facilitate rabies virus entry into cells. Other components of the cell membrane, such as gangliosides, may also participate in the entry of rabies virus. However, little is known of the role of these molecules in vivo. This review proposes a speculative model that accounts for the role of these different molecules in entry and trafficking of rabies virus into the nervous system.