Although hepatitis C virus (HCV) induces a strong humoral response within a few weeks after infection, only 25-30% of the infected individuals resolve the disease whereas the rest develop chronic hepatitis C. Among the latter group, 20% progress to cirrhosis, whereof 5% further evolve towards hepatocellularcarcinoma (HCC) (Zein 2003 ). The parameters determining the outcome of the infection remain unknown. Indeed, cellular and humoral immune responses work together to protect individual against infections, but are also suspected to induce liver damage associated with HCV infection (reviews (Gremion and Cerny 2005 ; Irshad et al. 2008 )). As antibodies (Ab) play a key role in immune response, they are believed to be involved in the outcome of HCV disease. It is well known that HCV elicits antibodies against both structural and non-structural proteins (Gabrielli et al. 1996 ; Gutierrez et al. 2011 ; Sillanpaa et al. 2009 ) and several studies have investigated a possible relationship between immuno-reactivity and clinical features. While a correlation between viremia and Ab titres has been established (Bassett et al. 1998 ; Lohr et al. 1996 ; Yuki et al. 1994 ), studies attempting to discriminate between acute and chronic phases or between responders and non-responders to treatment led to more controversial results (Araujo et al. 2011 ; Hilfenhaus et al. 1992 ; Lohr et al. 1996 ). However, most of these studies have focused on the occurrence of Ab directed against different viral proteins without regards to their quantification. Moreover, they are often limited by the use of recombinant large proteins or polypeptides as probes to analyse anti-HCV Ab, which allows for a global, rather than precise determination of their specificity. A more accurate analysis using smaller peptides as probes could increase the number of identified epitopes targeted by anti-HCV Ab while also indicating their specificity. Although allowing for the detection of linear epitopes only, this approach can be useful in the case of HCV infection since human Ab are mainly directed against linear epitopes of the virus (Sallberg et al. 1992 ). It also benefits from the recent development of microarray-based screening systems. This technology has found its first applications in gene expression screening, and DNA/RNA microarrays have been used in clinical research for predicting the outcome of the disease (van’t Veer et al. 2003 ; Zhao et al. 2006 ), response to treatment (Augustine et al. 2009 ) or for determining cancer class (Golub et al. 1999 ). More difficult to implement, protein microarrays have emerged later and still remain rarely used: the complexity and heterogeneity of the probes together with the necessity to preserve their conformational folding impair their benefit. Peptide microarrays are therefore a good alternative to protein arrays and are suitable for large screening. They have demonstrated their reliability in detecting specific Ab content in serums, allowing one to test both a wide range of potential epitopes and a large number of samples (Cherif et al. 2006 ; Maksimov et al. 2012 ; Perez-Gordo et al. 2012 ; Villiers et al. 2010 ; Villiers et al. 2011 ).

Using such a technology with a set of probes comprising 56 peptides spanning the entire HCV proteome, our aim was to analyse the HCV-elicited immune response in 45 HCV-positive donors. Both the occurrence and magnitude of each Ab response were determined using surface plasmon resonance imaging (SPRi) to detect Ab-peptide interactions. We investigated a possible relationship between immuno-reactivity and various clinical parameters such as state of the disease (chronic, cirrhotic, HCC), viremia, ALAT (alanine amino-transferase) level, and Metavir scores. A good discrimination between HCV-positive and negative donors validated our approach, and several correlations between Ab profiles and clinical parameters have been identified. In particular, we demonstrated that ratios between particular Ab levels allow for accurate separation of the three states of the disease.

The aim of this study was to analyse the specific humoral response developed upon HCV infection, taking advantage of the potentialities of microarray-based screening systems. Previous studies dealing with anti-HCV Ab were informative but restricted to a global determination of their specificity because of the size of the probes that did not allow for precise identification of the epitopes. Moreover, the clinical parameters associated with the pathology that were taken into account were often restricted to the viral load, the response to the treatment or to two states of the disease, usually acute and chronic (Araujo et al. 2011 ; Desombere et al. 2007 ; Sillanpaa et al. 2009 ). The use of peptide microarrays together with SPRi detection allows for more accurate identification of epitopes together with Ab quantification and is suitable for a large screening of both probes and serums (Cherif et al. 2006 ). Three main reasons guided us to use 20aa peptides as probes to determine the specificity of anti-HCV Ab. First, peptide microarrays are more stable than protein arrays and are thus more suitable for large screening (Villiers et al. 2011 ). Second, it is established that human Ab elicited by HCV infection are mainly directed against linear epitopes (Sallberg et al. 1992 ). Third, it is usually considered that linear B cell epitopes typically vary from 6 to 20aa in length with a mean value of 13aa (Sollner 2006 ).

The great immunogenicity of the Core protein (Figure 1), as revealed by using these peptides as probes, is in agreement with previous works using larger sequences (Araujo et al. 2011 ; Chen et al. 1999 ). This suggests that a large proportion of the Core epitopes are actually linear. This does not seem to be the case for the other viral proteins, especially envelope and NS2-NS3 proteins, since poor antibody responses against linear epitopes from these proteins are found in our study (Figure 1). Although we cannot exclude that this observation may partly result from the fact that the selected peptides do not cover the entire sequence of the protein (as opposed to Core-derived peptides), the low amount of Ab directed against these regions was also observed in other analyses using peptides as probes (Sallberg et al. 1993 ). In contrast, authors using polypeptides or recombinant proteins detected anti-E1, anti-E2 (Chen et al. 1999 ; Zibert et al. 1999 ), and anti-NS3 (Hwang et al. 1996 ; Yuki et al. 1994 ) in more than 50% of the tested serums. Moreover, it is known that E1-E2 provide several conformational epitopes for neutralizing Ab (Giang et al. 2012 ; Law et al. 2008 ). In addition, and despite the fact that peptide sequences were chosen in regions exhibiting relatively conserved sequences across genotypes, the presence of three hypervariable regions in E2 makes it difficult to identify common epitopes. In the Core protein, the immunodominant region lies within the first 90aa, which are probably more exposed than the others as suggested by the hydrophobicity index (Gravy) of the sequence (-1.442 (aa 1–89) versus 0.313 (aa 90–189)). This is due to the presence of two regions with a high density of basic residues (leading to hydrophilic regions) in the N-ter portion of the protein (Klein et al. 2005 ).

We first examined whether there was any correlation between the peptides recognized by serum Ab. Statistical studies by either PLS-DA or hierarchical clustering using Pearson’s correlation determine four groups of peptides (Figure 2). As expected, peptides that possess a 10-aa overlapping sequence tend to stand in the same group. However, the Core protein presents two independent regions with regards to its antigenicity, as previously observed (Nasoff et al. 1991 ), which distinguish the first 80 aa (N-ter) from the rest of the molecule. This suggests a difference in the structure and/or the behaviour of the N-ter compared to the C-ter regions of the Core protein, which is supported by the difference in the hydrophobicity index between them. Analysis of aa sequences of different HCV genotypes shows three domains in the Core protein, one of them (N-terminal region) spanning residues 1–118 that roughly corresponds to the most immunogenic portion of the protein (Hope and McLauchlan 2000 ). The observed correlation between the amounts of Ab against E1 and the C-ter part of the Core can be related to the correlation between circulating core and E1 levels (El Awady et al. 2006 ).

Then we looked for a relationship between Ab specificity or amount and disease state. The very good separation between HCV- and HCV + patients obtained using hierarchical clustering (Figure 3) and PLS-DA (95% correct classification, Table 1) validates the methodology used in our study, especially the normalization of the signals. However, this approach is not sufficient for discrimination between each state of the disease, which would provide a good prognosis tool. Despite different Ab profiles in the serums from HCC versus chronic + cirrhotic patients, Ab profiles fail in affording correct classification of all patient samples, even after merging different disease states.

However, the lack of a predictive model does not mean a lack of significant differences between Ab profiles with regard to the disease states. The Mann and Whitney test reveals that only 4 peptides out of the 56 tested are recognized to different extents by Ab from at least two categories of serums (Figure 5): serums from cirrhotic patients display a significant difference in Ab amount against C15 and NS3-6 compared to chronic patients and against NS4-1 compared to HCC patients. However only one peptide, NS3-3, is recognized differently by the Ab from the three disease states. Three of these peptides belong to group 2, whereas none of them belong to groups 1 and 4. It is noticeable that all these 4 peptides lead to a low average SPR signal, resulting in a low (C15 and NS4-1) or null (NS3-3 and NS3-6) percent positive patients when focusing on easily detectable antibodies (SPR signal threshold > 0.5, Figure 1). It is possible that the high immunogenicity of the peptides of group 1 dwarfs differences between disease states. On the other hand, too weak immunogenicity, as for peptides from the envelope, can lead to the same result. Overall, our data demonstrate a very weak specificity of the Ab response with regard to the disease state, which is unlikely to be due to the linear nature of the epitopes tested. Indeed, a study using chimpanzees and large polypeptides derived from Core and NS3/NS4 concludes that Ab patterns are similar in persistently infected animals and those that cleared viraemia (Hilfenhaus et al. 1992 ); also, acute and chronic infections are not discriminated in a study based on the measurement of Ab reactivity of human serums against polypeptides from Core and non-structural proteins (Araujo et al. 2011 ).

We wondered whether other clinical parameters could account for the establishment of particular Ab profiles. Correlations with several parameters are highlighted by our study (Table 1). First, Ab patterns are related to Metavir A when scores 1 + 2 are merged, providing a predictive model allowing for more than 75% correct classification. The interdependence between Ab profiles and some clinical parameters is even stronger when considering ALAT values (correct classification > 90% - threshold = 2 N) and viral load (correct classification 100% - threshold = 600 000 UI). Few studies deal with the correlation between ALAT and anti-HCV Ab; Lohr et al. (Lohr et al. 1996 ) observed that patients whose lymphocytes secreted anti-HCV in vitro had higher ALAT levels in serum, and Brillanti et al. (Brillanti et al. 1992 ) observed a positive correlation between the ALAT and anti-HCV IgM levels. However, our result, in agreement with Nikolaeva et al. (Nikolaeva et al. 2002 ), do not reveal any correlation between ALAT levels and anti-HCV titres. Further studies aimed at correlating anti-HCV Ab and viral load however they usually show a correlation between total anti-HCV Ab titres and viremic levels. This discrepancy between these studies and ours may be due to the probes used for anti-HCV quantification consisting of large recombinant proteins (Bassett et al. 1998 ; Gane et al. 1999 ; Yuki et al. 1996 ) rather than peptides.

Thus, we demonstrate that although HCV infection elicits different Ab profiles in patients, only a very few linear epitopes are significantly different in the response they trigger by serums from chronic, cirrhotic, and HCC states. All these results led us to hypothesize that the different stages of the disease may be less correlated with the absolute value of the immune response against each viral protein than with the relative value of Ab recognition of different epitopes. This prompted us to analyse the ratio of SPR signals grouped in pairs of peptides (Table 2). This original approach improves the sensitivity of the analysis by reducing the impact of differences in the amplitude of humoral immune responses observed between patients. First, it reveals that Ab ratio for one peptide pair (C11/NS3-3) is an accurate marker of each state of the disease. Anti-NS3-3 alone was already identified in this study to separate the chronic state from a group consisting of cirrhotic and HCC states (Figure 5). Furthermore, association with anti-C11 allows for distinguishing cirrhotic from HCC states (Table 2). Second, a large number of peptide pairs discriminate chronic and HCC (15 and 14 pairs, respectively) from the two other states. Interestingly, we observed that the majority of the peptide pairs allowing for separation of the chronic state are intergroup (66.7%) In contrast, only 28.6% of the ratios used to discriminate between HCC and other states involve peptides belonging to different groups. In the first case, peptides from the Core protein are often implicated (66.6%) while NS4-1 is the most represented peptide in the second case (85.7%). Cirrhosis is the most difficult state to identify. This is not surprising since it is an intermediate state between chronicity and HCC.

Although direct profiling of Ab specificity does not provide an accurate prognostic tool, our study demonstrates that humoral response against HCV linear epitopes is modified according to disease state. This relationship does not result in a significant change in the overall immune response against individual epitopes. However, relative amplitudes of Ab responses against selected pairs of epitopes appear to be relevant in discriminating disease states. Overall, SPRi system for high-throughput screening has allowed for a thorough analysis of Ab specificity against multiple epitopes directly from patients serums. Our data reveal a set of peptides derived from HCV related to the state of the disease. This method opens potential further studies aiming to decipher the host- and pathogen-dependent molecular processes underlying the infections and the evolution of the resulting diseases.

Ab: Antibody; ALAT: Alanine amino-transferase; HCC: Hepatocellularcarcinoma; HCC-OH: Alcoholic cirrhosis; HCV: Hepatitis C virus; PLS-DA: Partial least square determinant analysis; SPR: Surface plasmon resonance; SPRi: Surface plasmon resonance imaging.

The authors declare that they have no competing interests.

CB: has performed SPRi experiments and participates to data analysis. PA: has performed statistical analysis. FV: has performed statistical analysis. AM: has provided sera and clinical data from patients. AB: has performed statistical analysis. TL: SPR adviser. RC: SPR adviser. JPZ: clinical adviser. CLV: program designer for analysis of SPRi data. PNM: participation in experiment design. MBV: has designed experiments and has performed analyses SPRi data. All authors read and approved the final manuscript.