Various methods have been used for HCV genotyping, including genomic amplification and sequencing [10, 14-16], polymerase chain reaction (PCR) with genotype-specific primers [5, 8], restriction fragment length polymorphism of the PCR amplicons [13, 17], differential hybridization [18], and serologic genotyping [11, 19]. Genomic amplification and sequencing, followed by sequence comparison and phylogenetic tree construction for confirmation, is currently considered the gold standard; the genomic regions commonly used for this approach include the HCV core region, envelope 1, and nonstructural region 5 (NS5). This method, however, is expensive and labor intensive, making it difficult to study HCV genotypes in a large number of patients.

Several alternative methods have been proposed. Okamoto and colleagues [5] and Chayama and coworkers [8] have described PCR done with genotype-specific primers derived from the HCV core region and from NS5, respectively. These methods rely on the use of genotype-specific primers that anneal to sequences unique to specific HCV genotypes. Genotype-specific primers derived from the HCV core region allowed the differentiation of HCV types 1a, 1b, 2a, and 2b [5]; genotype-specific primers derived from NS5 allowed the differentiation of HCV types 1a, 1b, 2a, 2b, and 3b. Restriction fragment length polymorphism of the PCR amplicons relies on the presence of unique and genotype-specific nucleotide substitutions that are recognized and digested by restriction enzymes into fragments that can be separated be gel electrophoresis [13, 17]. The 5' untranslated region (5' UTR) is commonly used for this approach. Hybridization of PCR amplicon mounted on a solid phase using genotype-specific probes has also been used for HCV genotyping [18]. Finally, serologic approaches have also been used to determine HCV genotypes. These approaches rely on the different amino acid sequences encoded by different nucleotide sequences from different HCV genotypes. Patients infected with different HCV genotypes may therefore have antibodies directed to genotype-specific amino acid sequences. Polypeptides and synthetic peptides derived from nonstructural region 4 (NS4), as well as synthetic peptides derived from the HCV core region, have been used for serologic genotyping [11, 19]. These peptides are used to coat enzyme immunoassay plates, and specific binding by a serum specimen is determined by peptide competition. The HCV genotypes assigned by this method are commonly referred to as serotypes. Conventionally, the term "HCV serotype" refers to different viral types determined by a panel of cross-neutralization antibodies. Because recognition of genotype-specific peptides by a patient's antibody response reflects the difference in amino acid sequences and thus nucleotide sequences, the results of this approach are more appropriately referred to as "serologically defined genotypes" or "serologic genotypes."

The concordance among these different HCV genotyping methods has been questioned. In a previous study [20], we compared six different genotyping methods and found that most of them had high concordance with each other. The only method that we identified as unsuitable for U.S. patients was PCR with genotype-specific primers derived from the HCV core region, proposed by Okamoto and colleagues [11]. This method gave false-positive signals for HCV type 1b in patients who were actually infected with types 1a and 3a [20].

Because all molecular biological genotyping methods rely on PCR as a first step, those systems based on the 5' UTR, the most conserved genomic region, should be the most sensitive. Recently, a new genotyping system based on reverse hybridization of the labeled PCR amplicon derived from the 5' UTR (line probe assay, INNO-LiPA HCV, Innogenetics, Ghent, Belgium) was developed and has been used widely by investigators in Europe [21, 22]. Whether this line probe assay is also useful in patients with chronic HCV infection in the United States is unknown.

In this study, we sought to 1) assess the concordance of this line probe assay with other assays on a panel of well-characterized serum specimens to verify the line probe assay's validity; 2) determine the distribution of various HCV genotypes in a large sample of patients with chronic hepatitis C seen in tertiary referral centers in the United States; and 3) evaluate the clinical characteristics of patients infected with different HCV genotypes.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

We studied 438 patients from three different groups (Table 2). All patients were from the United States, had chronic hepatitis C, and were seen in tertiary referral centers. Most had been referred by their physicians or by a gastroenterologist or hepatologist for inclusion in experimental antiviral therapy programs. The first group consisted of 137 patients with chronic HCV infection whose serum specimens had previously been characterized with six different genotyping methods [20]. The specimens of these patients were studied to verify the validity of the line probe assay. The second group consisted of 248 patients from nine centers in the United States who had participated in a randomized, controlled study of interferon- therapy. One of these 248 patients was negative for the antibody to HCV (anti-HCV), and another had no serum specimen available, which left a total of 246 patients for this study. A careful review of the serum bank showed that 40 patients were included in both group 1 and group 2. To avoid duplication, we used only the specimen obtained from each patient before interferon- therapy was started. For patients for whom serum specimens obtained before treatment on the day of study entry were not available, specimens collected 1 or 2 months before treatment were used for HCV RNA quantitation and genotyping. A previous study [23] has shown that the viremia levels of patients with chronic HCV infection usually remain about the same over time. The details of the role of HCV genotype as a predictor of subsequent response to interferon- therapy will be discussed in another report (Lindsay KL and colleagues. In preparation). The third group consisted of 95 patients seen at the University of Florida in a prospective study of immune-mediated mechanisms of hepatocellular damage.

All patients studied were seropositive for anti-HCV and had had abnormal serum aminotransferase levels for at least 6 months. None had biochemical or serologic evidence of other causes of liver disease; in particular, all were seronegative for hepatitis B surface antigen. Duration of HCV infection was established by detailed history taking and was estimated by the clinical investigator (available in 258 patients).

Antibody to Hepatitis C Virus and Detection and Quantitation of Hepatitis C Virus RNA

We detected anti-HCV by using second-generation enzyme immunoassay (Ortho Diagnostics, Raritan, New Jersey or Abbott Diagnostics, North Chicago, Illinois). Serum specimens were tested for HCV RNA by reverse transcription "nested" PCR with primers derived from the highly conserved 5' UTR and quantitated by branched DNA (bDNA) signal amplification assay (Quantiplex HCV RNA, version 1.0, Chiron Corp., Emeryville, California) as described previously [24, 25]. All reverse transcription PCR assays were done in a single laboratory; the performance of the PCR assays has been previously verified to have specificity and sensitivity of 100% against a coded serum panel prepared by the Chiron Corporation.

Serum HCV RNA levels were measured by bDNA signal amplification assay (bDNA, Quantiplex HCV, version 1.0, Chiron Corp.). Because the bDNA assay underestimates HCV RNA levels in patients infected with HCV types 2 and 3, the appropriate correction factors (times 3 for type 2 and x 2 for type 3) were applied to obtain accurate levels of viremia [26-28]. The bDNA assay accurately measures HCV RNA levels for HCV types 1, 4, 5, and 6; thus, no conversion was necessary for these types.

Hepatitis C Virus Genotyping

The details of the six HCV genotyping methods used to characterize the serum specimens of group 1 have been described previously [20]. These methods are PCR with genotype-specific primers based on the HCV core region and the genomic region of NS5, restriction fragment length polymorphism based on the 5' UTR, direct sequencing of the NS5B region, and serologic genotyping based on NS4 recombinant and synthetic peptides.

The line probe assay was used to assess HCV genotyping as previously described [21]. Briefly, the 5' UTR was amplified using "nested" PCR with biotinylated primers. The labeled amplicon was allowed to hybridize with probes derived from various HCV genotypes mounted on a strip. After stringent washing, streptavidin labeled with alkaline phosphatase was used to trace the hybridized products, and nitroblue tetrazolium and 5-bromo-4-chloro-3-indoyl-phosphate were used as substrates. To ensure that this assay was consistently concordant with other tests, 342 specimens were tested by restriction fragment length polymorphism of the PCR amplicon generated from the 5' UTR, and 339 specimens were tested by a serologic genotyping assay based on the use of peptides derived from two regions of NS4 (amino acid positions 1691 to 1708 and 1710 to 1728) as capture antigens [20, 29].

The HCV genotype nomenclature used in this report is that proposed by an international panel [4] and is the most widely accepted of all nomenclature systems. We use "genotype" as a general term. The major types, designated 1, 2, 3, and so on are referred to as "types," and the subtypes a, b, c, and so on are referred to as "subtypes" (Table 1).

Hepatitis C virus types 2a and 2c cannot be distinguished on the basis of the 5' UTR, but they clearly separate into different subtypes by phylogenetic tree analysis of the subgenomic region of NS5 [30]. Therefore, it is possible that specimens genotyped as 2a were in fact 2c. In our study, the NS5B nucleotide sequences from three patients with HCV type 2a were determined by the dideoxy chain termination method and analyzed with both neighborjoining and maximum likelihood analyses. All three were specimens assigned as HCV type 2a (data not shown).

Liver Histology

Baseline liver biopsy specimens were available for 262 patients, and the investigators assessing these specimens were blinded. The Knodell score and its individual components (periportal, lobular, and portal inflammation and fibrosis) [31] and the histologic inflammatory index (summation of the lobular, periportal, and portal inflammation scores) [32] were assessed.

Statistical Analysis

All data were analyzed using SPS for Windows (SPS, Inc., North Chicago, Illinois). Once a specimen was included in the study, the data were used for the final analysis; this includes serum specimens that were negative on reverse transcriptase PCR. This approach is similar to the intention-to-treat analysis in clinical studies on therapeutic agents.

For the analysis of clinical associations, we used analysis of variance, which assumes normality, to compare the means for the variables across groups. For those cases in which the F test indicated statistically significant differences, pairwise multiple comparisons using the Student-Neuman-Keuls test were applied. The Mann-Whitney rank-sum test and the Spearman rank correlation coefficient were applied for those variables for which normality could not be assumed.

Results

Top Methods Results Discussion Author & Article Info References

Line Probe Assay

Amplification of the 5' UTR using the labeled primers and subsequent genotyping by the line probe assay have a sensitivity similar to that of our in-house PCR assay, which is also based on the 5' UTR (our in-house assay was determined to have a consistent sensitivity less than 1000 genomes/mL; in most runs, our detection limit was less than 500 genomes/mL). The line probe assay was repeated on 25 specimens, and the results were 100% reproducible on this small number of specimens. The line probe assay, which relies on differential hybridization of PCR amplicons derived from the 5' UTR, is probably more sensitive in identifying mixed infection than other genotyping methods, which amplify less conserved regions of the genome. Accordingly, if a specimen is found to have genotypes 1a and 3a by the line probe assay and 1a by another method, the results should still be considered concordant.

In the 137 specimens that had previously been genotyped by six other methods, the concordance of the line probe assay and the other methods was 100% (specificity of 100%). In the 342 specimens that were genotyped by both restriction fragment length polymorphism analysis of the 5' UTR and line probe assay, the concordance of the two methods was 98.2%. The results were discordant in 6 specimens (in 1 specimen, restriction fragment length polymorphism assigned type 1 and the line probe assay assigned type 2; in 5 specimens, restriction fragment length polymorphism assigned type 2 and the line probe assay assigned type 1). Of the 339 specimens that were serologic genotyped on the basis of the NS4 genomic region, the line probe assay and serotyping had a concordance of 98.5%. Five specimens were discordant (in 1 specimen, serologic genotyping assigned type 1 and the line probe assay assigned type 2; in 1 specimen, serologic genotyping assigned type 1 and the line probe assay assigned type 3; and in 3 specimens, serologic genotyping assigned type 2 and the line probe assay assigned type 1).

Three patients who were positive by PCR as determined by ethidium bromide staining showed no hybridization signals on the line probe assay strips. Overall, the line probe assay was able to assign HCV genotype in 417 of the 421 specimens (99.1%) that were positive for HCV RNA by PCR. If all 438 specimens tested were considered, the line probe assay assigned HCV genotype to 95.2%.

Distribution of Hepatitis C Virus Genotypes

On the basis of the line probe assay, the distribution of various HCV genotypes in patients with chronic hepatitis C seen in tertiary referral centers in various regions of the United States was evaluated (Table 3). Several important points emerged. First, 96.1% of the specimens were positive by PCR, indicating that these specimens were fairly well preserved. Second, types 1a and 1b were the most common HCV genotypes in this population. Third, the overall distribution of various HCV types in different geographic locations was similar.

Clinical Characteristics of Patients Infected with Different Hepatitis C Virus Genotypes

For patients infected with HCV type 1, genotyping by the line probe assay may show the following possible patterns: type 1a; type 1b; type 1a plus type 1b; and type 1 in which subtype a or b could not be defined. As a first step, we compared all of the subtypes within the major types, and no differences were seen in clinical characteristics between the subtypes within the major types. In particular, no differences in any clinical variables were seen between patients infected with HCV type 1a and those infected with HCV type 1b. Therefore, patients with each major HCV type were grouped together regardless of subtype (for example, patients with type 1a, type 1b, type 1a plus type 1b, and type 1 without specified subtype were all grouped together as HCV type 1). Patients with mixed infection were combined into a single group, and patients who had negative results on PCR were also considered as a group.

No difference in age was seen among patients infected with HCV types 1, 2, and 4. Patients infected with HCV type 3 were younger than patients infected with type 1 and type 2 (type 3 compared with type 1 [P = 0.002]; type 3 compared with type 2 [P = 0.004]; see Table 4). No difference in sex distribution was seen among patients infected with different HCV genotypes. Of the 258 patients in which the estimated duration of HCV infection was documented, no difference in duration was seen between patients infected with HCV type 1 and those infected with HCV type 2. However, patients infected with HCV type 1 had a longer estimated duration of infection than did patients infected with types 3 and 4 (P = 0.004 for type 3 and P = 0.049 for type 4). Of the 317 patients in whom the method of acquisition was documented, 108 of 216 (50%) of the patients with HCV type 1 infection had acquired HCV through the transfusion of blood products compared with 16 of 65 (25%) of the patients infected with another HCV type (P < 0.001; Table 4).

In this data set, no statistical differences were seen in serum alanine aminotransferase or aspartate aminotransferase levels among patients infected with different HCV genotypes, except that the few patients infected with HCV type 4 had lower serum alanine aminotransferase levels (70 ± 38 U/L [n = 5]) than patients infected with other HCV types (ranging from 154 ± 118 U/L [n = 24] for HCV type 3 to 163 ± 118 U/L [n = 313] for HCV type 1), but this did not reach statistical significance (P = 0.07; Table 4).

Of the 262 patients who had baseline liver biopsy specimens, 182 had HCV type 1, 35 had type 2, 12 had type 3, and 2 had type 4. Because so few patients with HCV type 4 infection had liver biopsy specimens available, analysis was done only on patients with HCV types 1, 2, and 3 (Table 4). No significant difference was seen in the total histologic inflammatory index or Knodell scores among patients infected with different HCV genotypes. Additionally, no differences were seen in individual histologic scores (lobular, periportal, and portal inflammation and fibrosis) among patients infected with different HCV types. The only exception was that patients with HCV type 2 infection had more periportal inflammation (mean rank, 124.4; n = 35) than did patients with type 1 infection (mean rank, 106.0; n = 182), but this difference was not statistically significant (P = 0.088).

In all patients, HCV viremia levels were documented by reverse transcriptase PCR and quantitated by bDNA assay (Table 4). Patients infected with HCV types 1, 2, and 3 had similar levels of viremia. However, patients infected with HCV type 4 had lower viremia levels than did patients infected with type 1 (P = 0.047).

When patients with mixed infection (across types [for example, 1a and 3a] and not within types [for example, 1a and 1b]) were considered as a group, they were older than patients infected with HCV type 3 (P = 0.018). No differences were seen in sex distribution, mode of HCV acquisition, or serum biochemical and histologic features between patients with mixed infection and patients infected with other HCV types (Table 4).

The demographic, biochemical, and histologic features of patients who had negative results on PCR did not differ from those of patients in other groups (Table 4).

Discussion

Top Methods Results Discussion Author & Article Info References

Our study makes several important points. First, for HCV genotyping, the line probe assay based on the 5' UTR is sensitive and reproducible and has good concordance with other tests. Second, the distribution of HCV types 1, 2, and 3 in patients with chronic HCV infection who are referred to tertiary centers in the United States is fairly uniform: Type 1 is the most common HCV type. Third, a higher proportion of patients infected with HCV type 1 than of patients infected with other HCV types appears to have acquired HCV through transfusion of blood products. Fourth, disease activity does not differ significantly among patients infected with HCV types 1, 2, and 3. Finally, patients infected with HCV types 1, 2, and 3 have similar serum HCV RNA levels, whereas patients infected with HCV type 4 have lower levels of viremia than patients infected with type 1.

The line probe assay is relatively simple and standardized. Recently, Smith and associates [30] reported that, theoretically, the line probe assay might assign HCV genotype incorrectly in patients infected with HCV types 3 and 4 and that restriction fragment length polymorphism based on the 5' UTR may be more reliable. Although this may be true in some geographic locations in which more genotypes are present, we show that the HCV genotype assignment produced by line probe assay is highly concordant with that produced by other tests, including restriction fragment length polymorphism, in this field study in the United States. This is probably because HCV types 1 and 2 are the most common HCV genotypes in the United States. In addition, the probes used in the line probe assay have been modified to accommodate the recently identified genetic variations of HCV.

Given that this line probe assay is based on the initial amplification of the 5' UTR, a highly conserved area of the HCV genome, it is not surprising that it is very sensitive. As reported previously [20], genotyping assays based on other genomic regions have lower sensitivity. This is believed to be because of the lower efficiency of amplification related to the higher genetic heterogeneity of these genomic regions. In this previous study of patients in the United States with chronic hepatitis C [20], genotyping based on the HCV core region as described by Okamoto and colleagues has a sensitivity of 89.2%, whereas genotyping based on NS5 has a sensitivity of only 83.5%. The specificity of this line probe assay is also excellent. The concordance of the assay with restriction fragment length polymorphism assay based on the 5' UTR, which we have previously shown to be very reliable [20], was 98.2%.

As of October 1995, the cost of each line probe assay strip was about U.S. $65.00. Adding the cost of PCR, reagents, and supplies and the institutional running cost for PCR and genotyping, excluding technician time, the total cost should be about $100 to $120 (estimated from the cost at the University of Florida). Genotype assignment is reported in a strip form with dark bands. We made photocopies of the strips, and the photocopies seemed to have sufficient quality for evaluation. This will allow investigators to compare results easily. We also sent the photocopies through a standard facsimile machine and found that most of the bands were adequately visible on the facsimile.

On the basis of this line probe assay, HCV type 1 was found to be the most common HCV genotype in patients referred to tertiary referral centers for inclusion in experimental antiviral therapy programs in the United States; it was assigned in 72% of cases. Types 2 and 3 were found in all geographic areas, with average overall distributions of 14% and 6%, respectively. Type 4 was identified in Florida. Recently, type 4 was also found in California (Zhou S and coworkers. In preparation), suggesting that our detection of HCV type 4 only in Florida might be related to the larger number of patients that we studied in that state. Because HCV type 5 is found mainly in South Africa and Belgium, and type 6 is found primarily in Hong Kong and Macau, all of which are developed countries or cities with heavy traffic to the United States, we anticipate that these genotypes will eventually be seen in the United States. We might have identified these types if we had studied more patients.

One potential pitfall of our study was that the persons we studied were referred to tertiary referral centers for consideration for inclusion in experimental antiviral therapy programs in the United States. We attempted to include patients from various geographic locations, and, as discussed above, the distribution of HCV genotypes was fairly similar in all study centers. Our data are important to clinicians and research scientists in the United States, because HCV genotype 1a is only prevalent in the United States, and data generated in other countries will not provide detailed information on the clinical characteristics of patients infected with this HCV type.

Another important issue is whether patients in different phases of infection (for example, healthy blood donors with normal liver test results who are seropositive for anti-HCV, patients with mild liver disease who are seen by primary care physicians, and patients referred for liver transplantation) have the same distribution of HCV genotypes. To date, no published data on the distribution of HCV genotypes in these groups are available. A recent analysis based on more than 100 patients with HCV-related cirrhosis who were referred to the University of California, San Francisco, for liver transplantation showed that these patients had a distribution of HCV genotypes similar to that seen in our study (Zhou S and coworkers. In preparation).

No differences were seen in any of the demographic, clinical, biochemical, virologic, and histologic features of patients with different HCV subtypes within each major HCV type. In particular, patients infected with HCV types 1a and 1b had similar profiles in all of the variables examined. We wish to emphasize that the duration of HCV infection was estimated on the basis of clinical history, and the data on the correlation of estimated duration of HCV infection with HCV genotype should be interpreted with caution. Several differences were found between patients infected with different HCV types. A higher proportion of patients infected with HCV type 1 than of patients infected with other HCV types had acquired HCV through the transfusion of blood products. It has been suggested that HCV type 3 infection is more prevalent among intravenous drug users in Scotland [29]. However, for the 16 patients in our study with HCV type 3 infection, intravenous drug use was not the predominant method of acquisition. This suggests that the association between HCV type 3 infection and intravenous drug use may not apply in the United States.

Patients with HCV type 3 infection were younger than patients infected with HCV types 1 and 2, and patients infected with types 3 and 4 had a shorter estimated duration of HCV infection than patients infected with type 1, as assessed from clinical history. There are three possible explanations for these observations. First, HCV types 3 and 4 may have been introduced into the United States more recently. Second, types 3 and 4 may be transmitted primarily within younger generations through intravenous drug use. Third, there might be a bias in the estimation of HCV infection in patients infected with different HCV genotypes. That HCV type 3 is more prevalent in India and Scotland and type 4 is more prevalent in the Middle East favors the first explanation. In contrast, the observation that the proportion of patients who acquired HCV through intravenous drug use is higher for patients with types 3 and 4 than for patients with type 1 favors the second hypothesis. However, we must emphasize that a significant proportion of patients with HCV types 3 and 4 did not acquire HCV through intravenous drug use, which suggests that other factors are involved. In fact, these two possibilities are not mutually exclusive, and both may be operative.

An important observation in our study was that disease activity did not differ among patients infected with different HCV genotypes. A previous study in fewer patients [20] suggested that HCV type 2 infection was associated with a higher level of disease activity. Given the absence of an association in this larger, present study, it is possible that the previous association was the result of a type II statistical error. Nousbaum and colleagues [33] reported that patient age and presence of cirrhosis were independently associated with HCV type 1b infection in France and Italy. There are at least three possible explanations for the difference between their data and our observations. First, they had only 4 patients with HCV type 2 and 20 patients with HCV type 3, and thus a type II statistical error may have existed. Second, their patients with HCV type 1b infection were older and had longer estimated durations of disease compared with their patients infected with HCV types 2 and 3. Types 2 and 3 may have been introduced more recently to France and Italy and, therefore, the effect of these HCV genotypes on liver disease may not have been obvious at the time the study was done. Third, a significantly higher proportion of patients with HCV type 1 had acquired HCV through blood transfusion. Other clinical or epidemiologic factors may have been associated with the development of liver disease in patients with HCV type 1b infection. As discussed earlier, the distribution of HCV genotypes in patients receiving liver transplants was similar to that detailed in our current study, and this further supports the hypothesis that all HCV genotypes are equally able to induce active and potentially progressive liver disease.

With appropriate adjustments for an accurate assessment of viremia in different genotypes, we found that patients chronically infected with HCV types 1, 2, and 3 had similar levels of viremia. Similar results were recently seen in a large study of blood donors infected with different HCV genotypes [27]. In contrast, in our study, the few patients with HCV type 4 had a lower level of viremia than patients with HCV type 1. The clinical and virologic implications of this low-level viremia in patients with HCV type 4 infection remains to be established.

Finally, it has been suggested that patients infected with HCV types 2 and 3 have a better chance of developing a long-term biochemical response to interferon- [33-36]. The patients in group 2 in our study received a complicated dose-adjustment protocol, and patients switched treatment arms if they did not show a complete biochemical response after 12 weeks of interferon- therapy. Hence, our analysis of response to interferon- therapy is complicated and is based on different criteria. These data, together with the treatment response data that involved the effect of changing interferon- dose, will be addressed in a separate manuscript (Lindsay KL and colleagues. In preparation).

In conclusion, this line probe assay is applicable to patients in the United States who have chronic HCV infection. Type 1 is the most common HCV genotype in patients with chronic hepatitis C who are referred to tertiary centers in the United States. Half of patients with HCV type 1 have acquired HCV through the transfusion of blood products. No difference in disease activity or viremia levels was seen among patients chronically infected with HCV types 1, 2, or 3. The distribution of various HCV genotypes in other clinical settings (such as blood donors seropositive for anti-HCV and patients with chronic HCV infection seen by primary care doctors) and the relation between HCV genotypes and response to interferon therapy remains to be established.

Appendix

Additional members of the Hepatitis Interventional Therapy Group are Robert P. Perrillo, MD (St. Louis Veterans Affairs Medical Center, St. Louis, Missouri); Eugene R. Schiff, MD (University of Miami, Miami, Florida); Henry C. Bodenheimer, MD (Rhode Island Hospital and Brown University, Providence, Rhode Island); Luis A. Balart, MD, and Frederic Regenstein, MD (Ochsner Clinic, New Orleans, Louisiana); Jules L. Dienstag, MD, and William N. Katkov, MD (Massachusetts General Hospital, Boston, Massachusetts); Carlo H. Tamburro, MD (University of Louisville, Louisville, Kentucky); John S. Goff, MD, and Gregory T. Everson, MD (University of Colorado, Denver, Colorado); Zachary Good-man, MD (Armed Forces Institute of Pathology, Washington, D.C.); and Janice Albrecht, PhD (Schering-Plough Research Institute, Kenilworth, New Jersey).

Ms. Prescott and Dr. Simmonds: Department of Microbiology, University of Edinburgh, Teviot Place, Edinburgh, EH8 9AG, United Kingdom.

Dr. Maertens: Innogenetics, Industriepark Zwijnaarde 7, Box 4, B9052, Ghent, Belgium.

Dr. Lindsay: Hepatitis Research, Division of Gastrointestinal and Liver Diseases, Department of Medicine, University of Southern California, Los Angeles, CA 90033-4581.

Dr. Mizokami: Second Department of Internal Medicine, Nagoya City University Medical School, 1-4 Kawasumi, Mizuho, Nagoya 467, Japan.