5′- and 3′-noncoding regions in flavivirus RNA

Publisher Summary

I. Introduction

The major human pathogens among the flaviviruses are yellow fever (YF) virus, the four serotypes of dengue (DEN) viruses, Japanese encephalitis (JE) virus, and the tick-borne encephalitis (TBE) viruses. Flaviviruses contain a positive-stranded RNA genome that is approximately 10,500 nucleotides in length. The RNA contains a type 1 cap at its 5′ terminus (Cleaves and Dubin, 1979) but is not polyadenylated (Wengler and Wengler, 1981), and it encodes a single, long open reading frame (ORF) (reviewed in Chambers et al., 1990). All evidence indicates that virus-specific proteins are derived by co- or posttranslational cleavage of the polyprotein encoded by the ORF. These include three structural proteins [capsid (C), premembrane (preM), and envelope (E)] and at least seven nonstructural (NS) proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5, in that gene order). Most cleavages of the polyprotein are mediated by the host cell enzyme signal peptidase or by a viral protease complex requiring a central hydrophilic domain in NS2B and the amino terminus of NS3 (Falgout 1991, Chambers 1993). Genomic RNA contains noncoding regions (NCR) 5′ and 3′ to the ORF. The 5′-NCR is about 100 nucleotides in length, and the 3′-NCR is 400 to nearly 800 nucleotides in length (Table I ). The conserved structural and nucleotide sequence elements of these NCRs and their function in RNA replication and translation are the subjects of this review.

The NCRs in genomic RNA must be involved in the initiation of negative-strand synthesis and in any processes related to switching from negative-strand synthesis to the production of progeny virion RNAs. NCRs may also play a role in the packaging of nascent virions and contain nucleotide sequences required to initiate translation. Of the viral proteins, NS1, NS2A, NS3, NS4A, and NS5 have been implicated directly or indirectly in RNA replication. The role of NS1 in this process is not defined, but mutations in NS1 have been shown to affect the initiation of minus-strand synthesis (Lindenbach 1997, Lindenbach 1999, Muylaert 1997). NS2A is a small hydrophobic protein localized to the replication complex that may target it to membrane organelles (Mackenzie et al., 1998). NS3 contains an RNA helicase and nucleotide triphosphatase activities in addition to its protease activity (Wengler 1991, Warrener 1993, Utama 2000), and NS4A may form a required association with NS1 during replication (Lindenbach and Rice, 1999). NS5 functions as an RNA-dependent RNA polymerase (Tan 1996, Ranjith-Kumar 2001) and also contains a potential methyl-transferase activity, possibly required for the capping of progeny viral genomes (Koonin, 1993).

The currently accepted model for flavivirus RNA replication is a semiconservative one. Input genome RNA must first be copied to form full-length minus strands in order to generate stable double-stranded replicative forms (RF). Free minus-strand genomic RNA is not detected (Chu 1985, Khromykh 1997), and the plus and minus RNA strands in the RF appear to be exact copies of each other, such that there are no overhanging unpaired nucleotides at the 3′ termini of either strand (Wengler and Wengler, 1981). Plus-strand RNA synthesis is thought to proceed in replicative intermediates, using RFs as a template. No subgenomic virus-specific RNAs have been detected in flavivirus-infected cells, in support of the concept that the genomic plus-strand RNA is the only virus-specific mRNA. The RNA replication process appears to be localized to perinuclear membrane organelles, probably in complexes containing nascent RNA and NS proteins in association with cellular proteins (Chu 1985, Chu 1987, Ng 1989, Westaway 1999).

Table I shows a list of the lengths of the respective 5′ and 3′ NCRs in representative flavivirus genomes. The total number of nucleotides in each genome is also shown for comparison. Flaviviruses have been speciated, and species have been subgrouped, by results of virus cross-neutralization assays (Calisher et al., 1989). The results of such assays depend on antigenic differences among the respective envelope glycoproteins of the different viruses. Seven subgroups were thus defined, but at least 15 viruses remained ungrouped, including YF virus. For example, the four serotypes of dengue viruses form one subgroup or “complex” among the mosquito-borne flaviviruses. Similarly, the Japanese encephalitis virus complex includes JE, West Nile (WN), and Murray Valley encephalitis (MVE) viruses. The tick-borne flaviviruses (e.g., TBE virus strains Sofjin and Neudorfl and Powassan virus, see Table I) form a complex that is serologically distinct from the mosquito-borne flaviviruses. Cell-Fusing Agent virus (CFA) was isolated from cultured mosquito cells and appears to have no serological relationship to other flaviviruses in the cross-neutralization assay, has no vector in nature, and causes no illness in humans or animals (Cammisa-Parks et al., 1992).

Flaviviruses have also been grouped by phylogenetic analysis, based on sequence data obtained for a 1000-nucleotide cDNA generated from the 3′ termini of the NS5 genes of several dozen isolates (Kuno et al., 1998). The hierarchical levels of relatedness defined by this technique were organized in descending order as follows: “cluster,” “clade,” and “species,” where a clade was defined as a group of viruses that share 69% or higher pairwise nucleotide sequence identity among the members. Results of the analysis for grouping flaviviruses in clades were quite comparable to those obtained by the cross-neutralization assay as it was used to define antigenic complexes or subgroups, suggesting a parallel evolution of genome sequences and antigenic character. These findings will be seen to be relevant as we discuss differences in 5′ and 3′ NCR nucleotide sequences among viruses of different antigenic subgroups or clades.

Data for viruses listed in Table I show that the 5′ NCRs of flavivirus RNAs are relatively short compared to the 5′ NCRs of other positive-strand RNA viruses. Lengths of the 5′ NCR sequences are between 89 nucleotides, for dengue serotype 2 (DEN2) strain New Guinea C (NGC), and 132 nucleotides, for TBE strain Sofjin. Tick-borne flaviviruses (strain Sofjin, strain Neudorfl, and Powassan) tend to have slightly longer 5′ NCRs than mosquito-borne viruses. For comparison, the 5′ NCR of a typical isolate of the hepacivirus, hepatitis C virus, is 341 nucleotides in length (Fukushi et al., 1994), whereas that of a picornavirus, poliovirus type 1, is 742 nucleotides in length (Rezapkin et al., 1998). This discrepancy in length of the 5′ NCRs of these otherwise related positive-strand RNA virus genomes is due primarily to the fact that the hepatitis C and poliovirus 5′ NCRs each contain an internal ribosome entry site (IRES) required for translation initiation (Wang 1995, Brown 1992, Bergamini 2000, Poyry 2001). IRESs are fairly large tRNA-like structures composed of hundreds of nucleotides. In contrast, translation of flavivirus RNAs is thought to initiate by ribosome scanning from the 5′ cap structure (Ruiz-Linares, 1989) so possibly there is an adaptive advantage related to having a short 5′ NCR.

DEN virus genomes had the shortest 3′ NCRs among the selection of flavivirus genome sequences surveyed, ranging from 384 nucleotides for DEN4, strain 814669, to 466 nucleotides, for DEN1, strain WP74. The 3′ NCRs in viruses of the JE group shown in Table I ranged from 573 to 631 nucleotides in length. The length of the YF vaccine strain 17D 3′ NCR was in between that of the DEN and JE complex viruses at 511 nucleotides. The 3′ NCRs of the TBE viruses, other than strains Neudorfl and Sofjin, vary greatly in length due to the presence of long poly(A) tracts in some genomes that are not present in others and due to deletions of sequences flanking the poly(A) tract within the variable region (see later for details). Many dozens of complete genome sequences of flavivirus have been entered in Genbank. In no case do the lengths of the 5′ and 3′ NCRs vary significantly from those noted for viruses listed in Table I.

II. Conserved LinearSequenceFeatures ofNCRs in FlavivirusRNA

A. Complementary Nucleotides at the 5′ and 3′ Termini of NCRs

The first complete flavivirus genome sequence to be published was that of the YF virus, strain 17D (Rice et al., 1985). The authors noted that the extreme 5′- and 3′-terminal sequences of the YF genome were partially homologous to those of the WN virus, which had been determined separately (Wengler and Wengler, 1981), and that for both genomes (i) the 5′-terminal two nucleotides of the plus-strand sequence (5′-AG) were complementary to the 3′-terminal two nucleotides (UC-3′) and (ii) there was a short (five-nucleotide) region of identity between a segment near the 3′ termini of plus-strand RNAs and the predicted 3′ termini of minus-strand RNAs for each of the two virus genomes due to the presence of a complementary sequence in the 5′ NCR of the plus strand (the sequence 5′-UGUGU in the 5′ NCR and the sequence 5′-ACACA in the 3′ NCR, as shown in Fig. 1 ). For the YF genome, the sequence ACACA initiated at the 13th nucleotide from the 3′ end of the RNA and spanned nucleotides –13 through –9 (reading in the 3′ to 5′ direction is indicated by a minus sign), and for the WN genome the ACACA sequence spanned nucleotides –11 through –6 (Fig. 1). The identical five-nucleotide sequences near the 3′ termini of both plus- and minus-strand RNAs were posited as recognition sites for the viral replicase that could be used in common during either positive- or negative-strand RNA synthesis (Rice et al., 1985).

Open in a separate window

Fig 1

Nucleotide sequences of 5′ and 3′ termini of a selection of mosquito-borne flavivirus positive-strand (+) genome RNAs are shown aligned with the nucleotide sequences of 3′ and 5′ termini of the respective minus (−) strands, predicted to result when the positive strand is used as a template to form a replicative intermediate during virus replication. Dotted lines indicate the intervening nucleotides of the respective full-length genomes. Flavivirus conserved 5′- and 3′-terminal dinucleotides and their complementary nucleotides are boxed. The semiconserved pentanucleotide sequence ACACA is shown in bold in the 3′ termini of positive- and negative-strand RNAs, respectively, where present. Genome sequences are as follows: YF, yellow fever (Rice et al., 1985); WNV, West Nile (Wengler and Wengler, 1981); MVE, Murray Valley encephalitis (Genbank number NC00943); DEN3, dengue serotype 3 (Genbank number NC001475); KUN, Kunjin (Genbank numbers {"type":"entrez-nucleotide","attrs":{"text":"L24511","term_id":"403463","term_text":"L24511"}}L24511 and {"type":"entrez-nucleotide","attrs":{"text":"L24512","term_id":"403464","term_text":"L24512"}}L24512); and JE, Japanese encephalitis (Genbank number NC001437).

An examination of a selection of genome sequences for other flaviviruses reveals that the complementary pairs at the ends of the RNA (5′-AG⧸UC-3′) are completely conserved among mosquito- and tick-borne viruses. The CFA virus genome is unique among the flaviviruses in that it terminates with GC-3′ and its 5′-terminal two nucleotides (5′-GU) are not complementary. The survey also revealed that the sequence ACACA and its complement are not well conserved in analogous positions in 3′ and 5′ NCRs, respectively, of flavivirus genomes (Fig. 1). For example, JE and Kunjin virus genomes do contain such sequences in the predicted locations, whereas the related MVE virus genome does not, and viruses of the DEN subgroup do not. Therefore, the significance of the “ACACA-type” sequences near the 3′ termini of plus and minus strands of the YF, WN, JE, KUN, and perhaps other flavivirus genomes must be regarded as uncertain, especially since the functionality of other better conserved segments of the NCRs in RNA replication has been demonstrated.

B. Cyclization Sequences

The most conserved linear sequence feature of the 5′ termini of flavivirus genomes is the cyclization sequence (5′CS in Fig. 2 ; Hahn 1987, Khromykh 2001). The motif is so named because of the existence of a complementary sequence in the 3′ NCR. Base pairing between 5′ and 3′ end sequences would in theory permit the formation of cyclized plus strands, perhaps during the initial phases of RNA replication. This conformation was initially predicted to permit the viral replicase to bind to both ends of the template RNA simultaneously in order to assure the generation of full-length copies (Hahn et al., 1987). Similar “panhandle” structures are thought to form, for example, during the replication of alphavirus (Hsu et al, 1973) and bunyavirus (Hewlett et al., 1977) RNAs and of phage QB RNA (Blumenthal and Carmichael, 1979).

Open in a separate window

Fig 2

Some nucleotide sequence features of 5′ and 3′ NCRs of mosquito-borne flavivirus genomes are represented. Horizontal lines represent nucleotide sequences of DEN2, DEN4, JE, MVE, WN, and YF genome 5′ NCRs (aligned in vertical order on the left) and 3′ NCRs (aligned in vertical order on the right). Upward vertical ticks in the 5′ NCR indicate the location of the 5′-most start codon for translation of the single ORF in the flavivirus genome. Downward vertical ticks indicate the location(s) of translational stop codons at the 5′ ends of 3′ NCRs. The flavivirus-conserved stem and loop secondary structure predicted to form at the 3′ end of all genomes is indicated by a loop at the end of the linear sequence (SL). The relative locations of the conserved eight nucleotide cyclization sequences, 5′CS and CYC, respectively, at the 5′ and 3′ ends of the genome are indicated by open boxes with cross-hatching. The relative location of the conserved sequence, CS2, in all genomes is indicated by a stippled box, and the location of the CS2 tandem repeat sequence, RCS2, in all but the YF genome is indicated by a second stippled box upstream from CS2. Similarly, the relative locations of the conserved tandem repeat sequences, CS3 and RCS3, are indicated by black, cross-hatched boxes in genomes of viruses of the JE subgroup. The locations of three tandem repeat sequences in the upstream portion of the YF genome 3′ NCR are indicated by open rectangles.

For mosquito-borne virus genomes, the 5′ cyclization sequence is known as the “5′CS” [indicating the 8 nucleotide sequence UCAAUAUG, spanning nucleotides +137 through +144 of the Kunjin virus (KUN) genome, for example, where the plus sign indicates nucleotides are numbered from the 5′ terminus]. The 5′CS actually lies within the ORF, about 34 to 40 nucleotides downstream from the start codon (Fig. 2). For tick-borne virus genomes, two potential 5′ cyclization sequences have been identified (Mandl 1993, Khromykh 2001), here designated “C1” and “C2” (Fig. 3 ). C1 and C2 differ markedly in linear sequence from each other and from that of the 5′CS in mosquito-borne virus genomes (Table II ). Tick-borne virus 5′-NCRs contain a very short ORF that is terminated with a stop codon within the NCR (Fig. 3). The most 5′-terminal potential cyclization sequence (C1; the 11 nucleotide sequence GGAGAACAAGA, spanning nucleotides +115 through +125 for TBE viruses) lies downstream from this mini-ORF and wholly within the 5′ NCR (Mandl et al., 1993). C2 (the 11 nucleotide sequence GGGGCGGUCCC, spanning nucleotides +164 through +174 for TBE) lies downstream from C1, within the true ORF, in a position analogous to that of the 5′CS in mosquito-borne virus genomes (Khromykh et al., 2001).

Open in a separate window

Fig 3

Some nucleotide sequence features of 5′ and 3′ NCRs of TBE virus strain genomes are represented. The 5′ NCR of TBE virus is represented on the left as a horizontal line. The start codon for translation of a very short ORF in the 5′ NCR is represented by a single upward vertical tick, numbered one (1); the start codon for translation of the long ORF in TBE genomes is represented by an upward tick, numbered two (2). Stop codons for the short ORF in the 5′ NCR and for the long ORF at the 5′ end of the 3′ NCR are indicated by a downward vertical tick. Four of six different possible nucleotide sequence motifs for 3′ NCRs of TBE viruses are indicated on the right as horizontal lines (adapted from Wallner et al., 1995). Motif A, strains Neudorfl and 263; B, strain Ljubljana I; C, strain 132; and D, strain RK1424. The flavivirus-conserved 3′ stem and loop structure is indicated by a loop in the horizontal line and is labeled “SL.” The location of the restriction endonuclease site AgeI in TBE genomic DNA is indicated by an arrow defining the boundary between the variable region and the core element in the strain Neudorfl 3′ NCR. Black boxes indicate the relative locations of cyclization sequences C1 and C2 in the 5′ NCR and C1′ and C2′ in the 3′ NCR. Relative locations of TBE virus–conserved repeat sequences R1, R1′, R2, and R3 in the 3′ NCR are indicated by cross-hatched or stippled boxes. IR, conserved inverted repeat. PU⧸PY⧸PR, conserved homopurine, pyrimidine, and purine-rich segments in 3′ NCR RNA, respectively. Dotted lines indicate the locations of deletions in the genomes of some TBE strains relative to the motif found in strain Neudorfl RNA. The notation “AAAn” indicates the location of a poly(A)tract in the strains Neudorfl, 263, and 132 genomes. The location of a severely truncated poly(A)tract in the strain Llubljana 1 genome is specifically indicated by an arrow.

TABLE II

Relative Locations of ProposedCyclizationSequences(CS)inGenomicRNA ofFlavivirus Species

Virus	Translation initiationa	Proposed CS and locationsb	Genbank accession No.	Reference for CS
Mosquito borne		5′-UCAAUAUG 3′-AGUUAUAC		Hahn et al. (1987); 5′CS⧸CYC in Fig. 2
Kunjin	97	5′ (137) (-104) 3′	{"type":"entrez-nucleotide","attrs":{"text":"D00246","term_id":"221966","term_text":"D00246"}}D00246, {"type":"entrez-nucleotide","attrs":{"text":"L24511","term_id":"403463","term_text":"L24511"}}L24511, {"type":"entrez-nucleotide","attrs":{"text":"L24512","term_id":"403464","term_text":"L24512"}}L24512	Khromykh et al. (2001)
JE	96	5′ (136) (-111) 3′	{"type":"entrez-nucleotide","attrs":{"text":"M10370","term_id":"173505","term_text":"M10370"}}M10370	Fig. 2
MVE	96	5′ (136) (-111) 3′	NC000943	Hahn et al. (1987)
WN	97	5′ (137) (-105) 3′	{"type":"entrez-nucleotide","attrs":{"text":"M12294","term_id":"11497619","term_text":"M12294"}}M12294	Hahn et al. (1987)
DEN1	81	5′ (118) (-103) 3′	{"type":"entrez-nucleotide","attrs":{"text":"M87512","term_id":"323660","term_text":"M87512"}}M87512	Fig. 2
DEN2	97	5′ (134) (-103) 3′	{"type":"entrez-nucleotide","attrs":{"text":"M20558","term_id":"323449","term_text":"M20558"}}M20558	Fig. 2
DEN3	94	5′ (132) (-103) 3′	{"type":"entrez-nucleotide","attrs":{"text":"M93130","term_id":"323468","term_text":"M93130"}}M93130	Fig. 2
DEN4	102	5′ (136) (-99) 3′	{"type":"entrez-nucleotide","attrs":{"text":"M14931","term_id":"6978317","term_text":"M14931"}}M14931	Fig. 2
YF	119	5′ (156) (-112) 3′	NC002031	Fig. 2
Cell-fusing agent	114	5′ (169) CCCCGUUCUGG	{"type":"entrez-nucleotide","attrs":{"text":"M91671","term_id":"336190","term_text":"M91671"}}M91671	Cammisa-Parks et al. (1992)
		3′ GGGGCA-GGUC (-134)
	114	5′ (149) GCCAGGG	{"type":"entrez-nucleotide","attrs":{"text":"M91671","term_id":"336190","term_text":"M91671"}}M91671	CS “D”⧸CS“C” in Khromykh et al. (2001)
		3′ CGGUCCC (-471)
Tick borne				Khromykh et al. (2001)
TBE	133	5′ (164) GGGGCGGUCCC	{"type":"entrez-nucleotide","attrs":{"text":"U27495","term_id":"975237","term_text":"U27495"}}U27495	C2⧸C2′ in Fig. 3
		3′ CCCCGGAGGGG (-193)
Powassan	112	5′ (136) GGGGGCGGUCC	{"type":"entrez-nucleotide","attrs":{"text":"L06436","term_id":"309916","term_text":"L06436"}}L06436	C2⧸C2′ in Fig. 3
		3′ CCCCCG–AGG (-189)
Louping ill	130	5′ (160) GGGGGCGGUCC	{"type":"entrez-nucleotide","attrs":{"text":"Y07863","term_id":"2058337","term_text":"Y07863"}}Y07863	C2⧸C2′ in Fig. 3
		3′  CCCCCGG AGG (-193)
TBE	133	5′ (115) GGAGAACAAGA	{"type":"entrez-nucleotide","attrs":{"text":"U27495","term_id":"975237","term_text":"U27495"}}U27495	Mandl et al. (1993) C1⧸C1′ in Fig. 3
		3′ CCUCUUGUUCU (−81)
Powassan	112	5′ (88) GGAGAACAAGA	{"type":"entrez-nucleotide","attrs":{"text":"L06436","term_id":"309916","term_text":"L06436"}}L06436	Mandl et al. (1993)
		3′ CCUCUUGUUCU (−81)
Louping ill	130	5′ (112) GGAGAACAAGA	{"type":"entrez-nucleotide","attrs":{"text":"Y07863","term_id":"2058337","term_text":"Y07863"}}Y07863	Gritsun et al. (1997)
		3′ CCUCUUGUUCU (−81)

Open in a separate window

^aNumbered position of first nucleotide in the AUG codon.

^bThe 5′ and 3′ CS (cyclization sequence) shown are base paired. The first nucleotide of each CS is shown in bold. Additional but nonconserved 5′⧸3′ trans complementary basepairs flanking the CSs are not shown. Positions of the first nucleotide of each CS in relation to the 5′ or the 3′ terminus (indicated by a minus sign) of each genome are shown in bold within parentheses. Reference is made to Fig 2, Fig 3 for mapping of the CSs in mosquito- and tick-borne virus genomes.

The 3′ extreme of the 8 nucleotide 3′ cyclization sequence (CYC; Fig. 2) in mosquito-borne virus RNAs is variably located 99 to 112 nucleotides from the 3′ terminus of the genome (Table II). This locus is just upstream from the conserved 3′ terminal stem and loop structure common to all flavivirus genomes (see later). The CYC sequence is included within the upstream portion of a longer sequence element, “CS1” (Fig 2, Fig 4 ) that is also semiconserved among mosquito-borne virus genomes (Fig. 5 ). For tick-borne viruses, the 11 nucleotide 3′ sequence complementary to C1 [C1′, the sequence CCUCUUGUUCU (Mandl et al., 1993)] is localized to the most upstream portion of nucleotides required to form the conserved stem–loop, spanning nucleotides –81 to –71 (Table II; Fig. 3). The 10 nucleotide 3′ segment complementary to C2 [C2′, the sequence CCCCGGAGGG (Khromykh et al., 2001)] is located further upstream from the 3′ terminus of the genome, spanning nucleotides –184 to –193, for example, in the TBE genomes.

Open in a separate window

Fig 4

Nucleotide sequences of 3′ NCRs of dengue virus genomes representing each of the four serotypes are shown. Shaded areas indicate sequence homology. Dashes indicate gaps introduced into a sequence in order to maximize its alignment with the other sequences. Nucleotides that form the long stem and loop within the flavivirus-conserved 3′-SL (see text) are indicated by a horizontal line above sequence data, labeled 3′-LSH. The location of the flavivirus-conserved pentanucleotide sequence, 5′-CACAG-3′, within the 3′-SL, is indicated by a line below sequence data. The conserved sequences, CS1, the cyclization sequence, CS2, and RCS2 are indicated by horizontal lines above relevant sequence data. The conserved small stem and loop structures, TL1 and TL2 (Proutski et al., 1997b), are also so indicated. Nucleotides presumed to base pair in the formation of the long stem in the 3′-SL and in the formation of the short TL1 and TL2 stems are numbered.

Open in a separate window

Fig 5

Nucleotide sequences of some conserved motifs in 3′ NCRs of mosquito-borne flavivirus genomes. Deletions in CS1 with respect to the sequence for MVE (Hahn et al., 1987) are indicated by a period. Nucleotide substitutions with respect to the sequences of CS1 and CS2 in the MVE genome are indicated by underlining the substituted nucleotides. Nucleotides comprising the 3′ cyclization sequence (CYC) in CS1 are shown in bold print. The stop codon that comprises the first three nucleotides of the most upstream tandem repeat in the YF genome 3′ NCR (Rice et al., 1985) is also shown in bold and indicated by an asterisk. Nucleotide sequence data are from Genbank (Table I) or from the aforementioned references.

Hahn and colleagues (1987) further predicted that the double-stranded panhandle structure potentially resulting from base pairing of the 5′ and 3′ cyclization sequences in the mosquito-borne virus genomes that were under analysis could be further stabilized by the presence of additional nonconserved complementary nucleotides upstream from the 5′CS at the 5′ end of the genome in a pyrimidine-rich segment of the genome and downstream from the 3′ CYC sequence at the 3′ end of the genome in a purine-rich region. Thus there were a total of 11 or 12 contiguous base pairings possible for cyclization of YF, MVE, WN, and DEN2 RNAs, resulting in a predicted thermal stability of from −9.1 kcal (for MVE RNA) to −12.3 kcal (for YF RNA). This was deemed to be sufficient free energy to cyclize flavivirus RNAs based on the previous observation that alphavirus RNAs could cyclize under physiological conditions (Hsu et al., 1973) and that the free energy of cyclization of alphavirus RNA measured thermodynamically was determined to be −13.5 kcal at 25°C (Frey et al., 1979). Additional but noncontiguous complementarity further upstream from the 5′CS and downstream from the 3′ CYC sequence was predicted to stabilize the panhandle further for some of the RNAs, resulting in an increase of potential thermal stability to −20.5, −25.5, and −33.7 kcal, respectively, for the panhandle form of the WN, MVE, and YF genomes (Hahn et al., 1987). The predicted thermal stability of the double-stranded regions potentially formed by base pairing of C1 with C1′ and of C2 with C2′ in the TBE virus genome was −19.8 and −30.9 kcal, respectively (Khromykh et al., 2001). Potential cyclization sequences have also been identified in the CFA genome (Cammisa-Parks et al., 1992; Table II).

Several bits of evidence demonstrate that base pairing of the cyclization sequences is required for RNA replication. First, in a study of the requirements for nucleotide sequences of the DEN4 3′ NCR for virus replication, Men and co-workers (1996) created a series of mutant DEN genomic RNAs by the transcription of DEN4 infectious DNAs containing internal deletions centered about a unique ApaI restriction endonuclease site located between the conserved tandem repeat sequences CS2 and RCS2 (see later) in wild-type DEN4 DNA (Fig 2, Fig 4). Eight mutant RNAs containing deletions ranging in size from 30 to 262 nucleotides, in which all or portions of CS2, RCS2, and CS1 were deleted, were evaluated for their infectivity. Although all the RNAs gave rise to viruses that were reduced in replication competence compared to wild type, only one RNA (mutant 3′d172–83) had a lethal phenotype. The deletion in this RNA was unique in that it included all of the 23 nucleotide CS1 segment in DEN4 RNA, whereas the maximum downstream extent of any of the other seven deletions spared the entirety of CS1. These results demonstrated that CS1, including the 8-nucleotide cyclization sequence, was indispensible for virus replication.

Second, Khromykh and Westaway (1997) evaluated the replication competence of several in-frame internal deletion mutant RNAs derived from a full-length infectious DNA copy of the Kunjin virus genome (“replicons”). They demonstrated that an RNA with a deletion starting at codon 20 of the capsid gene and extending to include downstream sequences encoding the premembrane protein and the entire ectodomain of the envelope glycoprotein (mutant C20rep) replicated efficiently in hamster kidney (BHK) cells, whereas an otherwise identical deletion mutation, in which the 5′ end of the deletion extended upstream to codon 2 of the capsid gene (mutant C2rep), was lethal for RNA replication. Similarly, another mutant RNA in which the capsid start codon and sequences downstream encoding the remainder of capsid, all of the premembrane protein, and the amino-terminal majority of the envelope protein (mutant ΔCME) was also replication defective. C20rep RNA contained the 5′CS for the Kunjin genome, whereas the mutants C2rep and ΔCME RNAs did not. This suggested that either the RNA sequence (including the 5′CS) within the ORF encoding the amino terminus of capsid or translation of the amino terminus of capsid was necessary for RNA replication. The results of follow-up experiments favored the former hypothesis, that nucleotide sequences comprising the 5′ terminus of the ORF, including the 5′CS, were required for RNA replication.

Subsequent studies specifically addressed the requirement for both the 5′CS and the 3′CYC sequence for RNA replication. You and Padmanabhan (1999) reported on an in vitro assay for RNA-dependent RNA polymerase (RdRP) activity in DEN2-infected mosquito cell extracts using exogenous subgenomic RNAs containing the 5′-terminal 230 nucleotides and⧸or the 3′-terminal 373 nucleotides of the DEN2 genome as templates. 5′-terminal DEN2 sequences containing the 5′CS sequence, either linked covalently to the 3′-terminal DEN2 RNA sequence or supplied in trans, were required for replication to initiate on the 3′-terminal RNA segment. Mutagenesis of either the 5′ or the 3′ cyclization sequence such that base pairing between them was abrogated resulted in total loss of in vitro RdRP activity, but when both sequences were mutagenized such that the mutant cyclization sequences were complementary, RdRP activity was detected. This demonstrated that cyclization sequences were required for synthesis of a covalently closed double-stranded hairpin molecule in vitro. Free progeny RNA molecules were not detected in this system. However, Khromykh and colleagues (2001) addressed the relevant question in vivo using the KUN replicon system. They initially created a 5′CS deletion mutation in KUN replicon RNA and demonstrated that the resulting RNA was no longer competent for replication. Next, they created substitution mutations in either the 5′CS or the 3′CYC sequence in order to disrupt base pairing between them. As seen in the in vitro study, abrogation of their complementarity by mutagenesis of either cyclization sequence was also lethal for RNA replication. However, when both cyclization sequences were mutagenized with respect to the wild-type sequence such that their capacity to base pair was maintained, RNA replication, including de novo minus and plus strand synthesis, was restored.

These results clearly demonstrated that complementarity, but not the respective wild-type nucleotides, of the 5′CS and 3′CYC sequences was necessary for flavivirus RNA replication to proceed. Future studies will likely be designed to prove that the input plus strand RNA actually does cyclize in vivo prior to or as a requirement for initiation of RNA replication and to determine how localization of the cyclization sequences, at nearly equal distances from the 5′ and 3′ termini of the genome, is related to their function in mediating the initiation of RNA synthesis. Moreover, it will be of interest to learn whether both pairs of complementary cyclization sequences present in the genomes of tick-borne viruses are required for the replication of tick-borne virus RNA or whether only one complementary pair is sufficient. Cellular and viral proteins that participate in the cyclization process are also yet to be identified. For further discussion of a model for panhandle formation related to additional complementarity of nonconserved sequences in the 5′ and 3′ NCRs (Khromykh et al., 2001), see later.

C. Additional Conserved Linear Sequences in 3′ NCRs of Mosquito-Borne Flaviviruses

1. The Pentanucleotide

The 3′ NCRs of mosquito-borne flaviviruses contain a number of conserved sequences that are localized primarily to the 3′-terminal portion of this region of the genome. The most 3′-terminal of these sequences is the pentanucleotide 5′-CACAG, which forms part of an unpaired region forming a closed loop within the conserved stem and loop structure formed by the 3′-terminal ∼95 nucleotides of all flavivirus genomes (Hahn et al., 1987). For example, in DEN virus genomes, the pentanucleotide sequence occurs between nucleotides –43 and –47 (Fig 4, Fig 6 ). This sequence and its location in the genome are completely conserved among mosquito- and tick-borne flaviviruses for which sequence data are available. It is also incompletely conserved in the 3′ stem and loop structure predicted to be present in CFA genome RNA; the fourth nucleotide is C instead of A (Cammisa-Parks et al., 1992). The functional requirement for the conserved pentanucleotide sequence in the virus life cycle is not known, but it has been suggested that the pentanucleotide may play an indispensible role in the binding of cellular or viral proteins to the conserved 3′ stem and loop structure (see later) during RNA replication (Khromykh et al., 2001).

Open in a separate window

Fig 6

Nucleotide sequences of 3′-SLs in WN strain E101 and DEN2 strain NGC genomes (Zeng et al., 1998) are shown in the expected conformation and numbered from the 3′ end of each genome. The flavivirus-conserved pentanucleotide sequence 5′-CACAG-3′ is indicated in bold underlined type. The “top” and “bottom” portion of the DEN2 3′-SL as designated in a previous study with respect to the WN 3′-SL is shown by a labeled horizontal dashed line (Zeng et al., 1998). Essential nucleotides of the putative binding site for the translation elongation factor, eF-1α, to the WN 3′-SL are circled (Blackwell and Brinton, 1997). Nucleotides potentially involved in pseudoknot formation in both DEN2 and WN 3′-SLs are linked by dashed lines (Shi et al., 1996a).

2. CS1

The next upstream conserved sequence in mosquito-borne flavivirus RNA is CS1, which has already been mentioned earlier in describing localization of the CYC sequence (Fig 2, Fig 4, Fig 5). The CS1 sequence is not present in 3′ NCRs of tick-borne flavivirus genomes, although, as will be shown, the characteristic 3′ cyclization sequence(s) in those genomes also occurs in the context of a larger conserved sequence motif (Wallner et al., 1995). CS1, as originally described by Hahn and colleagues (1987), consisted of a 25 nucleotide segment in the MVE 3′ NCR that was nearly but not totally conserved among viruses of the same subgroup and between subgroups. For example, CS1 for the MVE genome differs from that of the WN genome by one nucleotide substitution and the deletion of a single nucleotide, whereas CS1 in the JE virus genome differs from that of MVE by only a single nucleotide substitution. The sequence in the MVE genome differs from that found in the DEN1 and DEN3 genomes in length in that the DEN sequences appear to contain two discrete single nucleotide deletions and two discrete single nucleotide A to G substitution mutations with respect to the reference. DEN2 and DEN4 genomes contain the identical deletion mutations but only one of the two substitution mutations with respect to the MVE sequence. The CS1 sequence in the YF virus genome contains three substitution mutations and one deletion with respect to the MVE sequence. The 8-nucleotide CYC sequence is contained in the upstream half of CS1, between nucleotides –95 and –102 in DEN genomes. The CYC sequence itself is completely conserved among these and other mosquito-borne virus genomes.

3. CS2 and RCS2 and Analogous Upstream Tandem Repeat Sequences

The conserved tandem repeat sequences CS2 and RCS2 were also initially described by Hahn and co-workers (1987). CS2 and RCS2 are not found in 3′ NCRs of tick-borne flavivirus genomes. For mosquito-borne viruses, CS2 is located closest to the 3′ terminus of the genome, 12 to 22 nucleotides upstream from CS1 in the DEN, YF, and JE subgroup genomes (Fig 2, Fig 4, Fig 5). For the Genbank genome sequences of individual strains surveyed, the 20 nucleotide sequence of CS2 is completely conserved for viruses in the JE subgroup (MVE, WN, and KUN) and among all four serotypes of DEN viruses, except for a single internal nucleotide substitution (G to U) found in the DEN3 genome. The CS2 sequence in the YF 17D genome (Rice et al., 1985) deviated from that seen in genomes of viruses of the JE subgroup by two nucleotide substitutions, resulting in A to U and C to U changes in the sequence.

In the MVE genome, and similarly in genomes of the other JE subgroup viruses, RCS2 is 51 nucleotides upstream from the 5′ terminus of CS2 and represents an exact duplication of the 20-nucleotide sequence of CS2. In the DEN genome sequences, RCS2 is located 62 (for DEN1) to 68 nucleotides (for DEN4) upstream from CS2 (see Fig. 4). CS2 is not repeated upstream in the YF genome. Instead, the YF genome contains three unique tandem repeats at the 5′ terminus of its 3′ NCR. The first of these repeats is 40 nucleotides long in the YF strain 17D genome, and it initiates downstream from the third of three in-frame stop codons found at the 3′ terminus of the YF 17D ORF (Fig 2, Fig 5). The second and third iterations of the repeat are 44 nucleotides in length. Save for the fact that the second repeat is 4 nucleotides longer than the first, there are only 5 nucleotides different between the two segments. Similarly, the third repeat differs from the first by only 4 nucleotides in 40. The second repeat initiates 4 nucleotides downstream from the 3′ terminus of the first one, and the third repeat in turn initiates 6 nucleotides downstream from the 3′ terminus of the second.

Not all YF virus genomes contain an analogous set of three tandem repeat sequences, as found in vaccine strain 17D. A study of partial nucleotide sequence data for genomes of 13 YF field isolates (Wang et al., 1996) revealed that only the West African strain genomes contained a 511-nucleotide 3′ NCR, including three tandem repeat sequences. Genomes of strains from central and east Africa and from South America had shorter 3′ NCRs (443 to 469 nucleotides) due to the lower number of YF-specific repeat sequences. Central and east African strains had only two YF-specific repeats, and those from South America had only one copy of the “repeat” sequence. On the basis of overall nucleotide sequence relatedness, it was speculated that duplication of the repeated sequence took place in west Africa after introduction of the YF virus into South America. The resulting variation in lengths of 3′ NCRs of the genomes of YF virus isolates, due to duplication and⧸or deletion of redundant nucleotide sequences within the 5′ proximal portion of the 3′ NCR, is reminiscent of that observed among the genomes of different strains of TBE virus (see later).

The JE subgroup viruses have an additional pair of repeats upstream from CS2 and RCS2 in the 3′NCR (CS3 and RCS3 in Fig 2, Fig 4). For the MVE genome (Hahn et al., 1987), these sequences are each 28 nucleotides in length and are situated 120 nucleotides apart in the MVE 3′NCR. The most upstream of these repeat segments has its 5′ terminus 153 nucleotides downstream from the stop codon in the MVE ORF. There are only three nucleotides different between the sequences of CS3 and RCS3 in the MVE 3′ NCR.

4. Function of Conserved Tandem Repeat Sequences in Mosquito-Borne Flavivirus Replication

Relatively little is known regarding this subject. Perhaps more work will be done in this area in the future, as the use of full-length infectious DNAs to effect site-directed mutagenesis of the flavivirus genome has become more common. However, it is clear that deletion of conserved repeat sequences alters virus growth properties. As mentioned previously, Men and co-workers (1996) created a series of eight different mutant dengue genomic RNAs derived from DEN4-infectious DNAs containing internal deletions centered about a unique ApaI restriction endonuclease site located at nucleotide –171 in the DEN4 genome, about midway between CS2 and RCS2 (Fig 2, Fig 4). Thus, all or portions of CS2 and⧸or RCS2 were deleted in six of the resulting seven viable deletion mutant viruses. (The eighth mutant RNA contained a deletion of CS1, which proved to be lethal, as discussed previously.) The maximum 3′ boundary of any deletion that gave rise to virus was 113 nucleotides from the 3′ end of the genome, preserving CS1 and the 3′ stem and loop (SL). One DEN4 mutant containing a 121-nucleotide deletion that included all of RCS2 and most of the upstream portion of the 3′ NCR (mutant 3′d303-183) was only slightly impaired for replication in cultured monkey kidney cells compared to wild-type virus. In contrast, a mutant containing a 60-nucleotide deletion that included almost all of CS2 (mutant 3′d172-113) was reduced in peak titer by about 10-fold with respect to wild type in both experiments. This suggested that CS2 may be more critical for viability than RCS2. Interestingly, a mutant that contained a deletion of only 30 nucleotides (3′d172-143) between CS2 and RCS2, but sparing both conserved sequences, was more impaired for replication than either mutant 3′d172-113 or 3′d303-183 in that its peak titers in the two assays were reduced compared to wild type by about 100-fold. This indicates that nonconserved nucleotides between CS2 and RCS2 are important for virus viability, possibly because of a requirement for spacing between the two conserved sequences or possibly because this deletion removed a DEN and JE subgroup conserved SL structure (TL2) predicted by Proutski and colleagues (1997b). Mutant virus 3′d172-143 is currently under study as a potential DEN vaccine candidate (Durbin et al., 2001).

The DEN4 genome could tolerate deletion of more than half its 3′ NCR, provided the deleted sequences were upstream from nucleotide –113, without complete loss of viability. A deletion of 202 nucleotides, extending from the stop codon (at nucleotide –384) to nucleotide –183 (3′d384-183) produced a viable virus that formed plaques in monkey cells and was only a little more reduced in peak titer compared to the vaccine candidate, mutant 3′d172-143, in the two growth assays, and a 262-nucleotide deletion of all but 10 nucleotides between the stop codon and nucleotide –113 produced viable virus after transfection of mosquito cells. However, this virus did not form plaques in monkey cells.

Khromykh and Westaway (1997) addressed the requirement for CS3 and RCS3, found in the JE subgroup virus genomes, using the KUN replicon system. They started with replicon ΔME, which was stable and replicated efficiently in vivo. A 76-nucleotide deletion (KUN nucleotides 10,422 to 10,499) was created in ΔME DNA between the stop codon and the upstream boundary of RCS3 (Fig. 2) to create mutant ΔME⧸76 RNA. A second mutant RNA derived from ΔME DNA, ΔME⧸352, contained a 352-nucleotide deletion of KUN nucleotides 10,422 to 10,775, which included all of RCS3 and CS3, nonconserved sequences between these two boxes, and additional nucleotides extending upstream to overlap the upstream boundary of mutant ΔME⧸76 RNA. Whereas ΔME⧸76 RNA replicated efficiently in BHK cells, comparable in level to that of the “wild-type” ΔME replicon RNA, only very low-level replication of ΔME⧸352 RNA was detected above the threshold of sensitivity of the assay. Unfortunately, data are lacking relative to the phenotype of either deletion mutation in the context of a replication-competent KUN virus genome. Proutski and colleagues (1997b) offered the hypothesis that the phenotype of mutant ΔME⧸352 RNA was related to the deletion of nonconserved nucleotides that form a conserved secondary structure in all JE subgroup virus genomes.

III. SecondaryStructure ofNCRs

A. Secondary Structure of the 5′ NCR

Brinton and Dispoto (1988) determined the nucleotide sequences of the 5′ NCRs of WN virus and of seven different isolates of the mosquito-borne flavivirus, St. Louis encephalitis (SLE) virus. They then compared the sequences to those of YF, MVE, and DEN viruses. While only short regions within the 5′ NCRs were conserved among different subgroups, significant sequence homology was observed among viruses of the same subgroup, and an almost complete conservation of nucleotide sequence was observed among different strains of the same virus, i.e., the seven isolates of SLE virus and two different strains of WN virus. For each 5′ NCR sequence, secondary structures of similar size, shape, and predicted thermodynamic stability could be formed. In each case, the structures consisted of a stem with a small top loop and a larger side loop (Fig. 7 ). In most cases, a second short stem could be formed by 3′-terminal sequences of the 5′ NCR, including the predicted start codon for the long ORF and downstream nucleotides in the ORF. The predicted thermal stability of the long stem structures ranged from −17.3 kcal (for the SLE sequence) to −27.8 kcal (for the WN strain E101 sequence, as shown). The conservation of this secondary structure in unrelated flavivirus genomes was taken as evidence of its possible importance in replication.

Open in a separate window

Fig 7

Proposed secondary structure of the WN 5′ NCR (adapted from Brinton and Dispoto, 1988). Nucleotides are numbered from the 5′ terminus, excluding the cap, which is shown boxed. Nucleotides of the ORF, starting with nucleotide number 96, are shown in large, bold type (adapted from Brinton and Dispoto, 1988).

The relative importance of the predicted double-stranded regions in the 5′ NCR was studied by deletion mutagenesis of an infectious DNA copy of the DEN4 strain 814669 genome (Cahour et al., 1995). The linear sequence of the 5′ NCR is partially conserved between DEN and WN species (Fig. 7). For example, nucleotides 1–7, 11–16, 21–25, 55–60, and 69–77, numbering from the 5′ terminus of the WN strain E101 sequence, are present in the 5′ NCRs of both the DEN4 strain 814669 and the WN strain E101 genomes. Conservation of sequence and predicted secondary structure are much higher when any two DEN 5′ NCRs are compared. Nine distinct deletions of from 5 to 12 nucleotides each were created between nucleotides 50 and 87 of the DEN4 sequence in the context of a DEN4 infectious DNA. In addition, two deletions of 5 and 26 nucleotides, respectively, were created downstream (involving nucleotides 94 to 98; d94-98) and upstream (d18-43) from the central targeted domain. Nucleotide 98 is four nucleotides upstream from the start codon of the DEN4 ORF. Mutants d55-60, d62-68, d69-75, and d82-87 contained deletions of nucleotides that were base paired in the long stem region of the 5′ NCR, whereas mutants d50-54, d58-64, d76-81, d73-77, d76-87, and d94-98 contained deletions in the shorter stem or non–base-paired regions within the long stem. Mutant d18-43 contained a deletion of the upper loop and the adjacent upper portion of the long stem, as shown for the WN 5′ NCR in Fig. 7. Five mutations were deemed lethal, in that infectious virus could not be recovered after transfection of cultured monkey kidney cells with RNA derived in vitro by the transcription of mutagenized DNAs. These were mutants d55-60, d58-64, d62-68, d69-75, and d94-98. Nucleotides 55 to 75 in the DEN4 genome are analogous to nucleotides 60 to 80 in the WN 5′ NCR secondary structure shown in Fig. 7. In general, failure of the four contiguous deletion mutations (spanning DEN4 nucleotides 55 to 75) to yield viable viruses was ascribed to a requirement for those nucleotides for formation of the base pairs that constitute the bottom-most portion of the long stem and the short stem, which also requires nucleotides in the ORF for its formation.

The viable mutant viruses derived in this study were distinguishable from each other by plaque size in cultured mosquito and monkey cells, growth curves on these two cell types, relative levels of RNA in infected cells as assessed by dot-blot hybridization, and translation efficiency of the ORF in an in vitro system. Mutant d82-87 had a host range-restricted phenotype in that it uniquely failed to replicate in mosquito cells and in adult mosquitoes, apparently due to a defect in RNA synthesis. Translation of a portion of the DEN4 ORF containing the d82-87 5′ NCR was also least efficient among all the mutants tested. However, the cell-free translation assay was not predictive of the viability of mutant viruses; RNAs containing some of the lethal deletion mutations, e.g., d55-60, exhibited enhanced activity in translation compared to wild-type control RNA. Therefore, the dominant effect of deletion mutations on the viability of mutant viruses appeared to be at the level of RNA synthesis. The host range-restricted phenotype of mutant d82-87 also suggested that binding to RNA of host-specific factors, presumably cellular proteins, was a secondary determinant of the viability of mutant viruses.

Shi and co-workers (1996a) performed a computer-based analysis of the 3′ negative-sense (−) RNA predicted to be synthesized in the process of replication of the 5′ NCR of WN strain E101 virus. The presence of an SL secondary structure in 3′ (−) NCR RNA, formed by the sequence complementary to the SL predicted to occur in the 5′ NCR, was subsequently confirmed by spectroscopy, analysis of thermal melting curves, and selective RNase digestion of an in vitro-transcribed 75-nucleotide RNA representing the SL within the 96 nucleotide 3′ NCR of the WN-negative strand. The 75-nucleotide transcript was then shown to bind specifically to four discrete proteins in three RNA–protein complexes in lysates of BHK cells and in lysates of embryonic fibroblasts derived from C3H⧸RV and C3H⧸He mice. C3H⧸RV mice are homozygous for the flavivirus resistance gene Flv, whereas C3H⧸He mice are susceptible to flavivirus infection (Groschel and Koprowski, 1965). Resistance of mice to the pathologic effects of encephalitic flavivirus infection had been shown previously to relate to reduced production of progeny plus strand RNAs at the cellular level. Because RNA–protein complexes formed in lysates of C3H⧸RV cells were indistinguishable in size from those formed in C3H⧸He cells (and in BHK cells), an investigation was conducted to determine whether the complexes between C3H⧸RV proteins and WN 3′ (−) SL RNA were different in stability from those formed between C3H⧸He proteins and WN 3′ (−) SL RNA. Complexes 1 and 3 formed in lysates of C3H⧸RV cells were shown to have a significantly longer half-life than the analogous complexes formed in C3H⧸He cell lysates.

B. The 3′ SL

Analysis of the YF genome sequence originally revealed the presence of a potential SL structure formed by the 3′-terminal 87 nucleotides (Rice et al., 1985). Similar stem–loop secondary structures with similar predicted thermal stability have been subsequently proposed for the 3′ termini of all flavivirus genomes for which nucleotide sequence data are available (see, e.g., Hahn 1987, Irie 1989, Brinton 1986, Wengler 1986). In fact, similar structures are predicted to form at the 3′ termini of the genomes of the other Flaviviridae, the hepaci- and pestiviruses (Blight 1997, Deng 1993). The stem consists of about 30 hydrogen-bonded base pairs and is variably interrupted by bulges due to the predicted apposition of nucleotides that are unable to hydrogen bond. Sequence analysis also predicted a smaller adjacent stem–loop structure that is a consistent feature of the flavivirus 3′ NCR, upstream from the long stem–loop. Together the two adjacent structures involve 90 to 100 nucleotides at the 3′ end of the genome and are referred to collectively, for purposes of this discussion, as the 3′ SL (Fig. 6). The 3′ SL has a predicted thermal stability of −40 to −45.2 kcal, largely due to the free energy state of the long stem region.

The nucleotide sequence of the 3′ SL is only partially conserved among unrelated flavivirus species (e.g., Fig. 6), although conserved segments do exist. For example, as already noted, the distal loop region of the long stem contains the flavivirus-conserved CACAG box, and in general there is a similarity among the nucleotide sequences of the predicted loop regions, both the double loop that terminates the long stem and in the single loop atop the adjacent short stem (the sequence 5′ GANAGA-3′) (Shi et al., 1996b), nucleotides –83 to –89 in the DEN2 3′ SL and nucleotides –85 to –91 in the WN 3′ SL. In addition, there is similarity among flavivirus genomes at nucleotides –1 to –6, which includes the 3′ terminal 3′ UC dinucleotide that is conserved among all known flavivirus genome sequences except that of the CFA virus. For example, Fig. 6 shows the similarity in sequence between 3′ termini of the DEN genomes (the sequence 3′-UCUUGG-5′, conserved in all four DEN serotypes) compared to the 3′ terminus of the WN genome (the sequence 3′-UCUAGG-5′, conserved in KUN, JE, and MVE genomes). Upstream from these six nucleotides, forming one side of the lower half of the long stem, one finds the semi-conserved pentanucleotide ACACA, which has been discussed previously. The 3′ SL nucleotide sequence is best conserved among the closely related members of the same subgroup or clade, for example, among the four DEN serotypes and among species assigned to the JE subgroup.

Physical and biochemical evidence for the existence of the 3′-terminal long stem and loop in WN virus RNA was provided by Brinton and co-workers (1986). In their efforts to determine the nucleotide sequence of the genome at its 3′ end, they subjected virus RNA to digestion with five different ribonucleases. Comparison of the gel mobility of the end products of these reactions to that of a sample of RNA that was sequenced by selective chemical degradation revealed that most of the 3′-terminal 84 nucleotides were relatively resistant to one or more of the RNases. This suggested that the nucleotides in question were involved in a secondary partially double-stranded structure that conferred the observed resistance to degradation. Drawing on the previous prediction of a 3′-terminal SL structure in the YF virus genome (Rice et al., 1985), these workers hypothesized that the last 84 nucleotides of the WN genome were involved in a similar motif. WN virus RNA did not exhibit significant resistance to RNase digestion upstream from nucleotide –84.

Further analysis of interactions among nucleotides of the 3′ SL suggested the existence of a pseudoknot tertiary structure (Shi et al., 1996a), potentially formed by hydrogen bonding between nucleotides comprising one strand of the long stem (nucleotides –71 to –74 in the WN 3′-SL; Fig. 6) and nucleotides contained in the loop structure atop the adjacent short stem (nucleotides -86 to -89 in the WN 3′ SL). In order for the pseudoknot to form in the 3′ SL for most flavivirus species, one must posit the occurrence of G-A and G-U base pairs in some genomes (see DEN2 3′-SL; Fig. 6). Such atypical hydrogen bonding has been described for other RNA species (Heus 1991, Wimberly 1993), and pseudoknots have been described at the 3′-ends of other viral genomes (Pleij 1985, Dreher 1988, Jacobson 1993). Replication of phage QB RNA was shown to be dependent on interaction of the replication complex with a pseudoknot formed by internal base pairing of the 3′ terminus of the genome to a single-strand RNA segment 1200 nucleotides upstream (Klovins and van Duin., 1999).

WN and DEN3 3′ model RNAs that included the lower half of the long stem and the upstream adjacent small stem and loop (nucleotides –1 to –11 linked covalently via four U residues to nucleotides –71 to –96, for the WN sequence; Fig. 6) were synthesized in vitro (Shi et al., 1996a). Results from RNase probing, circular dichroism spectral analysis, UV-melting experiments, mutagenesis of the RNA sequence, and molecular modeling were consistent with the existence of the predicted pseudoknot in each of the two RNAs that were studied. A consequence of this prediction was that the RNA in a pseudoknot was predicted to form a cleft that could be functional in protein or metal binding. A potential weakness of this work, borne of necessity, was the fact that it was conducted using in vitro–transcribed RNA representing a small, discontinuous segment of the 3′ NCR. Thus, possible interactions between and among the segments that were studied and more upstream sequences in the genome could not be assessed.

1. Binding of Cellular Proteins to the 3′ SL

Because it likely contains RdRP promoter elements, the 3′ SL is likely to interact with both cellular and viral proteins required for viral RNA replication, which must initiate with the synthesis of full-length minus-strand RNAs. The host range of most flaviviruses, both in nature and in the laboratory in tissue culture, is very broad. For example, flaviviruses can replicate efficiently in mammalian, avian, insect, reptile, and amphibian hosts, as well as in cell cultures from these hosts (Brinton, 1986). Therfore, one might expect that host cell proteins required for flavivirus replication are highly conserved or that domains required for this function may be conserved within a variety of multifunctional proteins of disparate size in different cell types. Cell proteins that bind 5′ and⧸or 3′ NCRs of genomes of several different RNA viruses have been described, including those of mouse hepatitis virus (Furuya and Lai, 1993), bovine coronavirus (Spagnolo and Hogue, 2000), measles virus (Leopardi et al., 1993), Rubella virus (Nakhasi 1990, Nakhasi 1991), Sindbis virus (Pardigon et al., 1993), hepatitis A virus (Schultz 1996, Yi 2000), and hepatitis C virus (Luo 1999, Spangberg 1999). In many cases, the cited reports refer to binding of cellular translation factors to IRES elements found in the 5′ NCRs of hepaciviruses and picornaviruses, among others, but not present in the flavivirus 5′ NCR. For further information, the reader is referred to reviews of this subject with emphasis on positive-strand RNA viruses in this volume (Strauss 2002, Brinton 2001).

In general, the literature on binding of proteins to 3′ NCRs of positive-strand RNA virus genomes does not shed light on the functional requirements associated with protein binding. However, two reports [for mouse hepatitis virus RNA (Huang and Lai, 2001) and poliovirus RNA (Herold and Andino, 2001) suggest that cellular proteins may be involved in linkage of 5′ and 3′ termini of replicating RNA virus genomes through the formation of ribonucleoprotein (RNP) complexes. For poliovirus, formation of this RNP was required for the initiation of negative-strand RNA synthesis.

Cell proteins required for flavivirus replication have not yet been identified with certainty. However, the specific interaction of a radiolabeled RNA representing the nucleotide sequence of the WN 3′ SL with proteins in extracts of uninfected BHK cells was reported by Blackwell and Brinton (1995). (BHK cells are a common laboratory cell substrate for growth of the WN virus to very high titers.) Gel mobility shift assays, UV cross-linking, Northwestern blotting, and ion-exchange chromatography were initially used to identify 56-, 84-, and 105-kDa proteins that bound to a truncated in vitro–synthesized fragment of the WN and DEN3 3′ SLs. The unidentified 105-kDa protein species also had significant affinity for an RNA representing the sequence of the 3′ (negative-sense) SL (see earlier discussion). The hypothesis was advanced that this protein may play a role in the initiation of both plus- and minus-strand RNA synthesis. Similar specificity has been reported for other viral systems. For example, one of the cellular proteins that binds rubella virus RNA appears to have specific affinity for both minus- and plus-sense SL structures at the 3′-termini of the respective RNAs (Nakhasi et al., 1991).

Blackwell and Brinton (1997) subsequently identified the 56-kDa protein that bound the WN 3′ SL in BHK cell lysates as the translation elongation factor, eF-1α. The site for eF1-α binding was localized to nucleotides –62 to –65, the sequence 3′ACAC-5′ (Fig. 6). Li and Brinton (2001) made the possibly relevant observation that an in vitro–synthesized RNA containing the nucleotide sequence of the WN 3′ SL inhibited in vitro translation of WN RNAs containing 5′-terminal portions of the WN ORF and of foreign mRNAs. In contrast, an RNA containing nucleotides representing the 3′-terminal secondary structure of rubella virus RNA enhanced in vitro translation. These contrasting results may indicate that the affinity of the WN 3′ SL for translation elongation factors is significant for upregulating WN viral translation or downregulating translation of cellular mRNAs in vivo. Plant and animal elongation factors eF1-α, β, and γ are recognized components of replicases in other systems. Das and co-workers (1998) demonstrated that the polymerase of the negative-strand virus, vesicular stomatitis virus (VSV), which replicates in both insects and mammals, binds eF1-α tightly, and that the resulting complex binds eF1-β and γ. All three elongation factors are required for VSV replicase activity in an in vitro system. In earlier work, Blumenthal et al. (1976) demonstrated that RNA bacteriophage Qβ uses three cellular proteins as components of its replicase, the bacterial translation elongation factors, Ts and Tu, and ribosomal protein S1. Further work is required to establish with certainty the physiological significance of the observation that the WN 3′ SL RNA binds translation elongation factors.

In a more recent study, an in vitro–synthesized, 83-nucleotide RNA representing the JE virus 3′ SL nucleotide sequence was evaluated for binding of specific proteins in lysates of neonatal mouse brain (Ta and Vrati, 2000). Three RNP complexes thus derived were stable in the presence of high-salt buffer. UV cross-linking and Northwestern blotting analysis identified three proteins involved in RNPs, with apparent molecular masses of 32, 35, and 50 kDa. Gene products derived from a neonatal mouse brain cDNA library were screened for binding to the JE 3′ SL, and a 36-kDa protein, MOV34, was thus identified. Subsequent experiments showed that the MOV34 protein could also bind to the 3′ SL in native full-length JE virus genomic RNA. MOV34 is a 26S proteasome subunit protein that belongs to a family also thought to be involved in RNA transcription and translation. A computer-based comparison of the amino acid sequences of translation elongation factor eF1-α and MOV34 revealed no detectable sequence similarity between them, even when constraints for detection were set very low (L. Markoff, unpublished data). However, both of these proteins contain a high proportion of positively and negatively charged amino acids, which could effect an association with RNA and⧸or cell membrane organelles. The translation elongation factor sequence is 463 amino acids in length and includes 23 Asp residues, 31 Glu residues, 17 Arg residues, and 47 Lys residues for a net positive charge of +10. MOV34 is 309 amino acids in length and contains 18 Asp residues, 23 Glu residues, 11 Arg residues, and 23 Lys residues for a net negative charge of –8.

2. Binding of Viral Proteins to the 3′ SL and the Upstream Proximal Portion of the 3′ NCR

As mentioned previously, the viral nonstructural proteins NS3 and NS5 contain domains with helicase and polymerase activities, respectively, required for flavivirus RNA synthesis. It is logical to expect that the viral helicase and RdRP both bind the 3′ termini of plus-strand RNA in order to initiate negative-strand synthesis. Chen and co-workers (1997) made lysates of BHK cells infected with JE virus and first demonstrated that these lysates retained viral RNA synthetic activity. Next, they probed active lysates for proteins that could bind an exogenous 585-nucleotide in vitro–synthesized RNA representing the JE 3′ NCR by UV cross-linking. They focused on proteins that specifically formed in lysates of infected cells, as compared to uninfected cells. Two proteins, p71 and p110, were thus identified. Mapping experiments revealed that protein p110 required only the 3′-terminal 80 nucleotides of the JE genome for binding, whereas protein p71 appeared to have a more diffuse binding site on the 585-nucleotide probe. The two proteins were subsequently identified as NS5 and NS3, respectively. The results of immune precipitation assays suggested further that NS5 and NS3 formed a complex with each other in association with the 3′ NCR.

3. Nucleotide Sequence Specificity of the 3′ SL

As mentioned previously, the nucleotide sequence of the 3′ SL in the genomes of different species of flaviviruses is only partially conserved. Zeng and colleagues (1998) asked whether nucleotide sequences of the WN 3′ SL could replace those of the DEN2 3′ SL, in the context of the DEN2 genome. This was done by site-directed substitution mutagenesis of sequences encoding the 3′ SL in a DEN2-infectious DNA (Polo et al., 1997). RNA derived from “D2⧸WN-SL” DNA was infectious in that DEN2-specific immune fluorescence was observed in transfected cells, but the resulting virus was highly impaired for replication; it never achieved a titer greater than 100 pfu⧸ml in monkey cells. In contrast, DEN2 wild-type virus derived from RNA transcribed in vitro typically grew to titers >10E6 pfu⧸ml. This demonstrated that the nucleotide sequence, not merely the secondary structure of the 3′ SL, was a determinant of replication efficiency.

An attempt was then made to define which DEN2 3′ SL nucleotides were required for efficient virus replication by restoring DEN2 3′ SL–specific sequence elements in the context of the D2⧸WN-SL mutant DNA. “Top” and “bottom” portions of the DEN2 (and WN) 3′ SLs were defined arbitrarily according to previous work by Blackwell and Brinton (1995) (Fig. 6). Six additional DEN2 3′ SL mutant DNAs were created, and the growth phenotypes of the resulting viruses in LLCMK2 cells are summarized in relation to their 3′ SL structure in Fig. 8 . The mutations were “lethal,” meaning no positive immunofluorescence could be detected in transfected cells; “sublethal,” meaning the viruses replicated about as well as the parent mutant, D2⧸WN-SL; or “viable,” meaning peak titers in LLCMK2 cells were equal to or no worse than 10- or 100-fold reduced compared to the wild-type DEN2 virus. Results identified an 11-bp segment, comprising the uppermost portion of the bottom of the wild-type DEN2 3′ SL sequence, as being essential in viable virus mutants (DEN2 nucleotides 7–17 and 63–73 in Fig. 6). In addition, the study demonstrated that the DEN2-specific nucleotide sequence of the “top” portion of the DEN2 3′ SL was not required for efficient replication, but that its base-paired structure was essential. A mutant in which the entire top portion of the WN 3′ SL was substituted for that of the DEN2 3′ SL grew as well as the wild-type virus, but a second mutant, in which base pairing in the top portion of the long stem was abrogated by mutagenesis of the DEN2 nucleotide sequence, had a lethal phenotype. Defects in replication of all mutant viruses appeared to be at the level of RNA replication, suggesting a critical role of the 3′ SL for initiation of negative-strand RNA synthesis.

Open in a separate window

Fig 8

Chimeric DEN2⧸WN 3′ SL nucleotide sequences in DEN2 genomic DNA. The 3′ SL secondary structure is depicted as a line drawing. DEN2 nucleotide sequences are depicted as a thin line, and WN nucleotide sequences are depicted as a thick line. “Top” and “bottom” portions of the 3′ SL are indicated by horizontal dashed lines bisecting the long stem in the 3′ SL. The chimeric nucleotide sequences indicated were substituted for the wild-type nucleotide sequence of the 3′ SL in DEN2 infectious DNA, and the resulting mutant DNAs were used to derive RNAs that yielded DEN2 viruses with the indicated phenotypes after transfection of cultured monkey kidney cells (see text). Viable viruses all retained the DEN2-specific 11-bp segment comprising the top portion of the bottom section of the long stem in the 3′ SL. The DEN2mutF virus (asterisk) was notable for its host range-restricted phenotype in mosquito cells, where its replication was 100,000-fold reduced, compared to the wild-type DEN2 virus (Zeng et al., 1998).

Mutant D2⧸WN-SL(mutF) or “mutant F” virus (Fig. 8) was of special interest in that it displayed a viable phenotype in LLCMK2 cells but was severely impaired for replication in cultured mosquito cells. Thus, site-directed mutagenesis of either the 5′ NCR (Cahour et al., 1995) or the 3′ SL (Zeng et al., 1998) may result in the genesis of host range-restricted mutant DEN viruses. These findings support the hypothesis that host cell proteins play an important role in virus replication. Zeng and co-workers (1998) also demonstrated that DEN2 mutant F virus was defective in RNA synthesis at early times after infection of mosquito cells using an assay that did not distinguish between plus- and minus-strand synthesis. However, a more specific mechanism for the host range-restricted phenotype of DEN2 mutant F virus has not been elucidated thus far. For example, using an in vitro assay for RdRP activity (described previously), You and co-workers (2001) demonstrated that an RNA containing the DEN2 mutant F 5′ and 3′ NCR nucleotide sequence was not defective for replication in a mosquito cell extract compared to its activity in an extract of monkey kidney cells. Because this assay detected only negative-strand synthesis, perhaps the results indicate that DEN2 mutant F replication is defective at the level of progeny positive-strand RNA synthesis. Other efforts to detect differential binding of mosquito cell proteins to the DEN2 mutant F 3′ SL, using techniques pioneered by Blackwell and Brinton (1995), have thus been unsuccessful (Yu and Markoff, unpublished results).

Introduction of mutant F mutational changes (Markoff et al., 2002) into infectious DNAs for human virulent DEN1 and DEN4 viruses has resulted in the production of mutant viruses that have a host cell-restricted phenotype similar to that of the DEN2 mutant F virus. The DEN1 mutant F virus was subsequently shown to be attenuated and highly immunogenic in a monkey model for DEN virus infection of humans and is currently a candidate for DEN vaccine development.

IV. MutationalChanges in the 5′ or3′ NCR and the3′ SL MayResult inAttenuation ofVirusVirulence

Mutational changes, either in the ORF or in the 5′ and⧸or 3′ NCRs, could be expected to alter virus growth properties and therefore virulence. For example, Chiou and Chen (2001) reported on the properties and genome sequences of two naturally occurring isolates of JE virus found in an area of Taiwan where local residents have high levels of JE antibodies. One of these isolates, strain T1P1, exhibited 10,000-fold less neurovirulence in mice compared to the other strain, CH1932. A comparison of the complete genome sequences of these two strains revealed that they differed by seven nucleotides. Two of these sequence differences resulted in amino acid changes in NS3, and there were two others in the 3′ NCR, upstream from the 3′ SL. The other three nucleotide sequence differences between the two strains were “silent” ones in the M and NS5 genes. The inference was made that the nucleotide sequence differences noted in NS3 and in the 3′ NCR were responsible for the marked differences in neurovirulence between strain T1P1 and strain CH1932. No attempt was made to determine whether the 3′ SL mutational difference between strains T1P1 and CH1932 resulted in altered binding of virus- or cell-specific proteins to the 3′ NCR. See also discussions of the work of Cahour 1995, Men 1996, Mandl 1998, Pletnev 2001, Durbin 2001, and Zeng et al. (1998) in conjunction with Markoff et al. (2002) for other examples of attenuation of flavivirus virulence resulting from mutagenesis of the 5′ or 3′ NCR.

V. Hypothesis forHydrogenBonding between5′ and3′ NCRs

The existence in vivo of the the previously predicted secondary structures formed by cis interactions of nucleotides within the 5′ NCR (Brinton and Dispoto, 1988) and within the 3′ NCR (Rice 1985, Hahn 1987, Brinton 1986) must be reconsidered in light of a more recent disclosure of a possible interaction between nucleotides of the 5′ and 3′ NCRs, including the cyclization sequences, in trans to form double-stranded panhandle-like structures (Khromykh et al., 2001). Sequences representing the 5′-terminal 150 to 170 nucleotides, an intervening poly(A)tract, and the 3′ terminal 110 to 120 nucleotides of the KUN and other mosquito-borne virus genomes, the TBE genome, and the CFA genome were folded by computer using the “MFOLD” program (Mathews et al., 1999) and were based on the assumption of hydrogen bonding between the respective 5′ and 3′ cyclization sequences. Results of this analysis revealed that the 5′ and 3′ NCRs in all flavivirus genomes contained trans complementary sequences upstream and downstream from the conserved cyclization sequences, such that secondary structures of very similar shape and very high thermal stability could be formed between them.

Khromykh et al. predict an extensive trans interaction between the KUN genome 5′ and 3′ NCRs. Very similar structures were predicted for the JE, YF, DEN, TBE, and CFA genomes. The model predicts for the KUN genome that (1) 3′ nucleotides –1 to –8 hydrogen bond to 5′ nucleotides +45 to +52 and (2) 3′ nucleotides –11 to –14 and –68 to –106 hydrogen bond with 5′ nucleotides +77 to +150. As a result, (1) 3′ NCR nucleotides –15 to –67 remain available to form a 3′ cis stem and loop structure, a truncated 3′ SL composed of nucleotides previously comprising the top portion of the long stem and including the distal loop regions (Fig. 6), and (2) 5′ NCR nucleotides +1 to +44 and +53 to +76 are available in cis to form short stem–loop structures, conserving part of the 5′ SL predicted by Brinton and Dispoto (1988) (Fig. 7), including its distal loop. Therefore, the distal loop in the 5′ NCR could still form. The pseudoknot predicted by Shi et al. (1996a) and involving a portion of the 3′ SL conserved in the 5′⧸3′ panhandle structure could also still form, as the semiconserved nucleotides that form the loop atop the short stem in the conventional model of the 3′ SL secondary structure (5′-CANAG-3′; Fig. 6) are available for hydrogen bonding with the conserved remnant of the long SL in the model proposed by Khromykh and co-workers (2001). It is worthy to note that the predicted thermal stability of the panhandle exceeds those of the earlier predicted 5′ SL and 3′ SL by a wide margin. In addition, the predicted thermal stability of mutant KUN RNA 5′⧸3′ hydrogen-bonded complexes that lacked complementary cyclization sequences was very little different from that of the wild-type complex. This suggests that the requirement for the cyclization sequences relates to their specific functional necessity, e.g., in initiation of RNA synthesis or translation or packaging, rather than to their additive effect on thermal stability of the panhandle. One must view all of these hypotheses for base pairing with an open mind for two reasons: (1) Computer power to predict folding and base pairing in nucleic acids is increasing rapidly and therefore any of these hypotheses are subject to evolution based on the state of the art in computer programs and computer power and (2) the “wild card” in any computer-based prediction is the unknown but certain effect of cellular and possibly viral proteins on the secondary and tertiary structure of the RNA.

VI. AdditionalSecondaryStructurePredictions in the 3′ NCR

Proutski and colleagues (1997b) analyzed the distal 330 to 400 nucleotides of published flavivirus 3′ NCR sequences for predictable secondary structure using a genetic algorithm that purportedly simulates the natural folding pathway that takes place during RNA elongation and allows prediction of tertiary interactions as well as secondary structure. They then used a second program to detect compensatory mutations (meaning mutations in both strands of a base-paired region that conserve hydrogen bonding) in the 3′ NCRs of related viruses. The presence of compensatory mutations was taken as an indication that the base pairing in that region was likely to occur in vivo, as random mutational events conserved the double strandedness.

As a result of this analysis, three regions of secondary structure could be defined within what could be described as the core element (Wallner et al., 1995) of 3′ NCRs in mosquito-borne as well as tick-borne viruses (Fig 2, Fig 3). Region I consisted of nucleotides upstream from CS1, CS2, and RCS2 in mosquito-borne virus genomes or the 5′-most extreme of the core element in tick-borne virus genomes. This is a region of high variability in linear nucleotide sequences. However, all flavivirus sequences in region I formed very similar long hairpin structures with a branching stem–loop side structure or, in some cases, a bulge–loop on the 5′ side of the main hairpin. Downstream from this long hairpin in region I was a conserved shorter hairpin of similar configuration. Nucleotides in region II included those assigned to the CS2 and RCS2 boxes plus intervening sequences. Unlike region I, where a single consensus secondary structure was predicted, region II nucleotides exhibited differences in folding patterns that separated the viruses into three groups: (1) DEN and JE subgroup viruses, (2) tick-borne viruses, and (3) YF virus. Region II in genomes of viruses in the DEN and JE subgroups was shown to contain tandem repeat sequences (TL1 and TL2; Fig. 4) that potentially form conserved short SL structures. TL1 is located upstream from RCS2, and TL2 is located between RCS2 and CS2. The secondary structures in region II for tick-borne viruses were so different from those predicted for mosquito-borne viruses that no one element was conserved between the two groups. Region III consisted of nucleotides traditionally believed to constitute CS-1 and the 3′ SL. A 3′-terminal SL structure was predicted by this analysis, just as it was predicted by the MFOLD program used by Khromykh et al. (2001). As in the Khromykh model for panhandle formation, the length of the long stem at the genome 3′ terminus in the Proutski model (Proutski et al., 1997b) is shorter than that predicted by older methods due to the alternative hydrogen bonding of nucleotides that form the lower portion of the long stem shown in Fig. 6. The Khromykh model (Khromykh et al., 2001) and predictions for the secondary structure within the 3′ NCR put forward by Proutski and colleagues (1997b) are not incompatible if one posits that sequences not included in the 3′ SL in the Proutski model are base paired with nucleotides of the 5′ NCR.

The model for the secondary structure of the 3′-terminal 380 nucleotides of the YF genome was subsequently applied to the sequences of several different YF isolates and to the sequences of vaccine strains 17D-204 and 17DD (Proutski et al., 1997a). An association between secondary structure of the 3′ NCR and virus virulence was claimed in that all wild strain genomes were likely to fold similarly and in a significantly different way from the attenuated vaccine strains. In addition, strain 17DD had an intermediate folding pattern that was less different from that of wild strain genomes than was the strain 17D-204 3′ NCR. The authors noted that strain 17DD is more virulent in mice than strain 17D-204 and suggested that mutagenesis of the primary nucleotide sequence of the 3′ NCR to alter the secondary structure could be a useful strategy for vaccine development. The point may be well taken with respect to mice, but after more than 60 years of experience, there is no indication that the strain 17DD vaccine is any less attenuated in humans than is the strain 17D-204 vaccine (Monath, 1999).

The same group (Proutski et al., 1999) applied similar reasoning to an analysis of the results of Men et al. (1996) and Khromykh and Westaway (1997), which were described earlier, in which each of these groups of investigators created internal deletions in the 3′ NCRs of DEN4 and KUN virus genomes, respectively. Consideration of the resulting alterations of the 3′ NCR secondary structure resulting from deletions in the DEN4 genome, in particular, could help explain the observed discordance between the size of the deletion and the growth phenotype of the resulting mutant DEN4 viruses.

Olsthoorn and Bol (2001) also sought to detect a secondary and tertiary structure in the 3′ NCR, and they performed a computer analysis of 191 flavivirus genomic sequences representing four serogroups (DEN, JE, YF, and TBE), also using the MFOLD program. This program yielded a set of suboptimal structure predictions, which were then checked manually for the presence of pseudoknots and for phylogenetic consistency. The analyses confirmed the presence of conserved serogroup-specific secondary structures for each of the four serogroups in the 3′ NCR upstream from the 3′ SL. As in the work of Proutski and colleagues (1997b), the existence of conserved SL structures that involved CS2 and RCS2 was confirmed for viruses of the DEN and JE subgroups. These were paired “dumbbell”-shaped structures that included the conserved stem and the respective TL1 and TL2 loop sequences identified previously. Olsthoorn and Bol (2001) proposed that TL1 and TL2 loops could each form pseudoknots by hydrogen bonding with conserved tetranucleotide sequences pk1 and pk2 present in predicted linear regions of the DEN and JE subgroup 3′ NCRs. For example, in the DEN serogroup, nucleotides in the TL1 sequence (5′-GAAGCUGUA-3′; Fig. 4), which forms a loop within a secondary structure that includes CS2, could form hydrogen bonds (bold letters in the TL1 sequence) with conserved sequence pk1 (5′-ACAGC-3′), which lies downstream from CS2. Similarly, in the upstream dumbbell-shaped secondary structure that includes RCS2, the nucleotides 5′-GCUGC-3′ in TL2 could hydrogen bond with linear sequence pk2 (5′-GCAGC-3′), which also lies between RCS2 and CS2. For JE serogroup viruses, pk1 and pk2 were both located downstream from CS2. In the YF genomes, which lack RCS2 (Fig. 2), three pseudoknots were predicted, one of which was similar in its genesis to the more downstream pseudoknot in DEN and JE serogroup genomes. As in the analyses of Proutski et al. (1997a), the TBE genomes were folded very differently from those of the mosquito-borne viruses, resulting in the formation of a single predicted pseudoknot within the core element. The internally consistent results of this study and the compatibility of the results with the work of others (Proutski 1997a, Rauscher 1997) constitute additional evidence that pseudoknot formation may be required for proper function of the 3′ NCR in flavivirus RNA replication.

VII. Summary andFutureDirections

The flavivirus genome is a capped, positive-sense RNA approximately 10.5 kb in length. It contains a single long ORF, flanked by a 5′ NCR, which is about 100 nucleotides in length, and a 3′ NCR ranging in size from about 400 to 800 nucleotides in length. Replication is thought to occur by a semiconservative mechanism requiring double-stranded replicative form RNA to generate replicative intermediate RNA containing nascent single strands. The 5′ and 3′ NCRs must play a role in the initiation of negative-strand synthesis on virus RNA released from entering virions, switching from negative-strand synthesis to synthesis of progeny plus strand RNA at late times after infection, and possibly in the initiation of translation and in the packaging of virus plus strand RNA into particles. However, none of these mechanisms is as yet well defined. The presence of conserved and nonconserved complementary nucleotide sequences near the 5′ and 3′ termini of flavivirus genomes suggests that “panhandle” or circular RNA structures are formed transiently by hydrogen bonding at some stage during RNA replication.

Genomes of flaviviruses that differ according to vector (tick-borne viruses vs mosquito-borne viruses) and by antigenic subgroup may also contain unique or characteristic nucleotide sequence features in their noncoding regions. However, a few such features are conserved in all flavivirus genomes. These are as follows: (1) All vector-borne flavivirus genomes begin with the dinucleotide sequence 5′-AG and end with the complementary nucleotide sequence UC-3′. The functional significance of these sequences is not known but may play a role in cyclization of the genome during or as a prerequisite for replication. (2) While the nucleotide sequence of the 5′NCR is not well conserved among flavivirus species, the 5′NCRs of all flavivirus genomes do contain a predicted stable stem and loop structure. (3) All flavivirus genomes contain a 3′-terminal approximately 100-nucleotide sequence that is predicted to form a highly stable stem and loop structure, the 3′ SL. The actual nucleotide sequence of this domain is well conserved among viruses of the same subgroup but is much less well conserved across subgroups. The 3′ SL is known to bind the virus-coded RNA-dependent RNA polymerase protein NS5 and to be required for binding of a virus-coded helicase, NS3, to the 3′NCR. In addition, the 3′ SL apparently binds several cellular proteins required for virus replication. One of these proteins has been identified as the translation elongation factor eF1-α. Mutations of the 3′ SL can be shown to affect the host range of the virus, suggesting that binding of host cell proteins is critically important for replication. (4) Within the nucleotide sequence forming the distal loop of the 3′ SL in all vector-borne flavivirus genomes, there is a conserved nucleotide sequence, 5′-CACAG. The function of this conserved pentanucleotide sequence in replication is unknown.

Functionally analogous group- and subgroup-specific nucleotide sequence elements have been identified in the 5′ and 3′ NCRs of flavivirus genomes. For example, mosquito-borne flavivirus genomes contain conserved complementary 8-nucleotide sequences. One is located in the 3′ NCR, referred to as the cyclization sequence (CYC). The other is located in the 5′ terminus of the long ORF, just downstream from the start codon that initiates translation of the capsid protein gene segment, referred to as the “5′CS.” The CYC sequence and the 5′CS are thought to hydrogen bond, helping to form a panhandle structure during RNA replication. Tick-borne flavivirus genomes contain two analogous pairs of 5′⧸3′ complementary sequences, each 11 nucleotides in length. The 3′-terminal motifs in tick-borne virus genomes are herein referred to as C1′ and C2′, and the 5′-terminal motifs are referred to as C1 and C2. The CYC sequence in mosquito-borne virus genomes and C1′ are analogous in their position in the 3′ NCR, just upstream from nucleotides comprising the 3′ SL. However, the complement of C1′, C1, lies wholly within the 5′ NCR in tick-borne virus genomes, not in the ORF. C2′ lies upstream from C1′, and its complement at the 5′ end of the genome, C2, lies within the ORF, in a location analogous to that of the 5′CS in mosquito-borne virus genomes. The 5′ and 3′ complementary sequences are necessary for viral RNA replication, but the specific nucleotide sequence of, for example, the CYC sequence and the 5′CS in mosquito-borne virus genomes can be substituted by a random pair of complementary sequences 8 nucleotides in length without much loss of replication efficiency.

Mosquito-borne viruses of the dengue subgroup contain the shortest 3′ NCRs in this group (∼400 nucleotides), with a minimum of conserved features. CS1 is a 20-nucleotide segment just upstream from the 3′ SL. The 8-nucleotide CYC sequence comprises the upstreammost part of CS1. Deletion of CS1 nucleotide sequences was specifically lethal for DEN4 virus replication, but the functional requirement for the 12 nucleotides of CS1 that do not compose the CYC sequence is not known. CS2 is a 23-nucleotide conserved element that lies 12 to 22 nucleotides upstream from CS1. CS2 is one of a pair of repeated sequence elements in the 3′ NCR; the repeat sequence RCS2 lies 51 to 62 nucleotides upstream from CS2 in the genome of mosquito-borne viruses. All or parts of CS2 and RCS2 were dispensible for DEN-4 virus replication, but the respective mutant viruses were mildly to markedly reduced in replication efficiency.

3′ NCRs of viruses in the JE subgroup are larger in size than those of the DEN subgroup (∼500 to ∼600 nucleotides) and contain additional conserved repeat sequence motifs, in addition to all the conserved motifs described for the 3′ NCR in DEN virus genomes. These additional repeat sequences are referred to as CS3 and RCS3, respectively, in relation to their increasing distance from the 3′ terminus of the genome compared to CS2 and RCS2. CS3 and RCS3 are each 28 nucleotides in length and are separated by 120 nucleotides in the 3′ NCRs of JE subgroup viral genomes.

The YF virus genome is unique in several ways from that of the other mosquito-borne flaviviruses, including its organization of the 3′ NCR, which contains CS1 and CS2 but lacks the repeat segment RCS2 found in the other genomes of the mosquito-borne virus group. The YF strain 17D vaccine virus 3′ NCR also contains a series of three unique 40-nucleotide tandem repeats at the 5′ end of its 3′ NCR. Also, unlike other flavivirus genomes, the yellow fever 3′ NCR contains three in-frame stop codons for translation of the ORF, the last of which lies within the upstream boundary of the first of the tandem repeats.

For tick-borne flaviviruses, the 3′ NCRs of tick-borne encephalitis virus strains have been analyzed extensively. The genome of TBE strain Neudorfl virus genome contains the longest 3′ NCR in this group, at 764 nucleotides. This segment has been divided, by comparative analysis of the nucleotide sequences of strain Neudorfl and other TBE viruses and the results of mutagenesis experiments, into a “core element” and a “variable region.” The core element is directly analogous to the entirety of the DEN virus 3′ NCR and includes the most distal portion of the 3′ NCR, containing the 3′ SL. The cyclization sequence, C1′, lies within a mixed purine⧸pyrimidine-rich tract upstream from the 3′ SL, which is analogous in position and sequence to CS1 in mosquito-borne flavivirus genomes. Upstream from this PR segment exists consecutive pyrimidine- and purine-rich segments, each about 14 nucleotides in length. These could be said to be analogous to CS2 found in mosquito-borne virus genomes. The second cyclization sequence motif, C2′, lies upstream from C1′, also within the core element. The core element also includes one of a pair of long tandem repeats in the archetypal strain Neudorfl 3′ NCR, R3, which is 76 nucleotides in length.

The structure of the variable region, which comprises the upstream portion of the 3′ NCR in strain Neudorfl and related TBE virus genomes, is complex. The variable region contains a second iteration of the R3 sequence found in the core element. It also contains additional repeated sequences, R1 and R2. R1 is repeated three times in the strain Neudorfl genome. One of these repeats lies within the 3′ terminus of the ORF. The single repeat of R2 also lies within this same distal segment of the ORF. The most striking feature of the variable region in TBE genomes is the presence of a poly(A) segment that can be as long as 250 nucleotides. The function of the variable region remains a puzzle because there exist naturally occurring highly related TBE virus variants with much shorter 3′ NCRs compared to that of strain Neudorfl, completely lacking a variable region. In addition, the variable region nucleotide sequence can be completely removed from the strain Neudrofl genome by site-directed mutagenesis or by spontaneous mutation that occurs during repeated in vitro tissue culture passage, with no effect on virus replication in tissue culture or on mouse neurovirulence.

In recent years, computer-based methods have been used to predict the existence of additional secondary and tertiary structures within the 3′ NCR and to predict additional interactions between 5′ and 3′ NCRs based on the potential for hydrogen bonding of complementary segments in these two regions of the genome, in addition to that afforded by the known cyclization sequences. The significance of this work remains to be seen and requires better knowledge and understanding of the possible effects of viral and cellular proteins on folding of the RNA, which can only be gained by further in vitro studies with engineered RNAs and virus infectious clones.

References