|Home | About | Journals | Submit | Contact Us | Français|
Autoimmune hepatitis (AIH) is a chronic necroinflammatory disease of the liver with a poorly understood etiology. Detection of non-organ-specific and liver-related autoantibodies using immunoserological approaches has been widely used for diagnosis and prognosis. However, unambiguous and accurate detection of the disease requires the identification and characterization of disease-specific autoantigens. In the present study, we have profiled the autoantigen repertoire of patients with AIH versus those with other liver diseases, identifying and validating three novel and highly specific biomarkers for AIH. In Phase I we fabricated a human protein chip of 5,011 non-redundant proteins and used it to quickly identify 11 candidate autoantigens with relative small serum collection. In Phase II we fabricated an AIH-specific protein chip and obtained autoimmunogenic profiles of serum samples from 44 AIH patients, 50 healthy controls, and 184 additional patients suffering from hepatitis B, hepatitis C, systemic lupus erythematosus, primary Sjögren’s syndrome, rheumatoid arthritis, or primary biliary cirrhosis. Using this two-phase approach, we identified three new antigens, RPS20, Alba-like, and dUTPase, as highly AIH-specific biomarkers, with sensitivities of 47.5% (RPS20), 45.5% (Alba-like), and 22.7% (dUTPase). These potential biomarkers were further validated with additional AIH samples in a double-blind design. Finally, we demonstrated that these new biomarkers could be readily applied to ELISA-based assays for use in clinical diagnosis/prognosis.
Autoimmune hepatitis (AIH) is a progressive, necroinflammatory liver disease of cryptic etiology1 that affects people regardless of age and gender; however, some people are known to be genetically susceptible to this disease. Evidence suggests that the condition is associated with the human complement allele C4AQO and HLA haplotypes B8, B14, DR3, DR4, and Dw32. Nevertheless, the pathogenesis of the condition remains unclear. Patients exposed to effective treatment early in the course of the disease may have their symptoms suspended; thus, early and precise diagnosis is clearly important3. Despite the difficulty involved in differentiating AIH from non-AIH cases, diagnostic criteria were codified by an international panel in 1993 and revised in 19994, 5. One of the significant parameters in this scoring system is the presence of autoantibodies in the patient’s serum.
AIH has been divided into two types that differ in their immunoserological presentation. The AIH-1 type is characterized by the presence of smooth muscle autoantibodies (anti-SMA) and/or antinuclear autoantibodies (ANA). The targets of the anti-SMA autoantibodies are F-actin and actin, and the probable targets of ANA are the histidines, DNA-binding proteins, and the centrosome complex. The AIH-2 type is characterized by the presence of autoantibodies against liver and kidney microsomal antigens (anti-LKM) and/or liver cytosol antigen 16, 7. It was previously thought that a third AIH type also existed, one that was characterized by the presence of anti-soluble liver antigen/liver pancreas (anti-SLA/LP) autoantibodies. However, it has recently been acknowledged that this variety of AIH should be considered an AIH-1 type.
At present, indirect immunofluorescent (IIF) staining of rodent tissue and HEp-2 cells is the most frequently used method for detecting circulating ANA, SMA, and anti-LKM autoantibodies1; the antigens recognized by these autoantibodies are used as diagnostic antigens for profiling autoantibodies in patients’ sera. ELISA methods utilizing autoantibodies that target specific autoantigens are being used, or are under development, as a complementary approach to IIF diagnosis, with the goal of improving diagnostic accuracy8. As yet, however, there is no compelling evidence regarding the identity of the ANA targets. By screening a human liver lambda gt11 cDNA expression library9, Manns et al. have identified cytochrome P450db1 as one of the target antigens of anti-LKM autoantibodies. Using inhibition ELISAs and serum samples from controls and patients with various liver diseases to screen cDNA expression libraries for SLA autoantibodies, another group has identified a 50-kDa protein, cytosolic UGA-suppressor tRNA associated protein, as the target of the anti-SLA/LP autoantibody10. Asialoglycoprotein receptor (ASGP-R) is a liver-specific protein that is localized to the surface of the liver cell; using a radioimmunoassay, researchers have detected immunoreactivity against this protein in antisera from AIH patients11. Detection of autoantibodies against ASGP-R has proved to be of great value for the diagnosis of AIH patients who are seronegative for SMA, ANA, and other classes of autoantibodies12.
The use of conventional methods to screen expression libraries with AIH and control sera in order to identify AIH-specific autoantigens is a time-consuming, tedious process that tends to miss proteins that are in low abundance but robust in serological assays for AIH. Protein microarray technology is a newly developed analytical method that has been shown to have great potential for identifying autoimmune disease biomarkers13. In a previous publication, we reported the construction of a protein chip containing more than 1,000 recombinant human liver proteins and described the use of this chip to identify antigens in a high-throughput approach to monoclonal antibody production14.
We have now expanded the protein collection on this chip to 5,011 human proteins and used this chip to test the hypothesis that a panel of autoantigens can be used as capture molecules in a microarray to improve both the sensitivity and specificity of AIH diagnosis. To assess this hypothesis, we made use of the expanded chip to screen serum samples from healthy serum controls and individuals with AIH as a means of identifying novel AIH-specific autoantigens. In order to validate the specificity of the candidate autoantigens that we had identified in our screening assay, we developed an AIH-related microarray containing all the candidate autoantigens and probed this microarray with serum samples from patients with a variety of liver diseases, including AIH, hepatitis B (HB), hepatitis C (HC), systemic lupus erythematosus (SLE), primary Sjögren’s syndrome (pSS), rheumatoid arthritis (RA), and primary biliary cirrhosis (PBC). These microarray results demonstrated that three previously identified autoantigens (p62, CYP2D6, and ASGP-R [H2 subunit]) and three previously unrecognized autoantigens (RPS20, Alba-like protein, and dUTP diphosphatase) could specifically differentiate AIH patients’ sera from other sera. Using two algorithms developed in the current study, we were able to demonstrate that the six autoantigens could be used in combination to clearly diagnose AIH-positive serum samples; we further showed that these autoantigens could be effectively used in protein microarray assays as well as traditional ELISA-based assays.
Human serum samples were obtained between 2003 and 2006 from 278 individuals at the Peking Union Medical College Hospital: 44 AIH, 50 PBC, 41 HC, 43 HB, 11 SLE, 11 pSS, and 2 RA patients, as well as 26 with non-autoimmune diseases and 50 healthy subjects. All AIH patients were clinically diagnosed according to the scoring system established by the IAIHG in 1999. All of the AIH serum samples had been clinically characterized as positive for ANA (n=28), SMA (n=21) and/or SLA/LP (n=3) by indirect immunofluorescence assay (Euroimmune, Germany). An additional 41 serum samples from 15 AIH patients and 26 healthy controls were collected at Beijing Union Medical College Hospital in 2007 for validation experiments. The study was approved by the IRBs at Peking Union Medical College Hospital.
Human gene cloning was carried out as previously described14, with one modification: The samples used for the PCR templates were individual cDNA clones that had been sequence-identified and isolated from various human tissues. The resulting PCR products were subcloned into pEGH using a yeast homologous recombination strategy15. In brief, primers containing yeast recombinant sites were designed and used to PCR-amplify human coding regions, and the PCR products were co-transfected into yeast (Y258) with the linearized yeast expression vector pEGH containing the homology recombination sites. The resulting recombined plasmids were amplified and carried in the yeast. To monitor the success of the open reading frame (ORF) cloning, the resulting plasmid DNAs were extracted, and the inserts were amplified using generic primers flanking the recombination sites. Insert sizes were determined by gel electrophoresis.
Protein expression and purification were conducted as described previously15. The purified GST fusion proteins were eluted into printing buffer containing 30% glycerol in 50 mM HEPES (pH 7.0). The eluates were arrayed in 384-well titer dishes for printing. The quality of the protein purifications was verified by SDS-PAGE. Purified human proteins and control proteins were spotted in duplicate onto poly-L-lysine-coated slides at high density using a SpotArray-72 printer (Perkin Elmer). The printed human protein chips were kept horizontal at room temperature (RT) for 1 h before storage at 4° C.
Protein chips were first blocked by immersion in blocking buffer (PBST, 0.1% [v/v] PBS with 0.05% [v/v] Tween 20, and 1% [w/v] BSA) at RT for 1 h with gentle shaking. Subsequently, 200 µL of serum (diluted 1000-fold in PBS buffer) was added to the chip and incubated under a glass coverslip (LifterSlip, Erie Scientific Company, Portsmouth, NH, USA) to maintain the appropriate humidity. After an 1-h incubation at RT, the chip was subjected to three 10-min washes with PBST at 40° C, followed by a 1-h incubation with 1,000-fold-diluted anti-human IgG antibody labeled with Cy-5 (The Jackson Laboratory, Bar Harbor, ME, USA) in the dark at RT. The chip was again washed three times with PBST. After two rinses in double-deionized water, the chip was blown to dryness with compressed air and scanned with a BioCapital microarray scanner (BioCapital, Beijing, China). The binding signals were acquired and analyzed using the GenePix Pro 5.0 software (Molecular Devices, Sunnyvale, CA, USA).
Following image analysis with GenePix Pro 5.0, the signal intensity of each protein feature was determined by subtracting the median background value from the median foreground value. When the resulting signals were <0, the protein spots were assigned a minimum signal of 1. Since every protein was spotted in duplicate, the average of each duplicate was used as the final signal intensity for a given protein.
The total signal intensity was used to perform inter-chip normalization. The total signal intensity for a given chip (chip signaln) was divided by the mean value for all the chips (average (chip signaln)) to obtain an intensity factor for the chip (intensity factor (chipn)). Each signal for a given chip was then divided by the intensity factor (chipn) to generate a normalized signal.
The mean value of a given protein for all control serum samples was then calculated and termed the averageprotein i control value. Dividing the signal intensity value for each protein in the individual patient’s serum by the averageprotein i control value yielded the so-called “factor F” value. A protein was considered positive when its F value was >216. The number of each protein’s positive values for the patient sera was counted, and those proteins with a positive rate >60% for all the patient sera were considered AIH-specific autoantigen candidates.
All the AIH-specific autoantigen candidates identified on the human protein chip were printed together with controls (human IgG, glutathione-S-transferase [GST], and printing buffer) in duplicate within a 2.52 mm×1.44 mm area, and 12 identical probe areas were fabricated on a single poly-L-lysine-treated microscope slide. The printed microarrays were allowed to remain at RT for 1 h before storage at 4 ° C.
In order to probe the protein microarrays with human sera, a 12-hole rubber gasket (BioCapital, Beijing, China) was applied to each microarray to form 12 individual chambers. The probing procedure was basically identical to that performed without chambers, except that 50 µl of 1/1,000-diluted human serum was incubated in each microarray chamber. After the incubation, the microarrays were rinsed with wash buffer before the rubber gaskets were carefully removed. The chips were scanned using a BioCapital microarray scanner (BioCapital), and the signals were analyzed using GenePix Pro 5.0 software (Molecular Devices).
The mean signal intensities and standard deviations (SD) for 50 healthy controls were calculated, and the value for the meanhealthy+3×SD for each particular potential antoantigen was used as an immnoreactive cutoff value.
To determine the specificity and resolution of each of the potential AIH autoantigens, we applied two commonly used computational methods, logistic regression and discriminant analysis17, using open-source SPSS 12.0 software. In the case of both methods, two group analyses were conducted, for the AIH and non-AIH groups. Logistic regression is a generalized linear regression model for binary responses. The candidate features were selected by the model using a dual-direction stepwise search with Akaike’s information criteria, and the probabilities were calculated as follows: p = EXP (Σ(bixi)+c)/(1+EXP (Σ(bixi)+c)), where p is the probability of each case, i=1 to n; b is the regression coefficient of a given autoantigen; × is the signal intensity; and c is a constant generated by the model. For prediction, only those cases with a >0.5 probability were classified as AIH-positive. The model first performed t-tests for each of the 14 autoantigens and selected those with p values <0.01 for the model building.
Discriminant analysis is a linear regression model that predicts group membership by using a set of predictors. The formula is y=Σ(bixi)+c, where i=1 to n, x is the signal intensity, b represents the unstandardized canonical discriminant function coefficient, and c is a constant provided by the model. Those cases in which the y value was >0 were considered AIH-positive. The model first generated standardized canonical discriminant function coefficients for each autoantigen in order to evaluate its importance in prediction. The selected autoantigens were then used for model building.
Purified recombinant protein (100 µl at 2.5 µg/ml) was added to the wells of a 96-well plate and incubated for 1 h at 37 °C. The wells were blocked by discarding the protein solution and adding 200µl of blocking buffer (1% BSA), then incubating the samples for 1 h at 37 °C. The blocking buffer was then discarded, and the sera (1:40 dilutions) were individually added to the wells and incubated for 1 h at 37 °C. The serum dilutions were removed, and the wells were washed three times with 300 µl PBST each time; HRP-conjugated mouse anti-human IgG antibody (diluted 1:5,000) was then added, and the samples were incubated for 1 h at 37°C. The wells were again washed three times with 300 µl PBST each before the addition of 100 µl tetramethylbenzidine (TMB) substrate solution (10 µg TMB, 1 µl DMSO, 0.51 g trisodium citrate dihydrate, 0.184 g disodium hydrogen phosphate dodecahydrate, 10 µl H2O2, and 99 µl H2O). After a 5- to 10-min incubation at RT, the reaction was stopped by adding 100 µl sulfuric acid (4.5 N) per well, and the absorption at 450 nm (OD450) was measured using an automatic ELISA reader (Tianshi, Beijing, China).
We have previously reported the fabrication of a human protein microarray containing 1,058 liver-expressing proteins in full-length and demonstrated its application in high-throughput antigen identification14. To enable a broader use of this technology platform, particularly with regard to autoantigen identification, we first attempted to increase the protein content of the chips by adding proteins expressed in a wide range of tissues. First, we designed a primer set for PCR amplification of 5,670 human ORFs using cDNA templates obtained from human fetal brain, bone marrow, pancreas, and testicle as well as adult and fetal liver. Second, we used a yeast recombinant cloning strategy to clone these ORFs into a yeast expression vector, pEGH. A total of 4,053 ORFs were successfully cloned. Together with the 1,058 previously reported proteins, a total of 5,011 non-redundant human proteins (88.6 % full-length) were over-expressed and purified as GST fusion proteins in yeast.
In a pilot assay, we found that poly-L-lysine-coat slides produced a signal-to-background ratio comparable to that of other surfaces, such as FAST and Fullmoon, in serum profiling assays (data not shown). At the same time, we also tested a variety of different conditions in order to optimize the serum profiling assay, including a range of serum concentrations (200-to 5000-fold dilutions) and washing conditions (e.g., high salt versus low salt, and room temperature versus 40°C). On the basis of these experiments, we chose poly-L-lysine slides for chip fabrication, 1000-fold dilution as the most appropriate serum concentration, and 40°C as the chip-washing temperature (see the Materials and Methods for details).
We then fabricated a second generation of human protein chips by printing all the purified proteins onto poly-L-lysine-coated microscope slides. Because CYP2D6 and ASGP-R had already been identified as autoantigens in AIH patients18, 19, and p62 has been implicated in some PBC cases20, we included these three proteins on the chips as positive controls. Human IgG at a known concentration was also printed at the upper left corner of each block on the chips to serve as a control and landmark. The quality and quantity of the immobilized proteins on the chips were determined by probing with an anti-GST antibody, and 84.5% of the proteins produced signals that were significantly above background (Fig. 1).
To save time and expense with regard to the serum profiling assays, we carried out the assays in two phases: In Phase I, we selected 22 AIH and 30 control serum samples and probed the human protein chips with the samples individually at a 1000-fold dilution, followed by detection of bound human autoantibodies using a Cy-5-conjugated anti-human IgG antibody. We observed that a large number of proteins (e.g., in the hundreds) could be readily recognized by sera from both patients and healthy controls, and the serum profiles also showed obvious individual-to-individual variation within both the patient and control groups (Fig. 2), a phenomenon that has been reported in previous investigations21, 22.
To identify potential autoantigens that are specifically associated with AIH, we used GenePix software to acquire and process the protein chip data. Binding signals for each serum probing were first normalized using the overall chip signals (as detailed in Materials and Methods). Then, the value of each signal produced by a protein that reacted with the AIH serum was divided by the mean value of the protein’s signal from the healthy control serum samples (negative controls). The resulting ratio was defined as factor F, reflecting the differential autoimmunogenic potential of a given human protein, or autoantigen. At F values >2, all three positive control antigens (p62, CYP2D6, and ASGP-R) identified at least 13 of the 22 AIH-positive serum samples. Therefore, we decided to use this criterion to score the potential autoantigens. This scoring system identified 11 candidate autoantigens (Table 1), of which two (RPS20 and Alba-like) were ranked higher than all of the three previously known antigens. None of the patients showed a positive reaction with all of the candidate antigens.
In Phase II, we fabricated an AIH-specific protein chip that carried the 11 candidate autoantigens and the three positive control antigens, in addition to GST and BSA as negative controls. This strategy allowed us to fabricate protein chips in large number at low cost and to screen a large number of serum samples in a relatively short time (Fig. 3). To determine which candidate autoantigens were AIH-specific, we sequentially probed the AIH-specific chips with the serum samples used in Phase I and obtained data for samples from an additional 22 AIH, 50 PBC, 43 HB, 41 HC, 11 SLE, 11 pSS, and 2 RA patients using the same protocol as described above. As negative controls, we also included 26 serum samples from patients suffering from other types of severe diseases and 50 samples from healthy subjects. Representative probing images are shown in Fig. 3.
After the same process of data normalization, we identified AIH-specific autoantigens by using a calculation that was more stringent than the one used in Phase I. We first calculated the mean value (meanhealthy) and standard deviation (SD) for the individual signal intensities obtained for the 50 healthy control sera. Using a value of three times the SD above meanhealthy as a cutoff value, we scored the positives for each serum probing reaction and calculated the true- and false-positive values for each antigen presented on the chips (Table 2). Three known antigens (p62, CY2D6, and ASGP-R) and three new candidate antigens (RPS20, Alba-like, and dUTP diphosphatase [dUTPase]), showed a true positive value for the AIH samples that ranged from 20.5% to 47.7%, significantly higher than the values for the remaining eight candidate autoantigens. These six antoantigens also showed a slight cross-reactivity with PBC, HB, HC, and other control samples (Table 2), which is consistent with the behavior of autoantigens reported previously23, 24.
We next applied a logistic regression algorithm, a generalized linear regression analysis for binary responses, to analyze the serum profiling data. Of the 11 candidate and 3 known autoantigens, the top 6 selected by the software because of their low p values (<0.01) in t-tests were the same 6 antigens identified in the screening described above (Table 2). Using the signals for these six proteins, we created a logistic regression model. The formulas for the logistic regression and p value calculation were (Formula I):
where X1 to X6 represent the signals from RPS20, Alba-like, p62, CYP2D6, dUTPase, and ASGP-R, respectively. When our model was used to classify the 278 training samples, it achieved an accuracy of 91.7% (sensitivity of 68.2% and specificity of 96.2%, with the sensitivity and specificity defined as correctly classified positive and negative sera, respectively) (Table 3).
We next calculated the standardized canonical discriminant function coefficients and found that seven autoantigens had a positive contribution, and the six antigens identified by the microarray analysis and logistic regression algorithm were among them (Table 2). Using the signals for these six proteins, we created a second discriminant model, Formula II:
where X1 to X6 represent the signals from RPS20, Alba-like, p62, CYP2D6, dUTPase, and ASGP-R, respectively. When the y value was >0, the autoantigen was scored as positive. Using this model, we were able to classify the AIH-positive training cases versus the non-AIH cases with an overall accuracy of 91.7% (sensitivity of 84.1% and specificity of 93.2%) (Table 3).
These analyses indicated that both computational models identified RPS20, Alba-like, p62, CYP2D6, ASGP-R (H2 subunit), and dUTP diphosphatase as powerful potential autoantigens for use in differentiating AIH from non-AIH human serum samples, consistent with the results we had obtained by calculating positives using cutoff value. Thus, our combined analyses convincingly demonstrate the specificity of these six proteins for AIH.
Finally, we conducted box-whisker plot analysis to visualize the range of signal intensities for the six candidate autoantigens across the various groups of sera. Two autoantigens (RPS20 and Alba-like) identified in this study were not only able to differentiate AIH from non-AIH sera but also to differentiate AIH from PBC or HCV serum samples (Fig. 4). Because the previously identified autoantigens used for AIH diagnosis are often detected in sera from PBC and HCV patients23, RPS20 and Alba-like offer the important advantage that they can be used as more specific autoantigen biomarkers for diagnosing AIH.
In order to validate the specificity of the three autoantigens identified using our protein microarray approach and to assess their utility for AIH diagnosis, we conducted a double-blind experiment using 41 additional serum samples that were collected at the Beijing Union Medical College Hospital in 2007; 15 of these samples had been identified as AIH-positive. Using the protocol described above, we obtained the autoantigen profiles of all 41 serum samples using the AIH-specific chips. When we applied the logistic regression model, 9 of the15 AIH sera were diagnosed as positive, whereas none of the healthy controls were scored as AIH-positive. With the discriminant model, we were able to recognize 10 of the 15 samples as AIH sera and, again, none of the healthy controls were misdiagnosed. Therefore, the protein microarray approach was confirmed as being able to diagnose AIH with an accuracy of 85.4% (logistic regression model: 60% sensitivity and 100% specificity) or 87.8% (discriminant model: 66.7% sensitivity and 100% specificity) (Table 4).
To determine whether the autoantigens identified by the protein chip approach could be readily transformed for use as diagnostic biomarkers for AIH in a traditional ELISA format, we tested the three new autoantigens (i.e., RPS20, Alba-like, and dUTPase) in an ELISA assay and compared the results we obtained to those obtained for the three known autoantigens (CYP2D6, ASGPR2, and p62), using 15 AIH-positive and 15 healthy serum samples.
We then calculated the mean value and standard deviation of the OD450 values for the healthy serum samples in the ELISA tests. When we used a value of three times the SD of the negative controls as a cutoff, we found that 9 of the 15 AIH-positive sera were positive for at least one of the three new autoantigens, and 10 of 15 reacted with at least one of the three previously known autoantigens (Fig. 5). Four AIH-positive samples did not react with either of these groups of autoantigens. Consistent with previous reports, p62 had the lowest sensitivity (2 of 15); however, sample AIH10 was only reactive with p62, indicating that it is still useful as a biomarker. The other five autoantigens showed comparable sensitivities, ranging from 26.7% to 46.7%. It is also noteworthy that sample AIH7 was reactive only with a new biomarker, dUTPase. In summary, the new autoantigens can be readily applied to a traditional ELISA-based clinical diagnostic format and, together with the existing biomarkers, can be used to improve the accuracy and sensitivity of AIH diagnosis.
Proteomics technologies have been employed to profile autoimmune disease samples in order to identify specific autoantibodies for diagnosis and/or prognosis25. In recent years, we have witnessed a rapid development of the protein chip approach in this fast-moving field. Robinson and colleagues were among the first to develop autoanigen microarrays containing hundreds of known autoantigens that could be used to profile autoimmune diseases (not including AIH) 26. Others have screened random cDNA or phage-display libraries in a protein chip format27–30. Recently, Snyder and colleagues identified four biomarkers for ovarian cancer using a commercially available human protein chip of 5,005 proteins31. However, these approaches, although fruitful, have proved to be time-consuming, insufficiently comprehensive, or high in cost.
We now report the development of a new strategy that has allowed us to quickly narrow down the number of potential autoantigens to a manageable number. Our two-phase strategy involved probing a high-content human protein chip with fewer than two dozen patient serum samples, and then following this preliminary step with extensive serum profiling using a disease-specific (AIH) protein chip. By taking advantage of existing biomarkers, we were able to readily identify 11 potential autoantigens in Phase I, and then further confirm three of these proteins as newly identified AIH-specific autoantigens in Phase II in experiments using a large number of AIH-positive and control sera. Using two kinds of statistical algorithms, we were able to achieve similar diagnostic results, with high sensitivity and specificity. The effectiveness of our strategy was then validated using additional AIH samples in a double-blind format. Thus, our analyses confirm that this new strategy has considerable potential as an efficient and cost-effective method for identifying relevant biomarkers of AIH and, by extension, of other diseases.
The new autoantigens identified in this study may provide additional clues regarding the composition of anti-nuclear autoantibodies (ANAs). The Alba-like protein is a member of the Alba family, which includes the archaeal protein Alba and a number of eukaryotic proteins with no known function. The DNA/RNA-binding protein Alba binds double-stranded DNA tightly but without sequence specificity. It binds rRNA and mRNA in vivo and may play a role in maintaining the structural and functional stability of RNA and, perhaps, ribosomes. It is distributed uniformly and abundantly throughout the chromosomes32. The dUTP diphosphatase (dUTPase) has been identified in the form of two splice variants, with dUTP-N localized to the nuclei in HeLa cells33. Hence, both proteins are presumably representative autoantigen targets among the ANAs. Interestingly, the ribosomal protein RPS20 has previously been identified as an autoantigen in SLE patients, but with a relatively low sensitivity of detection (20%)34. In the present study, we also observed that one out of 11 serum samples (9%) from SLE patients recognized RPS20; however, this protein was a much more sensitive autoantigen (48%) in the case of the AIH patients in our study (Table 2).
Our new approach and relevant clinical findings provide strong evidence that advances in high-throughput protein microarray technology have the potential to rapidly identify novel biomarkers useful for autoimmune diseases and to improve the diagnosis and/or prognosis of AIH and other types of autoimmune disease.
This work was supported in part by funding from Ministry of Science and Technology of the People’s Republic of China (2006AA02A311 to L.W.), National Natural Science Foundation of China (30640084 and 30872331 to Y.L.), and the NIH (R01GM076102 to H.Z.). We thank Dr. Deborah McClellan for editorial assistance.