Nonrandom Distribution of Hepatitis C Virus Quasispecies in Plasma and Peripheral Blood Mononuclear Cell Subsets

Abstract

The existence of an extrahepatic reservoir of hepatitis C virus (HCV) is suggested by differences in quasispecies composition between the liver, peripheral blood mononuclear cells, and serum. We studied HCV RNA compartmentalization in the plasma of nine patients, in CD19⁺, CD8⁺, and CD4⁺ positively selected cells, and also in the negatively selected cell fraction (NF). HCV RNA was detected in all plasma samples, in seven of nine CD19⁺, three of eight CD8⁺, and one of nine CD4⁺ cell samples, and in seven of eight NF cells. Cloning and sequencing of HVR1 in two patients showed a sequence grouping: quasispecies from a given compartment (all studied compartments for one patient and CD8⁺ and NF for the other) were statistically more genetically like each other than like quasispecies from any other compartment. The characteristics of amino acid and nucleotide substitutions suggested the same structural constraints on HVR1, even in very divergent strains from the cellular compartments, and homogeneous selection pressure on the different compartments. These findings demonstrate the compartmental distribution of HCV quasispecies within peripheral blood cell subsets and have important implications for the study of extrahepatic HCV replication and interaction with the immune system.

Hepatitis C virus (HCV) infection induces chronic hepatitis, cirrhosis, and hepatocellular carcinoma worldwide. At all stages of liver disease, HCV exists as a heterogeneous population of distinct but closely related genomes referred to as quasispecies (26). HCV quasispecies composition is mainly assessed by focusing on a particular hypervariable region (HVR1) 27 amino acids long located at the NH₂ terminus of envelope glycoprotein 2 (E2) (11). The HVR1 quasispecies composition undergoes extensive variations during the course of chronic infection, possibly corresponding to the emergence of immune escape mutants. This has been proposed as a pivotal mechanism of HCV persistence (15, 16, 19, 46, 51, 53). Nevertheless, numerous authors have detected HCV RNA in peripheral blood mononuclear cells (PBMC) and the bone marrow cells of chronically infected individuals (23, 24, 32, 34, 56). Interference with cells of the immune system is a known mechanism by which viruses evade the host response and become chronically infective (for a review, see reference 38). The notion of the infection of extrahepatic tissues by HCV is still controversial, however. Indeed, detection of HCV RNA in PBMC could be due to the simple adsorption of viral particles. The specificity and sensitivity of methods used to detect the negatively stranded HCV RNA, the obligatory replication intermediate (30), are also controversial (9, 21, 22, 29, 37, 54). Infection of extrahepatic tissues is supported by the finding that HCV quasispecies composition differs according to the sample type, i.e., liver, serum, or PBMC (2, 8, 25, 35, 37). Some human lymphoid cell lines are susceptible to HCV infection in vitro (12, 31, 47, 48, 49). A few minority HCV strains in the inoculum continued to replicate in such cells, supporting the concept that some quasispecies are preferentially lymphotropic rather than hepatotropic. However, the possibility that the difference observed in quasispecies composition between compartments is due to chance has not been formally studied. Furthermore, this compartmentalization has been described in total PBMC but not in the constituent cell types. We thus studied the distribution of HCV quasispecies in plasma, in peripheral positively selected CD19⁺, CD8⁺, and CD4⁺ cells, and in negatively selected remnant cells.

MATERIALS AND METHODS

Patients.

Nine anti-HCV-positive patients were studied. Six were infected by genotype 1b, two were infected by genotype 4, and one was infected by genotype 2 (InnoLipa II Genotyping Kit; Innogenetics, Gent, Belgium). The estimated duration of the disease ranged from 7 years to more than 30 years. Five patients had compensated cirrhosis, while three had chronic hepatitis at various stages. One patient was tested 6 months after liver transplantation for decompensated cirrhosis. None of the patients had received antiviral therapy, and all were hepatitis B surface antigen negative and anti-human immunodeficiency virus (HIV) negative.

Cell sorting.

Mononuclear cells were obtained from 20 ml of EDTA-treated blood by centrifugation through a Ficoll density gradient. Cells were washed three times, counted, and immediately used for separation and purification following immunomagnetic-positive selection with antibodies directed against specific surface molecules. First, 10 × 10⁶ to 20 × 10⁶ PBMC were suspended in 1 ml of RPMI medium. Then, 25 μl of anti-CD19 coupled to magnetic microbeads (Dynabeads M450 CD19; Dynal, Oslo, Norway) was added. The cell sample was then incubated at 4°C with gentle rotation for 30 min. After magnetic separation, the remaining cells were subjected to CD8- and then CD4-positive sorting with 140 μl of anti-CD8 (Dynabeads M450 CD8; incubation time, 30 min) and 140 μl of anti-CD4 (Dynabeads M450 CD4; incubation time, 60 min). All sorted cells were washed five times in RPMI medium. Dynabeads were next removed by using Detachbeads CD19 or CD4/CD8 (Dynal), according to the manufacturer’s instructions, and then washed. Our positive purification protocol resulted in very pure (>95%) CD19⁺, CD8⁺, and CD4⁺ cell subsets as assessed by flow cytometry with paired and directly labeled monoclonal antibodies (B4-fluorescein isothiocyanate [FITC] for CD19-positive cells, T8-FITC, and T4-FITC) purchased from Coulter (Immunotech, Marseille, France) (data not shown). The negative fraction (NF) collected at the end of the last immunomagnetic selection step contained less than 1% of CD19⁺, CD8⁺, or CD4⁺ cells and up to 80% of CD45⁺ cells. Further characterization with FITC or phycoerythrin-labeled antibodies (Ortho-Clinical Diagnostics GmbH, Neckargemünd, Germany) showed 10 to 38% CD11c⁺ cells (mainly monocytes) and 13 to 38% CD16⁺ cells (mainly natural killer cells). The remaining 10 to 20% of cells were granulocytes (not shown). Cell subsets were stored at −80°C until use.

Extraction of nucleic acids, RT-PCR, and quantification of HCV RNA.

RNA was extracted from 140 μl of plasma by using the QIAmp viral RNA kit (Qiagen GmBH) and from cell subsets by using the RNeasy Minikit (Qiagen). For detection of the HCV genome, one-fifth of the extracted plasma RNA or 1 μg of cellular RNA was subjected to reverse transcriptase PCR (RT-PCR) by using Ready-To-Go RT-PCR beads (Pharmacia Biotech, Uppsala, Sweden). A 3-μl portion of the first PCR product was subjected to a second-round PCR with Ready-To-Go PCR beads according to the manufacturer’s instructions. Outer and inner primers for the 5′ noncoding region were as follows: SF1 (5′-TGCACGGTCTACGAGACCTC-3′) and SR1 (5′-GCCATGGCGTTAGTATGAGT-3′) (outer) and SF2 (5′-GTGCAGCCTCCAGGACCCCC-3′) and SR2 (5′-GGGCACTCGCAAGCACCCTA-3′) (inner).

Outer and inner primers for HVR1, which generate a 307-bp product (+956 to +1262 according to the numbering system of Choo et al. [4]) encompassing the C-terminal E1 transmembrane region and the N-terminal hypervariable E2 region, were described elsewhere (13). To check for the absence of PCR inhibitors, RT-PCR was performed with the following primers spanning exons 3 and 4 of the cyclin A gene (52): He4 (sense) (5′-GCGGAATTCGAGTCACCACATACTATGGAC-3′) and He3 (antisense) (5′-GCGCTGCAGTAACAGCATAGCAGCAGTGC-3′).

The molecular weight of the amplified products is 545 bp for the DNA fragment and 335 bp for the mRNA fragment. Amplification products were revealed by ethidium bromide staining after agarose gel electrophoresis. Plasma HCV RNA was quantified by using the noncompetitive PCR-based Amplicor HCV Monitor assay (Roche Molecular Systems, Branchburg, N.J.) as specified by the manufacturer.

Cloning and sequencing of PCR products.

HCV isolates were amplified from plasma or cell subsets by using the HVR1 nested primer set. PCR products were cloned in the pGEM-T Easy Vector System (Promega Corp., Madison, Wis.) and transformed into Escherichia coli JM109 competent cells (Promega). After overnight incubation at 37°C, insertion was checked by PCR with the HVR1 inner primer pair on white colonies. Each of these PCR products with the correct molecular weight was sequenced bidirectionally and automatically by using the HVR1 inner primer pair on an ABI377 sequencer with the ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction kit with Amplitaq DNA polymerase (FS; Perkin-Elmer/Applied Biosystems, Foster City, Calif.).

Sequence analysis.

Nucleotide and deduced peptide sequences of cloned products were aligned by using the CLUSTALW program, version 1.5 (50). As an index of HVR1 genetic complexity within a given compartment, the normalized Shannon entropy (Sn) (55) was calculated as follows: Sn = −Σi (pi ln pi)/ln N, where pi is the frequency of each sequence and N is the total number of sequences analyzed in each compartment. Sn theoretically varies from 0 (no diversity) to 1 (maximum diversity). Synonymous (d_S) and nonsynonymous (d_N) distances were calculated with the Jukes-Cantor correction in the Molecular Evolutionary Genetics Analysis (MEGA) package, version 1.01 (18). Statistical comparisons of distances were made by using the t test. Pairwise nucleotide distances were calculated by using the Kimura two-parameter method with a transition-to-transversion ratio of 2 (DNADIST from the Phylogeny inference package (PHYLIP), version 3.2. This matrix was used to determine both evolutionary relationships among sequences and correlation with compartmental distribution. Unrooted phylogenetic trees were constructed with the MEGA package by using the neighbor-joining algorithm, a cluster analysis method that fits sequences with high similarity scores such as HCV quasispecies. The statistical evaluation of the obtained topology was performed with 500 replications of bootstrap sampling. To determine whether sequences from a given compartment shared more genetic identity with each other than with sequences from other compartments, we used the Mantel’s test (43), which compares the Kimura two-parameter distance matrix to a compartment distribution matrix Mc of the same dimensions, where Mc(i, j) = 0 if sequence i is from the same compartment as sequence j and where Mc(i, j) = 1 otherwise. The Pearson correlation coefficient r² was computed for all pairs excluding the diagonal of both matrices (observed r²). The null distribution was constructed by permuting the rows and columns of the matrix Mc 10,000 times. The number of times, more than 10,000 permutations, where the observed r² was exceeded gave the exact P value of the correlation observed.

Antigenic analysis.

The antigenic analysis was done blindly with regard to the origin of sequence. Peptide sequences were analyzed by one of us (F.P.) according to their antigenicity, hydrophobicity, and polarity values. The antigenic grouping was performed manually and with Parker’s algorithm (40). Comparisons of antigenic grouping and compartmental distribution were made by using the χ² or F tests.

Nucleotide sequence accession numbers.

The GenBank accession numbers of the nucleotide sequences reported here are AF163152 to AF163253.

RESULTS

HCV RNA detection in plasma and cell subsets.

HCV RNA sequences were detected in the plasma and in at least one peripheral blood cell subset of the nine patients tested, regardless of the degree of viremia, the disease duration, and the genotype. Overall results are shown in Table 1, along with the respective genotypes and viral loads. Cyclin A mRNA was positive in all but one of the cell samples, rendering the negative result of HCV RNA in the CD8⁺ subset of patient F noninterpretable. HCV RNA was detected in CD19⁺ cells from seven of nine patients, in CD8⁺ cells from three of eight patients, in CD4⁺ cells from one of nine patients, and in the NF cells from seven of eight patients. Two patients (A and B) had HVR sequences in four compartments: plasma, CD19⁺, CD8⁺, and NF cells.

TABLE 1.

Detection of HCV and cyclin RNA in plasma, in CD19⁺, CD4⁺, and CD8⁺ cells, and in the NF cells (CD19⁻, CD4⁻, and CD8⁻) of PBMC

Patient (genotype)a	HCV load	Plasma		CD19+ cells			CD4+ cells			CD8+ cells			NF cells
		5′NC	HVR	Cyclin	5′NC	HVR	Cyclin	5′NC	HVR	Cyclin	5′NC	HVR	Cyclin	5′NC	HVR
A (1b)	480,000	+	+	+	+	+	+	−	−	+	+	+	+	+	+
B (1b)	999,000	+	+	+	+	+	+	−	−	+	+	+	+	+	+
C (1b)	<1,000	+	+	+	+	−	+	−	−	+	−	−	+	−	−
D (1b)	11,900	+	+	+	−	−	+	−	−	+	−	+	NDb	ND	ND
E (2)	1,510,000	+	+	+	−	+	+	−	−	+	−	−	+	+	+
F (1b)	391,000	+	+	+	−	−	+	−	−	−	−	−	+	+	−
G (1b)	207,742	+	+	+	+	+	+	+	−	+	−	−	+	+	+
H (4)	190,601	+	+	+	+	−	+	−	−	+	−	−	+	+	−
I (4)	13,981	+	+	+	+	+	+	−	−	+	−	−	+	+	−

^aClinical status: A and B, cirrhosis; C, transplanted with chronic active hepatitis; D, E, and F, cirrhosis and hepatocellular carcinoma; G, H, and I, chronic active hepatitis. 

^bND, not done. 

Analysis of HVR1 cloning and sequencing in two patients.

From the 307-bp PCR product, a final length of 273 bp for patient A and 240 bp for patient B was used to perform genetic analysis.

(i) Genetic diversity.

Totals of 48 and 53 HVR1 clones were obtained from patients A and B, respectively. The number of clones per compartment ranged from 3 to 19. The alignment of the C-terminal part of E1 clearly showed the existence of two main strains, one specific to each patient, thus excluding the possibility of sample carryover. Each patient harbored a mixture of genetically distinct but closely related variants (Table 2): the average within-patient genetic distance was 0.048 ± 0.029 in patient A and 0.057 ± 0.018 in patient B. Both patients’ CD8⁺ cells harbored fewer and less-divergent variants than their plasma and CD19⁺ cells. Indeed, the genetic diversity of the E1/E2 region, estimated by calculating the normalized amino acid entropy, ranged from 0.66 (CD8⁺) to 0.94 (CD19⁺) in patient A and from 0.48 (CD8⁺) to 0.98 (plasma) in patient B; the degree of genetic diversity, determined as the average within-compartment genetic distance, was lower in CD8⁺ than in CD19⁺ cells both in patient A (0.027 ± 0.021 versus 0.07 ± 0.047; P was not significant) and in patient B (0.009 ± 0.004 versus 0.06 ± 0.011; P < 0.001).

TABLE 2.

Genetic diversity of E1/E2 quasispecies from plasma and CD8⁺, CD19⁺, and NF cells in two patients as determined by E1 and HVR1 nucleotide substitution analysis

Patient and compartment	Na	Normalized amino acid entropy	Mean genetic distanceb	Type of nucleotide substitutionc
				E1			HVR
				dS	dN	dN/dS	dS	dN	dN/dS
Patient A	48	0.74	0.048 (0.029)
Plasma	13	0.68	0.065 (0.050)	0.070 (0.026)	0.016 (0.006)	0.228*	0.191 (0.058)	0.142 (0.032)	0.743
CD19+	16	0.94	0.070 (0.047)	0.082 (0.027)	0.019 (0.007)	0.232*	0.296 (0.092)	0.245 (0.051)	0.828
CD8+	16	0.66	0.027 (0.021)	0.061 (0.021)	0.009 (0.004)	0.147**	0.032 (0.014)	0.029 (0.012)	0.906
NF	3	0	0.030 (0.00)
Patient B	53	0.86	0.057 (0.018)
Plasma	19	0.98	0.067 (0.006)	0.021 (0.016)	0.021 (0.008)	1	0.281 (0.091)	0.3535 (0.054)	1.258
CD19+	17	0.94	0.060 (0.011)	0.024 (0.016)	0.014 (0.007)	0.583	0.096 (0.04)	0.2151 (0.039)	2.240*
CD8+	6	0.48	0.009 (0.004)	0.010 (0.010)	0.007 (0.005)	0.67	0	0.0202 (0.012)	ND
NF	11	0.77	0.104 (0.051)	0.083 (0.032)	0.034 (0.012)	0.41	0.247 (0.089)	0.2473 (0.089)	1

^aNumber of independent analyzed clones. 

^bThe within-sample genetic distances were calculated with the DNADIST module of the PHYLIP package and were based on a Kimura two-parameter matrix with a transition to transversion ratio of 2. 

^cd_N and d_S, the mean proportion of nonsynonymous substitutions per nonsynonymous site and the synonymous substitutions per synonymous site, were computed by the MEGA package with the Jukes-Cantor correction. ND, not determined. d_N and d_S were compared by using a t test (∗, P < 0.05; ∗∗, P < 0.01). 

(ii) Nucleotide substitutions.

The relative proportion of synonymous (d_S) and nonsynonymous (d_N) mutations can reveal positive selection pressure operating on a viral population if d_N is greater than d_S (36). We determined average d_N and d_S values for E1 and HVR1 sequences obtained from each compartment (Table 2). In a given patient, the sequences from all the compartments yielded d_N/d_S ratios of similar magnitude, suggesting a relatively homogeneous selective pressure in all the compartments concerned. All HVR1 variants from patient A had d_N < d_S, whereas those from patient B had d_N ≥ d_S. In this latter patient, the existence of positive selection pressure on the quasispecies originating from CD19⁺ cells was demonstrated.

In the absence of a reliable method to assess HCV replication in PBMC, we assumed that only the liver contributed to the plasma viral pool and that cellular sequences originated primarily from plasma. We compared the average proportion of nucleotide substitutions per nonsynonymous sites in sequences derived from CD8⁺ and CD19⁺ cells and from plasma. In patient A, greater divergence was observed between the CD19⁺ compartment and plasma (25 ± 5%) than between the CD8⁺ compartment and plasma (11 ± 2%; P < 0.01). Conversely, in patient B, greater divergence was observed between the CD8⁺ compartment and plasma (32 ± 5%) than between the CD19⁺ compartment and plasma (20 ± 2%; P < 0.05).

(iii) Genetic relationships among compartment quasispecies.

Bootstrapped phylogenetic trees obtained from viral sequences suggested significant phylogenetic grouping according to the compartmental origin of the clones. The strains harbored by CD8⁺ lymphocytes clustered close to one another in both patients. The CD19⁺, plasma, and NF variants also clustered in patient A, whereas patient B’s plasma, NF, and CD19⁺ sequences appeared to be melted (Fig. 1).

FIG. 1.

Phylogenetic tree, patients A (A) and B (B). Abbreviations: T, CD8⁺ lymphocytes; B, CD19⁺ lymphocytes; X, negative cell fraction (CD19⁻, CD8⁻, and CD4⁻); S, plasma. The phylogenetic grouping of sequences from all four compartments in patient A and from CD8⁺ cells in patient B was confirmed by a significant correlation between genetic proximity and compartment appurtenance (P < 0.0001, by Mantel’s test; see Materials and Methods). Variants in NF cells from patient B were not included in the same node of the phylogenetic tree but were overall more genetically identical to each other than to variants from any other compartment (P < 0.0001). These NF cells have at least three phenotypes: monocytes/macrophages, natural killer cells, and granulocytes.

To obtain statistical confirmation of genetic compartmentalization, we used Mantel’s test to search for a relation between pairwise Kimura two-parameter distances and compartment distribution. Significant genetic compartmentalization was observed for CD8⁺ variants in both patients, since 10,000 random permutations of the Mc matrix did not produce a correlation coefficient higher than that found with the observed distribution (P < 0.0001). Similarly, a significant compartmental structure, with a P value of less than 0.001, was observed for sequences harbored by CD19⁺ cells in patient A and by NF cells in patients A and B. For patient B, it is noteworthy that sequences from the NF cells were not included in the same node of the phylogenetic tree. These sequences segregated in a few clusters of the tree topology. The high P value of the Mantel’s test demonstrated formally that sequences from NF cells were closer to each other than to any sequences from other compartments. This was not the case for variants in CD19⁺ cells from patient B (P = 0.07). This statistical analysis of genetic distances unambiguously demonstrated that the compartmental sequence distribution was not due to chance: quasispecies harbored by CD8⁺ and NF cells, and in at least one patient by CD19⁺ cells, were significantly different from those detected in plasma.

(iv) Grouping of quasispecies according to amino acid sequences.

Blinded antigenic grouping of all E1/E2 sequences distinguished two main groups, corresponding to each of the two patients. Further grouping yielded four antigenic groups in patient A and five in patient B (Fig. 2). The F test showed that the global antigenic distribution in patient A was compartment-specific (P = 0.04) (Table 3). Each antigenic group corresponded statistically to a particular quasispecies origin: A1 overlapped all 3 variants from NF cells; A2 overlapped 15 of 16 variants (94%) from CD8⁺ cells; A3 overlapped 10 of 13 variants (77%) from plasma; and A4 overlapped 11 of 16 variants (69%) from B lymphocytes (P < 0.001). The global antigenic distribution among compartments was not statistically significant in patient B, although B5 contained five of six (83%) variants from CD8⁺ cells (P < 0.001). These cell-specific variants were rarely found in plasma: 1 of 13 and 2 of 13 plasma variants from patient A resembled the CD8⁺ and CD19⁺ variants, respectively, and none of the plasma variants from patient B was homologous to those found in CD8⁺ cells. Although specific antigenicity profiles were detected in groups A2 and B5 (mainly composed of CD8⁺ strains), no CD8⁺ cell-specific signature was detected: A2 and B5 antigenicity profiles differed mainly at the N terminus of HVR1 (Fig. 3).

FIG. 2.

Peptide alignment, patient A (A) and patient B (B). The lefthand column indicates the origin and number of identical clones: T, CD8⁺ cells; B, CD19⁺ cells; X, negative fraction cells; S, plasma. A consensus sequence was generated for each patient. Antigenic grouping was done manually and blindly with regard to the origin of the clones.

TABLE 3.

HVR1 quasispecies distribution according to the site of detection and antigenic grouping

Antigenic groupa(n)	Quasispecies distributionb (n [patient A], n [patient B]d)
	Plasma (13, 19)	CD8+ cells (16, 6)	CD19+ cells (16, 17)	NF cells (3, 11)
Patient Ac
A1 (3)	0	0	0	3*
A2 (19)	1	15*	3	0
A3 (13)	10*	1	2	0
A4 (13)	2	0	11*	0
Patient B
B1 (4)	3	0	1	0
B2 (16)	4	0	7	5
B3 (11)	6	1	3	1
B4 (15)	6	0	6	3
B5 (7)	0	5*	0	2

^aAntigenic grouping was done blindly with regard to the origin of the clones. 

^b∗, Antigenic groups statistically corresponding to a single compartment (P < 0.001). 

^cThe global antigenic distribution was compartment specific in patient A (P = 0.04). 

^dn, number of clones. 

FIG. 3.

Antigenic grouping of HVR1 variants from patient A (A) and patient B (B). Antigenic profiles were obtained by using the Parker’s algorithm. HVR1 amino acid sequences composing the signature of antigenic groups are aligned with the consensus sequence.

(v) Amino acid composition of HVR1.

McAllistair et al. (27) recently showed that amino acid replacements in HVR1 were highly constrained at some positions and confined to certain residues with similar characteristics at the majority of variable positions. We observed the same conserved positions in our two patients (positions 2, 6, 20, 23, and 26) in both plasma and cellular variants, a finding suggesting that cellular variants had the same structural constraints as plasma variants in the HVR1. We then matched HVR1 antigenic types against GenBank types by using the BLAST program. For each antigenic group, we entered the consensus sequence with the amino acid residues composing the type signature (Fig. 3). More than 700 hits were obtained for seven of nine type sequences, including the CD8⁺-specific groups A2 and B5. This indicates that the cellular HVR1 quasispecies did not differ substantially from the plasma variants in terms of their global sequence and structure. Two antigenic groups (A4 and B4) did not match any GenBank HCV sequence. They were composed mainly of variants from CD19⁺ cells and also of variants from plasma (15 and 30%) and NF cells.

DISCUSSION

This study demonstrates that HCV quasispecies are not randomly distributed among the different cell subsets composing PBMC in chronic HCV carriers but that they display a statistically significant compartmental structure.

Our results are consistent with the detection of HCV sequences, regardless of the infecting genotype and viral load, in B (CD19⁺) lymphocytes and phagocytic cells (monocytes/macrophages and granulocytes) and the occasional presence of HCV in T (CD3⁺) lymphocytes (24, 56). Quasispecies harbored by CD8⁺ lymphocytes are likely to be a minor fraction of the total PBMC variants. Detection of such quasispecies from total PBMC or CD3⁺ lymphocytes (in which CD4⁺ cells are in the majority) could thus be impaired. HCV detection was based on an equal number of cells in the different compartments, thereby increasing the probability of discovering such minor strains in minor cell subsets. The observation of two individual cases, both long-term infected by the HCV genotype 1b, may limit the strength of our conclusions on HCV quasispecies compartmentalization. Indeed, genotype 1 is associated with higher detection rates of HCV RNA in PBMC compared to other genotypes (24). Our data confirm by phylogenetic and statistical sequence analysis the previously reported compartmentalization between serum and total PBMC (23, 35, 37). However, further studies are required to assess HCV compartmentalization within lymphoid compartment in patients with different stages of disease and different genotypes. Given the high variability of HCV and the limitations of techniques for quasispecies analysis, the probability that a particular variant would only be detected in a given site is not nil. To overcome such difficulties, we applied Mantel’s test, an approach previously used to demonstrate HIV anatomic compartmentalization (43). The phylogenetic grouping was confirmed by a statistically significant compartmental structure: variants from CD8⁺ or NF cells were much closer to one another than to those from any other compartment, including plasma. This genetic grouping for NF sequences in patient B was established despite the absence of an obvious phylogenetic structure. Indeed, NF cells are composed of several cellular subsets, and each of them could harbor specific quasispecies clustering at different levels of the tree. Such compartmentalization was also evidenced in CD19⁺ cells from patient A by both phylogenetic and statistical approaches.

The clear compartmentalization of HCV variants shows that HCV PCR positivity on PBMC was not due to contamination by plasma-derived HCV RNA or to virus carried on the membrane in a nonspecific manner. Despite the very specific antigenicity patterns of CD8⁺ and, to a lesser extent, CD19⁺ quasispecies, cellular and plasma variants seemed to have the same structural constraints (27). Moreover, no stop mutations were observed, and most quasispecies found in all the compartments matched GenBank sequences. This suggests that the cellular quasispecies are not defective. The important issue of so-called lymphotropic variants was first raised by Shimizu et al. (47): the same inoculum produced identical variants in cultured T-cell lines and PBMC of an infected chimpanzee. The widespread cellular distribution of CD81, the putative HCV receptor, is consistent with HCV binding to cells other than hepatocytes (41). We failed to reliably detect negative-stranded HCV RNA in the cell subsets tested (data not shown), the specific methods (21) being poorly sensitive (37). Because HCV replication in the lymphoid compartment remains unproven, we must interpret our results according to both possibilities. If HCV productively infects certain peripheral blood cell subsets, the observed divergence would not only be related to the binding of these quasispecies to specific molecules but also to replication, even at a low level.

In both patients, the rate of HCV replication (if any) in CD8⁺ cells may be low given the low genetic diversity of CD8⁺ variants and the very low frequency of antigenically homologous variants in plasma. The quantification issue was not addressed in our study. From other studies (24, 56) it appears that the detection of HCV RNA is largely limited to B lymphocytes and monocytes/macrophages. The very low heterogeneity of sequences in CD8⁺ cells might be related to a very weak viral load. Indeed, quantitative differences among compartments may result in quasispecies complexity differences. However, quantitative factors do not explain the compartment specificity of sequences. In patient A, CD19⁺ and NF cells also harbored a major antigenic group that was scarce in plasma. If extrahepatic replication does not occur, it is conceivable that so-called lymphotropic variants are ancient serum quasispecies bound to long-lived plasma cells. However, as this process would be continuous, CD8⁺ or CD19⁺ cells should also harbor current plasma variants and should not therefore show such a significant compartmental structure. The limited number of analyzed sequences (3 to 19 per compartment) may raise the concern of sampling bias. The cloning approach yields only a partial view of the viral population. In this view, these cellular variants might represent minor plasma quasispecies sharing the ability to bind to CD8⁺, CD19⁺, or NF cells through different receptors in a cell-phenotype-specific manner. HVR1 sequences common to PBMC and liver but absent from serum can be detected (37). In the absence of analyzed liver sequences, we cannot exclude that some variants found in CD8⁺, CD19⁺, or NF cells may also be found in the liver. Indeed, some PBMC could be infected in the liver and then migrate to the bloodstream. In that case, different cellular subsets could have been infected by specific strains with particular lymphotropism.

The heterogeneity of HVR1 quasispecies has been linked to the presence of antibodies to HVR1 (6, 15, 16, 46), which are found in up to 80% of chronically infected patients (44). Circulating immune complexes are found in most chronically infected patients (5, 10, 14, 33). Moreover, the quasispecies composition differs between the immunoglobulin-free and the immunoglobulin-associated fractions (5, 14, 17, 20). Immune-complexed virions could thus bind to cells bearing immunoglobulin Fc receptors, such as phagocytes, monocytes, B cells, and activated T cells. However, HCV binding via Fc receptors cannot alone explain such a significant compartmental structure in CD8⁺, CD19⁺, or NF cells. Alternatively, free virions could bind to anti-HCV membrane antibodies on B cells (B-cell receptors). The evidence of positive selection pressure acting on quasispecies harbored by CD19⁺ cells in patient B supports antibody-mediated binding to these cells. Compartmentalization of HCV quasispecies in B lymphocytes could thus reflect the viral counterpart of the humoral immune response to HVR1. However, the striking similarity of the d_N/d_S ratios in each compartment of a given patient and the conservation of invariant sites of HVR1 in all the compartments suggest that common positive and negative selective pressures are exerted on all quasispecies, whatever their cellular origin.

Experimental and clinical studies support the hypothesis that neutralizing antibody response to HVR may be important for virus clearance (7, 57). At the onset of the disease, the quasispecies nature of HCV may lead to a lack or delayed production of neutralizing antibodies to minor variants which become predominant during chronic infection. In addition, though it is generally assumed that HCV elicits a strong humoral immune response, quantitative and qualitative defects of the antibody response may contribute to persistence (3). As we mentioned in the introduction, the infection of immune cells is a well-known mechanism of persistence, especially for noncytopathic viruses. For lymphocytic choriomeningitis virus (40) and hepatitis B virus (1), the lack or delayed production of neutralizing antibodies has been linked to the destruction of virus-specific B cells by virus-specific cytotoxic T cells. The noncytopathic infection of B cells by measles virus aborts humoral immunity (28). Our findings support the hypothesis of a specific interplay between HCV and B lymphocytes. Indeed, whatever the HCV replication status in PBMC, viral proteins are likely processed and presented by class I or class II major histocompatibility complex molecules. Furthermore, the HCV core protein has been shown to interact with the lymphotoxin-β receptor known to play a role in antigen display to B cells. Recently, CD81, which associates with CD21 and CD19 on B cells, has been characterized as a putative receptor for HCV. This suggests for HCV, in addition to the hypervariability advantages, an active strategy to achieve persistent infection.

In conclusion, the compartmental distribution of HCV quasispecies in different PBMC phenotypes is consistent with the existence of distinct viral receptors and, possibly, with the immune presentation of particular viral epitopes, leading to interference with the immune system. Further clinical and experimental studies on nonhepatocytic HCV tropism must take into account the cell phenotype and include the monocyte/macrophage and dendritic cell fractions.