Potential conflict of interest: Nothing to report.
See Editorial on Page 786
Autoimmune hepatitis (AIH) is a liver disease with circulating autoantibodies predominantly directed against widely held cellular components. Because AIH is a liver-specific disease, autoantibodies against plasma membrane antigens may be involved in its pathogenesis and have been reported; however, no definite identification has been described. We thus investigated the fine specificity of anti-hepatocyte plasma membrane autoantibodies in type 1 AIH (AIH-1) using a proteomic tool. A plasma membrane–enriched fraction was validated using enzymatic activity and western blot analysis experiments. Sera from AIH-1 patients (n = 65) and from 90 controls, that is, healthy blood donors (n = 40) and patients with systemic diseases (n = 20) or other liver diseases (n = 30), were studied by immunoblot performed with plasma membrane proteins resolved by either sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) or 2-dimensional (2D) electrophoresis. Proteins contained in the immunoreactive spots were identified by sequences provided by ion-trap mass spectrometry. Hepatocytes probed with sera were also studied using confocal immunofluorescence and immunoelectron microscopy. The more prominent bands stained by patient sera were located at 38 kDa, 48, 50, 52 kDa, 62 kDa, 70 kDa, and a 95-kDa double band. Six proteins with known potential plasma membrane expression were identified: liver arginase (38 kDa), cytokeratins (CK) 8 and 18 (48-52 kDa), heat shock proteins (HSP) of 60, 70, 90 kDa, and valosin-containing protein (VCP) of 92 kDa. The presence of anti-membrane antibodies was confirmed by immunofluorescence and immunoelectron microscopy. Conclusion: Overall, our data demonstrate that liver arginase, CK 8/18, HSP 60, HSP 70, HSP 90, and VCP represent potential candidate targets on liver membrane for autoantibodies in AIH-1. (HEPATOLOGY 2008;47:937–948.)
Autoimmune hepatitis (AIH) is an inflammatory disease characterized by circulating autoantibodies and histological features such as lymphoplasmocytic cell infiltrates. AIH type 1 (AIH-1) is defined by the presence in patient sera of autoantibodies directed against nuclear and/or smooth muscle antigens (especially actin filaments), and/or soluble liver antigen, whereas type 2 is characterized by antibodies to liver-kidney microsome type 1 (LKM1) and/or liver cytosol type 1.1 The pathogenesis of AIH is not fully understood, but it is an organ-specific autoimmune disease, suggesting that the target of immune destruction could be molecules expressed on the membrane of hepatocytes.2 Circumstantial evidence has suggested that liver destruction does not only implicate T-cell cytotoxicity but also involves antigens on the hepatocyte surface as targets in a reaction of antibody-dependent cellular toxicity.2–4 The membrane expression of CYP2D6 (which constitutes the microsomal molecular target of anti-LKM1 antibodies) has been well documented, but anti-LKM1 autoantibodies are only detected in sera from patients with AIH-2.5 Asialoglycoprotein receptor (ASGPR) or sulfated glycosaminoglycan may constitute attractive candidates, as they have been reported as targets for antimembrane antibodies in AIH-1. Antibodies against ASGPR or sulfatide have been found in diseases other than AIH, and their responsibility in hepatocyte destruction has not been established.6, 7 Although autoantibodies reacting with the surface of intact hepatocytes have been described previously using various techniques such as indirect immunofluorescence and immunoblotting assays, the specificity of these anti-plasma membrane antibodies has not as yet been clearly characterized.6, 8–11
Immunoscreening of a complementary DNA (cDNA) bank is one of the principal methods used to elucidate the molecular targets of autoantibodies. However, gene regulation may differ between the in vivo situation and the expression vector.12 Proteomic analysis, combining 2-dimensional gel electrophoresis for protein separation and then mass spectrometry (MS) for their identification, focuses on gene products.12, 13 Moreover, it is difficult to predict the localization of a gene product, unlike proteomic analyses performed on cellular compartments. In terms of autoimmunity, proteomic analysis has been performed to determine autoantigens in a variety of autoimmune diseases, including AIH.14–16
This study was designed to identify the plasma membrane autoantigen reacting with the autoantibodies present in sera from AIH-1 patients, using a proteomic approach.
We selected 65 sera from patients with well-documented AIH and fulfilling the diagnostic criteria defined by the International Autoimmune Hepatitis Group.17 Table 1 shows their clinical and serological features. Control sera (n = 50) were obtained from patients with chronic liver diseases [chronic hepatitis due to hepatitis C viral infection without anti-LKM1 antibodies (n = 10) and alcoholic liver disease (n = 10)] and from patients with systemic autoimmune diseases [systemic lupus erythematosus (n = 10) and rheumatoid arthritis (n = 10)]. Normal control sera were obtained from healthy blood donors (n = 10).
Table 1. Clinical and Serological Features of Patients with Type 1 Autoimmune Hepatitis Tested for Antimembrane Antibodies
All chemical reagents used were obtained from Sigma-Aldrich (St Quentin Fallavier, France), unless otherwise stated.
Plasma membrane fraction from the livers of Wistar male rats (Janvier breeding facility, le Genest Saint-Isle, France) was obtained as described elsewhere.18 Briefly, low-speed centrifugation separated the nucleus and larger structures in membrane sheets from vesicular and soluble elements. The plasma membrane fraction was then obtained after sucrose density gradient centrifugation, collected at the interface of 1.42 M and 0.25 M sucrose, and enriched by final differential centrifugation. Other fractions, that is, nucleus, mitochondria, microsome, and the soluble phase of the cell, were obtained as described elsewhere.15
One-Dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis Electrophoresis.
For sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), 100 μg protein per centimeter of gel was resolved in a Protean II slab gel vertical electrophoresis unit (Bio-Rad, Hercules, CA) with 4% polyacrylamide stacking gel and 10% resolving gel.19
The membrane protein fraction (1 mg) was diluted in DeStreak® solution (Amersham Biosciences, Uppsala, Sweden) containing urea, thiourea, CHAPS, and disulfide bridge reducing agent, supplemented with 0.5% (vol/vol) immobilized pH gradient buffer pH 3-10 (Amersham Biosciences). Samples were applied by in-gel rehydration for 16 hours. Isoelectric focusing was performed using the Multiphor II system (Amersham Biosciences) and 18-cm ready-made immobilized pH gradient strips, Immobiline DryStrip (Amersham Biosciences).20 Focusing was initiated at 50 V and the gradient increased to 3 kV to reach a total of 50 kVh. The strips were then equilibrated with a first buffer containing 6 M urea, 0.5 M Tris-HCl pH 6.8, 30% (vol/vol) glycerol, 2% (wt/vol) SDS, and 0.5% (vol/vol) dithiothreitol, and after 15 minutes with a second buffer containing the same chemical reagents except for dithiothreitol, which was replaced by iodoacetamide 4.5% (wt/vol). SDS-PAGE separation was carried out as described previously. After separation, proteins were either stained with colloidal Coomassie brilliant blue or transferred to a nitrocellulose membrane. The gels used for immunoblotting were silver-stained after transfer (PlusOne silver staining reagent kit from Amersham Biosciences).
Proteins resolved by 1-dimensional and 2-dimensional electrophoresis were blotted onto nitrocellulose membranes (Bio-Rad), in a transfer buffer containing 25 mM Tris, 192 mM glycine supplemented with 20% (vol/vol) methanol, using a Trans-Blot cell (Bio-Rad) and applying 60 V for 2 hours.21 Membranes were blocked in 5% (wt/vol) skimmed milk in phosphate-buffered saline for 1 hour and probed overnight at 4°C with either human sera diluted 1:100, monoclonal or polyclonal antibodies to liver arginase (rabbit, dilution 1:100), to cytokeratins (CK) 8/18 (guinea pig, dilution 1:400), to heat shock protein (HSP) of 60 kDa (HSP 60) 70 kDa (HSP 70) or 90 kDa (HSP 90) (rabbit, dilution 1:100), or to valosin-containing protein (VCP) (rabbit, dilution 1:100). These were obtained from SantaCruz Biotech, CA, with the exception of anti–CK 8/18 (Progen, Heidelberg, Germany) and anti-VCP (Cell Signaling Technology, Boston, MA).
During other experiments, sera diluted 1:100 were incubated with nitrocellulose filters blotted with 5 μg recombinant liver arginase, HSP 60, HSP 70, HSP 90, CK 8, and CK 18 (Alexis Biochemicals, San Diego, CA).
Next, after washing with phosphate-buffered saline 0.1% (vol/vol) Tween 20, the membranes were incubated with appropriate antibodies conjugated with horseradish peroxidase and diluted at 1:200, that is, anti–human immunoglobulin G (BioRad), anti–rabbit immunoglobulin G (Amersham Biosciences), anti–guinea pig Ig (Santa Cruz, Biotech). After 1.5 hours at room temperature, the visualization of peroxidase was performed with 0.06% (wt/vol) 4-chloro1-naphtol (Merck, Whitehouse Station, NJ) diluted in 200 mM methanol containing 0.033 mM H2O2.
Verification of the Enrichment of the Plasma Membrane Fraction.
5′-Nucleotidase enzymatic activity was used to test enrichment of the plasma membrane fraction compared with the starting homogenate.18 The abundance in the plasma membrane fraction of proteins derived from cellular organelles was assayed by western blot comparative analysis. Immunoblots performed with 100 μg protein per centimeter of gel of the starting homogenate, plasma membrane, and also mitochondria, microsome, and the soluble fraction of the cell, were probed with mouse antibodies specific to cytochrome C, an enzyme located in the mitochondrial intermembrane space (final dilution, 3 μg/mL) (Calbiochem, Merck Biosciences Darmstadt, Germany), with rabbit antibodies to calnexin, an endoplasmic reticulum protein (final dilution, 1:2000) (Calbiochem), with mouse antibodies to hnRNP A2/B1 localized in the nucleus, diluted to 1:2000 (a gift from Dr Praseuth, Laboratoire de Biophysique, Institut National de la Santé et de la Recherche Médicale U201, CNRS UMR 8646, Museum National d'Histoire Naturelle, Paris), and a human reference serum to cytosol histidyl t-RNA synthetase diluted 1:100 (Laboratoire d'immunologie, Hôpital Saint-Antoine, Paris).
Peptide Preparation Procedure.
Colloidal Coomassie blue–stained protein maps, silver-stained transferred 2-dimensional gels, and nitrocellulose membranes were scanned and superimposed using Adobe Photoshop® software to match immunoreactive spots on the immunoblots to the silver-stained gel and blue-stained map, where spots of interest were excised for in-gel digestion with trypsin for 16 hours at 37°C. The resulting peptides were desalted with ZipTip® C18 (Millipore).15, 22
LC-ESI-MS/MS and Data Analysis.
LC-MS/MS analysis was performed using an ESI ion trap mass spectrometer (LCQ Deca XP, ThermoElectron, San Jose, CA) coupled online with a capillary nano–high performance liquid chromatography system, that is, a PepMap C18 reverse phase (LC Packings, Dionex Amsterdam, Netherlands). Data were initially acquired over an m/z (mass-to-charge) range of 400 to 2000 Da to select the 3 most intense ions for further fragmentation (MS/MS scan). Proteins were identified by correlation between MS/MS spectra and rodent protein sequences present in the National Center for Biotechnology Information nonredundant protein sequence database, using the SEQUEST algorithm incorporated in Finnigan Bioworks 3.1 software. SEQUEST search results were initially assessed by examination of the cross-correlation and delta normalized correlation scores. As a general rule, a cross-correlation value higher than 1.5, 2.0, and 2.5, for 1+, 2+, 3+ charged peptides, respectively, and a delta normalized correlation value higher than 0.1, were accepted as a positive identification.23
Antimembrane antibodies from patient sera were depleted by incubating sera diluted at 1:100, for 12 hours, with 50 μg recombinant liver arginase, HSP 60, HSP 70, and HSP 90, CK 8, and CK 18 blotted onto nitrocellulose filters. Bound antibodies were eluted by a glycine buffer 0.05 M containing NaCl 0.5 M, Tween® 20 0.5%, and bovine serum albumin 0.01%. Absorption was repeated 3 times, and the sera were then tested on immunoblots performed with the plasma membrane fraction.
Immunofluorescence and Immunoelectron Microscopy.
To show that some of the plasma membrane proteins were autoantigens, we exposed nonpermeabilized hepatocytes (HEPG2 line) to AIH-1 sera. Hepatocyte viability was first assayed by trypan blue exclusion.
Hepatocytes were fixed for 20 minutes with 4% buffered paraformaldehyde, incubated for 30 minutes with 2% bovine serum albumin, and then with AIH-1 sera or controls (diluted 1:160) for 1 hour. Antibody binding was detected after 1 hour incubation at room temperature with an anti-human antibody conjugated with fluorescein (1:200) (Biorad). Cells were imaged using a Leica TCS SP2 confocal microscope equipped with an oil immersion plane apochromat X63 objective (numeral aperture, 1.32). To rule out any artifacts caused by the fixation process, some hepatocytes were cultured on a slide and directly exposed to sera at 4°C, and paraformaldehyde fixation was performed after incubation with the fluorescein-conjugated antibody.
For electronic microscopy, hepatocytes were fixed and incubated with sera, as described previously.
After blocking endogenous peroxidase with 3% H2O2, antibody binding was detected by incubation with horseradish peroxidase–labeled polyclonal anti-human immunoglobulin G (1:200) followed by diaminobenzidine chromogen solution (Dako). The cells were then fixed in 2.5% glutaraldehyde and postfixed in 2% osmium tetroxide, dehydrated in graded ethanol solution, and embedded in LX-112 resin. Ultrathin sections were prepared and examined under a JEOL jem-100S electron microscope. A monoclonal antibody against tetraspanin CD9, a plasma membrane protein (a gift from Dr. Le Naour, INSERM U602, France), served as a positive control for both immunofluorescence and immunoperoxidase.
Qualitative data were compared using Statview statistical software, the chi-squared test being applied with the Yates' correction if necessary. The minimum level of significance was set at P < 0.05.
Immunoblotting Patterns with Plasma Membrane Proteins Resolved by SDS-PAGE.
Figure 1 shows the reactivity to plasma membrane proteins of both patient and control sera. Sera from AIH-1 patients reacted more strongly, and significantly more frequently, than that in controls with several proteins. Twelve percent (8/65) of patients with type 1 AIH showed no reactivity with plasma membrane fraction. AIH sera without any reactivity on the immunoblot exhibited a similar immunofluorescence pattern with respect to antinuclear antibody, antismooth muscle antibodies, and anti–soluble liver antigen antibodies as AIH sera that were immunoblot-positive. A 38-kDa protein was the target most frequently recognized by 47 of the 65 AIH-1 sera (72%), compared with 21 of the 50 controls (41%) (P < 0.001). Band cluster at 48, 50, 52 kDa was stained by 27 (41%), 13 (20%), and 19 (29%) of the 65 AIH-1 sera, respectively, compared with 6 (12%) (P < 0.002), 0 (P < 0.007), and 5 (10%) (P < 0.02) of the 50 controls. A 62-kDa protein was recognized by 27 (41%) of the 65 AIH-1 sera compared with 4 (7%) of the 50 controls (P < 0.0001). Twelve (18%) of the 65 AIH-1 sera stained a 70-kDa protein versus 1 (2.5%) of the 50 controls (P < 0.04); and 13 (20%), a 95-kDa double band versus no controls (P < 0.006). Other autoantibody protein targets were also found in both AIH sera and controls, but the differences in frequencies were not significant. Analysis of the reactivity of different control groups showed that sera from patients with alcoholic liver disease, and particularly with hepatitis C virus infection, stained the 38-kDa band at higher frequencies (5/10 and 8/10, respectively). Reactivity to the 48-kDa and 52-kDa protein was also found with sera from patients with connective tissue diseases (3 of 10 for both systemic lupus erythematosus and rheumatoid arthritis with respect to the 48-kDa stained band, and 2 of 10 systemic lupus erythematosus and 3 of 10 rheumatoid arthritis for the 52-kDa band) (Table 2).
Table 2. Analyses of the Reactivity of Different Control Groups Compared With Patients With Autoimmune Hepatitis
Colloidal Coomassie blue–stained 2-dimensional patterns of the rat plasma membrane fraction are shown in Fig. 2. Seven sera from AIH patients were analyzed on 2-dimensional immunoblots. Representative patterns of AIH-1 sera on the 2-dimensional immunoblotting performed with plasma membrane proteins are also shown in Fig. 2.
MKIAA0098, other alias, chaperonin subunit 5 (mouse)
MKIAA0098, other alias, chaperonin subunit 5 (mouse)
Heat shock protein 60 (Rattus)
16, 17, 19
Heat shock protein 60
11, 2, 13
21, 5, 33
Similar to stress 70 protein (predicted) (Rattus)
Heat shock protein 70 (Mouse)
Unnamed protein product (Mouse) 99% identity with Valosin-Containing Protein
Heat shock protein 90 (Mouse)
18, 23, 24
The significantly stained band at 38 kDa on the 1-dimensional immunoblot corresponded on 2-dimensional immunoblots to 3 well-resolved spots between experimental pI 7.3 and pI 8.0. The 48-kDa to 52-kDa cluster corresponded to a smear of spots focusing between pI 5.1 and pI 5.7 for the 48-kDa to 50-kDa stained band, and to 4 immunoreactive spots focusing around pI 6.0 for the 52-kDa molecular weight. Other bands significantly stained on the 1-dimensional immunoblots corresponded on 2-dimensional immunoblots to several contiguous spots: around experimental pI 5.5 for 62-kDa targets, to 3 spots between pI 5.4 and pI 5.6 for the 70-kDa stained band, and to 2 immunoreactive groups for the 95-kDa double band, a spot at pI 4.5 and a smear of spots around pI 5.2.
Identification of the Protein Recognized by Autoantibodies.
Immunoreactive spots on an immunoblot were superimposed on the 2-dimensional protein profile on both the silver-stained transferred 2-dimensional gel and Coomassie blue–stained gel. Spots of interest were cut from this latter gel, and proteins were in-gel digested with trypsin. The resulting peptides were analyzed using ion trap mass spectrometry, which allows protein identification from fragmented peptides. The sequences deduced from parent ion fragmentation spectra were compared with the sequences of proteins present in the National Center for Biotechnology Information protein sequence database. The results are summarized in Table 3. Eight proteins were identified from the 26 spots, whereas 3 were not. The combination of cross-correlation and delta normalized correlation scores with the SEQUEST algorithm, the excellent correlation between observed and theoretical molecular masses and pI values, the number of matching peptides, and the percentage coverage between the peptide sequences obtained and the candidate targets led to clear identifications. Liver arginase corresponded to the three 38-kDa immunoreactive spots. Cytokeratins 8 and 18 were obtained from all immunoreactive spots in the 48-kDa to 52-kDa cluster. Spots at 62 kDa corresponded to HSP 60 and to chaperonin subunit 5, which is implicated in protein folding. Spots at 70 kDa corresponded to HSP 70. In the double band at 95 kDa, the more basic spot corresponded to HSP 90 and the acidic spot to a protein with 99% homology (using BLAST [basic local alignment search tool] research) with a VCP.
Comparative Immunoblotting Patterns of AIH Sera and Polyclonal Antibodies to Targeted Proteins.
The reactivity patterns of AIH-1 sera were compared with anti-liver arginase, anti-HSP60, anti-HSP70, anti-HSP90, and anti-VCP monoclonal or polyclonal antibodies using the plasma membrane fraction as the antigen. The polyclonal antibodies stained proteins with molecular masses that were the same when compared with those recognized by sera from patients with AIH-1 (Fig. 3).
Patient sera absorbed after incubation with recombinant liver arginase, or with HSP 60, HSP 70, HSP 90, CK 8, and CK 18 and blotted onto nitrocellulose filters, exhibited a reduction in the intensity of the corresponding band when tested on the plasma membrane fraction (Fig. 4). The specificities of antimembrane antibodies for the putative antigens were also proved by the fact that AIH sera stained the corresponding recombinant proteins blotted onto nitrocellulose filters (Fig. 4).
Associated Membrane Expression of Immunoreactive Proteins.
Significant membrane purification was deduced from the 5′-nucleotidase activity, which was 36-fold enriched in the plasma membrane fraction compared with the starting homogenate (specific activity of 19.10−1 mMol/minute in the plasma membrane fraction and 53.10−3 mMol/minute in the homogenate). To further validate the localization of immunoreactive proteins found with the cell plasma membrane, we compared the immunoreactive patterns of the starting homogenate, plasma membrane fraction, nucleus, mitochondria, microsomes, and cytosol probed with antibodies directed to proteins well localized in the cell. Cytochrome C specific to the mitochondria, hnRNPA2/B1 to the nucleus, and histidyl t-RNA synthetase in the cytosol were not detected in the plasma membrane–enriched fraction, indicating that contamination of the plasma membrane fraction by these organelles was below the limit of detection (500 pg) for the 4-chloro 1-naphtol thus used (Fig. 5). A very weak signal was detected for calnexin at the limit of detection.
The associated membrane expression of immunoreactive proteins was judged by immunostaining with confocal and electron microscopy. Nonpermeabilized hepatocytes from the HEPG2 line were exposed to sera and probed with fluorescent-conjugate antibody. Paraformaldehyde fixation was performed either before sera incubation or after conjugate antibody incubation. Granular staining of the cell circumference was observed after exposure to the AIH-1 sera, and a similar pattern developed after incubation with the monoclonal antibody against the plasma membrane protein, tetraspanin CD9 (Fig. 6). A discontinuous granular pattern with capping phenomena was observed with some cells. No labeling was detected after incubation with blot-negative control sera from blood donors.
Immunoelectron microscopy revealed clear labeling of the plasma membrane itself. Indeed, hepatocytes fixed with paraformaldehyde and probed by immunoperoxidase assays exhibited discontinuous peroxidase labeling of the plasma membrane after incubation with AIH-1 sera (Fig. 7). No staining was noted after exposure with blot-negative controls.
We were able to detect several types of reactivity by immunoblotting AIH-1 sera on liver cell plasma membrane proteins. These results were in line with those obtained by several other authors. Given the margin of error inherent in reading a molecular weight on a 10% SDS-PAGE, the molecular weights of some stained bands were similar to those described by Matsuo et al.,11 who reported prominent proteins at 81 kDa and 49 kDa, whereas we observed stained bands at 95 kDa and a cluster at 48 kDa to 52 kDa. We also reported a 62-kDa antigenic protein, whereas Swanson et al.9 mentioned a 60-kDa protein as the dominant stained band. In addition to the 95 kDa, 62 kDa, and 48, 50, 52 kDa bands, the large number of AIH-1 sera we used (n = 65) [compared with the number included in Swanson et al.'s study9 (n = 12)] made it possible to characterize 2 other bands at 38 kDa and 70 kDa as being recognized at statistically significant higher frequencies by AIH-1 sera when compared with controls. Nevertheless, fewer reactive antigenic proteins were detected during our study than by Matsuo et al.,11 who used immunoblotting with enhanced chemoluminescence, which is a more sensitive method than the 4-chloro 1-naphtol technique we employed.
In Matuso et al.'s study,11 2 other prominent bands at 136 kDa and 116 kDa were characterized; we were unable to focus on these 2 high molecular weight protein bands because of the weak (10%) percentage of polyacrylamide gel used. However, as Mastsuo et al.11 and Swanson et al.9 had reported, we could not identify the 26-kDa antigen described by Hopf.10
Using mass spectrometry, we were able to identify 8 putative targets for AIH-1 sera on liver cell plasma membranes. The target most frequently recognized by our AIH-1 sera corresponded to liver arginase with a molecular weight of 38 kDa. Patients with hepatitis C virus infection or alcoholic liver disease displayed a high frequency of antibodies to liver arginase (80% and 50%, respectively), thus confirming that levels of anti-liver arginase are increased in a context of liver disease.24 The 2 liver-specific CK, 8 and 18, were identified with prominent staining at 48 to 52 kDa. Patients with connective tissue diseases also displayed antibodies to CK (Table 2), as reported earlier.25 A third group identified as targets for AIH-1 sera belonged to the chaperonin protein family, that is, HSP 60, HSP 70, and HSP 90. This is the first report demonstrating that chaperonin subunit 5 is an autoantigenic protein. These proteins are produced under stress but also under normal, nonstressful conditions, and have numerous properties.26 The last target identified was VCP, an adenosine triphosphatase associated with a variety of cellular activities.27 Interestingly, VCP has been shown to occur in heterocomplexes containing HSP 90.28
Arginase, CK, VCP, and HSP (except HSP 60, which is mainly present in mitochondria) are known to be localized in the cytosolic compartment. VCP is described as being active in the cytosolic surface of the endoplasmic reticulum.27 A nuclear localization has also been demonstrated for HSP 70 and HSP 90. Probable contamination of the plasma membrane fraction by other organelles or cytosolic proteins should be considered, but a possible technique-dependent redistribution of cytoplasmic proteins to the plasma membrane during the cell fractionation procedure is improbable. Other authors9, 11 determined proteins with the same molecular weight as some of the targets we found, using for antigens on immunoblots plasma membrane proteins that had been obtained using aqueous 2-phase partition, a technique differing from that used in the current study, thus suggesting that the results were not technique dependent. Enzymatic assays indicated that the plasma membrane preparation we used was 36-fold purified, a rate similar to that described by Hubbard et al.18 The comparative immunoblotting pattern experiments performed with antibodies to proteins well assigned to a cellular compartment suggested that potential contamination by nuclear, mitochondrial, microsomal, or cytosolic proteins was below the limit of detection of the method employed.
Most of the proteins we identified are not known to have a cell surface localization or to possess a transmembrane domain or signal peptide sequence in their genomic structure. Surprising plasma membrane localizations had previously been reported by several proteomic experiments focused on HSP, VCP, and CK.29–31 Protein secretion independent of the endoplasmic reticulum–Golgi pathway has been suggested by an accumulation of evidence32 and is currently a matter of debate,29, 33 for example, exosome-dependant trafficking,32 HSP assistance to maintain membrane receptor integrity by accompanying the protein from its cytosolic site of synthesis, or the transport of improperly folded proteins, allowing the cell to remove an excess of improperly folded cytosolic proteins.34, 35 Experiments not using a proteomic approach have demonstrated an extracellular surface expression on liver membrane of CK18, with a function of a thrombin receptor.36 CK8 has also been demonstrated as having a possible cell surface location on certain epithelial cells, including hepatocytes.37 HSP 60 has been reported in plasma membrane preparations of human cell lines and primary cultured hepatocytes.38 HSP 70 and HSP 90 have also been detected on different epithelial cell surfaces,39, 40 demonstrated to constitute a receptor complex for Dengue virus,41 and to form a cluster with CD14 or TLR4 within the lipid microdomain.42 With respect to liver arginase, localization on the surface of the plasma membrane of rat hepatocytes was observed under immunocytochemical electron microscopic examination.43
However, we did not know whether the targets we found were inside or outside the plasma membrane. To further exclude any artifact, we confirmed the existence in AIH-1 sera of antibodies to the plasma membrane by immunoelectronic and confocal microscopy, which provided the membrane staining pattern of nonpermeabilized hepatocytes probed with test sera. These observations led us to think that some autoantibodies present in AIH-1 patient sera may be able to recognize proteins localized on the outer surface of the plasma membrane, thus confirming results of previous studies.6
We failed to identify anti-ASGPR antibodies as plasma membrane targets, as has been reported elsewhere.11 Actin may be found in plasma membrane preparations,44 and anti-actin antibodies were reported by Swanson et al.9 Because of the detergent properties of the SDS added to the electrophoresis preparation, conformational epitopes could not be detected, although anti-ASGPR and anti-actin cable antibodies are mainly directed against nondenatured conformational structures.45, 46
Interestingly, some of the targets reported in our study also have been described in autoimmune diseases. Anti-liver arginase antibodies were detected in 20% to 30% of AIH,24 and anti-CK8/18 are present in 30% of these patients.47 HSPs are well known to constitute targets in various autoimmune diseases, including AIH, in which anti-HSP 70 has been shown to be present in 52% of patients.48 Antibodies to a 95-kDa protein assimilated to VCP were present in 9.7% of AIH sera and 12% of primary biliary cirrhosis sera.49
These autoantibodies were also present in both disease-free controls and in sera from patients with autoimmune diseases other than AIH, albeit at lower frequencies and intensities. This previously reported observation9, 11 raises questions as to the role of natural autoantibodies in the pathogenesis of AIH. Natural autoantibodies belong to the physiological autoimmune system and can participate in controlling autoreactivity through, for example, the idiotypic network.50 A dysregulation for reasons as yet not elucidated may trigger an autoimmune disease, which in such a case appears to constitute an aberration of a normal phenomenon.51 This is consistent with the high levels of immunoglobulin found in type 1 AIH sera.
Overall, these results, obtained using a proteomics technique, demonstrate that several potential autoantigens, such as liver arginase, CK, and chaperon proteins, were located within the plasma membrane and acted as autoantigens for antibodies in AIH-1 patient sera. It has been reported that autoantibodies against liver arginase exhibit antibody-dependent cell-mediated cytotoxicity, as well as direct cytotoxic activity, perhaps by inhibiting arginase activity, causing a reduction in ornithine production and hence in polyamine formation, thus interfering with cell membrane stability.43 CKs play a role in cell protection, particularly against oxidative injury and resistance to Fas-mediated apoptosis.52 If antibodies gain access to CK, they can disturb this cell protection against different types of injury and increase the predisposition to apoptosis. At the least, HSP may stimulate immunity for several reasons: first, HSP are phylogenetically highly conserved, and a high degree of homology between HSP in pathogens and humans may generate cross-reactivities. Second, an interaction between HSP and another potential antigenic protein (apoptotic bodies, viral protein) may expose HSP to an autoimmune cascade.53 These results constitute the first step in developing further experiments to gain a clearer understanding of the pathophysiology of AIH.
The authors thank Professor Fernando Alvarez (Service de gastroentérologie, hépatologie et nutrition, Hôpital Sainte-Justine, Montréal, Québec, Canada) and Doctor Thierry Tordjmann (INSERM U757, Université Paris Sud) for critically reviewing the manuscript.