Distinct hepatitis C virus core and F protein quasispecies in tumoral and nontumoral hepatocytes isolated via microdissection

Authors

  • Rodolphe Sobesky,

    Corresponding author
    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Institut Pasteur, Laboratoire Associé au Centre National de Référence pour les Hépatites Virales B et C, Paris, France
    3. Université Paris-Sud, Faculté de Médecine, Orsay, France
    • INSERM U 785, Centre Hépato-Biliaire, Hôpital Paul Brousse, 12 Avenue Paul Vaillant Couturier, 94807 Villejuif, France
    Search for more papers by this author
    • fax: (33)-1-45-59-60-80

  • Cyrille Feray,

    1. INSERM, CIC-0004, Nantes, France
    2. Université de Nantes, Nantes, France
    Search for more papers by this author
  • François Rimlinger,

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Institut Pasteur, Laboratoire Associé au Centre National de Référence pour les Hépatites Virales B et C, Paris, France
    3. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author
  • Nicolas Derian,

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Institut Pasteur, Laboratoire Associé au Centre National de Référence pour les Hépatites Virales B et C, Paris, France
    3. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author
  • Alexandre Dos Santos,

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Institut Pasteur, Laboratoire Associé au Centre National de Référence pour les Hépatites Virales B et C, Paris, France
    3. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author
  • Anne-Marie Roque-Afonso,

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author
  • Didier Samuel,

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author
  • Christian Bréchot,

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author
  • Valérie Thiers

    1. Institut National de la Santé et de la Recherche Médicale (INSERM), U785, Villejuif, France
    2. Institut Pasteur, Laboratoire Associé au Centre National de Référence pour les Hépatites Virales B et C, Paris, France
    3. Université Paris-Sud, Faculté de Médecine, Orsay, France
    Search for more papers by this author

  • GenBank sequence accession numbers: EF652512 to EF652810.

  • Potential conflict of interest: Nothing to report.

Abstract

Hepatitis C virus (HCV) genetic variability may be involved in liver carcinogenesis. We investigated HCV core and corresponding putative F protein genetic variability in hepatocellular carcinoma (HCC) and cirrhotic nodules. Hepatocyte clusters from 7 patients with HCC and HCV1b-related cirrhosis were isolated via microdissection of HCC tissues and 2 nontumoral cirrhotic nodules. The HCV core complementary DNA was cloned and sequenced from each liver compartment and from the serum of 2 patients. Nucleotide diversity and synonymous and nonsynonymous substitutions were analyzed within and between compartments via phylogenetic analysis and Mantel's test. Liver HCV RNA accumulation was lower in HCC. Increased quasispecies diversity and complexity was observed with HCC in 6 of 7 patients. Mantel's test demonstrated marked compartmentalization of quasispecies between HCC and cirrhotic nodules in all 7 patients and also between the 2 nontumoral nodules in 5 of them. Synonymous–nonsynonymous substitution analysis indicated low selection against tumoral core quasispecies in all patients and a more selective pressure against F protein quasispecies in all compartments. In the 2 subjects analyzed, HCC and nontumoral hepatocyte quasispecies were only minor or undetected in serum. Conclusion: In tumoral hepatocytes, low-replicating hepatitis C quasispecies are compartmentalized and more diversified and are subjected to low selective pressure. Our study supports the importance of core genetic variability in hepatocellular carcinogenesis. (HEPATOLOGY 2007.) This is a corrected version of the abstract first published online on 12 October 2007 — the corrected version appears in print.

Cirrhosis due to hepatitis C virus (HCV) infection frequently leads to hepatocellular carcinoma (HCC).1 The HCV RNA virus encodes a single polyprotein, cleaved to release structural and nonstructural proteins.2 Mature HCV core protein (p21), produced within the cytoplasm from the N-terminal portion of the polyprotein by host signal peptidases, has a molecular weight of 21 kDa.3 The function of core protein is to assemble and package HCV RNA. A variety of cellular pathways, including transactivation, signal transduction, and apoptosis, may be triggered by core protein.4–6 Some transgenic mice expressing core as well as HCV full-length polyprotein develop liver steatosis followed by HCC.4, 7

The existence of an HCV protein encoded through alternative translation in an overlapping reading frame of the core sequence (known as F protein or alternative reading frame protein) has recently been hypothesized.8–14 Despite this debate, the main evidence for the expression of this putative protein, which has never been directly observed in native tissues, has been the specific immune humoral and cellular responses mounted against it in patients with chronic HCV infection.9, 13, 15, 16 Its functions remain unknown.14, 16

The high rate of HCV replication, combined with lack of an error correction mechanism and immune selection, has driven the emergence of closely related viral isolates. Thus HCV exists as a quasispecies in patient sera or tissues, and its genetic variability has been well characterized, especially with respect to the 5′ untranslated region, E2/hypervariable region 1, core, and NS5A.17–19 This genetic variability may have several important effects in terms of viral persistence and pathogenesis.20, 21 Some differences between serum and liver quasispecies have been sought but inconsistently described.22–24 The distribution of liver quasispecies may also differ when HCC and nontumoral cirrhotic liver are compared.25–30 Finally, some articles have shown that peripheral blood mononuclear cells or hepatic lymph nodes can harbor minor or major variants not found in the serum.31, 32 However, in none of these studies was serum, different parts of the liver, and HCC from the same subjects examined. Furthermore, statistical analysis of quasispecies compartmentalization that would clearly enable a demonstration of their nonrandom distribution and the detection of multiple clustering has rarely been applied.33–35 Finally, none of these studies used microdissection, and the analysis of compartmentalization in the liver was limited by the tissue heterogeneity.

The identification of specific viral variants in a given compartment raises the key question of whether viral proteins with different sequences might have different functional properties. In this respect, biological studies performed on HCC have shown that HCV variants identified in tumoral liver could possess distinct biological properties affecting cellular pathways when compared with nontumoral variants.26, 28 Thus, a detailed compartmentalization study based on tumoral and nontumoral cirrhotic nodules could reinforce the hypothesis of a selection of variants with modified functional properties in tumor and gain insight into the mechanism of HCV-induced carcinogenesis.

During this study of 7 patients, we compared quasispecies distribution of the HCV core gene and of putative F protein in tumoral and nontumoral hepatocytes isolated from HCC and from 2 distinct cirrhotic nodules using laser capture microdissection. We present definitive evidence for HCV compartmentalization in both tumoral tissue and cirrhotic nodules.

Abbreviations

dN, nonsynonymous substitution; dS, synonymous substitution; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; NT1, nontumor tissue cirrhotic nodule 1; NT1, nontumor tissue cirrhotic nodule 2; PCR, polymerase chain reaction.

Patients and Methods

Patients.

Seven patients infected with HCV genotype 1b and with cirrhosis and histologically proven HCC were selected (Table 1). None of them was positive for hepatitis B surface antigen or human immunodeficiency virus antigen, consumed excessive quantities of alcohol, or had genetic hemochromatosis or other causes of chronic hepatitis. None of them had been treated with interferon before surgery. Fresh tumor and nontumor tissue samples were collected from each patient after hepatectomy for HCC and then stored at −80°C until use. Informed consent was obtained from each patient before the study was initiated. Serum samples collected at the same time as surgery were available from 2 patients.

Table 1. Clinical Characteristics of Patients
Patient No.Age (Years)Metavir ScoreEdmondson GradeHCV RNA Titer (×103)
NontumoralTumoral
  1. HCV loads are defined as numbers of HCV RNA copies per 105 cell equivalents (see Material and Methods).

174A1F4II1847.7
259A2F4II369733
359A2F4II1026.5
464A1F4II1953.2
568A2F4I42240.3
670A2F4II2981.6
774A3F4I11614.9

Laser Capture Microdissection.

Under RNase-free conditions, 7-μm frozen sections were cut with a cryostat at −20°C using a clean disposable blade for each tissue block. Mirror sections were stained with hematoxylin-eosin and then analyzed to select specific areas for microdissection with the help of a pathologist. Adjacent frozen sections were rapidly defrosted for microdissection, stained with Histogene staining solution, and dehydrated in accordance with the manufacturer's instructions (Arcturus Engineering, Mountain View, CA). The parameters applied included a 30-μm laser with a 60-mW pulse and 30-millisecond duration. Approximately 8.103 to 10.103 hepatocytes were captured for each studied area: tumor tissue and 2 different fully microdissected cirrhotic nodules (NT1 and NT2) in nontumor tissue. Captured groups of hepatocytes were stored at −80°C until use. To ensure that nonspecific transfer did not occur during microdissection, the following procedures were applied: before microdissection, a Prepstrip tissue preparation (Arcturus) was placed on the slide to flatten the tissue and remove any nonadhering cell fragments before laser capture microdissection. After capture, the cap was cleared of nonspecifically adhering tissue using an adhesive pad (Capsure Cleanup Pad, Arcturus) and microscopically controlled to ensure that all nonspecific material had indeed been eliminated.

RNA Extraction, Full-Length HCV Core Gene Amplification, Cloning, and Sequencing.

Total RNA was isolated from caps using the Qiagen RNeasy Micro Kit (Qiagen, Courtaboeuf, France) and eluted in 14 μL of RNAse-free water. After complete laser capture microdissection, RNA integrity was assessed in 3 randomly chosen samples using a bioanalyzer (Agilent Biotechnology, Massy, France). Full-length core gene amplification was performed via reverse transcription nested polymerase chain reaction (PCR) as described previously26 with the following modifications: use of the GC Rich PCR system (Roche Diagnostics, Meylan, France) including a highly proofreading DNA polymerase increased the efficiency of full core amplification. The PCR products were purified using a PCR purification kit (Roche Diagnostics) and cloned into a plasmid (Topo TA cloning; Invitrogen, Cergy Pontoise, France). At least 10 clones were bidirectionally sequenced using the dideoxy chain termination method with a DNA sequencing kit (PerkinElmer, Courtaboeuf, France) and an ABI sequencer for each compartment.

HCV RNA Quantification in HCC and Nontumoral Liver.

Real-time RT-PCR was performed in a LightCycler (Roche, Grenoble, France) using the QuantiTect Probe RT-PCR Kit (Qiagen, Courtaboeuf, France) with a 20-μL reaction volume containing 5 μL of each sample. For HCV quantification, the primers and probe were derived from the standard test available, based on 5′ untranslated region (Roche Monitor). Serial 10-fold dilutions of a positive serum were used as standards for each run. To normalize the quantification of HCV genomes in liver tissues, we also quantified on the same extracts a human single copy gene, β-globin, in accordance with the manufacturer's instructions (“LightCycler-control kit DNA” Roche Molecular Systems). For this control, the RNeasy Micro Kit (Qiagen, Courtaboeuf, France) was used, omitting the deoxyribonuclease step.

Genetic Distances Within and Between Compartments

Nucleotide and deduced amino acid sequences of core and F proteins were aligned using MEGA software (version 3.1).36 The putative F protein was translated from core nucleotide sequences on the +1 frame. Pairwise nucleotide distances were calculated using (1) the Kimura 2-parameter method with a transition-to-transversion ratio of 2; (2) the Gobojori and Nei method, which computes the numbers of synonymous (dS) and nonsynonymous (dN) substitutions and the numbers of potentially synonymous and potentially nonsynonymous sites37; and (3) the Dayhoff matrix model for the distance between amino acid sequences. Positive dS-dN distance values reflect low positive selective pressure. These distances were used to define mean distances (± standard deviation) within and between compartments.

Phylogenetic trees were constructed using the neighbor-joining algorithm, a cluster analysis method that fits sequences such as those of HCV quasispecies that have high similarity scores.38 Phylogenetic trees were drawn using both nucleotide and amino acid sequences. Statistical evaluation of the topology obtained was based on 999 replications of bootstrap sampling.

Correlation Between Genetic Distances and Compartments.

Mantel's test was used to determine if sequences from a given compartment (NT1, NT2, T or serum) were genetically closer to each other than to sequences from other compartments.39 This test compares the Kimura 2-parameter or dS-dN distance matrix with a compartment distribution matrix (Mc) of the same dimensions, where Mc(i, j) = 0 if sequences i and j are from the same compartment, and Mc(i, j) = 1 in other cases. The Pearson correlation coefficient r2 was computed for all pairs, excluding the diagonals of both matrices (observed r2). The null distribution was constructed by permuting the rows and columns of the Mc matrix 999 times. The number of times that the observed r2 was exceeded during the 999 permutations yielded the exact P value of the observed correlation. Permute software written by Philippe Casgrain (www.bio.umontreal.ca/casgrain/fr/index.html) can calculate the significance of correlations between a genetic distance (Kimura or dS-dN) and 1 of the following phenotypic distances: tumor tissue versus NT1, tumor tissue versus NT2, and NT1 versus NT2; and when serum is available, serum versus tumor tissue, serum versus NT1, and serum versus NT2. It is thus possible to perform multivariate analysis to determine whether a correlation between a genetic distance and a phenotypic distance is independent of other phenotypic distances. These calculations were performed on core and F protein sequences.

Shannon Entropy.

As an index of core genetic complexity within a given compartment, normalized Shannon entropy was calculated as:

equation image

where pi is the frequency of each amino acid sequence and N the total number of sequences analyzed in each compartment. Sn theoretically ranges from 0 (no complexity) to 1 (maximum complexity). The mean Kimura distance, reflecting genetic diversity, was calculated between all pairs of variants from each compartment.

Univariate and Multivariate Analyses.

Univariate and multivariate analyses were performed using StatView software (Abacus Concept, Berkeley, CA). Fisher's exact test was used to compare 2 nominal variables, and a t test was used to compare nominal and continuous variables. A nonparametric Spearman correlation test was used to compare continuous variables.

Results

Quantification of HCV RNA in HCC and Cirrhotic Nodules.

The quantification of β-globin DNA showed that the quantities of amplifiable DNA were similar in tumoral and nontumoral liver and in the different patients (data not shown), thus confirming that similar amounts of cells were indeed being analyzed in each liver compartment. As shown in Table 1, liver HCV RNA, the quantification of which was expressed per 105 cells, was 8-fold to 186-fold less abundant in HCC than in cirrhotic nodules in 6 of 7 analyzed cases. In 1 case (P2), viral loads were similar in different compartments. The viral loads in HCC and nontumoral liver were not correlated.

Mean Genetic Distance and Entropy Within Compartments.

A total of 299 clones were sequenced in all 7 patients, and at least 10 independent clones were obtained for each of the 3 liver compartments (NT1, NT2, and tumor tissue), allowing identification of major hepatitis C quasispecies per area. Thirty-nine and 33 clones were derived from the serum in 2 patients. A total of 23 compartments (21 liver and 2 serum) were analyzed. Core sequences from tumor and nontumor compartments showed no significant difference in Shannon entropies.

Quasispecies diversity was assessed at the nucleotide level by comparing the mean Kimura parameter distance between liver compartments. In all but 1 patient, greater genetic diversity was observed in the quasispecies found in HCC than in those isolated from NT1 or NT2 nodules (P < 0.01) (Table 2). The dS-dN distance, a marker of selective pressure, was compared between liver compartments. The mean dS-dN distances defined in core were greater in tumoral quasispecies than in those infecting the NT1 and NT2 in 5 of 7 patients. This suggested that tumoral HCV quasispecies are subjected to weaker selective pressure than nontumoral quasispecies. Taken together, these findings suggest that core is submitted to weak selective pressure in HCC; moreover, tumoral HCC quasispecies are subjected to weaker selective pressure than nontumoral quasispecies.

Table 2. Mean Genetic Distance and Entropy Within Compartments
Patients and CompartmentsCoreF Protein
Nucleotide Distances (mean ± SD)*Amino Acid Distances (dS-dN ± SD)EntropyAmino Acid Distances (dS-dN ± SD) Entropy
  • *

    Nucleotide distance was computed using the Kimura 2-parameter model.

  • dS and dN distances were computed using the Nei-Gojobori method.

  • Normalized Shannon entropy at the amino acid level.

Patient 1     
 NT10.014 ± 0.0020.018 ± 0.0060.850.001 ± 0.0051
 NT20.022 ± 0.0040.039 ± 0.0090.95−0.006 ± 0.0070.95
 Tumor0.026 ± 0.0040.016 ± 0.0060.530.000 ± 0.0000.27
 Serum0.010 ± 0.0020.056 ± 0.0110.81−0.005 ± 0.0060.95
Patient 2     
 NT10.006 ± 0.0020.012 ± 0.0050.61−0.005 ± 0.0030.71
 NT20.005 ± 0.0010.008 ± 0.0040.63−0.005 ± 0.0010.73
 Tumor0.023 ± 0.0040.068 ± 0.0150.850.009 ± 0.0101
Patient 3     
 NT10.004 ± 0.0010.006 ± 0.0040.270.001 ± 0.0030.59
 NT20.005 ± 0.0020.009 ± 0.0050.4−0.003 ± 0.0030.8
 Tumor0.006 ± 0.002−0.003 ± 0.0020.760.006 ± 0.0060.65
Patient 4     
 NT10.002 ± 0.0010.002 ± 0.0020.25−0.005 ± 0.0030.84
 NT20.005 ± 0.0020.010 ± 0.0050.430.002 ± 0.0020.36
 Tumor0.010 ± 0.0020.014 ± 0.0050.65−0.001 ± 0.0030.85
Patient 5     
 NT10.001 ± 0.0010.034 ± 0.0090.95−0.004 ± 0.0050.95
 NT20.016 ± 0.0040.004 ± 0.0030−0.002 ± 0.0010.34
 Tumor0.024 ± 0.0040.065 ± 0.0120.7−0.011 ± 0.0050.88
Patient 6     
 NT10.003 ± 0.0020.003 ± 0.0010.270.000 ± 0.0000.27
 NT20.002 ± 0.0010.002 ± 0.0010.4−0.001 ± 0.0010.27
 Tumor0.004 ± 0.0010.004 ± 0.0010.680.007 ± 0.0050.45
Patient 7     
 NT10.001 ± 0.0010.001 ± 0.0010.14−0.001 ± 0.0010.27
 NT20.008 ± 0.0030.006 ± 0.0060.620.013 ± 0.0080.78
 Tumor0.003 ± 0.0010.014 ± 0.0050.49−0.003 ± 0.0030.82
 Serum0.009 ± 0.0020.002 ± 0.0020.520.000 ± 0.0010.32

Phylogenetic and Matrix Correlation Analyses Between Compartments.

Phylogenetic trees drawn from nucleotide and core amino acid alignments revealed a clustering of variants in HCC in 4 of 7 patients (Figs. 1A –B, 2). Interestingly, in one patient (Fig. 1B), a strict separation of quasispecies was observed as a function of each compartment. However, in 3 patients, a clustering of sequences was also found in non-HCC tissues, being observed in only one or both cirrhotic nodules (Fig. 3). This was observed for nucleotides and core amino acid sequences. A comparison was made of serum and liver compartments in 2 subjects. In the first patient, for core proteins, all serum clones (n = 33) except one clustered separately from liver compartments (Fig. 1B). The other patient exhibited a more complex pattern because serum quasispecies were scattered throughout the phylogenetic tree (Fig. 1A).

Figure 1.

Phylogenetic trees of liver compartments NT1 (orange squares), NT2 (purple squares), and tumor (black circles) and serum (green triangles) in (A) patient 1 and (B) patient 7, including nucleotide, amino acid, and F protein representation. Bootstrap values above 65% are presented.

Figure 2.

Phylogenetic trees of liver compartments NT1 (orange squares), NT2 (purple squares), and tumor (black circles) for patients 2 and 3, including nucleotide, amino acid, and F protein representation. Bootstrap values above 65% are presented.

Figure 3.

Phylogenetic trees of liver compartments NT1 (orange squares), NT2 (purple squares) and tumor (black circles) for patients 4, 5, and 6, including nucleotide, amino acid, and F protein representation. Bootstrap values above 65% are presented.

An analysis of quasispecies clustering is insufficient using phylogenetic tree analysis alone, particularly when bootstrap values are low or when multiple clustering exists. Correlating pairwise genetic and phenotypic distances through permutations (or using Mantel's test) is a clustering method that is entirely separate from phylogenetic tree analysis. Multivariate analysis using Mantel's test further demonstrated in all 7 patients analyzed that quasispecies were not randomly distributed between the compartments (Table 3). In 6 of 7 cases, HCC-derived core differed from those detected in other compartments. This was highly significant in most cases and independent of other possible phenotypic matrices (NT1 versus NT2, serum versus tumor tissue, serum versus NT1, serum versus NT2). In 5 of 7 patients, the correlation matrix also provided evidence that variants found in 1 nontumoral nodule were significantly distant from variants detected in the other nontumoral nodule, and this was also independent of other possible matrices (Table 3). Overall, phylogenetic trees and matrix correlation analyses revealed a clear clustering of core tumoral sequences. Matrix correlation analysis reinforced phylogenetic analysis by demonstrating almost exclusive tumoral compartmentalization; it also showed that such compartmentalization occurred among cirrhotic nodules.

Table 3. Matrix Correlation Analyses Between Compartments
Patient No.Compared CompartmentsP Value (Mantel's Test)
Nucleotide Distances*Amino Acid Distances
  • Comparison was made using Mantel's test between quasispecies isolated from tumorous hepatocytes (T) and nontumorous hepatocytes (NT1, NT2) from 2 separate cirrhotic nodules and from serum (S).

  • *

    Nucleotide distances were computed using the Kimura 2-parameter model.

  • Amino acid distances were computed using the Nei-Gojobori method.

1T versus NT10.0020.021
 T versus NT20.10.134
 NT1 versus NT20.0010.001
 S versus T0.0010.252
 S versus NT10.1030.162
 S versus NT20.0150.424
2T versus NT10.0010.001
 T versus NT20.0010.001
 NT1 versus NT20.0010.001
3T versus NT10.0020.003
 T versus NT20.0020.002
 NT1 versus NT20.1820.142
4T versus NT10.0010.001
 T versus NT20.0010.001
 NT1 versus NT20.0050.001
5T versus NT10.0070.011
 T versus NT20.0160.002
 NT1 versus NT20.3730.419
6T versus NT10.050.01
 T versus NT20.10.01
 N1 versus N20.0010.01
7T versus NT10.0030.003
 T versus NT20.0030.003
 NT1 versus NT20.0030.013
 S versus T0.0030.003
 S versus NT10.0030.003
 S versus NT20.0030.04

Amino Acid Profiles.

Only 5 stop mutations were observed among the 299 core sequences, and these clones were excluded from the study. No obvious hot spots of amino acid substitution were observed in the different compartments for core protein. Minor changes in hydrophobicity and antigenicity were observed but appeared to be randomly distributed along the core sequence. Of note was the fact that all the amino acid core variations observed in the 7 patients had already been described among the 356 previously reported HCV genotype 1b core sequences available in the HCV databases (European, American and Japanese).40, 41

Putative F Encoding Open Reading Frame Genetic Variability.

Whether or not the F protein is actually expressed in HCV-infected patients, either transiently or permanently, remains a subject of debate; however, F protein-specific humoral and cellular immune responses have been reported in patients with chronic HCV infection.9, 13, 15, 16 We thus took advantage of the present study to investigate the pattern of genetic variability in the F-encoding sequence. We studied F-derived population parameters and compared them with those observed for core protein. Core and F-derived amino acid entropy values were correlated (P = 0.0009; Spearman's test). In 20 of 23 cases, the mean dS-dN distances defined in F protein were smaller than those defined in core; in particular, F protein dS-dN was negative in 13 of 23 of the compartments studied, while core dS-dN was positive for core protein in 22 of 23 cases (Table 2). dS-dN values for core and F protein were not correlated. Thus, the difference in dS-dN values observed between core and F proteins was not simply due to the frameshift. Moreover, no dS-dN differences were observed between tumor and nontumor tissues.

Phylogenetic trees drawn using F protein also revealed a clustering of variants in HCC in 4 of 7 patients (Figs. 1A-B, 2), and multivariate analysis using Mantel's test further demonstrated in all 7 patients that quasispecies were not randomly distributed between the compartments. In 6 of 7 cases, HCC-derived F variants differed from those detected in other compartments. Finally, comparative analysis demonstrated a higher rate of amino acid mutations in F protein than in core, thus confirming the dS-dN analysis.

Discussion

We analyzed the distribution of HCV core and F variants using laser capture microdissection of tumor and nontumor hepatocytes selected in different cirrhotic nodules and in HCC specimens. Quasispecies analysis based on genetic and phenotypic matrix comparisons revealed that in adjacent tumoral hepatocytes, there was a frequent and strong clustering of low replicating HCV core variants that were subjected to weak selective pressure. The clustering of HCV core protein was also observed within cirrhotic nodules.

The persistence of HCV RNA in HCC tumor cells has been demonstrated elsewhere.42, 43 The level of HCV RNA replication in HCC compared with nontumor liver has been debated. However, most studies have noted an overall low expression level of HCV genome in tumor cells regarding both viral RNA and proteins.44–46 Consistent with this view, our results demonstrate a low level of HCV replication in tumoral tissue in all but 1 patient. However, measuring HCV RNA from a small number of hepatocytes (≈8 × 103-10 × 103) raises the theoretical problem of quasispecies distribution bias due to sampling in a small population of HCV RNAs. In this context, it is important to note that, although HCV replication was lower in HCC tissues than in nontumoral cirrhotic nodules in 6 of 7 subjects, the complexity and diversity of HCV were greater in HCC than in nontumoral cirrhotic nodules in almost all analyzed subjects. Furthermore, although the viral load was higher in cirrhotic nodules, the compartmentalization of HCV core and F quasispecies between 2 cirrhotic nodules was also observed. Therefore, sampling bias is not a likely explanation of our observations.

Our study based on microdissection did indeed demonstrate that adjacent hepatocytes are infected by variants, which diverge markedly from circulating quasispecies. Using Mantel's test, we were able to evaluate the probability of the clustering of quasispecies sharing a similar phenotype (for instance, cellular tropism). This test is more sensitive than phylogenetic analysis.47 In this respect, bootstrap values (which are traditionally considered as being significant when higher than 70%), evidenced clear clustering in a limited number of cases. Conversely, Mantel's test demonstrated the significant proximity of variants infecting compartments in most cases.

There is considerable evidence to suggest the existence of specific HCV quasispecies in different compartments such as the liver, serum B cells or monocytes.32, 33 In light of these results, it is probable that the circulating HCV population may be built up from a heterogeneous population of variants released from the hepatic and extrahepatic compartments. However, most studies have suggested a minor contribution of extrahepatic viral replication to the circulating HCV pool.48 Alternatively, a difference in HCV variant clearance rates may be responsible for the observed differences in the circulating and hepatic variants.

Another major finding offered by our study is the demonstration that the compartmentalization of HCV quasispecies differs in nontumoral cirrhotic nodules and HCC. In line with previously reported data, compartmentalization was observed consistently in tumoral hepatocytes.25, 27, 29, 30 Our results also showed that HCV replication is less efficient in tumoral hepatocytes and that HCC quasispecies are subjected to weak selective pressure, exhibiting stronger diversity and entropy than those derived from nontumoral tissues. Also, we demonstrated the compartmentalization of HCV core quasispecies between 2 cirrhotic nodules. Such a topological distribution of HCV quasispecies in distinct parts of cirrhotic liver, together with increased genetic diversity, has previously been observed in advanced liver diseases but not at the early stage of infection.24 Overall, it is possible to hypothesize on the basis of our results that in cirrhotic liver, tumoral or nontumoral hepatocytes infected by HCV are not efficiently reinfected by circulating virions. HCV quasispecies replicating in hepatocytes may inhibit either the entry, viral replication, or processing of circulating quasispecies in these hepatocytes; alternatively, cell-to-cell spread within the liver is also a possibility.

Furthermore, the simultaneous presence of different HCV quasispecies variants in cirrhotic nodules and HCC raises the question of the functional capacities of corresponding core proteins. The topological distribution of HCV in different parts of a cirrhotic liver may simply reflect tissue-specific evolution, in which cirrhotic nodules and tumor act as selective bodies for HCV variants. Hypothetically, the continuous production of HCV quasispecies might permit ongoing virus production in a changing cellular context due to the progression of liver disease. In this respect, the results obtained with different HCV replicons illustrate the importance of nonsilent mutations to fitting viral sequences with the cellular environment.49, 50 On the other hand, it is conceivable that accumulations of randomly distributed mutations in HCV core proteins confer different functional properties and perhaps alter the regulation of cellular genes influencing host cell behavior. In this respect, we previously showed that some tumoral core variants could have particular biochemical properties that reflected their adaptation to the tumoral environment. Thus, tumor-derived HCV cores were shown to activate double-stranded, RNA-dependent protein kinase and also inhibit transforming growth factor β–dependent signaling.26, 28 These hypotheses are not mutually exclusive. Indeed, coevolutionary events have been suggested during persistent HCV infection in a cell culture system that might lead to the selection of viral and host variants that could favor the survival of both.51 Thus, our observations may point to the selection of viral variants that display adaptative mutations to preneoplasic hepatocytes, such as those present in cirrhotic nodules.

Finally, our study also contributes to the general debate on HCV F protein. Our investigations revealed a stronger selective pressure on putative F than on core protein. F protein appears to be subjected to specific selective pressure that is distinct from the obligatory constraint induced by the frameshift. This pressure may reflect immune-driven selective pressure and/or the adaptation of F protein to the cellular environment.52 In this respect, F protein has recently been described as potentially interacting with the cytoskeleton.52 Thus, by demonstrating F variants that cluster in HCC, this study will make it possible to test whether or not different F protein mutants display distinct biological activities.

In conclusion, through the laser capture microdissection of adjacent tumoral and nontumoral hepatocytes, we have demonstrated the nonrandom topological distribution of HCV quasispecies within HCC and within nontumoral cirrhotic nodules. Our study supports the selection of HCV core and F protein quasispecies adapted to the different stages of liver carcinogenesis.

Acknowledgements

The authors thank Eric Marchadier and Elisabeth Dussaix for HCV quantifications and France Demaugre for fruitful and stimulating discussions.

Ancillary