INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

GB virus C (GBV-C)/hepatitis G virus (HGV) is a recently described agent which has been proposed as a member of the Flaviviridae family, distantly related to GBV-A and even more related to HCV (18, 33). The virus is able to infect humans parenterally and vertically, although its true pathogenicity is under extensive study (2, 22). Its viral genome is a positive-strand RNA that appears to be approximately 9.4 kb long and that includes a long open reading frame. Based on sequence homology with HCV, it has been proposed that the predicted HGV polyprotein contains two putative envelope proteins (E1 and E2), an RNA helicase, a serine protease, and an RNA-dependent RNA polymerase (18, 27). Initial suggestions of an apparent lack of core coding region have very recently been challenged, when the expression of sequences upstream of the E1 coding region was demonstrated (41).

Sequence analysis of the most 5'-terminal region shows a certain degree of heterogeneity among different isolates. Based on a study of 44 sequences from around the world, Muerhoff et al. (24) initially proposed the existence of three genotypes, i.e., 1, 2, and 3, the first two including two subtypes each. When such analysis was extended to a fragment of 374 nucleotides (nt) of the 5' untranslated region (UTR) from 83 isolates, four groups, namely, 1, 2a, 2b, and 3, were finally established (23).

Up to now, at least two attempts have been made in order to provide a rapid method for HGV typing (25, 29). However, the growing number of HGV sequences deposited in the GenBank database and the changing view regarding types and subtypes led us to improve our initial method. To reach this goal, all so-called complete HGV sequences deposited in GenBank at the time of submission of this manuscript (n = 22) and partial sequences from the 5' UTR were used for phylogenetic analysis and predictive restriction fragment length polymorphism (RFLP) patterns in this study. Since a consensus nomenclature about GBV-C/HGV grouping and subgrouping is still lacking and because the observed degree of genomic diversity at 5' UTR is not similarly represented within other genomic regions (34), we will use the terms group and subgroup instead of type and subtype. In this study, we propose a rapid method for GBV-C/HGV classification by 5' UTR cDNA amplification followed by enzymatic cleavage to obtain group- and subgroup-specific RFLP profiles.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

RNA extraction and cDNA synthesis. RNA was initially obtained from 200 µl of a reference plasma, PNF2161 (18), kindly provided by Margaret Gallagher (Centers for Disease Control and Prevention, Atlanta, Ga.). After reverse transcription (RT) and amplification conditions were settled, serum samples from 100 parenterally HIV-infected patients were studied. RNA was obtained from serum by guanidinium isothiocyanate-acidic phenol extraction (5), as previously described (28). RT was performed by incubating the template (equivalent to 100 µl of serum) in the presence of 200 pM outer antisense primer (GOA) 5' CCC GGC CCC CAC TGG TCC TTG 3' and 200 U of Moloney murine leukemia virus for 1 h at 37°C. Reverse transcriptase was inactivated by being heated to 95°C for 10 min, and cDNA was kept cold until PCR amplification, which was performed immediately.

PCR amplification. An aliquot of cDNA was amplified in a volume of 50 µl by using GOA and an outer sense primer (GOS) 5' CGG CAC TGG GTG CAA GCC CCA 3'. After a denaturation step (2 min 30 s at 94°C), 35 cycles of PCR, with 1 cycle consisting of denaturation (30 s at 94°C), annealing (30 s at 55°C), and primer extension (45 s at 72°C), followed. Nested PCR was performed in a volume of 50 µl after transfer of 5 µl from the first round of PCR to a mix containing 200 pM inner sense primer (GIS, 5' AGC CCC AGA AAC CGA CGC CTA 3') and an inner antisense primer (GIA, 5' TAT TGG TCA AGA GAG ACA TTG 3'). This second PCR was performed after a denaturation step as given above, followed by 40 cycles of PCR, with 1 cycle consisting of denaturation (30 s at 94°C), annealing (30 s at 53°C), and extension (45 s at 72°C).

cDNA sequencing. The specificity of the reaction was assessed by direct sequencing of PCR amplicons. Briefly, 3 nested PCR mixtures were pooled, purified in 7% polyacrylamide gels, eluted in 0.5 M NH₄ acetate, 0.01 M EDTA, and 0.1% sodium dodecyl sulfate, phenol-chloroform extracted, ethanol precipitated, and resuspended in 10 µl of sterile distilled water. Manual and/or automatic sequencing by the chain termination method (31) was performed alternately with [-³²P]dATP-labelled GIS or GIA primer or with either primer in the presence of fluorescent dye-labelled chain terminators, respectively.

Molecular evolutionary genomic analysis. Considering the location and length (325 to 329 bp) of the above-mentioned predicted amplicons, both reported complete (n = 120) and partial 5' UTR GBV-C/HGV sequences (n = 30) from GenBank, as well as from virus-infected Argentine patients (n = 21), were included for alignment and phylogenetic analysis. An initial phylogenetic analysis was performed on downloaded GenBank sequences. DNA alignments were generated with the Clustal X program (11, 13, 38). Evolutionary distances between sequences were determined with the DNADIST program (Kimura two-parameter method) of the PHYLIP package version 3.5c (8). The computed distances were used for the construction of phylogenetic trees by the neighbor-joining method of NEIGHBOR program. The robustness of the trees was assessed by bootstrap resampling with the programs SEQBOOT (to generate 1,000 reshuffled sequences), DNADIST, and NEIGHBOR. The consensus tree was calculated with CONSENSE. Bootstrap values of less than 70% were regarded as not providing strong support for the phylogenetic grouping. The tree was obtained from RETREE program (PHYLIP package) with the midpoint rooting option. Final graphical output was created with the program TREEVIEW. The MapDraw program (DNASTAR Package, Lasergene for Windows) was selected to detect group- and subgroup-specific restriction sites.

Nucleotide sequence accession numbers. The following sequences from GBV-C/HGV Argentine strains were deposited in GenBank: AF081562, AF081564, AF116325, AF116326, AF116327, AF116328, AF116329, AF116334, AF116335, AF116338, AF116339, AF116340, AF081561, AF116330, AF116332, AF116333, AF081563, AF116324, AF116331, AF116336, and AF116337.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Molecular genomic GBV-C/HGV analysis and predicted RFLP patterns. In order to determine whether restriction sites of GBV-C/HGV sequences deposited in the GenBank database might be useful in distinguishing groups and subgroups, 150 5' UTR sequences discussed above were selected. Of the 150 sequences, 120 encompassed the complete length of 325 to 329 nt extended between the 5' ends of GIS and GIA primers; 22 of them were regarded as full-length genomic sequences. Another group of 30 sequences (AF006957 to AF006986) did not have the first 37 nt (positions 10 to 46 according to the GBV-C numbering system) of our amplified products. Consequently, our sequence alignment used a fragment corresponding to positions 47 to 335, following the GBV-C numbering system.

View larger version (20K): [in this window] [in a new window]   FIG. 1.   Phylogenetic tree of 290-nt 5' UTR sequences deposited in GenBank. The GenBank accession numbers of the isolates are given in Table 1. A distance scale (in nucleotide substitutions per position) is shown.

View larger version (69K): [in this window] [in a new window]   FIG. 2.   Predicted RFLP patterns obtained by computer analysis of 150 5' UTR GBV-C/HGV sequences downloaded from GenBank. The 21 Argentine GBV-C/HGV sequences in this study also supported the proposed RFLP method. RFLP patterns with an asterisk belong to 290-bp partial sequences starting at position 47 according to the GBV-C numbering system. The numbers under restriction fragment length polymorphism indicate the expected sizes (in base pairs) of the fragments, and the numbers of sequences with a given pattern in group or subgroup 1, 2a, 2b, and 3 and the total number (n) are shown.

RT-PCR amplification. As a positive control, the specificity of the designed primers was assessed when only products of the expected size (no spurious bands) were observed after RT and subsequent amplification of the RNA obtained from the reference plasma PNF2161 (U44402). This system also allowed the detection of GBV-C/HGV RNA in 21 of 100 parenterally HIV-infected patients.

RFLP and cDNA sequencing. All Argentine isolates (n = 21) were grouped by the RFLP method described in this study. GBV-C/HGV strains were classified as group 1, 2a, 2b, and 3. The accuracy of the proposed methodology was subsequently substantiated by cDNA sequencing, since the phylogenetic analysis of the sequences obtained showed complete agreement with the observed RFLP patterns from each sample (Fig. 3 and 4). GBV-C/HGV sequences from Argentine strains deposited in GenBank, belonged to the following groups: group 1, AF081562; subgroup 2a, AF081564, AF116325, AF116326, AF116327, AF116328, AF116329, AF116334, AF116335, AF116338, AF116339, and AF116340; subgroup 2b, AF081561, AF116330, AF116332, and AF116333; and group 3, AF081563, AF116324, AF116331, AF116336, and AF116337.

View larger version (79K): [in this window] [in a new window]   FIG. 3.   RFLP characterization of GBV-C/HGV group 1 (lanes 4 and 5) and group 3 (lanes 2 and 3). AatII (lanes 2 and 4) or HinfI (lanes 3 and 5) were used. Lane 1 contains 50-bp ladder.

View larger version (64K): [in this window] [in a new window]   FIG. 4.   RFLP characterization of GBV-C/HGV 2a (lanes 4 and 5) and 2b (lanes 2 and 3) subgroups. AciI (lanes 2 and 4) and HinfI were used. Lane 1 contains 50-bp ladder.

Phylogenetic analysis. Three major groups, i.e., 1, 2 (including two subgroups) and 3 were observed (Fig. 1). This result was substantiated after 1,000 replicates for bootstrap analysis. Grouping into subgroups 2a and 2b was undoubtedly supported by 70 and 93% of all trees, respectively. In contrast, no sufficient support was observed for subgroups within group 1 (45 and 86% for the two observed branches) with the sole exception of sequence AF078048 from Singapore (40) which might represent a different (sub)group. RFLP profile from this sequence is shown among AatII restriction patterns in Fig. 2, followed by a "?" and in the phylogenetic tree from Fig. 1 as sequence 149.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Our sequence alignment confirmed isolate grouping of GBV-C/HGV strains included in a recently reported study by analysis of a partially overlapping 374-nt region (positions 79 to 447).

Considering all 150 sequences, three major groups (i.e., 1, 2 [including two subgroups], and 3) were observed as depicted in Fig. 1.

The initial visual and computational analyses, including both their alignment and phylogenetic relatedness, of 94 sequences allowed the detection of conserved regions to design universal primers, as well as group- and subgroup-specific restriction sites. When extended to additional 56 sequences (total number, 150) the same clustering was observed, although a few more RFLP patterns were evident. Although not closely related to other group 2a strains, sequence analysis of the reference isolate PNF2161 also supported the proposed methodology. Moreover, cDNA sequencing of 21 Argentine isolates confirmed GBV-C/HGV typing based on 5' UTR RFLP (data not shown).

Interestingly, sequence U75356 (Chinese isolate HGVC964 [44]) was putatively assigned to group 2a or 3 according to the length and subregion location analyzed at 5' UTR (23) (not included in the phylogenetic analysis shown). It was classified into a separate group by another phylogenetic tree when we analyzed its sequence, associated with the observed insertion of three consecutive nucleotides (data not shown). This sequence exhibited the predicted group 3 RFLP pattern, also in agreement with an analysis of its artificially combined E2-NS3-NS5b genome (23). Subsequently, sequence U75356 was described as the unique example of a putative genotype 4 by Smith et al. (34).

At least three items deserve special comments. First, although to a lesser extent than with 5' UTR HCV isolates (since genetic distances between GBV-C/HGV groups are most similar to those observed between HCV subtypes) GBV-C/HGV strains show sufficient heterogeneity to allow their clustering in at least three major groups, group 2 including two subgroups and group 1 possibly including at least one sequence (sequence 149 [Fig. 1]) clustered separately. Considering that such low level of heterogeneity might signify minor changes at the amino acid level in coding regions, antigenic diversity would seem to be unlikely reflected in clinical or biological differences of this viral infection. However, the effects of mutations at 5' UTR should be thoroughly investigated, since significant differences of biological properties could be associated with them, as has been shown with other viruses, i.e. poliovirus, HCV, etc. Second, of 150 previously reported sequences and of the 21 Argentine isolates shown here, almost no discrepancies between phylogenetic and RFLP analyses of the 5' UTR were observed with the unique exception of sequence AF078048 (40) which clustered with genogroup 1 but exhibited a group 3 RFLP profile (Fig. 1 and 2). As stated above, this isolate might represent a new (sub)group. Third, regarding a putative geographical distribution of genogroups, it should be noted that all groups and subgroups were observed in Argentine patients. These results extend the initial observations of Muerhoff et al, who documented the existence of two 2a and one 2b isolate from Argentina (23). GBV-C/HGV genotypic prevalence in Argentine patients and blood donors will be published elsewhere.

Based on a recent report showing that a 374-nt sequence from the 5' UTR truly represents differences in the whole genome (23), the methodology described here might be useful for a rapid screening of GBV-C/HGV isolates when epidemiological studies are undertaken. Indeed, selected 5' UTR regions from both studies partially overlap. Furthermore, another recent study also supports the use of the 5' UTR for GBV-C/HGV grouping (34). Mukaide et al. have also proposed RFLP typing of GBV-C/HGV variants based on the phylogenetic study of shorter 5' UTR sequences (183 nt). However, their analysis of 33 sequences was unable to substantiate group 2 subgrouping (25). Moreover, when we analyzed Argentine sequence AF116331, its phylogenetic grouping showed that it belongs to group 3, as also exhibited by our RFLP method (restriction patterns H1 and Aa2). However, according to the previously mentioned procedure (25), it should have been classified as HG type, assumed to correspond to group 2 following Muerhoff's nomenclature (23). Likewise, recently available sequences downloaded from GenBank (i.e., AF038810 to AF038814 and Y15266) exhibit RFLP patterns not initially described by Mukaide et al. In this regard, it should be stressed that other researchers have demonstrated that only sequences that are long enough could be used to properly perform phylogenetic analysis to obtain reliable data (23, 34).

In our proposed method, the use of GIS primer allowed the eventual detection of a HinfI cleavage site at position 38 of a given amplicon (nt 48 according to the GBV-C numbering system). Such a site appeared to be important to detect a high percentage of genogroup 2 sequences, while group 2 subgrouping is consistently determined by AciI digestion (Fig. 2). However, among 84 group 2a sequences downloaded from GenBank and characterized by phylogenetic relationships supported by bootstrap analysis, two (accession nos. AF038812 and AF038813; [20]) deserve special comments. These two sequences are shown with a dagger symbol in Fig. 2 among sequences producing H1 (HinfI) and Aa1 (AatII) restriction patterns. Therefore, such a profile is compatible with genotype 1. For properly assigning genogroup to these two unusual sequences, it was necessary to perform also AciI digestion, where an Ac2 pattern was observed, in contrast to all genuine group 1 sequences where this profile was not documented (n = 22).

A proposed algorithm of enzyme digestions is shown in Fig. 5. We suggest initially cleaving amplicons with HinfI. According to the pattern observed (H1 to H10), subsequent digestions could be performed either with AatII (usually to distinguish between groups 1 and 3) and AciI to differentiate subgroups 2a and 2b. From a practical point of view, the two subgroup 2a sequences discussed above, AF038812 and AF038813, appeared to add a third step in the algorithm of digestions to distinguish between group 1 and group 3 [i.e., (i) HinfI, (ii) AatII, and (iii) AciI]. However, it is worth mentioning that these two sequences belong to a single isolate (called S4) from which five clones were analyzed, only two (clones 4 and 5) showing the loss of a HinfI site due to the presence of a C at position 157, instead of G (20). Therefore, the possibility of picking an RT-PCR amplification error or the detection of nonpredominant quasispecies cannot be ruled out. Whether such sequences truly represent a feature from a proportion of Spanish patients remains to be determined. This methodology allowed proper assignment of 149 of 150 sequences downloaded from GenBank and 21 of 21 Argentine strains (n = 171; 99.4% accuracy).

View larger version (31K): [in this window] [in a new window]   FIG. 5.   Proposed algorithm for GBV-C/HGV typing by enzymatic digestion.

With regard to other sequences not included in this study, it should be mentioned that those entire polyprotein open reading frame sequences (AB003291 and AB003292) proposed to be new genotypes (36) could not be analyzed, since information about their sequence at 5' UTR does not show 111 nt from the most 5' end of the corresponding amplicon of our study. A similar exclusion criteriawith variable lengths in the region to be analyzedwas used for other several sequences (1, 21, 44).

To be useful, a typing method based initially on RT-PCR detection of genomic sequences must be able to detect most, if not all, strains. Therefore, universal primers are usually selected from conserved regions. This was the case when we initially designed primers used in this method. However, the increasing number of GBV-C/HGV sequences deposited in GenBank demonstrated that few mismatches were evident with some sequences. This appeared to be a problem with sequences like AF015870 to AF015872 and AF015876 to AF015878 (43) for which GOS primer exhibits a mismatch at its 3' end. Potential inefficacy in the RT-PCR amplification might be overcome by using primers with ambiguities at the critical nucleotide positions. As can be observed by visual and computer analyses, the mismatch at the most 3' end of the outer sense primer would be crucial to hybridize the cDNA synthesized by extension from the 3' end of the antisense primer, as shown in Fig. 6.

View larger version (17K): [in this window] [in a new window]   FIG. 6.   Sequence complementarity between GOS primer and GBV-C/HGV RNA and cDNA. AF015870-2, AF015870 to AF015872.

At first glance, it would appear that sense primer cannot efficiently amplify sequences like AF015870 to AF015872 and other related sequences. However, Kwok et al. (15) have elegantly demonstrated that this is not the case. They unambiguously demonstrated by experiments done in triplicate that despite an A at the 3' end of a primer, efficiency of amplification is 100% when the corresponding position at the template is occupied by a C. This effect is reciprocal (either the mismatch is in the template or in the primer). This research group, among many others, have also observed that a single internal mismatch (as very unfrequently observed with few GBV-C/HGV sequences) does not significantly affect yield of amplification (15, 35).

So far, whether GBV-C/HGV sequence variability influences the course of infection is unknown. However, such variation has been associated with a different outcome with the related HCV (3). In this regard, Japanese (42) and German (10) researchers have reported strain-specific association with fulminant hepatitis. Although such findings were not subsequently extended, the recent discovery of GBV-C/HGV in bone marrow, spleen, and liver (16) deserves thorough studies regarding cell tropism and potential influence of viral genomic variations in the infection of such tissues. While newly discovered sequences will probably warrant periodical updating of the proposed methodology, this study also contributes to the knowledge of the molecular epidemiology of viral hepatitis in Argentina (28, 30, 37).