Genetic Clustering of Hepatitis C Virus Strains and Severity of Recurrent Hepatitis after Liver Transplantation

Abstract

The influence of viral factors on the severity of hepatitis C virus (HCV)-related liver disease is controversial. We studied 68 liver transplant patients with recurrent hepatitis C, of whom 53 were infected by genotype 1 strains. Relationships between core sequences, serum HCV RNA levels, and fibrosis scores for each patient were analyzed in pairwise fashion 5 years after transplantation. We used Mantel's test, a matrix correlation method, to evaluate the correspondence between measured genetic distances and observed phenotypic differences. No clear relationship was found when all 68 patients were analyzed. In contrast, when the 53 patients infected by genotype 1 strains were analyzed, a strong positive relationship was found between genetic distance and differences in 5-year fibrosis scores (P = 0.001) and differences in virus load (P = 0.009). In other words, the smaller the genetic distance between two patients' viral core sequences, the smaller the difference between the two patients' fibrosis scores and viral replication levels. No relationship was found between genetic distance and differences in age, sex, or immunosuppression. In multivariate analysis, the degree of fibrosis was negatively related to the virus load (r = −0.68; P = 0.003). In the particular setting of liver transplantation, and among strains with closely related phylogenetic backgrounds (genotype 1), this study points to a correlation between the HCV genetic sequence and the variability of disease expression.

The influence of viral factors on the severity of hepatitis C virus (HCV)-related liver disease is controversial. HCV strains are usually classified into six genotypes and numerous subtypes on the basis of phylogenetic clustering (26). Many authors have sought correlations between the HCV genotype and various phenotypic characteristics, such as the level of viral replication (35), the severity of liver disease (3, 19), the risk of hepatocellular carcinoma (25, 32), the response to interferon therapy, and the severity of recurrent hepatitis after liver transplantation (5, 8, 9, 10, 11, 31, 36). The only clear relationship found to date is the relative inefficacy of recombinant alpha interferon therapy on HCV type 1 infection (17). Overall, these studies suggest that host factors are important in HCV disease progression and that few if any viral factors are involved. Viral taxonomy, or the choice of an arbitrary prototypic strain, may be inappropriate for studying relationships between the HCV nucleotide sequence and pathogenicity, i.e., for testing the hypothesis that genotype-specific mutations influence the course of the disease.

If phenotypic markers tend to converge as the genetic distance falls, then a relationship should exist between viral factors (e.g., the nucleotide sequence) and phenotypic markers of disease progression. Matrix correlation methods are commonly used to evaluate the degree of similarity between two variables that do not have the implicit assumptions inherent in most standard statistical tests, including normality and independence. These methods are particularly suited to evaluating the correspondence between two types of measured distance. We therefore used the Mantel test (16) to compare genetic distances between clinical HCV isolates and “phenotypic” distances, namely, differences in METAVIR fibrosis scores (2) and virus load, which are commonly used to appraise the severity of HCV infection (9, 17, 35, 36). We sequenced the regions coding for the core protein, which has numerous potential interactions with cellular or viral targets (1, 4, 7, 12, 15, 21, 24, 33, 34), and part of the nonstructural protein 5A (NS5A), which has been linked to the efficacy of interferon therapy on genotype 1 HCV infection (6).

MATERIALS AND METHODS

Patients.

Between January 1987 and January 1995, 178 patients with HCV-related liver cirrhosis underwent liver transplantation at our institution (Table 1). Sixty-eight of these patients were enrolled in this retrospective study on the basis of the following criteria: first liver transplantation, positive serum HCV RNA before and after liver transplantation, one biopsy during the fifth year posttransplantation, no episodes of steroid-resistant rejection or signs of rejection on the fifth-year biopsy, no use of anti-lymphocyte antibodies or chemotherapy, no antiviral therapy before the routine 5-year biopsy, and no surgical complications after transplantation. All the patients were hepatitis B surface antigen negative and anti-human immunodeficiency virus negative. None claimed excess alcohol consumption after liver transplantation.

TABLE 1.

Patients' characteristics and HCV genotypes

Variable	Value
Hepatocellular carcinoma	13/68a (19%)
Age (range)	51 ± 9.72 yr (25–68 yr)
Sex (F/M)b	15/53
Histological follow-up	78 ± 34 mo
French/Italian	28/40
Genotype 1b	45 (66%)
Genotype 1a	8 (12%)
Genotype 2	7 (10%)
Genotype 3	4 (6%)
Genotype 5	4 (6%)
Cyclosporine/tacrolimus	41/27
Age of donor	38 ± 21 yr (12–69)
Sex of donor (F/M)	30/48

^aNumber of patients with carcinoma/total. 

^bF, female; M, male. 

Nucleotide sequencing and HCV genotyping.

RNA was extracted from 140 liters of serum by using the QIAmp viral RNA kit (Qiagen GmbH, Hilden, Germany). To detect the HCV genome, 10 μl of the extracted RNA was subjected to reverse transcription-(RT)-nested PCR with Ready-To-Go RT-PCR beads (Pharmacia Biotech, Uppsala, Sweden). A 523-nucleotide fragment of the core gene was amplified with the following primers: 5′ outer set, 5′GTGGTACTGCCTGATGGTG3′ (nucleotides 283 to 304), 3′ outer set, 5′GGCAATCATTGGTGACGTGG3′ (nucleotides 965 to 943); 5′ inner set, 5′CGAGTGCCCCGGGAGGTCTCG3′ (nucleotides 308 to 329); and 3′ inner set, 5′GGAAGATGGAGAAAGAGCAACC3′ (nucleotides 874 to 852). The hybridization temperatures were 60 and 62°C for outer and inner PCR, respectively. The PCR primers used to amplify the NS5A gene (amino acids 2209 to 2248) have been described elsewhere (23). Direct automated sequencing was performed on both strands using the PCR inner primers. Core sequencing was performed on sera sampled after transplantation (mean, 60 months ± 6 months; range, 50 to 68 months) for all patients and also before transplantation for 35 patients (mean, 2 months before transplantation ± 1 month; range, 0 to 4 months). Genotyping was performed routinely on amplification products of the 5′ noncoding region of pre- and posttransplantation samples, using commercial hybridization techniques (InnoLipa II; Innogenetic, Ghent, Belgium) according to the manufacturer's instructions.

Quantification of serum HCV RNA.

Serum HCV RNA was quantified in posttransplantation sera by means of Taqman real-time PCR on the ABI PRISM 7700 sequence detector (Perkin-Elmer, Applied Biosystems) and with the Gold RT-PCR kit (Perkin-Elmer, Foster City, Calif.) using previously described primers and internal probes (18). Tenfold dilutions of a control serum, quantified with a noncompetitive PCR-based assay (Amplicor Cobas HCV Monitor test; Roche Molecular Systems, Branchburg, N.J.), were used as standards for each Taqman run. The estimated sensitivity of the Taqman assay was 100 copies/ml.

Histology.

Histological evaluation of the liver was based on 5- to 30-mm-long liver specimens obtained by percutaneous biopsy with 0.8- to 1.4-gauge needles. The specimens were fixed in 10% formalin buffer and stained with hematoxylin-eosin. Biopsy specimens with signs of rejection were excluded, and histological results were classified in four categories as follows: normal, lobular hepatitis, chronic hepatitis (defined by piecemeal necrosis or fibrosis), and cirrhosis. The METAVIR score was used to classify biopsy specimens, using simplified scores for fibrosis (F0, no fibrosis, to F4, cirrhosis) (2).

Nucleotide distance analysis.

Three types of nucleotide distance were computerized. Rates of synonymous (d_S) and nonsynonymous (d_N) substitutions per synonymous and nonsynonymous site were calculated with the Molecular Evolutionary Genetics Analysis (MEGA) software package version 1.01 (13). The Kimura two-parameter distance (d) (transition/transversion ratio, 2) was calculated with the DNADIST module of the PHYLIP package version 3.2. Phylogenetic trees were constructed by the neighbor-joining method from a Kimura two-parameter distance matrix, and bootstrap values were determined from 1,000 bootstrap resamplings of the original data (NEIGHBOR and SEQBOOT in the PHYLIP package).

Statistical analysis.

The fibrosis score, age, genotype, genotype 1b, sex, serum HCV RNA (logarithmic value), geographic origin, and immunosuppression were analyzed using conventional linear multiple regression (Statistica; Statsoft, Tulsa, Okla.).

Mantel's test was used to determine if two patients infected by genetically close HCV strains had an increased chance of having similar levels of fibrosis or virus load relative to two patients infected by more distant strains. This generalized regression permutation procedure is commonly used to compare two distance matrices. Genetic distance matrices consisted of pairwise Kimura two-parameter distances (d) and also pairwise distances at synonymous (d_S) and nonsynonymous (d_N) sites. The phenotypic distances between each pair of patients were calculated for a set of quantitative and qualitative variables. Distances between qualitative markers (fibrosis score, level of HCV viremia, and age of donor and recipient) were the absolute differences in these variables. Qualitative markers, genotype, genotype 1b (yes-no), recipient and donor sex, recipient origin, and initial immunosuppression, were arranged in similarity matrices in which the distance was 0 if the variables were identical and 1 otherwise. The correlation between each distance matrix and each phenotypic matrix was evaluated using the Pearson correlation coefficient (R₀), which ranges from −1.0 for a perfect negative correlation to 1.0 for a perfect positive correlation between two matrices. As variables arranged into matrices are not independent (e.g., the distance between cases 1 and 3 is not independent of the distance between cases 1 and 2, because case 1 is involved in both), the significance of the correlation is determined by a permutation test (27). The rows and columns of one matrix were randomly permuted 10,000 times, and the Pearson correlation was calculated for each permutation. The measure of significance (P) is given by the ratio N/10,000, where N is the number of times that R₀ is exceeded by correlation coefficients calculated with permuted matrices. The Mantel test was computerized on R4 software written by P. Legendre and P. Casgrain (http://www.fas.umontreal.ca/biol/casgrain/fr/labo/R/index.html), which also calculates the Smouse-Long-Sokal partial Mantel statistic, a partial linear correlation of two matrices after the linear effect of a third matrix is removed. This test shows whether a correlation between two variables is independent of a third variable.

RESULTS

Sequence analyses.

The sequences corresponding to the first 170 amino acids of the core protein were determined in all 68 patients after liver transplantation. Direct sequencing was used, and the sequences obtained represent, for each patient, the consensus of different quasispecies sequences. No stop codons, deletions, or insertions were observed in any of the consensus core sequences. Core sequences were available before transplantation for 35 patients. The mean homology with posttransplantation strains was 96% ± 2% (range, 85 to 100%) and 98% ± 2% (range, 92 to 100%) at the nucleotide and amino acid levels, respectively. In this study, we analyzed the sequences obtained during the fifth year after liver transplantation. NS5A sequences were available after liver transplantation in 39 of 45 patients infected by genotype 1b strains.

Multivariate analysis of variables influencing fibrosis and viral replication after liver transplantation.

Among the 68 patients, 49 (72%) developed histologically confirmed recurrent lobular hepatitis C a mean of 11 (± 14; range, 1 to 60) months after liver transplantation. Five years after transplantation, 8 of 49 patients were cirrhotic (F4), 9 were F3, 6 were F2, and 22 were F1. The remaining 23 patients had no fibrosis (F0) and had mild portal inflammation. In multivariate regression analysis, the fibrosis score correlated negatively with the serum HCV RNA level measured at 5 years (Table 2 and Fig. 1).

TABLE 2.

Multiple regression analysis of the 5-year fibrosis score

Independent variable	Dependent variable valuea
	Coefficient (95% CI)	P
Genotype 1b	0.41 (−0.23–1.11)	NS
Male (recipient)	0.13 (−0.07–0.28)	NS
Male (donor)	0.09 (−0.11–0.21)	NS
Age (recipient)	−0.010 (−0.025–0.005)	NS
Serum HCV RNA	−0.68 (−1.025–−0.345)	0.003

^aCI, confidence interval; NS, not significant. 

FIG. 1.

Levels of HCV RNA measured by Taqman PCR (log-transformed values) and the fibrosis score at the time of the 5-year routine biopsy after liver transplantation. The error bars indicate standard deviations.

Relationship between genetic and phenotypic distances.

Figure 2 shows the phylogenetic tree and 5-year fibrosis scores (F) observed in corresponding patients. The core Kimura distances (d) had a bimodal distribution, corresponding to strains belonging to the same or distinct genotypes (data not shown). Mantel's test showed no correlation between the genetic distance, d, and the phenotypic distance when all 68 patients were considered. However, when comparisons were made among the 53 patients infected by genotype 1 strains, the genetic distance, d, and also d_S and d_N, correlated with the phenotypic distance calculated for fibrosis scores and HCV viremia (Table 3 and Fig. 3). In this subgroup, correlations were far stronger for the Kimura distance, d, than for d_S or d_N. The relationships among the genetic, fibrosis, and viremia distance matrices were independent, as shown by applying partial Mantel tests to three matrices (Table 4). In other words, the shorter the genetic distance between two genotype 1 strains, the more similar the severity of liver fibrosis and the level of HCV viremia in the corresponding two patients. The correlations observed with d_S or d_N matrices were dependent on the d matrix. No relationship was found between d, d_S or d_N and matrices calculated for age (donor or recipient), sex (donor or recipient), geographic origin (recipient), or immunosuppression (Tables 3 and 4).

FIG. 2.

Phylogenetic tree constructed by the neighbor-joining method from a Kimura two-parameter distance matrix of HCV core sequences (left) in 68 liver transplant patients with recurrent hepatitis C and corresponding 5-year fibrosis scores (right). The apparent clustering between close core sequences and fibrosis scores was confirmed by Mantel's test. Bootstrap percentages (>50%) of 1,000 bootstrap replicates are given along the appropriate branch.

TABLE 3.

R₀ and P value of Mantel's test between genotypic (matrix A) and phenotypic (matrix B) matrices in 53 liver transplant patients infected by HCV genotype 1

Matrix A	Matrix B	R0	P = N/10,000a
d	Fibrosis	0.291	0.001
d	HCV RNA	0.024	0.009
d	Sex	0.010	0.421
d	Age	0.041	0.240

^aN is the number of times that R₀ is exceeded by correlation coefficients calculated with permuted matrices (10,000 random permutations). 

FIG. 3.

Mean phylogenetic Kimura distances between strains defined in the core region compared to pairwise differences in fibrosis scores (A) and levels of serum HCV RNA (B) measured during the fifth-year posttransplantation in 53 patients infected by genotype 1. Each box indicates d ± 1.96 standard error.

TABLE 4.

Smouse-Long-Sokal partial Mantel test results between genetic and phenotypic matrices in 53 patients infected by HCV genotype 1^a

Matrix A	Matrix B	Matrix C	R0	P
d	Fibrosis	HCV RNA	0.204	0.004
d	HCV RNA	Fibrosis	0.212	0.010
Fibrosis	HCV RNA	d	0.190	0.010

^aThe partial linear correlation of matrix A with matrix B was calculated after the linear effect of matrix C was removed. d, Kimura distance; 10,000 permutations. 

No correlation was observed between genetic distances calculated in the NS5A region (39 patients studied) and the above-mentioned phenotypic matrices, although the core and NS5A distance matrices correlated with each other.

Alignment of core protein sequences.

Alignment of core amino acid sequences failed to reveal unambiguous motifs or mutations corresponding to a particular phenotype.

DISCUSSION

This study of patients with recurrent hepatitis C after liver transplantation shows that the closer the genetic distance between two HCV strains, the more similar the levels of fibrosis and viral replication in the corresponding patients.

Mantel's test has been validated for the comparison of genetic and phenotypic and distance matrices in studies of human evolution (28, 29, 30) and in problems of spatial correlation (14). It has also been used in virology to demonstrate the cellular tropism of quasispecies of human immunodeficiency virus type 1 (20) and HCV (22). Mantel's test applied to all pairwise comparisons of strains avoids an arbitrary classification of HCV strains according to a prototype strain. However, the choice to define the fibrosis distance between two patients as the arithmetic difference between their METAVIR scores was arbitrary (2). The other possibility was to use similarity matrices, in which the distance is 0 if the two patients belong to the same phenotypic class (cirrhosis or no cirrhosis, for instance) and 1 in other cases. We tested such similarity phenotypic matrices and obtained very similar results (data not shown). The METAVIR fibrosis score was considered a continuous variable in this study, as in others (2, 17).

The correlations between genetic and phenotypic matrices were found within the main subgroup of patients infected by genotype 1 strains but not when strains of different genotypes were compared. In other words, our results strongly suggest, at least for genotype 1 isolates, that different HCV strains have different degrees of intrinsic pathogenicity. Correlations were found with core but not with NS5A sequences, although the two distance matrices correlated with each other. The fact that the sequenced NS5A region was shorter than the sequenced core region may explain this discrepancy, but it is also possible that the core region directly influences viral pathogenicity while the NS5A region does not. A pathogenic effect directly mediated by the core protein cannot be inferred from our data, as alignment of core amino acid sequences failed to identify clear clustering of amino acid sequences with particular phenotypes. The weak correlations observed between phenotypic matrices and the d_N matrix, which reflects the rate of amino acid substitutions, also argues against a role of the core protein sequence in the observed phenotypic differences. Thus, our data suggest that the nucleotide sequence of the core region, rather than its primary amino acid sequence, is involved. This highlights the importance of the phylogenetic background. Similar findings have been made with the so-called interferon sensitivity-determining region (ISDR) located in the nucleotide sequence of the NS5A protein. The ISDR model is mainly predictive of the response to interferon of Japanese HCV type 1b isolates, results for non-Japanese HCV type 1b isolates being far less clear cut. Similarly, in the field of liver transplantation, published data on the possible detrimental influence of genotype 1b on HCV recurrence tend to conflict, especially between European and non-European series. Our results suggest that these divergences may be linked to the heterogeneous distribution of particularly pathogenic genotype 1b strains among the different series. We did not find worse 5-year fibrosis scores in genotype 1b-infected patients than in patients infected by other genotypes. This may be because patients infected by genotype 1b strains were receiving antiviral therapy for severe disease and were therefore ineligible for this study.

It is noteworthy that the levels of HCV RNA observed in our study are low relative to previously published values. However, most papers report high levels of replication in the first or second year following transplantation, while all the patients in this series were evaluated at 5 years and had mild immunosuppression. The negative relationship between the levels of fibrosis and HCV replication confirms previous observations in a similar population (5) and could be explained by a host response that both induces fibrosis and reduces viral replication. In this series and others (8, 9, 10, 11), liver transplantation enhanced the fibrogenic properties of HCV: 5 years after infection, 12% of our patients were cirrhotic. This accelerated course of liver disease suggests that the putative intrinsic pathogenicity of HCV has a greater influence in transplanted patients than in immunocompetent patients. For these reasons, liver transplantation is a useful model with which to examine genotypic and phenotypic relationships during hepatitis C infection.