Intrahepatic Hepatitis C Virus Replication Correlates with Chronic Hepatitis C Disease Severity In Vivo

Abstract

The role of viral factors in the pathogenesis of chronic hepatitis C is unknown. The objective of the present study was to characterize markers of hepatitis C virus (HCV) infection and replication in liver biopsy specimens obtained from 65 genotype 1-infected subjects, including 31 who were coinfected with human immunodeficiency virus (HIV), and to analyze associations between intrahepatic viral markers and hepatitis C disease severity. The percentages of liver cells harboring HCV genomes (%G) and replicative-intermediate RNAs (%RI) were evaluated using strand-specific in situ hybridization, while HCV core and NS3 antigens were assessed by immunocytochemistry. HIV-positive and HIV-negative subjects had similar mean grades and stages of liver disease and had similar indices of HCV infection and replication in liver, even though coinfected subjects had significantly shorter mean disease duration (P = 0.0003). Multivariate analysis showed that %G was not associated with grade or stage of liver disease (P = 0.5 and 0.4, respectively), while %RI was strongly associated with liver inflammation (P < 0.001), liver fibrosis (P < 0.001), and serum alanine aminotransferase levels (P = 0.01). NS3 antigen (but not core) was more frequently detected in HCV RI-positive versus RI-negative specimens (P = 0.028). These findings demonstrate a link between HCV proliferation and hepatitis C disease severity and suggest similar pathogenic mechanisms in HIV-positive and HIV-negative individuals.

Hepatitis C is a chronic progressive disease of the liver that is caused by infection with hepatitis C virus (HCV) (9). Although 15 to 20% of individuals infected with HCV spontaneously clear the virus in the acute phase, up to 85% develop persistent viremia. Approximately 60% of those infected develop clinically overt chronic hepatitis, and in this group, the rate of liver disease progression is generally slow but variable (34a). Approximately 20% of patients with chronic hepatitis C develop cirrhosis within 20 years of infection, and those with cirrhosis are at risk of clinical decompensation and developing hepatocellular carcinoma. Finally, chronic hepatitis C occurs in up to 30% of individuals with human immunodeficiency virus (HIV) infection, and evidence to date suggests that HCV-associated liver disease is more severe in subjects with HCV/HIV coinfection than in subjects with HCV monoinfection. The mechanisms driving HCV disease acceleration in coinfected subjects are presently unknown (17).

As with other members of the Flaviviridae family, HCV replicates by enzymatically converting its positive-strand RNA genome into a complementary or minus-strand replicative-intermediate RNA (RI RNA) and then copying the minus-strand RNA to produce new progeny plus-strand RNA. Nascent HCV genomes are then packaged into virions that are released from infected cells by unknown mechanisms. For positive-strand RNA viruses such as HCV, the RI RNA is a highly specific index of active viral replication (26). Based on in vitro experiments using the HCV replicon model, HCV replication events are thought to occur within a perinuclear membranous web structure in infected cells following assembly of nonstructural proteins into a replication complex (14).

Because HCV has a strict host range, infecting only humans and chimpanzees, studies of HCV pathogenesis rely heavily on observations made during natural infection in humans. Pharmacodynamic analysis of HCV RNA levels following single or sequential doses of interferon in humans has led to the conclusion that HCV replicates at a very high rate in vivo, producing an estimated 10¹² virions per day (23, 33). Previous investigations of the distribution of HCV RNAs in liver and its relationship to both serum viral loads and liver disease activity have produced remarkably inconsistent findings, as reviewed by Gowans (16), most likely because of the difficulty in developing and standardizing methods for detecting and quantifying RNA molecules in situ. Because of such limitations, the extent of HCV infection and replication within liver remains controversial. Furthermore, the relationship between intrahepatic HCV and liver disease activity is unknown, although based on serum HCV RNA data, it is widely believed that HCV replication is not associated with liver injury (reviewed in reference 29).

To begin to address such questions, our group previously developed a highly sensitive and specific in situ hybridization (ISH) assay for detecting both HCV positive strands (genomes) and negative strands (RI RNAs) in tissue biopsy specimens by using complementary strand-specific RNA probes (HCV riboprobes) synthesized in vitro from recombinant HCV templates (6, 7). The high specific activity of HCV riboprobes is a reflection of the high affinity of RNA-RNA hybrids, resulting in a dramatically better signal with very clean background compared to more widely used DNA probes. Importantly, all tissues are fixed and frozen rapidly and in an identical manner at the time of biopsy, cellular RNAs are assessed by ISH on all specimens to ensure stability of RNA in the biopsy specimen, all ISH experiments are conducted using well-characterized riboprobes of known specific activity, and cell lines expressing subgenomic regions of HCV coding and noncoding strands are used as positive and negative controls in all experiments. In our initial study of liver biopsy specimens from immunosuppressed liver transplant recipients, we found an interesting association between the percentage of intrahepatic cells positive for HCV RI RNA and the degree of liver fibrosis (6). Using similar experimental methods, Agnello and colleagues reported that HCV infection and replication are more widespread in livers of subjects with chronic hepatitis C than previously appreciated (1), which is consistent with our previous report (7).

The present study describes strand-specific ISH analysis of HCV infection and replication in liver biopsy specimens from 65 subjects with chronic hepatitis C, 31 of whom had HIV coinfection, and investigation of the relationships between indices of intrahepatic HCV infection and replication, HIV infection status, and three important indices of liver injury (serum alanine aminotransferase [ALT] levels, hepatic inflammation, and hepatic fibrosis). We further characterized HCV core and nonstructural 3 (NS3) antigens in the same liver biopsy specimens from 25 cases to begin to characterize the relationships between intrahepatic HCV infection, replication, and viral protein synthesis during chronic hepatitis C in vivo. We hypothesized that intrahepatic HCV replication is correlated with liver disease severity and is increased in those with HIV coinfection and that NS3 antigen is detected more frequently in specimens with detectable HCV replication than in those without.

MATERIALS AND METHODS

Patients.

Sixty-five HCV genotype 1-infected subjects with and without HIV coinfection were recruited from clinics at Harborview Medical Center. Of the HCV/HIV-coinfected subjects, 10% were naive to prior antiretroviral therapy (ART), 36% had a history of prior ART but either were not being treated at the time of study or had recent initiation of highly active ART (HAART) (within 6 months of the liver biopsy), and 54% were on HAART for at least 6 months prior to liver biopsy. All but one subject had a parenteral risk factor for HCV (injection drug use [IDU] [n = 62] or blood transfusion [n = 2]). Exclusion criteria included recent history of any acute infection, positive hepatitis B surface antigen, or any history of HCV treatment. The University of Washington Human Subjects Committee approved the study protocol beforehand, and written informed consent was obtained from each patient prior to enrollment. To establish the estimated time of infection, subjects were questioned regarding the year of blood transfusion or IDU onset, if applicable. An adapted Skinner questionnaire regarding recent and lifetime alcohol use (36) was kindly provided for our use by the HALTC Trial Group. In blood transfusion recipients (prior to 1992), the year of transfusion was assumed to be the year of HCV infection. In subjects admitting a history of IDU, the year of HCV infection was assumed to be 2 years after the onset of IDU (19).

Laboratory testing.

Clinical laboratory data (serum ALT, CD4 cell count, and HCV RNA quantification) were collected in all cases within 30 days of liver biopsy and in the vast majority of cases within 7 days of liver biopsy. Retesting to confirm prior HIV antibody testing was initiated after enrollment for the 34 HIV-negative subjects. These subjects were recontacted and asked to consent to HIV testing; 32 of 32 retested subjects were repeatedly negative for HIV antibodies. The two subjects who could not be reached had been HIV negative in the past and had denied interim risk behaviors. HCV RNA was detectable and quantifiable in serum samples from all 65 subjects. HCV RNA levels were determined by third-generation branched-DNA assay (VERSANT HCV RNA 3.0; Bayer Diagnostics, Tarrytown, NY) (lower limit of detection, 600 IU/ml) prior to 2002 and by in-house real-time reverse-transcriptase PCR (lower limit of detection, 50 IU/ml) after 2002. In a study of 199 HCV-infected patients, the two assays correlated very well (r² = 0.81) (10). HCV genotype was assigned using restriction fragment length polymorphism analysis of the 5′ noncoding region (11).

Histology.

Liver biopsies were reviewed by a single pathologist who was blinded to HIV status and all other data. Grade (0 to 4) and stage (0 to 4) were assigned according to the system described by Batts and Ludwig (3).

In situ hybridization for genomic and replicative-intermediate RNAs. (i) Generation of cDNA clones and riboprobes.

HCV 5′ nontranslated region, core, and envelope 1 (E1) genes for both HCV genotypes 1a and 1b were amplified by reverse transcriptase PCR and cloned into plasmid pCR2 (Invitrogen, San Diego, CA) (6). To generate digoxigenin (DIG)-labeled riboprobes, RNAs were synthesized by runoff transcription with T7 or T3 polymerase in the presence of DIG-UTP (Boehringer Mannheim Biochemicals, Indianapolis, IN). DIG-labeled riboprobes were broken down by alkaline hydrolysis to an average size of 100 nucleotides. The final riboprobe was precipitated and dissolved in 0.1% sodium dodecyl sulfate. Newly synthesized DIG-labeled riboprobe was evaluated against a known, standard DIG-labeled RNA (Boehringer Mannheim Biochemicals, Indianapolis, IN), and the concentrations of riboprobes were determined by Northern dot blot hybridization.

(ii) Generation of control cell lines for ISH.

DNA containing the HCV genotype 1a core plus E1 gene and HCV genotype 1b core plus E1 gene were subcloned into the eukaryotic expression vector pTRE2hyg in both sense and antisense orientations to generate control cell lines expressing either HCV positive-strand or negative strand RNA. HeLa Tet-off cells were transfected by electroporation, and positive cell lines were selected by culturing in the presence of hygromycin B (Calbiochem, La Jolla, CA) (6).

(iii) ISH assay.

Frozen tissue sections (6 μm), after being heat thawed and fixed in 10% neutral buffered formalin, were treated with 0.2 N HCl and proteinase K (1 μg/ml) and soaked in equilibration solution followed by prehybridization solution (Novagen, Madison, WI) at 50°C for 1 h. Approximately 15 μl of DIG-labeled riboprobes was applied to each slide at a final concentration of 2 to 4 ng/μl in hybridization buffer. For analysis of HCV RNA, mixtures of core and genotype-specific E1 riboprobes were used as HCV antisense (negative-strand) or sense (positive-strand) riboprobes. Stability of cellular RNA was determined by performing ISH with beta-actin antisense riboprobes as a positive control. Assay specificity was ensured as follows. The positive and negative control HeLa Tet-off cell lines were included in all experiments and had to yield the expected level of positive or negative staining for runs to be considered valid and data analyzed. Fewer than 10% of runs had to be repeated due to failure of the positive controls, while the negative control HeLa cell line was negative in all experiments. Huh7 cells were also routinely assayed in our ISH experiments and were negative for both HCV genomes and RI RNAs in all experiments, confirming assay specificity in human liver. Finally, clinical liver biopsy specimens from patients with chronic hepatitis or cirrhosis not caused by HCV (i.e., hemochromatosis, alcoholic liver disease, primary biliary cirrhosis, primary sclerosing cholangitis, nonalcoholic steatohepatitis, or autoimmune hepatitis) were assayed by ISH during method development, with zero false-positive results. However, such tissues were included in only a subset of ISH experiments due to limited sample availability and thus were not a part of our routine experimental controls.

(iv) Immunological detection of riboprobes in tissue.

During the hybridization steps, tissue sections were covered with siliconized coverslips, sealed with rubber cement, and incubated at 50°C in a humidified chamber for 18 to 22 h. After hybridization, sections were treated with RNase A (Novagen, WI) (20 μg/ml in 2× SSC [1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate]) and subsequently washed in 50% formamide and SSC at 50 to 65°C for 30 min. Tissue sections were soaked in a commercially prepared blocking buffer (Vector Laboratories, Burlingame, CA) for 30 min at room temperature, followed by incubation with anti-DIG-alkaline phosphatase conjugate (1:250 dilution) at 4°C overnight in a humidified chamber. Sections were washed twice with 100 mM Tris buffer at room temperature. Vector red substrate (Vector Laboratories, Burlingame, CA) supplemented with 1.25 mM levamisole (Sigma, St. Louis, Mo.) was added and left for 30 min before the reaction was terminated with 10 mM Tris-HCl buffer (pH 8.0)-1 mM EDTA, followed by methyl green counterstaining.

(v) Interpretation of ISH results.

A single liver biopsy was obtained for the present ISH study in each of the 65 cases (n = 65 specimens). Specimens were adequate for analysis of both positive- and negative-strand HCV RNAs in 63 cases, while 2 cases had sufficient liver tissue for only positive-strand ISH analysis. A minimum of three different tissue sections from each biopsy were stained separately using the HCV sense and antisense riboprobes. On average, approximately 10 sections were assayed by ISH in each case, representing between 20% and 50% of the total biopsy specimen in most cases, although specimen volume was insufficient for negative-strand RNA testing in two cases. Three different observers performed independent blinded evaluations of all ISH experiments, counting numbers of positive and negative cells in all fields and calculating percentages of cells positive for HCV genomic or RI RNAs. All specimens were counted in an identical manner for the current study. Data were averaged and rounded to the nearest 10th percentile for statistical comparisons. Investigators interpreting the results were blinded to HIV status and all other clinical data.

(vi) ICC.

HCV core and NS3 proteins were studied in liver biopsy specimens by direct immunostaining in 25 samples from HCV-positive individuals in our cohort study (17 of whom were also HIV positive), based on sample availability. Snap-frozen, formalin-fixed sections were incubated overnight at 4°C with alkaline phosphatase-conjugated mouse monoclonal antibodies against HCV core (Affinity BioReagents, CO) and NS3 (Vision Biosystem, MA) antigens. For color detection, Vector red substrate (Vector Laboratories, Burlingame, CA) supplemented with 1.25 mM levamisole (Sigma, St. Louis, Mo.) was added and left for 30 min before the reaction was terminated with 10 mM Tris-HCl buffer (pH 8.0)-1 mM EDTA. Methyl green (Vector Laboratories, CA) was used to counterstain the sections. HCV replicon cells (4) (a gift from C. Rice, Rockefeller Institute, NY) served as positive controls for immunocytochemistry (ICC) experiments, while both Huh7 cells lacking HCV replicon and liver biopsies obtained from HCV-negative subjects served as negative controls.

(vii) Statistical methods.

The data were summarized using the appropriate descriptive statistics, and groups were compared using the chi-square test or Student t test for unequal variances where appropriate. Associations were assessed using the Spearman correlation coefficient or linear regression with robust variance estimates. Regression diagnostics were performed to evaluate for outliers and data points with leverage or influence. Nonparametric tests were utilized to confirm the validity of the findings. A P value of <0.05 was considered significant. Statistical analyses were conducted using Stata version 8.0 (Stata Corporation, College Station, Texas).

RESULTS

ISH analysis of HCV plus- and minus-strand RNAs in control cell lines.

Permanent cell lines expressing subgenomic regions of HCV sense and antisense RNAs were established from genotype 1a and 1b clinical isolates for use as positive and negative controls in our strand-specific in situ hybridization assay. Figure 1A and C illustrate positive staining of plus and minus HCV RNAs expressed in HeLa cells, using probes of opposite strand polarity and the same genotype. Red punctate signals were evenly distributed throughout the cell cytoplasm, with occasional signal over nuclear regions (green) that likely reflects cytoplasmic signal due to folding or inclusion of cytoplasm above nuclei during sectioning. Figure 1B and D show negative staining of HeLa cells expressing plus- and minus-strand HCV subgenomes, respectively, with HCV riboprobes of the same polarity demonstrating strand specificity of the riboprobes. Figure 1E and F demonstrate negative staining using the same probes applied to HeLa control cell lines lacking HCV RNA expression.

FIG. 1.

Detection of HCV plus-strand and minus-strand RNAs in HeLa cell lines by strand-specific ISH. Panels A and C demonstrate detection of genotype 1a positive- and negative-strand RNAs, respectively, in positive control HeLa cells. In panel A, negative-strand (AS) riboprobes were used to detect positive-strand RNAs in situ, while in panel C, positive-strand (S) riboprobes detected negative-strand RNAs in situ. The experiment displayed in panels B and D used the same positive control HeLa cell lines as in panels A and C; however, control cells were reacted with HCV riboprobes of the same strand polarity as the target RNAs, resulting in negative results. RNA signals are red, and cell nuclei are green. Magnification, ×40. Panels E and F show negative control HeLa cell lines (HCV RNA negative) stained with HCV AS and S riboprobes.

ISH analysis of HCV genomes and replicative-intermediate RNAs in human liver biopsy specimens.

After establishing reliable controls for ISH experiments, our next objective was to test for evidence of HCV genomes and RI RNAs in a panel of 65 liver biopsy specimens obtained from HCV genotype 1-infected patients (Table 1). Thirty-one samples from patients with HCV/HIV coinfection and 34 samples from patients with HCV monoinfection were analyzed by strand-specific ISH. Coinfected subjects were significantly younger, had a significantly shorter duration of HCV infection, and had significantly lower CD4 counts than HCV-monoinfected subjects (Table 1). There were no significant differences between the two groups with respect to gender, race, recent or lifetime alcohol use, ALT levels, HCV genotype or serum level, and either grade or stage of liver disease.

TABLE 1.

Characteristics of HCV genotype 1-infected subjects

Characteristic (unit)	Valuea for:		P value
	HIV-positive subjects (n = 31)	HIV-negative subjects (n = 34)
Age (yr)	40.1 ± 6.3	44.9 ± 6.4	0.003
Gender (% male)	74.2	58.8	0.19
Race (%)			0.14
Caucasian	80.0	82.4
African American	22.6	11.8
Hispanic	6.5	0
Native American	0	5.9
HCV duration (yr)	14.8 ± 7.5	22.5 ± 8.5	0.0003
Alcohol use in lifetime (g/day)	51.2 ± 91.5	65.8 ± 79.4	0.50
Alcohol use in last year (g/day)	11.4 ± 33.7	16.9 ± 40.2	0.56
ALT (U/liter)	108 ± 115	96 ± 81	0.64
HCV genotype (% 1a)	77.4	76.5	0.93
HCV RNA level (log10)	6.3 ± 0.7	6.0 ± 0.7	0.12
CD4 cell count (cells/μl)	449 ± 326	990 ± 309	<0.0001
HCV disease grade (0-4)	2.3 ± 0.7	2.0 ± 0.8	0.11
HCV disease stage (0-4)	2.1 ± 0.8	2.0 ± 0.8	0.61

^aValues are expressed as means ± standard deviations unless stated otherwise.

HCV genomic RNAs were detected within mononuclear cell cytoplasm by antisense riboprobes in 56 (86%) of liver biopsy specimens from the 65 HCV-infected subjects. Furthermore, 67% of specimens (42 of 63) were positive for HCV RI RNA when stained using sense riboprobes. Negative controls included in all experiments (see Materials and Methods) never showed positive signals for HCV genomes or RI RNAs, further demonstrating specificity of the ISH assay in human tissue specimens. Representative ISH data are illustrated in Fig. 2, showing detection of HCV genomes (Fig. 2A) and RI RNAs (Fig. 2B) in parallel sections of a hepatic lobular region. Figure 2C and D show negative ISH results in a different liver tissue sample. In Fig. 2A, positive signal for HCV genomes was uniformly present in virtually 100% of hepatocytes at a high signal density in all fields analyzed. The same was true of RI RNA in parallel tissue sections (Fig. 2B), except the signals appeared more punctate and the overall signal density was lower for RI RNA than for genomic RNA.

FIG. 2.

Detection of HCV genomes and replicative-intermediate RNAs in the lobular region of a human liver biopsy specimen by strand-specific ISH. Biopsies fixed and frozen in OCT buffer were stained by ISH using the same HCV riboprobes as in Fig. 1, as described in Materials and Methods. Panels A and B show detection of HCV genomes with an antisense (AS) riboprobe and of HCV replicative-intermediate RNA with a sense (S) riboprobes, respectively. RNA signals are red, and cell nuclei are green. Magnification, ×40. Panels C and D illustrate negative results when staining a liver biopsy specimen obtained from an HCV-negative individual with HCV S and AS riboprobes. Magnification, ×10.

HCV infection and replication in the portal region of a liver specimen is demonstrated in Fig. 3. Staining of genomes with antisense riboprobes (Fig. 3A) and of RI RNAs with sense riboprobes (Fig. 3D) is shown at a magnification of ×10, while Fig. 3B and E show ×100 magnifications of the marked portal region, demonstrating abundant HCV genomes and RI RNAs, respectively. HCV RNA signals were not restricted to hepatocytes in this case but were observed within several different cell types that morphologically resembled inflammatory mononuclear cells, sinusoidal epithelial cells, and biliary epithelial cells. A focal pattern of HCV genome accumulation observed in a different liver specimen is illustrated in Fig. 3C, while Fig. 3F illustrates perinuclear localization of punctate signals for RI RNA. Overall, HCV RNA staining varied between cases, and in many instances the signal varied between different fields of the same biopsy specimen, although cases with uniformly diffuse staining patterns throughout the entire specimen were not uncommon. Individual specimens also varied with regard to the presence and degree of positivity in portal versus lobular regions, and staining patterns were not restricted to areas of active inflammation.

FIG. 3.

Detection of HCV genomes and replicative-intermediate RNAs in portal and lobular regions of liver biopsy specimens. Panels A and D show positive staining of HCV genomes and replicative-intermediate RNAs, respectively, in a portal region under ×10 magnification (serial sections from the same specimen). Panels B and E show ×100 magnifications of the indicated areas of the biopsy specimen, demonstrating abundant HCV genomes and replicative-intermediate RNAs in cells with morphological features not typical of hepatocytes. Panels C and F show ×100 magnifications of hepatocytes in a different specimen, demonstrating a focal pattern of HCV genome accumulation (panel C) and perinuclear localization of punctate signals for RI RNA (panel F, arrows).

The percentage of cells positive for HCV replicative-intermediate or genomic RNA does not differ between HIV-positive and HIV-negative subjects.

To compare the degrees of HCV infection and replication in liver biopsy specimens from subjects with HCV monoinfection versus HCV/HIV coinfection, we blindly quantified the percentages of cells positive for HCV genomes (%G) and replicative-intermediate RNAs (%RI) in each specimen by using a standardized procedure described in Materials and Methods. Compared with liver biopsy specimens from HCV-monoinfected subjects (n = 31), specimens from subjects with HCV/HIV coinfection (n = 34) did not differ with regard to either grade or stage of liver disease, even though the coinfected group had a significantly shorter HCV disease duration (14.8 ± 7.5 years versus 22.5 ± 8.5 years; P = 0.0003) (Table 1). On average, 38.7% of cells stained positive for HCV replicative-intermediate RNA (median, 40%; range, 0 to 100%). There was no difference in mean %RI between the HIV-positive (mean ± standard deviation, 38.9% ± 32.7%) and HIV-negative (38.4% ± 37.0%) groups (P = 0.96). The mean percentage of cells staining positive for HCV genomic RNA was 51.6% (median, 60%, range, 0 to 100%). Again, there was no difference the mean %G between the HIV-positive (50.6% ± 31.7%) and HIV-negative (52.5% ± 33.4%) groups (P = 0.96). Finally, HAART did not influence intrahepatic HCV replication: the mean %RIs in the HAART versus no-HAART coinfected subjects were 37.6% versus 41.3%, respectively (P = 0.78), nor did HAART affect the intrahepatic distribution of HCV genomes (%G).

Data on both %G and %RI were available for 63 of 65 subjects. These results were highly correlated (Spearman rank r = 0.5750; P < 0.0001) (Fig. 4). This strong association remained when HIV-positive (r = 0.6192; P = 0.0002) and HIV-negative (r = 0.5159; P = 0.0025) groups were considered separately. Importantly, however, there was a wide range in the infection-to-replication ratio (%G divided by %RI) in individual biopsy specimens in both HIV-positive and HIV-negative patient groups (mean ± standard deviation, 1.47 ± 1.35 and 1.26 ± 1.14, respectively), indicating that even though %G and %RI correlated well for the study population, replication and infection were not detected in consistent proportions of cells within individual biopsy specimens. The variance data were consistent between laboratory observers and were thus not due to experimental imprecision but reflected distinguishable biological patterns, as discussed below.

FIG. 4.

Association between percentages of cells staining positive for RI RNA and genomic RNA by strand-specific ISH in 63 biopsy specimens.

HCV replication is strongly associated with liver disease and is a better predictor of liver disease than HCV infection.

We next asked if there were significant relationships between indices of HCV infection or HCV replication and either grade or stage of liver disease in the 65 liver biopsy specimens. The data in Table 2 compare mean grade and stage of liver disease according to qualitative positivity for HCV genomes (G) or RI RNAs (RI) in the strand-specific ISH assay. While there was no significant difference in mean grade or stage of disease for G-positive versus G-negative specimens (P = 0.96 for grade and P = 0.22 for stage), both the mean grade and mean stage of liver disease were significantly higher in specimens testing positive for RI RNA than in negative specimens (P = 0.018 and P = 0.006 for grade and stage, respectively). On univariate linear regression analysis, the percentage of cells staining positive for RI RNA was also highly associated with both grade (Fig. 5A) and stage (Fig. 5B) on liver biopsy (P < 0.001 for both). In this analysis, the percentage of cells staining positive for genomic RNA also predicted grade (Fig. 5C) and stage (Fig. 5D), but these associations were of borderline statistical significance (P = 0.047 and 0.035, respectively).

TABLE 2.

Correlation between HCV infection and replication and grade and stage of liver disease in 65 cases

Parameter	Infectiona			Replicationb
	Value (mean ± SD) for:		P value	Value (mean ± SD) for:		P value
	G-positive subjects (n = 56)	G-negative subjects (n = 9)		RI-positive subjects (n = 42)	RI-negative subjects (n = 21)
Grade	2.23 ± 0.74	2.22 ± 0.44	0.96	2.33 ± 0.72	1.95 ± 0.50	0.018
Stage	2.18 ± 0.75	1.88 ± 0.60	0.22	2.27 ± 0.74	1.81 ± 0.51	0.006

^aInfection is defined as positive detection of HCV genomes.

^bReplication is defined as positive detection of HCV replicative-intermediate RNA.

FIG. 5.

Panels A and B: dot plots illustrating the association between percentage of hepatic cells staining positive for HCV replicative-intermediate RNA and histologic grade (A) and stage (B) of liver disease. Each dot represents an individual case, and the hatched lines indicate the mean RI values. Liver grade and stage were assigned scores of 0 to 4 using the scoring system of Batts and Ludwig (3). Panels C and D: dot plots illustrating the association between the percentage of hepatic cells testing positive for HCV genomic RNA and histological grade (C) and stage (D) of liver disease. Each dot represents an individual case, and the hatched lines indicate the mean G values.

Because there were relatively few subjects with grade and/or stage 4 disease, analyses were repeated after combining grades 3 and 4 and stages 3 and 4. In these analyses, there was no significant change in the coefficients or P values when %RI was the predictor, but the associations between genomic RNA and grade and stage were less significant (P = 0.066 and 0.053, respectively). In addition, excluding subjects who had negative staining by ISH did not alter the results (data not shown). Finally, %RI predicted grade and stage in both HIV-positive (P = 0.002 and 0.021, respectively) and HIV-negative (P = 0.007 and 0.003, respectively) subjects, while %G predicted grade and stage in the HIV-positive group only (P = 0.05 and 0.01, respectively). Thus, it appeared that %RI was a better predictor of liver disease severity overall. This conclusion was confirmed when measures of both %RI and %G were included in a multivariate statistical model: %G was no longer predictive of grade or stage (P = 0.29 and 0.42, respectively), while %RI remained predictive of both outcomes (P = 0.001 for both).

HCV replication, but not infection, is correlated with serum ALT levels.

There was a positive association between the percentage of cells positive for RI RNA and serum ALT (Spearman rank r = 0.3104; P = 0.01). However, there was no association between the percentage of cells positive for genomic RNA and ALT (P = 0.34) (data not shown).

Indices of HCV infection and replication are not correlated with serum HCV RNA levels, CD4 count, or other variables.

There was no association between indices of either intrahepatic HCV infection (%G) or intrahepatic HCV replication (%RI) and serum HCV RNA levels (P = 0.8 and 0.9 for %RI and %G, respectively). There also was no association between %G or %RI and the CD4 count whether considering all subjects (P = 0.8 for both measures) or HIV-positive and HIV-negative subjects separately. Finally, the quantity of alcohol use in the last 6 or 12 months did not correlate with %RI or %G, although the data do not disprove an association between heavy alcohol use and HCV replication, since only one subject had a history of continuous heavy alcohol consumption during the 12-month period prior to liver biopsy. In fact, 48% of our population had been abstinent from alcohol for at least 12 months prior to liver biopsy, and the mean daily alcohol intake for the entire study population was only 14.3 g per day, which represents only minimal alcohol consumption.

Immunocytochemical staining of HCV core and NS3 antigens.

Based on the abundance of HCV RNA in infected liver specimens, we expected to be able to detect HCV antigens in situ using ICC methods. Figure 6A and B demonstrate ICC staining of positive control replicon-HeLa cells for HCV core (structural) and NS3 (nonstructural) antigens, respectively. ICC staining of negative control Huh7 cells lacking the HCV replicon with anticore and anti-NS3 antibodies showed negative results (Fig. 6C and D).

FIG. 6.

Identification of HCV core and NS3 antigens within liver tissues and HCV replicon cell lines by immunocytochemistry. Panels A and B: Huh7-replicon cells stained with anticore and anti-NS3 antibodies, respectively, followed by detection with Vector red (VR; red signal). Panels C and D: Huh7 cells without replicons stained with same antibodies, followed by Vector red detection (no red signals). Panels E and F demonstrate anticore and anti-NS3 staining of a liver biopsy specimen obtained from an HCV/HIV-coinfected individual, whereas panels G and H show negative staining using the same antibodies on a different specimen. Magnification, ×40.

Figure 6E and F demonstrate positive staining of HCV core and NS3 antigens in an infected liver tissue specimen that had strong signals for both HCV genomes and RI RNAs. Negative staining of liver is shown in Fig. 6G and H. Table 3 summarizes the relationships we observed between HCV antigen and RNA detection in 25 liver biopsy specimens, 18 of which were positive for HCV genomes and 15 of which were positive for RI RNAs. Overall, 20 of 25 specimens (80%) were positive for core antigen, and the percentage of core-positive specimens was not significantly different in G-positive versus G-negative specimens (P = 0.11) or in RI-positive versus RI-negative specimens (P = 0.36). On the other hand, specimens that were positive for RI RNA were more likely to be NS3 positive than specimens that were negative for RI RNA (13 of 15 versus 4 of 10, respectively; P = 0.028). Furthermore, replication-negative specimens were more likely to be core positive (70%) than to be NS3 positive (40%). Finally, the percentage of cells positive for HCV genomes in a specimen (%G) was associated with positivity for both core and NS3 antigens (P = 0.05 and P = 0.036, respectively), while the percentage of cells positive for RI RNA (%RI) was not associated with core positivity (P = 0.149) but showed a trend towards an association with NS3 positivity (P = 0.057). The data suggest that while core and NS3 are equally associated with HCV infection, NS3 antigen is more closely associated with HCV replication than is core antigen.

TABLE 3.

Correlation between positive staining for HCV antigens (core and NS3) and HCV genomes or replicative-intermediate RNAs

Parameter	Total (n = 25)	Genomes			Replicative-intermediate RNAs
		G positive	G negative	P value	RI+	RI−	P value
No. core positive/total (% positive)	20/25 (80)	16/18 (89)	4/7 (57)	0.11	13/15 (87)	7/10 (70)	0.36
No. NS3 positive/total (% positive)	17/25 (68)	14/18 (78)	3/7 (43)	0.16	13/15 (87)	4/10 (40)	0.028

DISCUSSION

The host-virus relationships driving chronic hepatitis C disease are presently undefined. In the present report we describe a highly significant association between intrahepatic HCV replication and liver disease severity in vivo in a study focusing on 65 patients with HCV genotype 1 infection, all of whom were naive to previous interferon-based treatment of hepatitis C. The data show a highly significant correlation between the percentage of hepatic cells with evidence of HCV replication (%RI) and three indices of liver disease activity and/or severity, namely, inflammation, fibrosis, and serum ALT, while the percentage of liver cells harboring HCV infection (%G) was either marginally related or unrelated to these clinical indices of liver injury. Conversely, we found no relationship between either HIV infection status or alcohol use and HCV replication activity in liver, although with regard to the latter observation, the mean alcohol consumption in our study population was to low to draw meaningful conclusions. These results represent the first direct evidence linking HCV replication to liver injury, although the study neither proves a causal relationship nor disproves the hepatitis C immunopathogenesis hypothesis. More definitive longitudinal studies with examination of serial liver biopsies are needed to address the key question of whether widespread HCV replication precedes or occurs concomitantly with progression of liver injury in vivo and whether patients with stable or improving liver disease have reductions in intrahepatic HCV replication indices. Such longitudinal studies are presently ongoing with several patient cohorts, including our own HCV/HIV coinfection cohort, the human liver transplant model, and the HALTC clinical trial.

In a previous study using methods similar to our own, Agnello and colleagues (1) also documented widespread intrahepatic HCV infection, but they failed to find an association between %RI and severe disease, perhaps because their study was underpowered relative to ours (19 versus 65 cases, respectively). Previous ISH studies by Kojima et al. (22), Rodriguez-Indio et al. (34), de Lucas et al. (12), and Negro et al. (31) each failed to find associations between %G and liver disease activity, confirming our results relative to HCV infection and disease. Only Negro et al. measured HCV RI RNAs in their respective liver biopsy study. Finally, two studies using independent methods to quantify HCV genomes in liver both found no correlation between total intrahepatic viral load (i.e., genomes per biopsy) and any indices of liver disease (13, 32). Therefore, we feel confident in our conclusion that the extent of HCV infection in liver (either %G or hepatic genome concentration) is not associated with liver disease severity.

We also describe an interesting and statistically significant association between positive staining for HCV RI RNA and NS3 antigen in liver biopsy specimens. Specifically, specimens that were positive for RI RNA were more likely to be NS3 positive than specimens that were negative for RI RNA, and specimens negative for RI RNA were more likely to be negative for NS3 antigen than core antigen. In contrast, HCV core antigen was equally positive in RI RNA-positive and RI RNA-negative specimens. Our findings confirm that core antigen is closely associated with HCV infection, while NS3 antigen is more tightly associated with HCV replication activity. At least two previous studies have reported that detection of HCV nonstructural antigens in liver is associated with more severe liver injury (18, 21), which further supports to our hypothesis that HCV replication is linked to liver injury.

Discrepancies between our present results and previously reported studies of HCV RNAs in liver are likely due to differences in experimental methods as well as study design. For example, a quantitative PCR study by Negro and colleagues (32) showed no correlation between total hepatic genome or RI RNA concentrations and either liver injury or HCV antigens. The fact that our study and theirs measured different indices of HCV replication (total liver RI RNA versus percentage of cells positive for RI RNA) may help explain the discrepancies in results. However, another important difference is that while our study controlled for possible genotype effects by including only HCV genotype 1-infected subjects, fewer than 50% of the subjects in the study by Negro et al. were infected with genotype 1. Indeed, when the latter investigators restricted their analysis to genotype 1-infected subjects, associations between RI RNA concentrations and HCV antigen levels approached statistical significance and thus were in better agreement with our present study. Finally, all our subjects were naive to prior treatment of HCV infections, which is an important distinction since treatment may alter intrahepatic HCV replication even in nonresponders.

Our ISH study of 65 cases and the quantitative PCR study of 98 biopsies by Haydon et al. (20) both found no correlation between intrahepatic HCV infection markers and serum viral load. In contrast, the quantitative PCR study by Negro et al. and the branched-DNA study by De Moliner et al. both reported that total hepatic genomic RNA levels correlated with serum viral load (13, 32). Previous ISH studies of 19 cases (1) and 10 cases (15) also reported positive associations between %G and serum viral load. Since our study was restricted to subjects with HCV genotype 1 infection, we again speculate that these disparate findings may be due to laboratory methodology and/or study design (e.g., sample size, inclusion of subjects with different HCV genotypes, methods of tissue preservation, and methods used to assess intrahepatic HCV infection). Thus, the relationship between hepatic and serum viral loads remains uncertain. All studies to date, including our own, have failed to find any association between intrahepatic HCV replication and serum viral load, which suggests the possibility that replicated viral RNAs may be sequestered and not packaged in some cases.

In a quantitative ISH study using computer-assisted digital image analysis, we previously reported that the ratio of total viral genomes to total RI RNA molecules in liver (G to RI ratio) ranged from 20:1 to 3:1 (7). Consistent with this finding, we report here a wide range in the percentages of cells harboring either HCV genomes (%G) or RI RNAs (%RI), as well as a wide range in the ratio of infected cells to replication-positive cells (%G to %RI ratio). The variation in data was observed on a patient-to patient basis as well as a within-biopsy basis but was not a result of variance in data collection by the three blinded counters, indicating that the variance has biological meaning. Based on these data, we conclude that HCV infection and replication do not go hand in hand. The variable distribution of HCV genomes and RI RNA molecules within individual biopsy specimens suggests that intracellular pools of RI RNA may be under different regulation than intracellular pools of HCV genomes and further suggests to us that intrahepatic spread of HCV infection is regulated differently than is intrahepatic HCV replication. Since spontaneous mutations within the NS5A region of the HCV replicon are known to drastically alter in vitro replication rates (4), it is certainly possible that HCV replication might be under either viral or cellular control in vivo. Whether or not this is the case, our current data argue that diffuse HCV replication is more detrimental to the infected host liver than replication that is restricted to a lower percentage of cells, regardless of the total burden of viral nucleic acid in liver.

The present data have several implications relevant to hepatitis C pathogenesis. One implication is that livers with widespread HCV replication differ from livers with restricted HCV replication with respect to the activity of protective versus pathogenic host responses, at either the tissue or cellular level. For example, inflammatory or noninflammatory cytokine responses may differ, and such differences may be critical for regulating the proliferative capacity of HCV within infected cells, similar to the relationship discovered for HBV in the murine transgenic mouse model (8). Although it is possible that increased or unrestricted HCV replication activity causes hepatocellular damage, it is also very possible that unrestricted replication is a consequence of host-induced liver injury (i.e., immunopathogenesis) leading to a loss of critical cellular control mechanisms that ultimately leads to enhanced HCV replication. Since there are no animal models that can recapitulate hepatitis C disease at present, it is difficult to study such complex interactions using conventional pathogenesis models.

There is general agreement that a wide variety of intrahepatic cells harbor HCV genomes, including biliary epithelial cells, sinusoidal epithelial cells, vascular endothelial cells, and mononuclear cells of apparent lymphocytic and monocytic lineage. The study by Agnello et al. (1) and our own study both found evidence of HCV RI RNA in biliary epithelial cells by ISH, a finding which is supported by the report of Loriot et al. (27) that HCV can replicate in vitro in cultured human biliary epithelial cells. We and several other groups (reviewed in reference 16) have reported evidence of HCV negative-strand RNAs within the cytoplasm of infiltrating mononuclear inflammatory cells in liver, while Sung and colleagues (37) reported evidence that HCV can infect and replicate in human lymphoma cells both in vivo and in vitro; thus, the evidence that HCV is lymphotropic is strong. Paradoxically, several well-conducted studies have found evidence of HCV genomes and unique quasispecies variants in circulating peripheral blood mononuclear cells while finding no evidence of viral replication in these infected cells using highly sensitive PCR-based methods (5, 24, 25, 28). One hypothesis to explain this paradox is that while circulating mononuclear cells represent a latent reservoir for HCV infection, viral replication activates once such cells become associated with tissue stroma. Efforts to develop reliable methods and models for confirming this hypothesis are ongoing.

In the present study, subjects who were coinfected with HIV had a significantly shorter duration of hepatitis C infection on average, but despite this they had similar degrees of liver injury on biopsy. While one could speculate that HIV plays a direct role in accelerating HCV disease, it seems more probable that loss of regulatory immunity contributes to the increased rate of disease progression. However, neither HIV status nor CD4 count was associated with %G or %RI, and neither influenced the significant relationship between HCV replication and liver disease. It is also intriguing that HCV-associated liver injury is greatly accelerated in the liver transplant setting, where up to 20% of allografts develop advanced liver injury or cirrhosis within the first few years posttransplant (35). Although host immunosuppression may be one determinant of liver disease in this population, up to 50% of subjects remain disease free after transplant, raising the question of why some individuals develop recurrent disease and others do not. We previously reported that detection of HCV nonstructural antigens in liver biopsy specimens correlated with disease severity in this population (18). In our initial ISH study in the transplant population, indices of intrahepatic HCV replication also correlated with both inflammation and fibrosis (6), suggesting similar pathogenic mechanisms of hepatitis C during natural infection and in immunosuppressed liver transplant recipients and HIV-coinfected individuals.

In summary, the present study used optimized molecular methods to investigate the intrahepatic distribution of HCV markers as they relate to degree of hepatic injury in a cohort of patients with chronic hepatitis C with or without HIV coinfection. While the study presents the first evidence linking HCV replication to progressive liver injury including fibrosis, it does not answer the question as to whether increased viral replication is the cause or the effect of chronic hepatitis C disease progression. Further longitudinal study of HCV replication and disease in human cohorts is necessary to substantiate this association until reliable nonhuman models of HCV disease are available.