Identification of new autoantigens by protein array indicates a role for IL4 neutralization in Autoimmune Hepatitis

Autoimmune hepatitis (AIH) is an unresolving inflammation of the liver of unknown cause. Diagnosis requires the exclusion of other conditions and the presence of characteristic features such as specific autoantibodies. Presently, these autoantibodies have relatively low sensitivity and specificity and are identified via immunostaining of cells or tissues; therefore, there is a diagnostic need for better and easy-to-assess markers. To identify new AIH-specific autoantigens, we developed a protein microarray comprising 1626 human recombinant proteins, selected in silico for being secreted or membrane associated. We screened sera from AIH patients on this microarray and compared the reactivity with that of sera from healthy donors and patients with chronic viral hepatitis C. We identified six human proteins that are specifically recognized by AIH sera. Serum reactivity to a combination of four of these autoantigens allows identification of AIH patients with high sensitivity (82%) and specificity (92%). Of the six autoantigens, the interleukin-4 (IL4) receptor fibronectin type III domain of the IL4 receptor (CD124), which is expressed on the surface of both lymphocytes and hepatocytes, showed the highest individual sensitivity and specificity for AIH. Remarkably, patients' sera inhibited STAT6 phosphorylation induced by IL4 binding to CD124, demonstrating that these autoantibodies are functional and suggesting that IL4 neutralization has a pathogenetic role in AIH.

Autoantibodies can just be an epiphenomenon secondary to a chronic inflammatory milieu (1) but can also play a direct pathogenetic role, such as anti-thyroglobulin antibodies in Hashimoto's thyroiditis (2).
Autoimmune Hepatitis (AIH) is a chronic necro-inflammatory disease of unknown etiology which affects predominantly women with an incidence of 1-2 per 100,000 per year and a prevalence of 10-20/100,000 (3,4). AIH is subdivided into two major types on the basis of autoantibodies reactivity (5). Antibodies to nuclei (ANA) and/or to smooth muscle (SMA) characterize type 1 AIH, whereas antibodies to a liver-kidney microsomal constituent (anti-LKM) define patients with type 2 AIH. Since detection of these autoantibodies is done by immunofluorescence on rodents multi-organ sections (liver, kidney, stomach), there are problems such as standardization and interpretation of the immunostaining patterns (6). To overcome these methodological problems, the International Autoimmune Hepatitis Group established an international committee to define guidelines, develop procedures and reference standards for more reliable testing (7,8). Although ELISA and bead assays with purified or recombinant autoantigens are under development (9), they actually represent a complementary rather than an alternative approach to traditional immunofluorescence. Moreover, serological overlap is frequently observed between AIH and other non-autoimmune liver diseases such as chronic viral hepatitis (10). Therefore new highly specific markers represent an unmet medical need for a more accurate diagnosis and classification of AIH.
Beside the potential diagnostic application, the discovery of novel AIH autoantigens could provide insights on the disease pathogenicity mechanism. Although some AIH target-autoantigens have been identified and characterized, little is known on their pathogenetic role, and other autoantigens are probably still unknown. Autoantibodies, to be considered pathogenetic, must have at least two features: (i) the target-autoantigen should be either expressed on the plasma membrane of target cells or secreted by cells, i.e. should be exposed to autoantibodies, (ii) binding of the autoantibodies to the target antigen should disturb a cellular function directly or indirectly. A possible pathogenetic role in AIH has been put forward for autoantibodies specific for Cytochrome P450 2D6 (CYP2D6) or Asialoglycoprotein receptor 1 (AGPR-1), which are both present on the hepatocytes cell membrane (10).
Protein microarrays are a powerful technology as they allow the simultaneous screening of thousands of analytes (11). In the present study, to identify new autoantigens with potential diagnostic and/or pathogenetic role in AIH, we printed a microarray with 1626 human proteins whose main features were to be either secreted or membrane associated, i.e., potentially exposed to autoantibody recognition. We used this microarray to screen panels of sera from patients with AIH and identified six new protein antigens that are recognized with high sensitivity and specificity. One of these six autoantigens is FNIII domain of interleukin-4 receptor (CD124) and, interestingly, patients autoantibodies specific for CD124 neutralize IL4 signaling, so suggesting a possible pathogenetic role for IL4 neutralization in AIH. Human proteins -selection, expression and purification -Genes whose translated products carry a secretion signal peptide or at least one transmembrane domain were selected, cloned and expressed in a high throughput system as histidine-tagged products as described (12). A total of 1626 polypeptides were cloned and expressed in E. coli. Of these, 1121 were cloned as protein fragments and 505 as full length proteins (Supplementary Table 1-2). The recombinant proteins were affinity-purified from the bacterial insoluble fraction by Immobilized metal ion affinity chromatography (IMAC, GE).

Experimental Procedures
Representative SDS-PAGE gels of a panel of purified proteins are shown in Supplemental Fig. 1.
Human, viral or bacterial proteins were used as biological or technical controls in the microarrays (Supplementary Figure 2): HCV Core protein and Non-structural proteins NS3 (from HCV genotype 1), NS3-4a (from HCV genotype 2) NS5b (from HCV genotype 1), Tetanus toxin and H1N1 antigen were produced in house by subcloning the corresponding genes in E. coli strain DH5 and expressing them in BL21(DE3); Bovine Serum Albumin (BSA), Human Serum Albumin, Human Glutathione-S-Transferase and Protein A from Staphylococcus Aureus were purchased from Sigma.
Protein quality control -Purified recombinant proteins obtained as described above, were analyzed by SDS-PAGE (Criterion PAGE system Bio-Rad) followed by Coomassie Blue staining of the gel to assess their integrity and purity ( Supplementary Fig. 1).
Protein purity was assessed by BioRad ChemiDocTM XRS, Quantity One ® software.
Proteins showing purity levels > 70% were used for protein array preparation.
To further analyse the quality of the purified proteins, we performed western blot analysis on the purified proteins with an anti-His monoclonal antibody (anti-His mAb).
More in detail, the proteins were resolved on 4-12% pre-cast SDS-PAGE gradient Tricine gels under reducing conditions, and electroblotted onto nitrocellulose membranes (Bio-Rad), according to manufacturer's instructions. The membranes were blocked with 5% non-fat milk in PBS with 0.1% Tween 20 (TPBS) for 1 h at room temperature, incubated with -His mAb (GE-Healthcare) diluted 1:1000 in 3% non-fat milk in TPBS for 1 h at room temperature, and washed three times in TPBS. The secondary HRPconjugated antibody (-mouse immunoglobulin/HRP, GE-Healthcare) was diluted 1:1000 in 3% non-fat milk in TPBS and incubated for 1 h at room temperature. The proteins were visualized by Enhanced Chemiluminescence (Super Signal West Pico Chemiluminescence Substrate, Thermo Scientific, USA) and detected with LAS-3000 (Fujifilm, USA).
Protein microarray printing -Protein Microarrays were generated by spotting the 1626 affinity-purified recombinant proteins (0.5 mg/ml in 6M Urea) in 4 replicates on nitrocellulose-coated slides (FAST slides, GE-Healthcare) using Stealth SMP3 ® spotting pins (TeleChem International, Sunnyvale, California) and a Microgrid II microarray contact printer (Biorobotics), resulting in spots with a diameter of approximately 130 μm.
As experimental positive control, a curve of human IgG at 11 different concentrations (from 0.001 to 1 mg/ml) was spotted on the arrays in 8 replicates (in 6M Urea) ( Supplementary Fig. 3A). Several spots of buffer alone were also printed and used to assess possible non-specific signals due to cross contamination. A quality control of the spotting procedure was performed on 10% of randomly chosen slides. The percentage of proteins successfully spotted on the slides was assessed by hybridizing the arrays with an -His mAb, followed by an Alexa-647 conjugated α-Human IgG secondary antibody and estimating the number of spots with a MFI value significantly above background. A distance matrix was calculated by TIGR Multiexperiment Viewer (version MeV4.7) software (13) to evaluate the system reproducibility ( Supplementary Fig. 3B).
The spotted microarrays were allowed to remain at room temperature for 1 h before storage at 4 °C until use. deviations) was chosen as the lowest signal threshold for scoring a protein as positively recognized by human sera. For each protein, a Coefficient of Variation (CV%), was calculated on four replicate spots, for intra-assay reproducibility (14). Each antigen was checked for displaying a CV% correlated to its MFI on the basis of standard IgG curves.
If the CV% value was not within the expected range the antigen was not considered for further analysis.
Recognition frequency was defined as the percentage of sera reacting with a particular antigen in protein array with a MFI ≥ 4.000, and it was calculated for each group of sera. TIGR Multiexperiment Viewer (version MeV4.7) software (13) was used to perform an unsupervised bi-dimensional hierarchical clustering.

Dissociation-Enhanced Lanthanide Fluoroscence ImmunoAssay (DELFIA ® ) assays -
The DELFIA ® assay is a time-resolved fluorescence method that can be used to study antibody binding to solid-phase proteins or peptides. The purified recombinant proteins were used at a concentration of 20 μg/ml (15) in 6M Urea to coat DELFIA ® plates (PerkinElmer). Plates were then blocked for 1 hour at 37°C with a blocking reagent (PerkinElmer). The serum samples, diluted 1:300 in PBS with 1% BSA (Sigma), and 0.1 %Tween 20 (Sigma) were incubated on the plates for 1 hour at 37°C. Plates were then washed 5 times with washing buffer (PerkinElmer) and then incubated 30 min at room temperature in the dark with Europium-labeled α-human IgG serum (1:500 in diluting buffer, PerkinElmer). After extensive washing, plates were left at room temperature for 10 min and then read on a Infinite F200 PRO instrument (Tecan).
Fluorescence intensity values higher than the mean of buffer plus 3 standard deviations were considered as positive.
Surface staining of IL4R on HeLa cells -To assess recognition of native IL4R by human sera, full length IL4R was overexpressed in HeLa cells. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2mM Lglutamine and 1% penicillin-streptomycin. The human cDNA clone of full length IL4R, Flow cytometry with FACS Canto II analyzer (Becton Dickinson) and data were processed with the program FlowJo (Flow Cytometry Analysis Software). To confirm that the mechanism was mediated by antibodies and not by other factors present in the serum, patients and healthy donors sera were depleted from anti-IL4R(FNIII) antibodies and tested in Stat6 phosphorylation assay. To do this, 2 mg of purified protein were blotted on nitrocellulose membrane and then incubated with patients or healthy donor sera. After this incubation sera were collected and tested as described above.
Soluble IL4R detection -The amount of soluble IL4R was determined with Abcam sIL4R Elisa kit. Briefly, sera were diluted 1:4 in diluent buffer provided by the kit and incubated in 96 well plates coated with anti-sIL4R. Biotinylated monoclonal antibody specific for sIL4R was added to the wells and plates were incubated 1 h at room temperature. After washing, streptavidin-HRP was added to the wells and plates were incubated for additional 30 min. 3,3',5,5' -tetramethylbenzidine (TMB) substrate was then added and OD at 450 nm was read on a Infinite F200 PRO instrument (Tecan).
Statistical analysis -Results of Protein Microarray and DELFIA ® experiments were analyzed using the two-tailed 2 test, the Student's t-test, the Fisher's exact tests or the ANOVA test. The Benjamini-Hochberg correction for multiple testing was used for the analysis of microarray data. Statistical analysis was carried out with the use of TIGR Multiexperiment Viewer or GraphPad Software Inc. Predictive analysis of microarray (PAM) was performed using the statistical package PAM 1.51 with the statistical tool R (http://wwwstat.stanford.edu/~tibs/PAM/index.html) (16). PAM executes a sample classification training routine from expression data via the nearest shrunken centroid procedure to find markers that discriminate best between AIH patients from HD. Data were log-transformed, mean centered and standard deviaton scaled. After training of the PAM classifier, we performed a 10-fold cross validation, in order to check the accuracy of the model and better select the threshold, as the one giving the lowest misclassification error. To evaluate the performance of autoantigens combinations in discriminating AIH patients from healthy donors, logistic regression analysis was performed with R. We created logistic regression models with signals of 4 autoantigens (IL4R(FNIII), AL137145, C17orf99, APCDD1L) or with signals of two known AIH autoantigens as control (CYP2D6 and AGPR-1). The probabilities were calculated as follows: p= exp ((Σ(b i x i )+ c)/(1+Σ(b i x i )+ c), where p is the probability of each case, i= 1 to n; b is the regression coefficient of a given autoantigen, x is signal intensity and c is a constant generated by the model. ROCR package was used to obtain the ROC curves of the models and the Area Under Curve (AUC) values (17).

Results
Design and construction of a microarray with secreted or membrane associated human proteins -To identify self-antigens recognized by antibodies from patients with AIH, we developed a microarray by printing 1626 recombinant products (Supplementary Table 1-2) that corresponded to 1371 distinct human proteins ( Table 1). 1329 of the 1371 proteins were selected through bioinformatic analysis of the whole human genome as hypothetical and/or poorly characterized by the available annotation or published information and either carrying a signal peptide (23% of them) or at least one transmembrane domain (61% of them, 75% of which assigned to the plasmatic cell membrane and the remaining 25% assigned to mitochondrial or endoplasmic reticulum membranes) (12). Fortytwo of the 1371 proteins had a well known immunological function, CD number assigned and were all surface exposed (Supplementary Table 2).
Proteins were cloned as either full length products (31%) or protein domains (69%) with a N-terminal Histidine tag, expressed in E. coli, purified by affinity chromatography and checked for purity and integrity ( Supplementary Fig. 1). Protein arrays were prepared by printing onto nitrocellulose-covered glass slides four replicates of each protein along with several controls (Supplementary Fig. 2). Replicates were randomly distributed to get optimal signal reproducibility. The final protein microarray layout consisted of 24 grids of 304 spots each, for a total of 7296 spots. Quality of microarrays was assessed by probing 10% of the slides with an anti-His mAb and by determining the number of immobilized proteins with signal intensity significantly above background.
About 90% of proteins fulfilled this criterion (data not shown). Moreover, a high correlation among signal intensities of different slides was observed, indicating high experimental reproducibility (Supplementary Fig. 3B). A representative picture of the array is shown in Fig. 1A.
In summary, we obtained high quality microarrays comprising more than 1600 human proteins which can be used as tool for the identification of autoantigens recognized by sera of patients with any disease of interest.  Table 2.
First, we compared autoreactivity of patients sera against healthy donors sera. AIH sera displayed a higher reactivity toward self proteins than HD sera as documented both by the intensity (mean fluorescence intensity, MFI) of recognition signals (Fig. 1B) and by recognition frequencies (Fig. 1C). Autoantigens recognized by AIH sera were then ranked according to (i) the recognition frequency and (ii) the MFI. Self proteins were regarded as potential autoantigens if they were recognized with MFI significantly higher in patients than in healthy donors sera (T test, p val <0.01) and if they were recognized by less than 5% of the healthy donors sera and by more than 50% of patients sera (Fisher test, pval <0.01). In this way we identified 33 proteins that allowed good discrimination of the two populations of sera, as shown in the unsupervised hierarchical clustering analysis in Fig. 2A. We then asked whether sera from patients with HCV liver disease displayed the same autoreactivity pattern. We therefore tested the same microarray with sera from 110 patients with chronic HCV infection ( Table 2). Fig. 2B shows the MFIs of the 33 autoantigens with sera from AIH, HD and HCV, and indicates that 16/33 autoantigens react preferentially and significantly with sera from AIH patients (T test, p val <0.01).
In order to confirm the identities of the proteins, the 16 recombinant autoantigens of interest (Supplementary Table 3) were resolved by SDS-PAGE and the prevalent bands were excised from gels, digested with trypsin and subsequelly analyzed by MALDI-TOF mass spectrometry. All antigen identities were confirmed either by Peptide Mass Fingerprint or fragmentation of selected ions (details, Mascot scores and MS spectra are provided as supplemental data).
We therefore conclude that 16 autoantigens, identified by protein microarray, are differentially recognized by AIH patients compared to healthy donors and chronic HCV patients.

Validation of selected autoantigens with an independent sample set of sera confirms
six of the sixteen proteins are new potential AIH biomarkers -In order to validate protein microarray results with a different assay and larger panels of patients sera, we used Dissociation-Enhanced Lanthanide Fluorescence ImmunoAssay method (DELFIA ® ) to screen an independent Validation set of sera comprising 50 AIH patients, 50 healthy donors and 74 patients with chronic viral hepatitis (50 HCV and 24 HBV) ( Table 2). DELFIA ® assay was therefore used to assess the IgG response both to the 16 autoantigens that were selected with the Discovery set and to CYP2D6 and AGPR-1, two benchmark protein autoantigens in AIH (18,19).
All sixteen antigens displayed higher mean fluorescence intensity compared to HD and chronic viral hepatitis patients (Supplementary Fig. 4) and six of these sixteen antigens displayed also significantly higher recognition frequency by AIH patients than by healthy donors and viral hepatitis patients (Fig. 3). These six antigens, showed high sensitivity (from 42 to 70% of positive AIH patients) and specificity (from 96 to 100% of negative HD). Interestingly, individual sensitivity was comparable to that obtained in our assay with CYP2D6 and AGPR-1, two benchmark protein autoantigens in AIH, while individual specificity was higher for our six candidates (Table 3).
We then performed a classification of the validation samples with a nearest shrunken centroid algorithm (Predictive Analysis of Microarrays, PAM), which identified a minimal set of five predictors ( Supplementary Fig. 5), and classified AIH samples with an accuracy of 94%. Importantly, these classifiers corresponded to the top antigens previously sorted out on the basis of the sensitivity and specificity values, validating the selection of these proteins as best antigens for further characterization.
We next assessed the discrimination power of combinations of the autoantigens. Fig.   4A shows the seroreactivity of the validation sample set to all six autoantigens and reveals that two of the antigens (LOC646100 and METRNL) have reactivity that overlap with IL4R(FNIII) domain. We therefore assessed whether the combination of only four antigens (IL4R(FNIII), AL137145, C17orf99, APCDD1L) performed better than individual antigens. Fig. 4B shows that the 4 antigens combination achieved 82% sensitivity and 92% specificity, thus performing better than the individual antigens as well as combination of the two known autoantigens AGPR-1 and CYP2D6 (SE=68% and SP=68%). Moreover, Fig. 4C compares the ROC curves of logistic regression models obtained with combination of our four autoantigens and combination of the two known autoantigens, and indicates that our four autoantigen combo is superior to AGPR-1 and CYP2D6 combination.

IL4R is the target of autoantibodies capable of neutralizing IL4 activity -To address
the biological significance of the newly identified autoantigens, we checked the public databases for any information regarding their putative function. We found that five of the six autoantigens have poorly known function. Indeed, three of them (AL137145, LOC646100 and C17orf99) are completely uncharacterized secreted proteins; METRNL is a secreted protein annotated as Meteorin-like protein precursor and APCDD1L is a membrane protein annotated as Protein APCDD1-like precursor (Adenomatosis polyposis coli down-regulated 1 protein-like). One of the proteins is instead a domain of a well known membrane protein, as it corresponds to the fibronectin type III (FNIII) domain of the Interleukin 4 receptor α chain (IL4R or CD124), which is expressed in lymphocytes in association with cytokine receptor common gamma chain (Type I IL4 Receptor) as well as in hepatocytes in association with the Interleukin 13 receptor alpha1 chain (Type II IL4 Receptor) (20). Interestingly, as shown in Table 3, IL4R(FNIII) is the antigen displaying the highest individual sensitivity (70%) and specificity (100%) for AIH. Therefore we decided to focus our attention on the possible functional role of autoantibodies targeting IL4R.
Firstly, we confirmed antibody specificity by titrating sera from four different patients who displayed high reactivity for CD124 and found they were able to recognize the receptor domain up to a 1:800 dilution (Fig. 5A). Secondly, we asked how patients sera compared to commercially available neutralizing anti-CD124 goat polyclonal antibody in their ability to recognize both the human CD124 domain that we used in our protein array, and a recombinant form of the human CD124 expressed in insect cells. Fig. 5B shows that patients sera and the goat anti-CD124 antibodies both recognize the two forms of CD124 in Western blot. Finally, we assessed the capability of human sera to recognize the native IL4R when expressed on mammalian cell surface. We therefore overexpressed full length IL4R in HeLa cells and then incubated cells with sera of AIH patients and HD and with anti-CD124 antibody as control. Fig. 5C shows that sera of patients but not of healthy donors recognize native IL4R on HeLa cell surface.
As the FNIII domain of CD124 is involved in the interaction with IL4, we then asked whether these autoantibodies neutralized the interaction of IL4 with its receptor. As binding of IL4 to its receptor results in the specific phosphorylation of Stat6, we assessed by flow citometry whether patients sera inhibited IL4-mediated Stat6 phosphorylation. Fig. 6A and 6B show that, when cells expressing CD124 are pre-incubated with sera from AIH patients, but not with control sera, there is a dramatic reduction of the Stat6 phosphorylation that follows exposure to IL4. To confirm that the inhibition of Stat6 phosphorylation was an antibody-mediated mechanism, sera depleted of anti-IL4R (FNIII) antibodies were tested for their ability to block IL4 signaling. Fig. 6A shows that sera depleted of anti IL4R antibodies no longer inhibit IL4 induced Stat6 phosphorylation. Thus we have demonstrated that AIH patients sera neutralize IL4 in vitro because of anti-IL4R antibodies. Noteworthy, the inhibition of Stat6 phosphorylation correlated with the signal intensity of the anti-IL4R reactivity we detected in the patient sera by DELFIA ® (Supplementary Fig. 6), and it is dilution-dependent ( Supplementary   Fig. 7). To rule out that the neutralisation of IL4 activity observed with patients sera was due to a competitive effect of circulating soluble form of IL4R (sIL4R) (21), we tested the possible presence of sIL4R by ELISA assay in the sera of 20 AIH patients and 20 healthy donors. No significant differences of sIL4R were observed in the sera of patients and HD (Supplementary Fig. 8). Interestingly, sera from AIH patients under immunosuppressive therapy have a strong reduction of anti-CD124 antibodies as measured by both quantitative titer assessment with DELFIA ® (Fig. 6C) and qualitative neutralization of Stat6 phosphorylation (Fig. 6D).
From all the above we conclude that patients with AIH have autoantibodies to CD124 and that these antibodies neutralize IL4 signaling.

Discussion
This study illustrates the steps and outcome of a custom protein array approach to tackle new autoantigens in Autoimmune Hepatitis. Using an array of about 1600 human recombinant products, we report the identification of several human proteins recognized by autoantibodies that are present in sera of patients with AIH. We suggest that these autoantibodies may serve both for the improvement of diagnosis and for the development of new immunotherapeutic agents that could interfere with these autoantibodies. In particular, antibodies to IL4R, the autoantigen recognized with the highest sensitivity and specificity by AIH patients sera, inhibit IL4 signal transduction, demonstrating these autoantibodies are functional, suggesting a pathogenetic role for the inhibition of IL4 signaling and possibly opening new therapeutic perspectives for AIH.
In recent years microarrays have become precious tools for biomedical research as they are very suitable to screen great numbers of samples, with very low amount of biological material, in a very short time (22). A limitation of protein microarrays is that, due to labor intensive protein production processes, they often cover only defined protein families with known relevance to a given scientific question, although important efforts to overcome this limitation have been recently done and arrays with thousands of proteins are now available (11,23). In this study we printed our custom array with a relatively small group of proteins that are known to play key roles in generation and differentiation of immune responses in health and disease, in combination to a functionally unbiased expression library. Our custom protein microarray is made of a library of more than 1600 recombinant products that were expressed in E. coli because of the costs and of the production and purification issues related to the handling of thousands of proteins. Expression of human proteins in bacteria is not ideal for functional studies, because post-translational modifications are generally lost and because proteins are mostly recovered in denaturing conditions. However, we aimed at identifying polyclonal antibodies specific for linear epitopes of human proteins, as it has been reported that linear epitopes are often recognized by autoantibodies in many autoimmune diseases (24,25).
The proteins printed in our microarray are recombinant products that are either membrane-associated or secreted proteins, the great majority of which are poorly characterized on the basis of both current annotation in public databases and scientific publications (12). The rationale for the use of this specific protein subset was that we were interested in a functionally unbiased search of new autoantigens among thousands of proteins; on the other hand, we wanted to focus on a functionally well known small group (40) of cell bound proteins that are exposed to the extracellular environment and that play crucial functions in regulating immune responses in health and diseases. It is worth noting that although membrane-associated and secreted proteins play a crucial role in many cell recognition and communication processes (26), most of the autoantigens identified so far in autoimmune diseases are intracellular components (11,27).
We used this protein microarray to screen a panel of sera from patients with AIH, a disease for which there is an unmet medical need for both new and highly specific biomarkers and for more specific therapies. Indeed, the diagnosis of AIH is a complex process made by exclusion of other factors leading to chronic hepatitis (including viral, toxic, genetic and metabolic causes) and by detecting autoantibodies in indirect immunofluorescence assays on tissue sections from rodents. However, this technique has several limitations, the major of which is the strong dependence on the operator expertise (6). Moreover, serological overlap of AIH with other liver diseases (such as chronic viral hepatitis or drug-induced hepatitis) (10) is frequently observed. The distinction between autoimmune and viral hepatitis is also important from a therapeutic point of view, since the immunosuppression used in AIH can increase virus replication in chronic viral hepatitis, while treatments used to eradicate viral infection, such as interferon-alpha, may lead to exacerbation of AIH (10).
Here we show that sera from patients with AIH recognize with high specificity and sensitivity five self proteins of unknown function and one known self protein (IL4R or CD124). This recognition pattern was first detected by protein array on fiftteen patients sera and subsequently confirmed in 96-well plate assay with sera from fifty patients with AIH, whereas sera from healthy donors and from patients with chronic viral hepatitis did not show significant recognition of these antigens. This autoreactivity pattern was mainly observed in type 1 AIH. Indeed, the great majority of AIH sera tested (100% of sera used in the discovery phase and 86% of sera used in the validation phase) were type 1 AIH, as reported in Table 2.
Interestingly, while individual performance of our new autoantigens for AIH was comparable to that we obtained with CYP2D6 and AGPR-1, two benchmark protein autoantigens, combination of four of the six autoantigens was superior to sensitivity and specificity of the benchmark autoantigens alone or in combination. Indeed we achieved an accuracy of 87% (SE= 82% and SP%= 92%) in discriminating AIH patients from healthy donors with the new antigens compared to the 68% accuracy of the combined benchmarks (SE=68%, SP=68%). Moreover, we achieved an accuracy of 68% (SE= 82% and SP=61%) in discriminating AIH from viral hepatitis patients compared to the 46% accuracy of the combined benchmarks (SE= 68%, SP= 31%) (data not shown).
Finally, compared to the work of Song and colleagues (11), who recently reported the identification by protein array of 3 new AIH autoantigens with similar performances, we used an ELISA-like approach to validate our candidates, that could be easily translated into the standard laboratory practice. Therefore, we set the stage for the development of a new serological assay that is easy to perform, is highly specific for AIH, and that could significantly contribute to improve AIH diagnosis.
Of the five proteins with unknown function, one is a membrane protein (APCDD1L) and four are secreted proteins (AL137145, LOC646100, C17orf99, METRNL). The one known autoantigen (IL4R(FNIII)) is the Fibronectin type III domain of the alpha chain of the IL4 Receptor (IL4R or CD124). Obviously we concentrated our functional investigation on autoantibodies targeting this well known receptor, but studies are in progress to address the structure and function of the other five poorly known autoantigens.
IL4 is a cytokine mainly produced by CD4+ Th2 lymphoctes, basophils, mast cells and eosinophils (20). It plays a key role in several aspects of lymphocytes differentiation and function and has been described to be involved in many autoimmune as well as inflammatory diseases (28). In particular, IL4 is reported to inhibit Th1 and Th17 differentiation, which are T cell subsets implicated in many autoimmune diseases including Autoimmune hepatitis (29,30). Altered IL4 expression has been reported in several liver diseases including chronic hepatitis C, drug induced hepatitis and liver transplantation (31). However its exact pathogenetic role in these diseases is still controversial: for some authors it has a protective effect (32,33) while others have reported that higher IL4 expression in the liver is detrimental, causing hepatocytes apoptosis (34,35). IL4 exerts its action by binding its receptor (CD124), which is present on many cell types including lymphocytes and hepatocytes, (20) and activating specific signaling cascades. The IL4 Receptor consists of a signaling alpha chain that binds IL4 with high affinity and a trans-activating low affinity chain which can be, according to the cell type, the common gamma chain (immune system cells) or the IL13 receptor alpha1 chain (epithelial cells). After IL4 binding to both immune and epithelial cells, the two chains form a heterodimer and initiate a phosphorylation cascade, which ends up in the specific activation of Stat6. Upon phosphorylation on Tyr641, Stat6 translocates to the nucleus and in turn activates the transcription of specific genes (36).
Interestingly, the FNIII domain of the CD124 alpha chain, which we found to be recognized with very high frequency (70%) by AIH patients sera, contains two binding sites for IL4 (37). Indeed, our data show that not only AIH patients sera recognize IL4R, but also IL4-mediated Stat6 phospohrylation is inhibited in vitro by AIH patients sera and not by HD or HCV sera. This finding points to the importance of IL4 pathway in autoimmune hepatitis and to the pathogenetic role that autoantibodies against the IL4 Receptor may play in AIH. Such antibodies might neutralize IL4 activity directly on hepatocytes and thus intefere with a potential anti-inflammatory role of STAT6, which has been put forward to explain the IL4 ability to reduce hepatic ischemia/reperfusion injury (32). Autoantibodies that neutralize IL4 activity might favour liver inflammation also indirectly by favouring development in lymphoid tissues and recruitment to the liver of T cell subsets, such as Th1 and Th17, involved in autoimmunity and inflammation (30).
Indeed, IL4 has been shown, in experimental and clinical situations, to be capable of ameliorating the effects of tissue-damaging autoimmunity (20). On the other hand, based on data published in the literature (34,35), one could also theorize that IL4 is detrimental in AIH and that neutralising anti-CD124 antibodies reflected an attempt of our immune system to buffer a negative role of IL4. Should this be true however, immunosuppressive therapy could be detrimental, as sera of patients after steroids treatment display a strong reduction of anti-CD124 antibodies as measured by both DELFIA ® and neutralization of IL4 activity (Fig. 6).
In conclusion, we describe the identification of a new panel of six autoantigens that are very specific and sensitive biomarkers of AIH. Compared to other similar approaches (11) our strategy based on a protein array enriched in "external" proteins, i.e. proteins physiologically more exposed to the immune system, allowed us to identify autoantigens which not only can be used as diagnostic biomarkers, alone or in combination, but could also give insights into some of the pathogenetic mechanisms involved in this autoimmune disease. Indeed, our results demonstrate that autoantibodies to CD124 (IL4R) have a neutralising effect on IL4 activity and suggest that these antibodies could have a pathogenetic role by favoring an inflammatory mileu leading to liver damage.