Antibody Repertoire Analysis of Hepatitis C Virus Infections Identifies Immune Signatures Associated With Spontaneous Clearance

Abstract

Hepatitis C virus (HCV) is a major public health concern, with over 70 million people infected worldwide, who are at risk for developing life-threatening liver disease. No vaccine is available, and immunity against the virus is not well-understood. Following the acute stage, HCV usually causes chronic infections. However, ~30% of infected individuals spontaneously clear the virus. Therefore, using HCV as a model for comparing immune responses between spontaneous clearer (SC) and chronically infected (CI) individuals may empower the identification of mechanisms governing viral infection outcomes. Here, we provide the first in-depth analysis of adaptive immune receptor repertoires in individuals with current or past HCV infection. We demonstrate that SC individuals, in contrast to CI patients, develop clusters of antibodies with distinct properties. These antibodies' characteristics were used in a machine learning framework to accurately predict infection outcome. Using combinatorial antibody phage display library technology, we identified HCV-specific antibody sequences. By integrating these data with the repertoire analysis, we constructed two antibodies characterized by high neutralization breadth, which are associated with clearance. This study provides insight into the nature of effective immune response against HCV and demonstrates an innovative approach for constructing antibodies correlating with successful infection clearance. It may have clinical implications for prognosis of the future status of infection, and the design of effective immunotherapies and a vaccine for HCV.

Keywords: antibody repertoire; hepatitis C virus; immune signature; infectious disease; neutralizing antibodies.

Figure 1

Scheme of workflow. The workflow included the following steps: collection of blood samples from SC, CI, and healthy individuals, sequencing of total B-cell repertoires, T-cell repertoires, and HCV-specific B-cell repertoires, analysis of repertoires and identification of antibody clusters and TCR sequences associated with viral clearance, construction of an antibody phage display library, isolation of a panel of HCV-binding antibody sequences that associate with cleared infections, and integration of all data to construct HCV-broadly neutralizing antibodies associated with clearance.

Figure 2

Characterization of B-cell repertoires in SC, CI, and healthy individuals. (A) The number of unique sequences per sample after pre-processing. (B) The CDR3 length distribution. (C) The IGHV gene distribution. Only functional V genes that were in the 15 topmost frequent in at least one sample are shown. (D) The IGHJ gene distribution. (E) Feature combinations whose abundance differ between the SC and CI groups are presented for sequence clusters grouped by identical IGHV and IGHJ and by high CDR3 similarity, which were significantly more abundant in either SC or CI cohort (|#samples_SC-#samples_CI| >3 samples). Sequence logos CDR3 of these clusters are presented in Supplementary Figure 4.

Figure 3

Machine learning model used to stratify between SC and CI. (A) Accuracy was based on the B cells' repertoire. Original labels represent clustered sequences by identical IGHV and IGHJ and the high similarity of the CDR3 amino acid sequence. For validation purposes, the model was trained and applied on randomly labeled data. (B) Prediction model based on the T cells' repertoire. The training for the T-cell repertoires model is very similar to the B-cell model, except that the data were clustered solely by CDR3 amino acid identity. (C,D) The top 10 clusters used by the model to stratify between the cohorts. (C) In B-cell clusters. (D) In T-cell clusters. Sequence logos of the CDR3 of the B cell clusters are presented in Supplementary Figure 6.

Figure 4

Isolation of HCV-specific B cells from resolved and chronic HCV infection. (A) HCV-specific B-cells isolated from six CI and three SC individuals, as compared with control healthy individuals. The fold enrichment of HCV-specific B cells from each sample was calculated compared with the number of B cells isolated from a healthy individual, as demonstrated in Supplementary Figure 7. (B) HCVcc-neutralization assays using supernatants of cultured B cells from healthy, SC, and CI samples after two 2 weeks of activation in vitro. (*P < 0.03, **P < 0.003, ***P < 0.0001, ****P < 0.00003, t-test). (C) Dendogram of CDR3s from HCV-specific B cells, generated based on Levenshtein distances. Each color of the CDR3 sequence corresponds to an individual. (D) Mutation numbers in IGHV genes in the general repertoire compared with the HCV-specific repertoire. Each specific sequence was randomly matched to a non-specific sequence with the same IGHV and IGHJ genes. The sequences were grouped by isotype and mutations were compared by Mann Whitney test (IGA p = 3.488873e-07, IGG p = 6.849511e-08, IGM p = 3.764229e-04). (E) Mutation number in the IGHV genes in the specific repertoire for SC and CI (IGA p = 0.000574, IGG p = 0.435930). (F) Conserved amino acids in CDR3 from the HCV-specific repertoire (binders) compared with the general repertoire (non-binders). For each specific sequence, a non-specific sequence was randomly matched. Sequences were then grouped by IGHV, IGHJ, and CDR3 length. Cases where CDR3 amino acids were very conserved for binder sequences but not for non-binders are shown.

Figure 5

Identification of HCV-specific antibody sequences associated with HCV infection clearance. (A) Binding of the phage-displayed antibodies to the rE2 protein (5 μg/ml) by ELISA. Each bar indicates the mean fold change ± SD in the OD compared with BSA binding, from three independent experiments. (B) Violin plot of the distances between HCV-specific sequences and the healthy, CI and SC repertoires. (C,D) Phylogenetic trees of the two closest clusters to scFv SC11 (C), and SC28 (D).

Figure 6

Construction and characterization of antibodies correlated with infection clearance. (A) Binding of antibodies RMS28 and RMS11 to the rE2 protein (5 μg/ml) compared with the phage display antibodies SC28 and SC11 by ELISA, using 16 μg/ml Ab. Each bar indicates the mean fold change ±SD in binding, compared with BSA, from three independent experiments. (B) Binding of antibodies RMS11 and RMS28 to the rE2 protein (5 μg/ml), compared with a well-defined panel of nAbs and a non-specific control antibody RO4 by ELISA, using 16 μg/ml Ab. Presented are mean OD (450 nm) values ±SD, from three independent experiments. (C,D) HCVcc neutralization assays were carried out with genotypes G1-G7 using 20 μg/ml of antibodies RMS11 (C) and RMS28 (D). The percent neutralization was calculated as the percent reduction in FFU compared with virus incubated with an irrelevant control antibody (RO4). Presented are means of % neutralization ±SD from three independent experiments.