Study serum specimens.
This study was performed using unlinked anonymous serum or plasma specimens and specimens obtained commercially from blood donor anti-HCV seroconversion panels. Ninety-nine plasma samples were obtained from 24 donor panels; the number of samples per panel varied between 1 and 11. Ten panels were acquired from Zeptometrix (Buffalo, NY) (batch numbers 6211, 6212, 6213, 6214, 6215, 9041, 9047, 9054, 9055, and 9058), 5 from NABI (Boca Raton, FL) (batch numbers 10, 20, 30, 40, and 60), 4 from Serologicals (Clarkston, GA) (batch numbers 4812B, 4813, 4814, and 4814B), 3 panels (batch numbers 908, 920, and 921) from BBI (West Bridgewater, MA), and 2 from Profile Diagnostic (Sherman Oaks, CA) (batch numbers RP006 and RP040). The samples were taken within 62 days after the last anti-HCV-IgG-negative result: 23 samples between 1 and 10 days, 25 samples between 11 and 20 days, 17 samples between 21 and 30 days, 17 samples between 31 and 40 days, 12 samples between 41 and 50 days, 3 samples between 51 and 60 days, and 2 samples between 61 and 62 days. These samples are here called the “acute” group. Of the 24 batches, the HCV genotype could be determined for 11; batches 6212, 6213, 6214, 6215, 9041, 9047, 9058, and 920 belonged to genotype 1, and batches 9054, 9055, and 921 belonged to genotype 3. Genotyping could not be determined for the remaining batches due to insufficient samples and/or low HCV RNA titers.
The chronic hepatitis C patients (the “chronic” group) consisted of anti-HCV-IgG-positive plasma samples from 141 blood donors: 64 samples were from BBI, and 77 were from the American Red Cross (9
). All samples were confirmed to contain anti-HCV IgG by the Ortho recombinant immunoblot assay (RIBA) and HCV RNA by reverse transcriptase PCR (9
). The HCV genotype was determined for 96 samples; 73 samples (76%) belonged to genotype 1, 16 samples (17%) to genotype 2, and 7 samples (7%) to genotype 3. In addition, a control group of 30 human serum samples negative for all markers of infection for hepatitis A, B, and C viruses (BBI, West Bridgewater, MA) was included in the study.
Recombinant HCV antigens.
Eight recombinant HCV proteins purchased from RPC Diagnostic Systems Inc., Nizhniy Novgorod, Russia, were used as antigens. Five of the proteins were derived from the NS3 region: NS3#201 (amino acids [aa] 1192 to 1459) from a genotype 1b strain and NS3#207, NS3#208, NS3#210, and NS3#215 (aa 1356 to 1459) from genotype 1a, 1b, 2c, and 1c strains, respectively. The other 3 proteins originated within the HCV core region (aa positions 1 to 100) of a genotype 1b HCV strain: a mosaic protein containing immunodominant regions of the NS4 protein (aa 1691 to 1710, 1712 to 1733, and 1921 to 1940) derived from HCV genotype 1, 2, 3, and 5 strains (1
) and an NS5 protein (aa 2061 to 2302).
All purified recombinant HCV proteins were coupled to individual color-coded carboxylated xMAP microspheres (Luminex Corp, Austin, TX) according to the manufacturer's protocol (7
). Briefly, 12.5 × 106
microspheres (assigned 8 individual identification [ID] numbers) were activated for 20 min in 100 mM monobasic sodium phosphate, pH 6.2, with 50 mg/ml N
-hydroxysulfosuccinimide and 50 mg/ml 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (Pierce, Rockford, IL). The recombinant proteins (200 μg each) were added to the resuspended microspheres, bringing the total volume to 500 μl with phosphate-buffered saline (PBS), pH 7.4, and vortexed. After incubation for 2 h by rotation at room temperature, the coupled microspheres were washed twice with PBS-TBN (PBS, pH 7.4, 0.1% [wt/vol] bovine serum albumin [BSA], and 0.05% sodium azide) and resuspended in 1 ml of PBS-TBN, after which the yield was determined by counting the microsphere suspensions using a hemacytometer. The microspheres were protected from light and stored in PBS-TBN at 4°C until use.
Multiplex HCV antibody assay.
The anti-HCV-IgG multiplex assay was performed by mixing each one of the different antigen-coupled microspheres to a working concentration of 2,000 beads per ID number per well, as described above in the coupling methods. Serum or plasma samples were diluted 1:100 in PBN (PBS, pH 7.4, 0.5% [wt/vol] BSA, and 0.02% sodium azide) for measurement of antigen-specific IgG binding activity. Briefly, 25 μl per diluted sample was incubated with the microsphere mixture in a 96-well filter microtiter plate (Millipore, Billerica, MA) in the dark at room temperature by shaking for 1 h. The plate was washed twice with wash buffer (0.05% Tween 20 in PBS). The microspheres were resuspended in 25 μl of PBN and 25 μl of a 1:200 dilution of 1.1 mg/ml of Biotin-SP-conjugated AffiniPure goat anti-human IgG (Jackson Immuno Research, West Grove, PA). After incubation in the dark at room temperature by shaking for 30 min, 25 μl of a 1:5 dilution of 1 mg/ml of streptavidin-R-phycoerythrin (SA-PE) (Molecular Probes, Eugene, OR) was added. The plate was incubated for 15 min in the dark at room temperature on a plate shaker and washed. The microspheres were then resuspended in 150 μl of PBN. Measurements were obtained using the Luminex 100 instrument (Qiagen, Valencia, CA) with the MasterPlex CT v 1.0 software program (MiraiBio, San Francisco, CA) installed. All tests were performed in duplicate, and the mean fluorescence intensity (MFI) values were averaged. The coefficient of variation (CV) was calculated from MFI values obtained by testing eight replicates of five sera in the same run for intra-assay variation or in 2 separate assays for interassay variation, and the values were averaged for each antigen. The intra-assay CVs varied between 2% and 12%, and the interassay CVs varied between 2% and 22%. The positive cutoff value was determined as the MFI value for 30 normal human serum samples plus 2 standard deviations; the signal-to-cutoff (S/CO) ratio was determined as the MFI divided by the cutoff.
Since multiple samples were available from a majority of donors in the acute group, the mean S/CO values were calculated per antigen, per donor. Comparison of IgG anti-HCV responses to each of the 8 HCV antigens between acute and chronic groups was performed after the normality assumption was verified. Since all antigens, except the core antigen, fit a log-normal distribution, the geometric mean or mean (normal distribution) and variances were calculated for each group. The F test was performed to test the equality of the two group variances. If the F test generated a significant outcome, the t test assuming unequal variances was used. Otherwise, the t test assuming equal variances was used. A P value of <0.001 was considered to be statistically significant. Analyses were performed using the SAS software program (SAS Institute, Cary, NC).
To facilitate the classification of samples as belonging to acute or chronic HCV infection, a multivariate logistic regression model was built in the MATLAB software program (6
). A multivariate model using the reactivities to all 8 antigens as covariates had a deviance of 24.95. The P
value of the reactivity to the NS5 antigen was 0.65. Consequently, this covariate was removed from the model, yielding a deviance of 25.39. The P
value of the likelihood ratio test for the two models was Pr[χ2
(1) > 0.44] > 0.1, confirming that the NS5 antigen is not significant for the model. Similarly, the covariate corresponding to the response to the antigen NS3#207 was removed, yielding a deviance of 25.44. The coefficients of the resulting model are given in Table .
Coefficients of multivariate logistic regression model
To study the predictive ability of the model, a leave-one-out cross-validation was performed. A panel from the acute group and one sample from the chronic group were removed in turn, following which a multivariate logistic regression model using the six covariates was built from the remaining data. The model was then used to predict the membership of the panel or sample set aside in the acute or chronic group. In the case of a panel with more than one sample, the accuracy was calculated as the percentage of correctly classified samples out of all samples in the panel. The calculated accuracies were averaged by group.