The detection of antibody to hepatitis C virus (HCV) is an important assay for the identification of individuals infected with this virus. However, the confirmation of antibody positivity remains problematic. Currently none of the screening or confirmatory assays provide quantitation of the antibodies present. A microsphere assay was designed to provide improved confirmation.
Microspheres of 3.6 μm in diameter coated with NeutraAvidin were used to capture biotinylated HCV recombinant proteins. A phycoerythrin goat anti-human immunoglobulin G (IgG) was used to detect specific antibody captured to the microsphere. A human IgG calibrator was designed that was internal to each sample in the microsphere assay.
Detection of HCV-specific antibody using these microspheres was straightforward in most samples, with the lower detection limit set at 0.01 μg equivalents of human IgG per milliliter. In antibody-positive samples, the HCV antibody levels ranged from 0.09 to 55 μg equivalents of IgG per milliliter. Forty-nine of the 54 samples (91%) previously identified as having an indeterminate serologic pattern were negative in the microsphere assay.
It is currently estimated that there are 3 million individuals in the United States infected with the hepatitis C virus (HCV) (1). Testing for infection with HCV is performed by using an enzyme immunoassay (EIA) on serum or plasma and is a commonly performed test. Recent studies in managed care populations have indicated that 0.7% of enrolled members are tested for antibody to HCV and that 6.7% of those tested are reactive (2). Samples, which are reactive in the EIA, are sometimes reflexively tested using a supplemented assay. The supplemental test most often used is the recombinant immunoblot (RIBA). Samples can be reactive in the screening EIA but negative or indeterminate in the supplemental assay, and studies have suggested that approximately 8% to 17% of samples, which are reactive in the EIA, cannot be confirmed with the supplemental assay (3, 4). Studies have suggested that the semiquantitation of the HCV antibody present, as evaluated by the magnitude of the signal:cutoff ratio from the EIA, might help identify true positive samples (3–5). Although the RIBA performs acceptably, significant numbers of samples provide inconclusive staining patterns and the quantity of antibody is not available, which may be helpful in predicting the likelihood that the sample is a true antibody positive (3, 5). In an effort to provide semiquantitative antibody levels and to improve the detection of antibody specific to HCV, recombinant HCV proteins were used to develop a microsphere immunoassay for analysis by flow cytometry. The basic assay format has previously been shown to improve the detection and resolution of samples, which were reactive with the EIA but indeterminate with the RIBA (6).
MATERIALS AND METHODS
NeutraAvidin-coated microspheres of 3.6 μm in diameter were used (Cyto-Plex microspheres, Duke Scientific Corp., Palo Alto, CA, USA) and designated as L5, L7, and L10. These three separate populations of microspheres have distinct and increasing levels of a red dye used to distinguish the populations in the FL3 channel (650LP) of the flow cytometer. These microsphere populations were coated separately with the HCV recombinant proteins or human immunoglobulin G (IgG). The L5 microspheres were coated with HCV NS3, the L7 microspheres with different concentrations of biotinylated human IgG, and the L10 microspheres with HCV capsid.
Dyed microspheres, which provide five separate fluorescent populations (Cyto-Cal microspheres, Duke Scientific Corp.), were used daily to monitor the fluorescence signal in the FL2 and FL3 channels. The Cyto-Cal microspheres have a stable fluorescence signal and are dyed to fluoresce at multiple wavelengths. The same instrument settings, including forward scatter (FSC) and side scatter (SSC) gains and phycoerythrin (PE; FL2) and red dye (FL3) photomultiplier tube voltages, were used for each experiment. The applicability of using the same instrument settings was verified by demonstrating that the Cyto-Cal microsphere fluorescence was in the same FL2 and FL3 channel for each experiment.
The microspheres were analyzed on a FACScan flow cytometer (Becton Dickinson, San Jose, CA, USA). The same instrument settings for FSC, SSC, FL2, and FL3 were used for all experiments. The Cyto-Cal microspheres and control sera were used to monitor interassay variability. Data were analyzed by using CellQuest Pro software. The instrument was setup using CaliBRITE microspheres (Becton Dickinson) and fluorescence compensation for FL2 and FL3 was set on the flow cytometer. The FL3 signal was used to identify the red-dyed microspheres using a 650LP filter. The PE (FL2) signal was used to identify the presence of IgG using a 585/42 filter.
PE-conjugated goat F(ab′)2 anti-human IgG and donkey anti-rabbit IgG were used (Jackson ImmunoResearch, West Grove, PA, USA). Each antibody was titrated and used at the optimum dilution in the HCV microsphere assay. Each antibody was absorbed to reduce cross-reactivity with IgG from the opposing species.
Biotin-conjugated human IgG (Jackson ImmunoResearch) was used with the L7 microspheres as an internal assay calibrator. The human IgG microspheres were used to express the HCV antibody in microgram equivalents of human IgG per milliliter.
Production of Polyclonal Rabbit IgG to HCV
The recombinant capsid and NS3 proteins, before biotinylation, were used for the production of rabbit antibody (DX-Sys, Inc., Mountain View, CA, USA). Briefly, four rabbits were used, two injected with the capsid and two with the NS3 protein. At days 30 and 60, a 60-ml sample of blood was taken from each rabbit. The serum was collected and IgG purified by using a protein A Sepharose column. Briefly, diluted serum in phosphate buffered saline (PBS) was added to protein A Sepharose (Sigma Chemical Co., St. Louis, MO, USA) and proteins other than IgG were washed through the column using PBS. A 0.1 M glycine buffer, pH 2.7, was used to elute the rabbit IgG, which was dialyzed against PBS after collection. The concentration of rabbit IgG was determined, and the final rabbit IgG preparations were diluted for use in the microsphere assay.
Production and Biotinylation of HCV Proteins
Two gene products from the Hutchinson strain of HCV, the nucleocapsid (capsid) and the NS3, were expressed in Escherichia coli as recombinant proteins, as previously described (7–9). Briefly, the capsid was obtained as a full-length polypeptide (120 amino acids) and was expressed without any carrier protein. The capsid was stored in 4 M urea/0.2 M NaCl in 20 mM acetate buffer, pH 6, at −20°C until used.
The NS3 protein (105 amino acids) was expressed with a hexahistidine tag at its N-terminus. The hexahistidine tag allowed for more effective purification using an affinity column containing an iminodiacetic acid chelate saturated with Ni2+ ions (10). The purified protein produced a single band on analysis by sodium dodecylsulfate gel electrophoresis in 16% acrylamide gel and was stored at −20°C in 50 mM carbonate buffer, pH 8.6, and 0.2 M NaCl until used.
The HCV capsid protein was dialyzed against 25 mM sodium carbonate/0.1 M NaCl in 4 M urea at pH 8.6 for 12 h. An aliquot at 1.12 mg protein/ml was mixed with the LHC biotin in dimethylsulfoxide (2 μmol/ml, Sigma Chemical Co.) and incubated for 5 h at room temperature. This preparation was diluted to 1.2 ml in 25 mM sodium carbonate/0.1 M NaCl in 4 M urea, pH 8.6, and dialyzed against this buffer for 48 h with three intermittent buffer changes. The dialyzed material was used for coating the L10 microspheres as described below.
The hexahistidine-NS3 polypeptide was dialyzed against 25 mM sodium carbonate/0.1 M NaCl at pH 8.6 for 12 h. An aliquot at 1.85 mg protein/ml was mixed with the LHC biotin in dimethylsulfoxide (2 μmol/ml, Sigma Chemical Co.) and incubated for 5 h at room temperature. This preparation was diluted to 1.2 ml in 25 mM sodium carbonate/0.1 M NaCl, pH 8.6, and dialyzed against the same buffer for 48 h with three intermittent buffer changes. The dialyzed material was used for coating the L5 microspheres as described below.
Procedure for Coating Microspheres With Biotinylated Proteins
The microspheres were incubated with the biotinylated HCV proteins for 60 min at room temperature at the concentration of 51 μg of NS3 protein per 106 L5 microspheres and 30 μg of capsid protein per 106 L10 microspheres. After 60 min the microspheres were washed three times in PBS and then mixed together in equal volumes.
The L7 microspheres were incubated with the biotinylated human IgG at 10, 0.1, 0.01, and 0.001 μg/ml for 60 min at room temperature. After 60 min, the microspheres were washed three times in PBS, mixed in equal volumes, and added to the coated L5 and L10 microspheres.
The final microsphere preparation contained an L5 NS3 population, an L10 capsid population, and four separate L7 human IgG populations. This mixture was stored at 2°C to 8°C until used.
A 100-μl aliquot of each diluted serum sample was added to a tube. A 100-μl aliquot of the microsphere mixture was added and mixed. The tubes were incubated for 30 min at room temperature and washed three times in PBS, and the microsphere pellet was resuspended in a 100-μl aliquot of PE-goat F(ab′)2 anti-human IgG. The tubes were incubated and washed as before and the final microsphere pellet was resuspended in 200 μl of PBS for flow cytometric analysis. All samples were tested at a 1:100 dilution. Samples, which were over the 10-μg equivalent of IgG per milliliter of calibrator, were retested using 1:1,000 and 1:4,000 dilutions. HCV antibody levels were expressed in microgram equivalents of human IgG per milliliter.
Human Serum Samples
Excess samples, which were submitted for HCV serologic testing, were coded in a blind fashion and were used for this study. All samples were screened by using a HCV EIA (Abbott Laboratories, Inc., Abbott Park, IL, USA) that was approved by the U.S. Food and Drug Administration, and supplemental testing was performed by using the RIBA (Ortho-Clinical Diagnostics Inc., Raritan, NJ, USA). Three groups of samples were tested, those identified as antibody negative (n = 103) were negative by the screening EIA with a signal:cutoff ratio less than 1.0, those identified as antibody positive (n = 118) were positive by the screening EIA with a signal:cutoff ratio greater than 3.8, and those identified as indeterminate (n = 54) were positive in the EIA with a signal:cutoff ratio in the range of 1.0 to 3.8. All indeterminate samples were tested by the RIBA.
The three microsphere populations selected designated, L5, L7, and L10, were well resolved in the FL3 channel of the flow cytometer, as shown in Figure 1. This separation allowed for clear distinction of the different populations, thereby allowing for their use in a multiplex assay. When excited at 488 nm, the dye used has a broad emission above 650 nm with a peak at 700 nm. Although the fluorescence in FL3 was affected slightly by the PE (FL2) signal, compensation settings could be adjusted to control for this effect. These data were collected using compensation set on the flow cytometer during acquisition. The hardware compensation settings used throughout these experiments consisted of subtracting 5% of the PE (FL2) signal from the red dye signal (FL3), which is the dye used to distinguish L5, L7, and L10. With this compensation setting of FL3 minus 5% FL2, the mean FL3 log channel number from these microspheres did not vary by greater than 5% when the intensity of the PE (FL2) signal varied over 3 logs. The red dye present in FL3 did not have spectral overlap into the PE (FL2) channel, and as such compensation for the red dye signal into the PE channel was not required. The FL3 signal, as expressed as log channel numbers, from each population of microspheres was stable for the 8 months of the study as demonstrated by the mean with percentage of coefficient of variation (CV) of the L5, L7, and L10 microspheres: 525 with 3%, 1,208 with 6%, and 4,580 with 4%, respectively. These represent analysis of a single data point from each experiment from 37 different experiments.
A five-microsphere set (Cyto-Cal) was run each day to verify instrument performance in the PE (FL2) and red-dye (FL3) channels. The instrument settings remained constant over the 8 months of this study and the log FL2 and log FL3 channels from the Cyto-Cal microspheres were plotted. The percentage of CV of the FL2 and FL3 signals ranged from 3% to 9% using the same instrument settings over 8 months. The use of the Cyto-Cal microspheres allowed for an objective determination that the instrument was performing similarly between experiments. A sample of this fluorescence display from the PE (FL2) channel is shown in Figure 2. Any variation in the PE (FL2) signal when using the HCV-coated and -reacted microspheres was then assumed to be due to assay variability and not to instrument variability.
The L7 microspheres were used to construct the internal human IgG calibrators. The mean of each fluorescence peak was used to construct a calibration curve from which microgram equivalent of human IgG per milliliter for the serum samples was derived. Figure 3 shows the PE (FL2) display of the L7 internal calibrator human IgG microspheres. Four separate peaks are seen, with each representing a different concentration of human IgG coated onto the microspheres. The microgram equivalent of human IgG per milliliter from these microspheres was used as the reporting unit for antibody to the HCV proteins.
Various ratios of biotin:HCV protein from 2:1 through 1:20 were evaluated for their effect on the detection of antibody in the microsphere assay. The optimum biotin:HCV protein ratio was selected by using the various biotinylated HCV proteins in the microsphere assay and evaluating the PE (FL2) signal from an antibody positive serum sample. The biotin:HCV protein ratio selected was the ratio that produced the highest mean PE (FL2) signal with the lowest CV of the fluorescent peak when using the antibody-positive sample. The optimum ratios were 1:14 for biotin:NS3 protein and 1:1 for biotin:capsid protein (data not shown).
Frozen aliquots of an antibody-negative and an antibody-positive serum sample were stored at −20°C as controls for the assay. An aliquot of each sample was tested each time the HCV microsphere assay was performed. Table 1 lists stability data of the avidin-coated microspheres, the HCV-coated microspheres, and the antibody-reacted HCV microspheres. The avidin-coated microspheres when stored at 2°C to 8°C did not show any significant change in ability to bind the biotinylated HCV proteins over 8 months as demonstrated by the reaction with the frozen aliquots of the negative and positive serum samples. The microspheres once coated with the HCV proteins were stable, with no significant change in performance over 8 days, which was the maximum time evaluated. The antibody-reacted HCV microspheres were stable for 24 to 48 h before analysis; after 48 h, the CV of the PE (FL2) peak in antibody-positive samples increases with a decrease in mean intensity. The decrease in mean with an increase in the CV suggests that some of the antibody, which was bound to the HCV protein on the microsphere, has detached.
Table 1. Stability of the HCV Microsphere Assay
Avidin-coated microspheres were stored at 2°C to 8°C and subsequently coated with HCV and reacted with an aliquot of the frozen antibody-negative and -positive samples to evaluate stability. Data shown are PE (FL2) mean log channel numbers from the antibody-positive sample to NS3 (L5) and capsid (L10), respectively.
HCV-coated microspheres were stored at 2°C to 8°C and subsequently reacted with an aliquot of the frozen antibody-negative and -positive samples to evaluate stability. Data shown are PE (FL2) mean log channel numbers from the antibody-positive sample to NS3 (L5) and capsid (L10), respectively.
HCV-coated and reacted microspheres were stored at 2°C to 8°C and analyzed daily for 5 days to evaluate stability. Data shown are PE (FL2) mean log channel numbers from the antibody-positive sample to NS3 (L5) and capsid (L10), respectively.
Figure 4 shows a typical antibody-negative and antibody-positive sample in the microsphere assay. The L5, L7, and L10 microspheres are well separated and the antibody positive sample is easily distinguished from the antibody negative sample. The singlet microsphere population was gated by FSC and FL3 and the PE (FL2) log fluorescence was displayed and used for the determination of the microgram equivalent of IgG per milliliter. The percentage of singlet microspheres ranged from 40% to 60% and did not significantly vary between antibody-negative and antibody-positive samples. The percentage of singlet microspheres was greater than 90% before coating with the HCV proteins or human IgG.
Table 2 shows the range of HCV antibody results expressed in microgram equivalents of human IgG per milliliter seen in antibody-negative and antibody-positive serum samples. The lower detection limit of the assay was determined to be 0.01-μg equivalent IgG/ml. There was complete separation in the microgram equivalent IgG per milliter values between the antibody-negative and antibody-positive samples. Most of the indeterminate samples were negative by the microsphere assay, with 49 of the 54 (91%) antibody indeterminate samples having results lower than 0.01 μg equivalent IgG/ml. Five of the 54 had levels between 0.01 and 0.15 μg equivalent IgG/ml for NS3 and 0.01 to 0.06 μg equivalent IgG/ml for capsid. Although these five samples had levels higher than that seen in antibody-negative samples, these values were below or on the low end of that detected with antibody positive samples. Three of these five samples had weak reactivity to the NS3 band in the RIBA, one of the five had weak reactivity to the capsid band in the RIBA, and one of five had no reactivity in the RIBA.
Table 2. Range of Antibody-Negative and Antibody-Positive Samples Expressed in Microgram Equivalents of Human IgG per Milliliter
Negative (n = 103)
Positive (n = 118)
Indeterminate (n = 54)
The IgG preparations made from rabbits immunized with the NS3 and capsid HCV proteins were used to estimate the level of specific antibody binding in the microsphere assay. The IgG preparations were serially diluted and incubated in the HCV microsphere assay followed by the PE donkey anti-rabbit IgG. The PE (FL2) mean log channel that was generated by using the rabbit IgG was correlated to the estimated concentration of rabbit IgG specific for HCV. This estimated concentration was used to verify that the L7 human IgG microsphere calibrator was providing results in microgram equivalents similar to what was expected from the rabbit IgG preparations (data not shown).
The detection of specific antibody to HCV is an important assay in the identification of individuals infected with HCV. Widespread screening and confirmation of samples is well addressed using the commercially available assays (11). However, the increased rate of false-positive antibody test results coupled with the data indicating that the concentration of HCV specific antibody can help to indicate the likelihood of antibody positivity suggest that an assay for better quantitation would be useful (3, 6, 11). As an attempt at the semiquantitation of antibody levels and as a possible improvement to the resolution of low levels of specific HCV antibody, the microsphere assay was developed.
The use of different microsphere populations, which are uniquely identified by a characteristic such as size and/or fluorescence, has been previously described (12). The use of stable red-dyed (FL3) microspheres, which are coated with NeutraAvidin, provided for relatively easy multiplex assay development. The microspheres used in this study were provided with a dye that when excited at 488 nm has a broad emission, beginning at 650 nm with a peak at 700 nm. This dye does not have fluorescence spectral overlap into the channel (FL2) used for the reporter, PE. The three separate microsphere populations were well resolved by evaluating their FSC and FL3 signals and singlet microspheres were analyzed for the presence of antibody as evidenced by the PE fluorescence signal in FL2. Although the analysis of HCV antibody was based on singlet microspheres, there were a significant percentage of aggregates in all studies performed. These aggregates appeared to be unrelated to the presence of antibody to HCV but might have been due to charge interactions with the coating protein. Evaluation of conditions, which can improve the stability of singlet microspheres, will help decrease the confounding affects on analysis due to aggregated microspheres.
For effective capture onto the avidin-coated microspheres, the protein must be biotinylated. For the biotin labeling of the HCV recombinant proteins, it was found that low ratios of biotin:HCV protein produced optimum antibody signals in the microsphere assay. However, the optimum ratio was different between the two different HCV proteins used. With NS3 and capsid, when the biotin concentration was increased, the antibody signal, when using an antibody-positive sample, was reduced presumably due to blocking of epitopes on the HCV protein by the biotin reagent (data not shown). The HCV proteins, once biotinylated, were stable for at least 1 year when stored at 2°C to 8°C (data not shown) and the avidin-coated microspheres were stable for the 8 months during this testing. The coating of the microspheres with the biotinylated proteins was accomplished within 2 h and after coating the assay time was approximately 90 min. The stability of the reagents and the relative ease in performing the assay allowed for the routine performance of the microsphere assay.
In an attempt to provide a more standardized assay unit, a human IgG calibrator was developed. The use of different concentrations of biotinylated human IgG on the L7 microspheres provided a calibration curve, which was internal to each serum sample tested. This allowed for the reporting of the antibody binding in semiquantitative units. Although these human IgG microspheres were used to report values in human IgG equivalents, the concentration of human IgG was the coating concentration used and not the concentration of IgG per microsphere. Although this value is not directly transferable to the concentration of HCV specific antibody in a sample, the evaluation of the rabbit IgG anti-HCV preparation correlated with the human IgG equivalents, suggesting that this calibration could be used as a semiquantitative result. For the comparison of rabbit anti-HCV to the human IgG calibrators, an estimate of 1% HCV-specific IgG was assumed in the rabbit serum. Serial dilutions of the rabbit IgG anti-HCV preparations produced fluorescence results in the microsphere assay, which were consistent with the human IgG calibration curve. Although there are data to support the estimate of 1% of IgG, which is specific for the immunizing antigen used, the range of concentration can vary greatly depending on the type of antigen, adjuvant if used, immunization site and schedule, the host used, and the time after immunization that the blood is collected (13).
By using the human IgG calibrators, the HCV antibody concentration in the antibody-positive samples was shown to cover a broad range of human IgG equivalents. However, there was complete separation in the human IgG equivalents of the samples identified by the screening EIA as positive from those identified as negative. Most of the samples, which were identified as being indeterminate, probably represent false-positive reactions in the screening EIA because most of these samples were negative in the microsphere assay. This is in agreement with previous studies, which have also shown high rates of false-positive reactions in samples that, when tested by the EIA, have a signal:cutoff ratio in the range of 1.0 to 3.8 (3–5, 11). Although those indeterminate samples, which were over the 0.01-μg equivalent IgG/ml cutoff, may represent false-positive reactions in the EIA and the microsphere assay; alternatively, they may have low levels of specific antibody to HCV. Additional testing for HCV RNA levels may help to resolve the interpretation of antibody indeterminate samples (11).
The reagents used in this study demonstrated acceptable stability to allow their use in periodic testing without increased variability due to reagent degradation. The microspheres once coated with the biotinylated proteins and then reacted with the serum samples and the goat anti-human IgG PE reagent were best analyzed within 24 h, although they appear to be stable up to 48 h. Analysis after 48 h showed a decrease in the mean and increase in the CV of the PE (FL2) signal from antibody-positive samples. This change suggests that some of the bound antibody has detached from the microsphere. Because the data suggest the HCV-coated microspheres are stable for at least 1 week (Table 1) and the biotin-avidin affinity is higher than that of the human antibody to HCV, it is most likely that the decrease in mean and increasing CV are primarily due to antibody detaching from the microsphere.
The three different microspheres selected allowed for the detection and semiquantitation of HCV-specific antibody using two HCV proteins and an internal human IgG calibrator. Although microsphere aggregates can be somewhat confounding, gating on singlet microspheres can provide for reproducible and discrete data. Fluorescence compensation was required for the fluorescence spectral overlap of the PE reporter into the red dye (FL3) channel; however, the compensation required was minimal and was accomplished by adjustments on the flow cytometer during sample acquisition. This microsphere assay was reproducible, demonstrated a wide dynamic range, and was performed using a standard flow cytometric setup including the standard data analysis package. For larger studies, microsphere handling devices and improved data analysis programs will certainly allow more widespread use of this assay.
The author acknowledges the guidance and assistance of Mack Fulwyler, Ph.D., in the work on microsphere assays, the support and guidance of Daniel Stites, M.D., appreciation to Torsten Helting, Ph.D., for providing the HCV proteins, and helpful discussions on the contribution of noise and photon counting statistics from Eric Chase, Ph.D.