1. Introduction

All viruses in the genus Flavivirus (family Flaviviridae) possess a single-stranded, positive-sense RNA genome of approximately 11 kb. The genome usually encodes a single open reading frame (ORF) that is flanked by 5' and 3' untranslated regions (UTRs) of ~100 and ~400–700 nt, respectively. The ORF encodes a large polyprotein that is co- and post-translationally cleaved to generate three structural proteins, designated the capsid (C), premembrane/membrane (prM/M) and envelope (E) proteins, and seven nonstructural (NS) proteins in the gene order: 5'–C–prM(M)–E–NS1–NS2A–NS2B–NS3–NS4A–2K–NS4B–NS5-3'. The genomes of some flaviviruses appear to encode an additional protein as a consequence of ribosomal frameshifting as discussed later in this review.

Despite their similar genomic organizations, flaviviruses possess fundamental differences in their host ranges and transmissibilities. Most recognized flaviviruses are transmitted horizontally between hematophagous arthropods and vertebrate hosts and are therefore considered to be dual-host viruses. Dual-host flaviviruses can be further divided into mosquito/vertebrate and tick/vertebrate viruses. Examples of mosquito/vertebrate flaviviruses include dengue virus (DENV), yellow fever virus, Japanese encephalitis virus (JEV) and West Nile virus (WNV), all of which are human pathogens of global concern. Flaviviruses of localized public health concern include St Louis encephalitis virus (SLEV) and Murray Valley encephalitis virus (MVEV). Tick/vertebrate flaviviruses associated with serious human disease include tick-borne encephalitis virus, Langat virus and Powassan virus. Not all flaviviruses cycle between arthropods and vertebrates; some have a vertebrate-specific host range while others appear to be insect-specific. Vertebrate-specific flaviviruses, also known as No Known Vector (NKV) flaviviruses, can be divided into two groups: those isolated exclusively from rodents (e.g., Modoc virus; MODV) and those isolated exclusively from bats (e.g., Rio Bravo virus). It has been suggested that NKV flaviviruses are maintained in nature by horizontal transmission among hosts. Insect-specific flaviviruses (ISFs) can also be divided into two distinct groups (Figure 1). ISFs in the first group are phylogenetically distinct from all other known flaviviruses and, for the purpose of this review, will be referred to as classical ISFs (cISFs) since they were discovered first. ISFs in the second group phylogenetically affiliate with mosquito/vertebrate flaviviruses and, for the purpose of this review, will be referred to as dual-host affiliated ISFs (dISFs). This group is apparently not monophyletic. There has been a dramatic increase in the number of ISFs discovered in the last ten years and this is partly due to advances in the methods available for virus detection.

In addition to the growing number of ISFs recently described, cISF-like sequences, designated cell silent agent (CSA), have been found integrated in the genomes of Aedes spp. mosquitoes. One sequence was shown to encode a NS1-NS4A-like transcript. Partial E, NS4B and NS5-like sequences were also identified. The occasional integration of flavivirus sequences into the mosquito genome could occur due to the reverse transcriptase and integrase activities of co-infecting or endogenous retroviruses, and indeed the integration of viral sequences into the host genome has since been documented for many other RNA viruses and host species. Most integrated sequences are highly fragmented or have internal stop codons but several encode intact ORFs. Importantly, some genome-integrated sequences may be transcribed and therefore, the detection of flavivirus-like RNA in an organism is not necessarily proof that the organism carries an active flavivirus infection.

2.1. Discovery, Geographic Distribution and Natural Host range

Classical ISFs have a ubiquitous geographic distribution; these viruses have been isolated from mosquitoes in every continent with the exception of Antarctica (Table 1). Additionally, cISF-like sequences have been detected in sandflies and midges in several Mediterranean countries as discussed later in this section. The first cISF to be discovered was cell fusing agent virus (CFAV) after its isolation from Aedes aegypti cell cultures over 40 years ago. This discovery received little attention as illustrated by the fact that 17 years passed before another article on the virus was published. CFAV has since been isolated from or detected in field-collected mosquitoes in Indonesia, Mexico, Puerto Rico and Thailand. CFAV has also been isolated from laboratory colonies originally established from mosquitoes collected in the United States. Additionally, CFAV-like sequences have been identified in field-collected mosquitoes in Argentina (Genbank Accession No. DQ335466, DQ335467 and DQ431718) but the sequences are too short (87–110 nt) for reliable analysis.

The next cISF to be discovered was Kamiti River virus (KRV) after its isolation from mosquitoes in Kenya in 1999 (Table 1). A KRV-like sequence has also been detected in mosquitoes in Argentina (Genbank Accession No. DQ335465) but, due to the limited amount of sequence data obtained in the study (a 124-nt region of the NS5 gene was sequenced), additional information is needed to determine whether KRV occurs in this country. The third member of the cISF group to be reported was Culex flavivirus (CxFV). The virus was first isolated from Culex spp. mosquitoes in Japan and Indonesia in 2003–2004 and later isolated from Culex spp. mosquitoes in Argentina, Brazil, China, Guatemala, Mexico, Taiwan, Trinidad, Uganda and the United States.

In the last ten years, several other cISFs have been described and these include: Aedes flavivirus (AEFV) in Japan, Italy, Thailand and the United States, Aedes galloisi flavivirus (AGFV) in Japan, Calbertado virus (CLBOV) in Canada and the USA, Nakiwogo virus (NAKV) in Uganda, Nienokoue virus (NIEV) in Cote d’Ivoire (Genbank Accession No. NC_024299), Palm Creek virus (PCV) in Australia and Quang Binh virus (QBV) in Vietnam and China (Table 1). Several other novel cISFs have also been identified. However, after performing nucleotide sequence alignments, it is apparent that some of these viruses have been assigned multiple names due to their simultaneous discoveries by independent research groups as discussed below.

Three apparently novel cISFs were reported in the literature within in space of a few months: Hanko virus (HANKV) after its isolation from Ae. caspius in Finland in 2005, Ochlerotatus flavivirus (OcFV) after its isolation from various Aedes and Culex spp. mosquitoes in Spain, Italy and Portugal from 2007 to 2010 and Spanish Ochlerotatus flavivirus (SOcFV) after its isolation from Ae. caspius in Spain in 2006 (Table 1). Huhtamo and colleagues sequenced the entire ORF of their virus; the two other research groups sequenced a 238 to 917 nt region of the NS5 gene. Pairwise nucleotide sequence alignments of the 163-nt region shared by representative HANKV, OcFV and SOcFV sequences (Genbank Accession Nos., JQ268258, GQ476991 and JF707790 respectively) revealed that these viruses are 91% to 96% identical. It has been proposed that flaviviruses with >84% nucleotide sequence identity should be classified within the same species. Although these alignments were performed using short sequences in the relatively highly conserved NS5 region, according to the above criterion, HANKV, OcFV and SOcFV are likely to be the same virus species. For the remainder of this review, the virus will be referred to as HANKV because Huhtamo and colleagues performed the most comprehensive sequence analysis. The following year, an article describing an apparently novel cISF designated Ochlerotatus caspius flavivirus from Portugal (OCFVPT) was published. The authors sequenced almost all of the OCFVPT genome and reported that it has 89% nucleotide identity to the corresponding region of HANKV. In accordance to the criterion for flaviviral species demarcation, this virus is HANKV and not an unrecognized cISF species.

Another cISF has been assigned multiple names: Culex theileri flavivirus (CTFV or CxthFV), Spanish Culex flavivirus (SCxFV) and Wang Thong virus (WTV). Culex theileri flavivirus (CTFV) was isolated from Cx. theileri in Portugal in 2009–2010 (Table 1). The same virus was independently discovered by another research group after its isolation from Cx. theileri in Portugal and Spain in 2007–2010, and coincidently given the same name but different acronym (e.g., CxthFV). SCxFV is the name assigned to several isolates obtained from Cx. theileri and Cx. pipiens in Spain in 2006. WTV was detected in Cx. fuscocephala in Thailand on an unspecified date (Genbank Accession No. AY457040). Parreira and colleagues sequenced almost the entire genome of their virus; the other groups sequenced a 159 to 917 nt region of the NS5 gene. Pairwise sequence alignments of the 140-nt region shared by representative CTFV, CxthFV, SCxFV and WTV sequences (Genbank Accession Nos. HE574574, EU716420, JF707811 and AY457040, respectively) revealed that these viruses are 91% to 100% identical. Therefore, according to the criterion for flaviviral species demarcation, CTFV, CxthFV, SCxFV and WTV are the same virus. For the remainder of this review, the virus will be referred to as Culex theileri flavivirus since this name was chosen by two of the four research groups that made the discovery. The acronym selected by Parreira and colleagues (e.g., CTFV) will be used because these researchers performed the most comprehensive sequence analysis.

Quang Binh virus or a novel QBV-like virus (designated Yunnan Culex flavivirus; YNCxFV) was isolated from 10 pools of Cx. tritaeniorhynchus and one pool of Anopheles sinensis collected in the Yunnan Province of China in 2009 (Table 1). The genome of one isolate was completely sequenced and the ORF was reported to have 83.0% nucleotide identity to the corresponding region of the prototypical QBV isolate. Because this figure is close to the >84% value for flavivirus species demarcation, the authors opted for a conservative approach and considered their isolate to be a strain of QBV. The authors also pointed out that their isolate was obtained from the same mosquito spp. and geographic region as QBV (Yunnan Province borders Vietnam) and that additional testing (i.e., neutralization assays) was required before the isolate could be considered the prototypical member of a novel species. For the purpose of this review, the entire genomic sequences (as opposed to the entire ORFs) of the prototypical QBV and YNCxFV isolates (Genbank Accession Nos. FJ644291 and KC464457, respectively) were aligned to shed more light on their genetic relatedness. The two sequences have 83.7% nucleotide identity which is even closer to the threshold value establish by Kuno et al. This analysis also revealed that the two genomes are of the exact same length (10,865 nt). Thus, YNCxFV could very well be a divergent isolate of QBV rather than a novel cISF species. A comparison of the lengths of the non-coding regions revealed that the 5' UTR of YNCxFV is one nucleotide shorter than the corresponding region of QBV (111 nt vs. 112 nt) while its 3' UTR is one nucleotide longer (674 nt vs. 673 nt). Some cISFs display strain-specific differences in the lengths of their non-coding region lengths; for example, the 5' and 3' UTRs of CxFV Toyama 1431 strain (Genbank Accession No. AB701775) are both one nucleotide shorter than the corresponding regions of CxFV Tokyo strain (Genbank Accession No. AB262759). Because it is unclear whether YNCxFV is an unusual isolate of QBV or a distinct virus species, the conservative approach opted by Zuo and colleagues will be used for the remainder of this review and their virus will be referred to as QBV.

A potentially novel cISF, designated Aedes vexans flavivirus (AeveFV), was isolated from Ae. vexans in Italy and the Czech Republic in 2008–2009. A short (131 to 263 nt) region of the NS5 gene was sequenced and shown to have no more than 80% nucleotide identity to the corresponding region of its closest relative, consistent with the discovery of a new virus. However, the sequences are short and more comprehensive sequencing experiments are needed to determine whether AeveFV is an unrecognized virus. Likewise, a potentially novel cISF, designated Czech Aedes vexans flavivirus (Czech AeveFV), was isolated from Ae. vexans in the Czech Republic in 2009 but the corresponding sequences are too short (209 to 217 nt) for reliable analysis. Another potentially novel cISF, designated Aedes cinereus flavivirus (AeciFV), was detected in Ae. cinereus in the U.K. in 2010. Comprehensive sequence alignments were not performed (the virus was compared to only five other cISFs) and the sequence has not been deposited into the Genbank database. Additional information is needed to determine whether AeciFV is a novel cISF. Although AeveFV, Czech AeveFV and AeciFV could very well represent novel cISFs, they are not listed in Table 1; the table is restricted to viruses for which more than 300 nt of sequence data are available.

Two 917-nt sequences corresponding to a cISF designated Culex pipiens flavivirus were detected in mosquitoes in Portugal in 2009–2010 (Genbank Accession No. HE997068-HE997069). The sequences are 97% identical to the corresponding region of CTFV and therefore, the species name of Culex pipiens flavivirus should be discontinued. Multiple 165-nt sequences corresponding to a virus denoted as Culicinae flavivirus were also identified in mosquitoes in Portugal (Genbank Accession No. EU716415–EU716419 and EU716421–EU716424). This species name should also be discontinued because these sequences are 90%–99% identical to the corresponding region of HANKV. Other species names that appear in the NCBI taxonomy and Genbank databases that should be discontinued for similar reasons are Mediterranean Culex flavivirus, Mediterranean Ochlerotatus flavivirus and mosquito flavivirus (for example, Genbank Accession No. JF707854, JF707806 and KF882513).

Although cISFs have been isolated exclusively from mosquitoes, cISF-like sequences have been detected by molecular methods in other dipterans indicating that cISFs may not have a mosquito-restricted host range. Novel cISF-like sequences of 157 nt were detected in male Phlebotomus perniciosus in Algeria in 2006–2007. PCR products were not detected when the reverse-transcription step was excluded suggesting that a novel virus, rather than CSA, was identified. Virus isolation experiments were not performed because the sandflies had been preserved in guanidinium thiocyanate. Classical ISF-like sequences were also detected in sandflies in Spain but once again virus isolation experiments were not attempted. In another study, a 6567-nt cISF-like sequence was identified by RNA deep sequencing in chironomids (non-biting midges) in France, although again isolation of virus particles was not attempted.

2.2. In Vitro and in Vivo Replication Potential in Vertebrates and Arthropod Cells

Classical ISFs have not been isolated from any vertebrates in nature and cannot replicate in any vertebrate cell lines that have been tested; thus, these viruses are assumed to possess a vertebrate-incompetent replication phenotype. The most comprehensive in vitro host range studies were performed with PCV which was shown to lack the ability to replicate in hamster (BHK-21), human (SW-13), monkey (Vero) and porcine (PS-EK) cells, and CxFV which cannot replicate in avian (DF-1), hamster (BHK-21) or monkey (Vero) cells. Most other cISFs have been demonstrated to lack the ability to infect hamster (BHK-21) and/or monkey (Vero) cells. Attempts to infect suckling mice with AeFV and CxFV by intracerebral inoculation were unsuccessful.

Every described cISF possesses the ability to replicate in Ae. albopictus (C6/36) cells. Some cISFs induce cytopathic effect (CPE) and form plaques in C6/36 cells whereas others do not. Another determinant of whether CPE occurs is the passage history of the virus. PCV does not induce CPE in C6/36 cells after the first or second passage but often morphological changes (i.e., syncytia and vacuolation in cells) are observed by the fourth passage. Moderate CPE was periodically observed in C6/36 cells inoculated with CxFV that had been passed at least twice whereas CPE was usually absent in cells infected with the original inoculum or virus passed once. CxFV isolates from Japan do not plaque in C6/36 cells unlike isolates from Guatemala. CxFV and CFAV both reach maximum titers of approximately 107 plaque forming units (pfu)/mL in C6/36 cells while KRV produces a maximum titer of 108 pfu/mL in this cell line. The replicative potentials of select cISFs have also been assessed in other mosquito cell lines including CFAV which replicates in Ae. albopictus (AA23) and Ae. aegypti (A20) cells and KRV which replicates in Ae. pseudoscutellaris (AP-61) and Ae. aegypti cells.

In vivo experiments have been performed to characterize the replicative potential and tissue tropisms of CxFV in Culex spp. mosquitoes. CxFV establishes a systemic infection in Cx. pipiens, as indicated by the detection of viral RNA in all tissues examined (salivary glands, ovaries, testes, head, fat bodies and midguts). The presence of CxFV RNA in the salivary glands is interesting because, due to the vertebrate-incompetent replication phenotype of this virus, establishment of a salivary gland infection does not appear necessary for its persistence in nature. CxFV was not detected in the saliva of Cx. quinquefasciatus infected with CxFV alone but was detected in the saliva of mosquitoes co-infected with CxFV and WNV.

2.3. Transmission

Vertical transmission is defined as the process by which an infected female directly transmits a pathogen to her progeny. The detection of cISFs in mosquitoes of all life stages, including adults of both sexes, indicates that vertical transmission is a major mechanism by which these viruses persist in mosquitoes in nature. The initial isolations of KRV were made from Ae. mcintoshi larvae and pupae, CxFV RNA has been detected in Cx. pipiens egg rafts, larvae, adult males and adult females and AeFV was isolated from a pool of male Ae. albopictus reared to adults from field-collected larvae. Additionally, the first isolate of CFAV was obtained from the C6/36 cell line which was derived from Ae. albopictus larvae.

One mechanism of vertical transmission is transovarial transmission (TOT), defined as the process by which progeny of infected females are directly infected in the egg stage within the ovary before release and subsequent insemination. Experiments performed with field-infected Cx. pipiens revealed that TOT is an efficient mechanism for CxFV persistence. Filial infection (FI) and TOT rates of 97.4% and 100%, respectively were reported. These values are considerably greater than the <1% FI and vertical infection rates typically reported in mosquitoes infected with dual-host flaviviruses. Viral dissemination to the ovaries is necessary for TOT to occur. Accordingly, CxFV RNA was detected in the ovaries of F1 produced from field-infected Cx. pipiens. Interestingly, TOT did not occur when uninfected laboratory-colonized Cx. pipiens were infected with CxFV by needle inoculation. One explanation for the different TOT rates between the experimentally and naturally infected Cx. pipiens could be that mosquitoes with lifelong infections (i.e., vertically infected mosquitoes) are more susceptible to TOT than mosquitoes infected as adults. The mosquitoes did not possess an ovarian infection barrier because CxFV RNA was detected in their ovaries.

Vertical transmission of KRV has been demonstrated in laboratory-colonized Ae. aegypti. The FI rate in the F1 produced by the infected mosquitoes was 3.9% while the TOT rate was not reported. One likely explanation for the dramatically lower FI rate in this study as compared to the FI rate of 97.4% reported for CxFV is that there is no direct evidence to indicate the Ae. aegypti is a natural host of KRV. The virus has only been isolated from Ae. macintoshi in the field and vertical transmission is presumably more efficient in the natural mosquito host.

The contribution of venereal transmission in cISF persistence was investigated by allowing CxFV-infected male Cx. pipiens to mate with uninfected females. Reciprocal mating experiments were also performed. Virus was transmitted to 2.4%–5.3% of the mosquitoes indicating that venereal transmission has a minor role in CxFV persistence. Horizontal transmission among larvae and non-sexual contact transmission among adults were considered unlikely modes of CxFV maintenance. Efficient per os transmission of KRV has been reported for Ae. aegypti. In these studies, 62.4% of mosquitoes that fed on infectious blood were positive for KRV by virus isolation in cell culture. Virus was also isolated from 90.2% of second instar larvae exposed to KRV-infected C6/36 cells. Efficient per os infection has also been reported for Ae. aegypti exposed to Eilat virus (EILV), an insect-specific alphavirus, via infectious blood meal. Infection and dissemination rates of 63%–78% and 8%–26% were observed. Ae. albopictus, An. gambiae and Cx. quinquefasciatus were also susceptible to EILV infection, albeit at a lower rate. Studies need to be performed to assess whether efficient per os infection occurs in mosquitoes exposed to cISFs via natural food sources (i.e., nectar).

Some cISFs exhibit seasonal activity. These findings could be considered unexpected if vertical transmission was the sole mechanism for their persistence in nature. CxFV was detected in mosquitoes in Texas, U.S.A from November to March but not April to August, even though mosquitoes were abundant at these times. CLBOV was not detected year-round in Cx. pipiens and Cx. tarsalis in Colorado, U.S.A. These findings could indicate that another mode of transmission has a major role in cISF persistence. Alternatively, these findings could be a consequence of sampling biases, small sample sizes or limitations in viral detection methods.

2.4. Competitive Interaction between cISFs and Dual-Host Flaviviruses

Superinfection exclusion (or homologous interference) is the process by which host cells infected with one virus do not support productive replication of the same or similar virus. This phenomenon has been observed during infections by a broad range of viruses and can occur in both vertebrate and invertebrate hosts. Data regarding the abilities of cISFs to induce superinfection exclusion of dual-host flaviviruses in mosquito cells has been variable. Prior infection with PCV significantly reduced WNV and MVEV replication in C6/36 cells. In contrast, prior exposure to CxFV had no effect on WNV replication in C6/36 cells. In another study, WNV titers were significantly lower in CxFV-infected C6/36 cells compared to uninfected C6/36 cells at earlier, but not later, time points. The in vitro growth kinetics and yields of JEV and DENV did not differ significantly in Cx. tritaeniorhynchus cells persistently infected with CxFV when compared to cells without pre-existing CxFV infections although JEV superinfection induced severe CPE. Taken together, the above data indicate that cISFs can suppress the in vitro replication of dual-host flaviviruses in mosquito cells under some circumstances.

Vector competence experiments were performed with two colonies of Cx. pipiens, one persistently infected with CxFV and the other not, in order to evaluate the effect of CxFV on WNV transmission in this mosquito spp.. At 7 days p.i., a significantly lower percentage of CxFV-infected mosquitoes (72%) had disseminated WNV infections compared to single-virus infected mosquitoes (94%). Infection and transmission rates did not differ significantly. At 14 days p.i., WNV infection, dissemination and transmission rates did not differ significantly between the two groups. These data indicate that CxFV can suppress the in vivo replication of WNV early during infection. However, it should be noted that the mosquito colonies used for these experiments are from different geographic locations (Colorado and Iowa) and therefore, their differential susceptibilities to WNV infection could to due to factors other than co-infection with CxFV. In another study, sequential infection experiments demonstrated that prior infection with CxFV had no significant effect on WNV infection, dissemination or transmission in Cx. quinquefasciatus. Coinfection experiments demonstrated that the ability of WNV to be transmitted by Cx. quinquefasciatus after simultaneous inoculation with WNV and CxFV was strain-specific. A significantly higher percentage of co-inoculated Honduras Cx. quinquefasciatus transmitted WNV compared to mosquitoes inoculated with WNV alone. In contrast, the percentage of co-inoculated Sebring Cx. quinquefasciatus that transmitted WNV did not differ significantly from the single-virus infected control group. These experiments indicate that cISFs can enhance the transmissibility of dual-host flaviviruses under some circumstances. In this regard, a positive ecological association between CxFV and WNV was reported in field-collected Culex spp. mosquitoes in Chicago, U.S.A. in 2006. WNV-positive mosquito pools were four times more likely to be positive for CxFV compared to spatiotemporally matched WNV-negative pools.

2.5. Genome Sequencing and Phylogeny

Complete genome sequences are available for five cISFs: AeFV, CFAV, CxFV, KRV and QBV (Table 2). The prototypical isolates of these viruses possess 5’ UTRs of 91 to 113 nt, consistent with the lengths of the 5’ UTRs of most other flaviviruses. The 3’ UTRs of CFAV, CxFV and QBV are also of the expected size. However, the 3' UTRs of AeFV and KRV are unusually long. The 3' UTR of AeFV consists of 945 nt while the 3' UTR of KRV consists of 1205 nt which is approximately twice the length of a typical flavivirus 3'UTR. It has been proposed that the unusually long KRV 3'UTR resulted from an almost complete duplication of a precursor sequence. According to the Genbank database, the complete genome of NIEV has also been sequenced (Genbank Accession No. NC_024299). However, the 3' UTR of this virus is remarkably short (167 nt) and therefore, we consider it likely that the sequence is truncated at the 3' end. Of the remaining cISFs, complete polyprotein ORF sequences are available for CTFV, HANKV, NAKV and PCV. Limited sequence data are available for AGFV and CLBOV; 556 and 946 nt of their NS5 genes have been sequenced, respectively.

The codon and dinucleotide usage preferences of cISFs are consistent with their apparent vertebrate-incompetent replication phenotype. Vertebrates and invertebrates preferentially have certain codon and dinucleotide usage biases, and studies performed with RNA viruses have shown that their preferences often mimic those of their hosts. Vertebrates display a strong under-representation of UpA and CpG, and over-representation of UpG and CpA. Mosquitoes also display a strong under-representation of UpA but have no bias for CpG depletion or for UpG and CpA excess. A comparison of the dinucleotide usage preferences of representative viruses from the cISF, NKV and dual-host groups (CxFV, MODV and WNV) revealed that MODV and to a lesser extent WNV demonstrate a CpG decrease while CxFV has no bias against this dinucleotide. All three viruses demonstrate an underutilization of UpA. Similar observations were reported when the dinucleotide usage preferences of CFAV, CxFV and KRV were compared to that of multiple NKV and dual-host flaviviruses.

Classical ISFs are phylogenetically distinct from all other known flaviviruses (Figure 1). These viruses currently separate into two main clades (Figure 2). Clade 1 is composed of cISFs usually associated with Aedes spp. mosquitoes (AEFV, AGFV, CFAV and KRV). Subclade 2 contains Culex-associated viruses (CLBOV, CTFV CxFV, NIEV and QBV) in addition to NAKV and PCV, which were isolated from Mansonia and Coquillettidia spp. mosquitoes, respectively. HANKV is highly divergent from both clades 1 and 2 and may be regarded as forming a third clade. Although this virus has been detected in Culex spp. mosquitoes, it is more frequently associated with Aedes spp. mosquitoes.

2.6. Ribosomal Frameshifting

Programmed -1 ribosomal frameshifting (-1 PRF) is the process by which specific mRNA sequences induce a proportion of ribosomes to shift -1 nt and continue translating in the new reading frame. PRF is utilized by many RNA viruses to control gene expression and to increase the number of protein products that can be expressed from a limited number of mRNA transcripts. The eukaryotic -1 frameshift site usually consists of a ‘slippery’ heptanucleotide fitting the motif X XXY YYZ (where XXX normally represents any three identical nucleotides although certain exceptions such as UCC, GGA, GUU and GGU also occur; YYY represents AAA or UUU; Z represents A, C or U; and spaces separate zero-frame codons), followed by a 5 to 9 nt ‘spacer’ region and then a stable RNA secondary structure such as a pseudoknot or stem-loop.

Previous studies have provided evidence that all known cISFs utilize -1 PRF to express a novel overlapping gene (designated fifo) in the NS2A-NS2B regions of their genomes. However, CTFV, HANKV, NIEV and PCV had not been discovered when this conclusion was reached. To investigate whether -1 PRF is a universal feature of cISFs, the nucleotide sequences of all cISFs for which NS2A-NS2B data are available (as of 22 January 2015) were analyzed. All sequences were shown to possess a heptanucleotide sequence that conforms to the requirements of a ‘slippery’ -1 PRF motif (Figure 3A) and a downstream fifo ORF ranging from 221 (NIEV) to 293 (CxFV) codons. The exception to this is the laboratory-adapted isolate of CFAV; the fifo ORF of this isolate is disrupted by three premature termination codons suggesting that the gene is dispensable for in vitro replication. In addition to the slippery heptanucleotide frameshift site sequence, all cISF sequences were found to contain a potential stem-loop (CxFV and HANKV clades) or pseudoknot (CFAV clade) structure at the appropriate spacing downstream of the slippery heptanucleotide (Figure 3A and Figure 4A). It is interesting to note that this overlapping gene is unique to viruses in the cISF group; it is not encoded by the genomes of any other flaviviruses. Nevertheless, -1 PRF is utilized by various other flaviviruses. Apparently all viruses in the Japanese encephalitis (JE) serogroup, except for SLEV, utilize efficient -1 PRF to produce a larger NS1-related protein (NS1’) and to reduce synthesis of the 3'-encoded non-structural proteins relative to the proteins encoded upstream of the frameshift site. -1 PRF has also been predicted to occur in many dISFs (see Section 3f) and in Wesselsbron and Sepik viruses.

Immunoflorescence assays performed using polyclonal antibodies raised against two predicted CxFV FIFO antigens detected a protein product in CxFV-infected, but not mock-infected, C6/36 cells. However, bands of the expected size were not detected when lysates from CxFV-infected cells were analyzed by Western blot using these same antibodies. Thus, it is not known whether fifo is expressed as a frameshift fusion simply with the N-terminal few amino acids of NS2A (i.e., NS2AN-FIFO) or, as in the case of the JE serogroup NS1' protein, as a fusion also with NS1 (i.e., NS1-NS2AN-FIFO). It is also possible that the fusion products are internally cleaved. Additional work is needed to investigate the expression and functional relevance of the fifo product in cISFs.

2.7. Predicted Polyprotein Cleavage Sites

The predicted proteolytic cleavage sites of all cISFs for which complete polyprotein ORF data are available are shown in Table 3. For the most part, these sites conform to the rules established for dual-host flaviviruses although there are some exceptions. Studies performed with dual-host flaviviruses have revealed that a host signal peptidase mediates cleavage between C/prM, prM/E, E/NS1 and 2K/NS4B and that these junctions typically conform to predicted signalase cleavage sites. Similar sites were identified at the predicted C/prM, prM/E, E/NS1 and 2K/NS4B junctions of most cISFs. The NS1/NS2A cleavage site of dual-host flaviviruses is signalase-like with respect to the '-1, -3' rule, but an upstream hydrophobic domain is absent. Previous work has demonstrated that, at least in DENV, cleavage requires translation of substantial parts of NS2A and it has been proposed that the hydrophobic domains in NS2A lead to a conformation that presents the NS1/NS2A cleavage site to an endoplasmic reticulum-resident host protease that may be signalase. In the dual-host flaviviruses, NS1/NS2A cleavage usually occurs after a Val-X-Ala site. Although the predicted cISFs cleavage sites generally lack one or other of these residues, they are still compatible with the signalase '-1, -3' rule. In dual-host flaviviruses, the cellular protease furin cleaves prM to generate the mature form of the protein. Furin normally cleaves after the motif Arg-X-Lys/Arg-Arg but cleavage can also occur after Arg-X-X-Arg. As for dual-host flaviviruses, the predicted pr/M junction of every cISF is preceded by RXKR or RXRR, except for HANKV and NAKV which contain only the minimal furin cleavage site RXXR.

The virion C/anchor, NS2A/NS2B, NS2B/NS3, NS3/NS4A, NS4A/2K and NS4B/NS5 junctions of dual-host flaviviruses are cleaved by the viral NS2B/NS3 serine protease, which normally cleaves after two basic amino acid residues (KR, RR, RK) or sometimes after QR at the P2 and P1 positions, followed by a small amino acid (G, A or S) at the P'1 position. The corresponding cISF cleavage sites are not always obvious and frequently appear to deviate from these motifs. It should be noted that only two cISF cleavage sites have been experimentally determined (viz. CFAV anchorC/prM and prM/E) both of which are signalase rather than viral protease cleavage sites. Alignment between cISF and dual-host or dISF flavivirus sequences at the NS2B/NS3, NS3/NS4A, NS4A/2K and NS4B/NS5 junctions suggests that cleavage in cISFs occurs between R at the P1 position and G, A or S at the P'1 position, but there seems to be substantial flexibility at the P2 position (Table 3). The exact cleavage site at the virion C/anchor junction was difficult to predict due to a cluster of basic amino acids; most cISFs contain a potential Arg-Gly/Ala/Ser (P1-P'1) cleavage site at the C-terminal end of the cluster of basic residues, while a few may use alternative motifs in this region. Prediction of the NS2A/NS2B cleavage site in cISFs was particularly problematic: while some species (e.g., CxFV) contain several Arg-Gly/Ala/Ser (P1-P'1) motifs in the critical region between two predicted transmembrane regions (aligning to the corresponding NS2A/NS2B junction in dual-host and dISF flaviviruses), other species contained no such motifs in this region. A conserved Arg residue was annotated as a potential cleavage site in Table 3, notwithstanding that in CFAV, KRV, AEFV and NAKV it is followed by Asp, Asp, Thr and Lys, respectively, while in several species (including AEFV) there are closely spaced alternative cleavage sites. It should be noted that, due to the additional constraints imposed on this sequence region in the cISFs as a result of the overlapping fifo ORF, it is possible that NS2A/NS2B cleavage in the cISFs may have evolved to take place at a somewhat different location than we have inferred from comparison to other flavivirus sequences.

In some instances, the predicted cleavage sites listed in Table 3 are different from those reported by others. For example, the NS2A/NS2B cleavage sites that we proposed for AEFV, CFAV and KRV are located 4 to 25 residues upstream of the sites previously proposed for these viruses. Our predicted NS4B/NS5 cleavage sites are located one residue upstream of the sites originally proposed for AEFV, CFAV and KRV. Our analysis was performed using additional cISF sequences, thus facilitating the identification of conserved sites. Nevertheless, our data are subjective and experimental data (e.g., amino acid sequencing) are needed to conclusively identify the cleavage sites in the polyproteins of cISFs.

3.1. Discovery, Geographic Distribution and Natural Host Range

Dual-host affiliated ISFs are not as well characterized as cISFs. The first dISF reported in the literature was Nounané virus (NOUV) after its isolation from Uranotaenia mashonaensis in Côte d'Ivoire in 2004 (Table 4). At least eight other dISFs have since been discovered: Barkedji virus (BJV) in Senegal (Genbank Accession No. EU078325) and Israel, Chaoyang virus (CHAOV) in South Korea and China, Donggang virus (DONV) in China (Genbank Accession No. NC_016997), Ilomantsi virus (ILOV) in Finland, Lammi virus (LAMV) in Finland, Marisma mosquito virus (MMV) in Spain and Italy, Nanay virus (NANV) in Peru and Nhumirim virus (NHUV) in Brazil. Evidence indicates that another dISF may occur in Kenya. A region of the NS5 gene of the virus was sequenced and shown to have greatest (77%) nucleotide identity with the corresponding region of CHAOV. The sequence has not been deposited into the Genbank database and therefore cannot be compared with those of the more recently discovered dISFs in order to establish whether it is indeed a novel virus. Collectively, dISFs have been isolated from or detected in at least 12 species and five genera of mosquitoes (Table 4). Dual-host affiliated ISFs have not been isolated from or detected in sandflies, midges or any other non-mosquito dipterans. It is unclear whether dISFs truly have a mosquito-restricted host range because other dipterans are rarely screened for these viruses. Additional work should be done to address this issue.

3.2. In Vitro and in Vivo Replication Potential in Vertebrates and Arthropods

By definition, dual-host affiliated ISFs have not been isolated from any vertebrates in nature and data from laboratory experiments indicate that they possess a vertebrate-incompetent replication phenotype. Suckling mice intracerebrally inoculated with ILOV, LAMV or NANV displayed no signs of illness and viral RNA was not detected in brains harvested from these animals. The most comprehensive in vitro experiments were performed by Huhtamo and colleagues who demonstrated that LAMV does not replicate in avian (primary chicken), canine (MDCK), hamster (BHK-21), human (SH-SY5Y, Hep, MRC-5, SW13, HEK293), monkey (Vero, BGM, MA104), mouse (Neuro 2A, L929), porcine (PK-15), snake or toad (XTC) cells. Selected mammalian cells were also incubated at lower temperatures (27 °C and 33 °C) and the virus was still unable to replicate. The replicative abilities of several other dISFs have also been assessed in multiple vertebrate cell lines including ILOV which could not replicate in canine, hamster, monkey, mouse, human, snake or toad cells and NOUV which could not replicate in avian, hamster, human, monkey or porcine cells. It has been suggested that transient replication of MMV occurs in Vero and BHK-21 cells. Cytopathic effect (CPE) was observed and viral RNA was detected at 5 to 7 days post-inoculation. However, neither CPE nor viral RNA was evident after the first passage and thus, the cells could not support prolonged MMV replication. One limitation of this study is that the authors did not show an increase in viral genome copy number by quantitative RT-PCR. The experiments were performed using conventional RT-PCR and therefore it is not clear whether the detected viral RNA was due to active virus replication or the presence of viral RNA in the original inoculum. The CPE may have been caused by an unrecognized virus in the original inoculum.

The replicative abilities of several dISFs have been assessed in multiple arthropod cell lines. LAMV and ILOV replicate efficiently in Ae. albopictus (C6/36 and AA23) and Ae. aegypti (AE and A20) cells. NHUV replicates efficiently in Ae. albopictus (C6/36 and C7/10) and Cx. quinquefasciatus cells but not Ixodes scapularis tick (ISE6) cells. CPE was first observed in NHUV-infected C6/36 cells at 3 and 6 days p.i. after the initial inoculation and second passage, respectively. MMV, NANV and NOUV are also known to infect C6/36 cells, with CPE observed in cells inoculated with the latter two viruses. CHAOV also replicates in C6/36 cells but there are conflicting reports as to whether or not it induces CPE. In vivo experiments have not been performed to assess the infection, replication and dissemination abilities of dISFs in mosquitoes or any other arthropods.

3.3. Transmission

The mechanism by which dISFs are maintained in nature is not known. Their phylogenetic placement with mosquito/vertebrate flaviviruses indicates that these viruses are (1) dual-host viruses with an unidentified vertebrate host or (2) insect-specific viruses that recently lost the ability to infect vertebrates. Given that the replicative abilities of several dISFs have been assessed in numerous vertebrate cell lines with none of the cell lines able to support prolonged virus replication, the second explanation appears more likely. However, there are no other experimental data to support this. All dISFs have been isolated from adult female mosquitoes or from mosquitoes of unspecified genders and life stages. To shed light on whether dISFs are maintained by vertical transmission, immature and adult male mosquitoes need to be screened for these viruses. Moreover, in vivo mosquito infections similar to those described in 2c need to be performed to determine the importance of vertical, horizontal and mechanical transmission in dISF maintenance.

3.4. Competitive Interaction between dISFs and Dual-Host Flaviviruses

Data on the ability of dISFs to suppress the replication of dual-host ISFs in mosquito cells are limited. In vivo experiments have not been performed and only one study has assessed the in vitro superinfection exclusion potential of a dISF. Briefly, prior or concurrent inoculation of mosquito (C6/36) cells with NHUV significantly suppressed the replication of WNV, JEV and SLEV. The effect was most pronounced in cells inoculated with WNV and SLEV; the peak viral titers of these viruses were reduced 106- and 104-fold, respectively. JEV exhibited an 80-fold reduction in peak titer. Titers were significantly reduced as early as 1 day p.i. (WNV and JEV) and 2 days p.i. (SLEV).

3.5. Genome Sequencing and Phylogeny

Complete genomic sequences are available for three dISFs: CHAOV, DONV and NHUV (Table 5). Complete polyprotein ORF sequences are available for ILOV, LAMV and NOUV and nearly all of the BJV ORF has also been sequenced. Partial E and/or NS5 sequences are available for MMV and NANV. The 5’UTRs of CHAOV, DONV and NHUV are similar in length (99, 113 and 102 nt, respectively) to those of most other flaviviruses. The 3’ UTR of NHUV (451 nt) is within the size range usually reported for flaviviruses while the 3’ UTRs of CHAOV and DONV (326 and 343 nt, respectively) are slightly shorter than expected.

A comparison of the dinucleotide usage preferences of NHUV to representative flaviviruses from the cISF, NKV and dual-host groups (CxFV, MODV and WNV, respectively) revealed that the CpG usage of NHUV was similar to that of WNV whereas, as mentioned in Section 2.5, CxFV has no strong bias against this dinucleotide. These findings could indicate that NHUV is not a mosquito-specific virus but instead a dual-host virus with a yet-to-be-determined vertebrate host. Another explanation is that dISFs lost their ability to replicate in vertebrates relatively recently and still possess dinucleotide usage preferences that partly reflect the dual-host precursor viruses from which they evolved. In support of this interpretation, on average dISFs have higher CpG usage than dual-host flaviviruses though lower usage than cISFs (Figure 5).

As already noted, dISFs phylogenically affiliate with mosquito/vertebrate flaviviruses (Figure 1).

These viruses cluster within the mosquito-borne flavivirus clade. CHAOV, LAMV, DONV and ILOV form a distinct clade. NHUV, BJV and NOUV form another clade. However it is not clear that these viruses can be grouped together into a single monophyletic clade. Thus, if dISFs are truly insect-specific viruses that lost the ability to infect vertebrates, this topological arrangement suggests that loss of vertebrate host happened separately for each of these two clades.

3.6. Ribosomal Frameshifting

For members of the dISF clade comprising CHAOV, LAMV, DONV and ILOV, -1 PRF has been predicted to occur within the NS2B-encoding region of the genome. A slippery heptanucleotide and 3' predicted stem-loop structure are phylogenetically conserved within the group (Figure 3B and Figure 4B) and the sequence elements have been shown to stimulate frameshifting in a reporter construct. Frameshifting would result in an alternative version of NS2B with a different C-terminal tail encoded by the -1 reading frame.

In NOUV, a potential frameshift site (U UUU UUA with a 3'-adjacent predicted stem-loop structure) was identified in the NS2A-encoding region, and the sequence was shown to stimulate frameshifting in a reporter construct. Frameshifting at this site would yield a truncated version of NS2A. The site is not conserved in BJV or NHUV, but a different potential frameshift site (again U UUU UUA with a 3'-adjacent predicted stem-loop structure) occurs in NHUV at a site around 25 codons upstream (again within NS2A). Between these two sites, BJV has a potentially slip-prone U CCU UUU sequence, but only a weak 3'-adjacent stem-loop structure was predicted. It remains unclear whether the NOUV, NHUV and BJV sites are chance features or functionally relevant -1 PRF sites.

3.7. Predicted Polyprotein Cleavage Sites

The predicted proteolytic cleavage sites in the polyproteins of all dISFs for which complete or almost complete ORF data are available are shown in Table 6. It is interesting to note that the predicted cleavage sites of these viruses more closely resemble those of dual-host flaviviruses than those of the cISFs. For instance, nearly all dISFs appear to utilize a Val-X-Ala cleavage site at the NS1/NS2A junction consistent with dual-host flaviviruses. The exception to this rule is BJV which has Thr instead of Val at the -3 position. In contrast, cISFs tend to use more varied signalase-like motifs at this site. Dual-host flaviviruses usually contain residues KR, RR, RK or QR immediately upstream, and a G, A or S residue immediately downstream of the virion C/anchor, pr/M, NS2A/NS2B, NS2B/NS3, NS3/NS4A, NS4A/2K and NS4B/NS5 cleavage sites. Every dISF conforms to this rule at all of the aforementioned junctions with the exception of BJV at the NS2B/NS3 junction where cleavage is predicted to occur between QK and A.

4. Closing Remarks and Future Research

Historically, ISFs have received limited attention due to the assumption that they have minimal, if any, impact on human health by virtue of their apparent insect-specific phenotype. However, there are several important reasons why these viruses should not be neglected. Comparative studies on insect-specific and dual-host flaviviruses will provide unique insight into why some flaviviruses are major human pathogens while others possess a vertebrate-incompetent replication phenotype. For instance, the generation and subsequent characterization of chimeric viruses of insect-specific and dual-host flaviviruses will result in the identification of the genetic elements that modulate the differential host ranges, transmission cycles and tropisms of these viruses. The characterization of chimeras of NKV and dual-host flaviviruses will build upon these studies via the identification of elements that preclude flaviviral replication in the mosquito host. Several groups have begun to address this issue through the characterization of NKV and dual-host flaviviral chimeras but no ISF/dual-host flaviviral chimeras have been described.

Some ISFs possess the ability to enhance or suppress the replication of dual-host flaviviruses in co-infected mosquitoes; thus, these viruses warrant further investigation due to the indirect impact that they have on human health. Another reason why it is important to increase our understanding of ISFs is because these viruses have the potential to be used to control mosquito populations and impede disease transmission. Genes that are essential for mosquito development, reproduction or longevity, or dual-host flaviviral infection, dissemination or transmission, could be silenced via the release of recombinant ISFs engineered to express specific genes or sequences. Only one ISF infectious cDNA clone has been described and most ISFs have not been fully sequenced. The availability of additional genomic sequence data and infectious cDNA clones would assist in this area of research. However, caution is required if recombinant ISFs are to be released in the environment because, as noted above, some cISFs possess the ability to enhance the replication of dual-host flaviviruses in coinfected mosquitoes. A final reason why ISFs should be further investigated is because they may have the potential to evolve into major human pathogens.

The identification and characterization of ISFs is dependent upon the availability of reliable detection tools, including ISF-specific monoclonal antibodies (MAbs). ISF-specific MAbs are available for one virus although some polyclonal antibodies raised against dual-host flaviviruses cross-react with ISFs. The production of additional ISF-specific MAbs would assist in the future identification and characterization of these viruses. Future research should also investigate the mechanism(s) by which dISFs are maintained in nature. Although the significance of vertical (i.e., transovarial and venereal) and mechanical transmission in cISF persistence has been assessed, similar studies have not been done with dISFs. The screening of immature or adult male mosquitoes could easily be incorporated into dISF identification and surveillance studies, and the resulting data would provide important insight into the role, if any, of vertical transmission in dISF persistence. Research should also be done to investigate the significance of the overlapping fifo ORF uniquely encoded in the NS2A-NS2B region of the cISF genome, and shorter predicted ORFs in similar regions of some dISF genomes. Because fifo is not encoded by any vertebrate-infecting flaviviruses, this gene could have a key role in the apparent ability of ISFs to be maintained in nature in the absence of a vertebrate host. For instance, fifo could have a pivotal role in TOT. Research is also needed to determine whether ISFs infect non-mosquito dipterans.

ISFs are presumably under-represented in the literature and sequence databases compared to their vertebrate-infecting counterparts. One explanation for this presumed under-representation is because some ISFs cause minimal (if any) CPE in mosquito cells and none possess the capacity to replicate in mammalian cells. Therefore, ISFs could easily remain undetected in virus surveillance studies in which mosquitoes are screened for viruses by suckling mouse brain inoculation or by virus isolation in Vero or C6/36 cells. Many dual-host flaviviruses were initially discovered as a result of their association with outbreaks of human disease. In contrast, ISFs are not associated with disease outbreaks due to their vertebrate-incompetent replication phenotype.

To summarize, additional research on ISFs is desirable because it will provide unique insight into medically important flaviviruses and has the potential to lead to the development of efficient vector and disease control strategies. Despite the dramatic increase in the number of ISFs identified in recent years, as well as the sudden increase in ISF-related research, it is likely that many undiscovered ISFs occur in nature and there are clearly many areas of ISF research that have not been pursued.