THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1989 by The American Society for Biochemistry and Molecular Biology, Inc. Vol. 264, No. 31, Issue of November 5, pp. 18768-18774 1989

Printed in i! S A .

Gastric Parietal Cell Antigens of 60-90,92, and 100-120 kDa Associated with Autoimmune Gastritis and Pernicious Anemia ROLE OF N-GLYCANS IN THE STRUCTURE AND ANTIGENICITY OF THE 60-90-kDA COMPONENT*

(Received for publication, April 10, 1989)

Isabela Goldkorn, Paul A. Gleeson, and Ban-Hock TohS From the Department of Pathology and Immunology, Monash Uniuersity Medical School, Prahran, Victoria, Australia 3181

Thirty-four human sera containing parietal cell autoantibodies (PCA) specifically immunoprecipitated two antigens, with apparent molecular masses of 60- 90 kDa and 100-120 kDa under nonreducing condi- tions and 60-90 kDa and 120-150 kDa under reduc- ing conditions, from porcine gastric membrane ex- tracts. A third antigen of 92 kDa was only observed in immunoprecipitates analyzed under reducing condi- tions. By immunoblotting, 24 of the 34 PCA-positive sera reacted with only the 60-90-kDa antigen, three reacted with a broad 60-120-kDa smear, one reacted only with a 92-kDa antigen and six did not react. Reactivity with the 60-90-kDa antigen was observed with gastric membranes from dog, pig, rat, and rabbit. Twenty PCA-negative sera did not react with these components by immunoprecipitation or immunoblot- ting. PCA reactivity with the 60-90-kDa antigen was abolished when the gastric membranes were (a) di- gested with Pronase, (b ) reduced with 100 mM dithio- threitol, (c) treated with sodium periodate, or (d) di- gested with N-glycanase. The 60-90-kDa and 100- 120-kDa components were insensitive to neuramini- dase treatment. N-glycanase digestion of lZ5I-labeled antigens purified by immunoprecipitation and prepar- ative sodium dodecyl sulfate-polyacrylamide gel elec- trophoresis collapsed the 60-90-kDa antigen to a sharp 34-kDa band; the 100-120-kDa component was unaffected. These observations suggest that (i) parietal cell antigens comprise three components of 60-90,92, and 100-120 kDa; (ii) the epitopes differ in confor- mational sensitivity; (iii) the 60-90-kDa antigen is a conserved molecule comprising a 34-kDa core protein extensively glycosylated with N-linked oligosaccha- rides; (iv) sialic acid residues are not present in the 60-90- and 100-120-kDa molecules, and (v) the car- bohydrate and protein moieties of the 60-90-kDa mol- ecule are required for antibody binding.

Autoimmune gastritis in pernicious anemia is characterized by mucosal atrophy, parietal cell loss, and submucosal B lymphocyte infiltration (Type A chronic atrophic gastritis) (1-3). Parietal cell autoantibodies (PCA)’ present in the sera

* This work was supported by grants from the National Health and Medical Research Council of Australia and by the Anti-Cancer Council of Victoria. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

$ To whom correspondence should be addressed. The abbreviations used are: PCA, parietal cell autoantibodies;

PBS, phosphate-buffered saline; SDS-PAGE, sodium dodecyl sulfate- polyacrylamide gel electrophoresis; ELISA, enzyme-linked immuno- sorbent assay; BSA, bovine serum albumin.

of most patients with this disease are a useful diagnostic marker (4). The autoantibodies correlate with the presence of atrophic gastritis; whenever the autoantibody is found gastric biopsy reveals gastric atrophy even in the absence of perni- cious anemia. Furthermore, the antibody titer correlates with disease severity (5), and passive transfer of PCA to rodents results in gastric atrophy (6), suggesting that the antibody might be implicated in the pathogenesis of atrophic gastritis.

The parietal cell autoantigens are associated with the gas- tric microsomal fraction (7) and have been localized using human PCA-positive serum to microvillar membranes of pa- rietal cell intracellular canaliculi (8). The canaliculi, abundant in acid-secreting parietal cells, are implicated in hydrochloric acid secretion (9), a role supported by the localization of the gastric proton pump at this site (10). Nonsecreting parietal cells contain few canaliculi, most of which have been appar- ently replaced by tubulovesicles (11, 12).

The parietal cell autoantigens are cell specific, but not species specific; parietal cell autoantibodies react with the acid-producing cells of a variety of animal species (13, 14). The conservation of the antigens suggests that the molecules may have an important physiological role, which may be related to acid secretion by these specialized cells.

A previous study by Dow et al. (15) identified a 65-75-kDa molecule as the major parietal cell autoantigen (15). However, recently Karlsson et al. (16), using pooled sera, have reported that the major parietal cell autoantigen is a 92-kDa-compo- nent. Both these studies used immunoblotting for antigen identification. Immunoblotting, however, may not detect an- tigens with conformational epitopes. In the present study, we have compared the reactivity of 34 PCA-positive sera by immunoprecipitation and by immunoblotting, biochemically characterized the parietal cell antigens, and determined the contribution of carbohydrate and protein toward antibody binding.

EXPERIMENTAL PROCEDURES

Sera Sera were from the departmental clinical immunology laboratory

and from B. Ungar, Royal Melbourne Hospital. Intrinsic factor au- toantibody was detected by radioimmunoassay (17); all other auto- antibodies were detected by indirect immunofluorescent tests on frozen tissue sections (18). The sera comprised 34 PCA-positive sera, including 16 from patients with pernicious anemia. Twenty PCA- negative sera comprised 2 sera positive for thyroid microsomal au- toantibody, 3 for smooth muscle autoantibody, 3 for anti-nuclear autoantibody, 1 for intrinsic factor autoantibody, and 11 from healthy individuals.

Preparation of Gastric Mucosal Membranes

Porcine stomachs (corpus) were washed with PBS and surface mucus removed by scraping with a glass slide. The mucosa was

18768

Gastric Parietal Cell Autoantigens 18769

dissected from the muscularis, cut into fragments, and agitated in PBS to remove remaining mucus. Mucosal tissue was recovered by centrifugation at 150 X g for 10 min and homogenized in PBS, pH 7.2, containing 0.25 M sucrose, 2 mM EDTA (Ajax Chemicals, Aus- tralia) and 2 mM phenylmethylsulfonyl fluoride (Sigma) (100 g tissue/ 100 ml buffer) in a Waring Blendor for 3-5 min. The homogenate was subjected to sequential centrifugation in the same buffer at 400 X g for 10 min, 10,000 X g for 30 min, and 100,000 X g for 50 min to give the crude membrane pellet, The pellets were resuspended in the buffer and stored at -70 'C for up to 4 months. Membrane prepara- tions of pig liver and kidney were similarly prepared. All procedures were carried out at 4 "C. Protein concentration of membrane prepa- rations, estimated by the Bio-Rad Protein Assay (19), was adjusted to 10 mg/ml.

ELISA Wells of flat-bottomed microtiter plates (Flow) were coated with

50 pl of gastric liver or kidney membranes (750 ng protein/well) in 0.05 M carbonate-bicarbonate coating buffer, pH 9.6, containing 0.02% NaN,. The plates were incubated for 4 h a t room temperature and washed three times with PBS, 0.05% Tween 20 (Sigma). Dilutions of sera in PBS, 0.05% Tween 20 (50 pl in duplicates) were added to wells and incubated (30 min at room temperature). The wells were washed six times with PBS, 0.05% Tween and incubated for 30 min with 50 p1 of peroxidase-conjugated rabbit anti-human immunoglob- ulins (DAKO, Denmark) diluted 1:800 in PBS, 0.05% Tween. En- zymic activity was determined with 100 p1 of 0-phenylenediamine (Sigma; 10 mg/ml), 3% HzOz for 30 min at room temperature in the dark. The reaction was stopped with 50 p1 of 8 M HzS04 and absorb- ance measured at 492 nm using a micro ELISA autoreader and buffer as blank.

Immunoprecipitation Membrane preparations, solubilized in 0.5% Triton X-100 (Sigma)

in PBS for 1 h at 4 "C were centrifuged at 100,000 X g for 30 min. Membrane extracts (50 pl, 0.2 mg/ml) were iodinated with 5 pl of Nal*'I (Amersham Corp., activity 10 mCi/100 pl) and 10 p1 of chlor- amine-T (1.0 mg/ml) for 1 min at room temperature (20). Iodinated extracts were transferred to a PD-10 column (Pharmacia), prewashed with PBS containing 0.1% bovine albumin (BSA, Sigma), 0.1% NaN3 and 0.5% Triton X-100. The column was then washed with 2.5 ml of PBS/BSA/NaN3/Triton X-100 and 3.5 ml of PBS/BSA/NaN3/Tri- ton X-100 added. The effluent, containing the iodinated membrane extract, was collected.

Iodinated extracts (300 pl) were precleared with 50 pI of 10% Staphylococccls aureus Cowan 1 strain (CSL, Melbourne, Australia), (washed with 50 mM Tris-HC1, pH 8.3) and centrifuged in an Eppen- dorf centrifuge for 30 s. The supernatants were reacted with 10 p1 of test or control serum for 1 h at 4 "C or room temperature. Immune complexes were precipitated with 50 p1 of S. aureus for 2 min and centrifuged as above. The precipitates were washed three times with 50 mM Tris-HC1, pH 8.3, containing 0.6 M NaCl and 0.5% Triton X- 100 (21), solubilized by boiling for 3 min in SDS sample buffer, and analyzed by 10% SDS-PAGE (22). The gels were stained with 0.2% Coomassie Blue R-250 (Bio-Rad), dried, and autoradiographed at -70 "C using Kodak X-Omat AR x-ray film and DuPont Lightning- plus intensifying screens.

Preparative SDS-PAGE Immunoprecipitates of '251-labeled gastric membrane extracts were

subjected to 10% SDS-PAGE, stained with Coomassie Blue, and autoradiographed. The regions containing the 60-90- and 100-120- kDa antigens were excised, rehydrated in 40 mM Tris-HC1 buffer, pH 7.8, for 30 min at room temperature, and equilibrated for 30 min in 50 mM NaHC03 containing 0.1% SDS. The proteins were electropho- retically eluted at 50 V for 24 h (23). Eluted proteins were methanol- precipitated a t -20 "C overnight and collected (50,000 X g, 45 min).

Zmmunoblotting (Western Blotting) Proteins in membrane preparations were solubilized by boiling for

3 min in SDS sample buffer. The solubilized proteins, nonreduced or reduced with 100 mM dithiothreitol, were separated by 7.5 or 10% SDS-PAGE (200-300 pg protein/well) and electrophoretically trans- ferred to nitrocellulose membranes (Schleicher & Schuell) (24). The membranes were blocked with PBS, 3% casein for 30 min, cut into strips, and incubated with human sera (diluted 1:100 in PBS, 3%

casein) for 1 h at room temperature. Unbound antibody was removed by lo-min sequential washes, once with PBS, twice with PBS con- taining 0.05% Tween 20 and 0.05% Nonidet P-40 (Sigma), and once with PBS. The membranes were incubated for 1 h either with horse- radish peroxidase-conjugated rabbit anti-human immunoglobulins, diluted 1:100 in PBS, 3% casein, or with iodinated Protein A. Bound conjugate was detected by color reaction with 4-chloro-l-naphthol/ H202, and bound Protein A by autoradiography. Molecular weight standards (BioRad) on nitrocellulose membranes were stained with 1% Ponceau S.

Enzymic and Chemical Treatments of Gastric Membranes Delipidation-Lipids were extracted from three volumes of gastric

membrane suspensions (5 mg/ml) with methanol/chloroform (8:4, v/ v) (1 h, room temperature) and centrifuged at 15,000 X g, 15 min. The pellet was resuspended in n-butanol-saturated H20 (at room temperature) and recentrifuged. The lipid-extracted fractions were tested by dot-blot assay (25) by spotting samples onto nitrocellulose membranes, blocking with PBS, 3% casein and reacting with a PCA- positive serum as for immunoblotting. The final pellet was also tested with the PCA-positive serum by immunoblotting.

Pronase Digestion-Delipidated gastric membranes were lyophi- lized and resuspended to original volume in 0.1 M Tris-HC1, pH 8.0, 1 mM CaClZ, and 0.2% NaN3. The preparation was boiled for 5 min to inactivate interfering enzymes and incubated at 37 "C for 24 or 72 h with 3% (wt/wt) Pronase (ex Streptomyces griseus, Behring Diag- nostics) (26). The Pronase-digested material was boiled for 10 min to inactivate the enzyme and analyzed by immunoblotting.

The enzyme and antigen preparations were tested for proteolytic activity on 1% agar gel containing bovine casein (Bio-Rad substrate Gel Tablets). Control samples were Proteinase K (Boehringer) and buffer blanks.

Periodate Oxidation-Gastric membranes (10 mg/ml in 0.1 M so- dium acetate buffer, pH 5.6) were incubated for 2 h at 4 "C in the dark with or without 20 mM sodium m-periodate (Sigma). The mem- brane samples were washed and tested for PCA reactivity by ELISA. Gastric membrane preparations were separated by 10% SDS-PAGE and electrophoretically transferred to nitrocellulose. The nitrocellu- lose membranes were then similarly treated with sodium periodate and tested for PCA reactivity by immunoblotting. As a control, BSA was similarly treated, probed with rabbit anti-BSA antiserum (diluted 1:20 in PBS, 3% casein) and incubated with peroxidase-conjugated swine-anti-rabbit immunoglobulins (DAKO), diluted 1:lOO in PBS, 3% casein.

N-Glycanase Treatment-Gastric membranes were boiled for 3 min in 0.5% SDS and incubated at 37 "C with 0.1 M Tris acetate buffer, pH 8.6, containing 10 mM EDTA and 3.5% (v/v) Nonidet P-40 (in a ratio of 3 volumes of Nonidet P-40 containing buffer/2 volumes 0.5% SDS) with or without N-glycanase (Genzyme, England) (final con- centration 4.6 units/ml) for the appropriate time (27). The samples were reacted with a PCA-positive serum by immunoblotting. Elec- troeluted '251-labeled antigens were similarly treated and analyzed on SDS-PAGE followed by autoradiography. N-Glycanase was tested for activity on fetuin (fetal calf serum type 111, Sigma) and bovine pancreatic ribonuclease B, type XII-B (Sigma), and the digests were analyzed by SDS-PAGE.

Neuraminidase Treatment-Gastric membranes were resuspended at 10 mg/ml in 50 mM sodium acetate buffer, pH 5.5, containing 1 mM CaClZ, and incubated with or without neuraminidase from Vibrio cholerae (Behringwerke, final concentration 0.2 milliunits/pl) for 3 h a t 37 "C (28) and samples were taken and analyzed by immunoblot- ting. The enzyme was tested for activity by incubation with the glycoprotein transferrin and the digest analyzed by SDS-PAGE.

RESULTS

Parietal Cell Autoantibody-Thirty-four human sera, in- cluding 16 from patients with pernicious anemia, gave positive PCA reactions by indirect immunofluorescence on frozen tissue sections of mouse stomach. These sera gave specific cytoplasmic staining of gastric parietal cells and did not stain liver and kidney tissues. Twenty sera from healthy individuals and from patients with autoimmune disorders other than pernicious anemia were used as negative controls and were PCA-negative by immunofluorescence.

An ELISA was developed to determine the reactivity of

18770 Gastric Parietal Cell Autoantigens

PCA with gastric membranes. Microtiter wells were coated with a gastric membrane preparation as described under "Ex- perimental Procedures," and the binding of antibody to the membrane-coated wells was determined. Fig. 1 shows a typical ELISA with three of the sera. Other PCA-positive sera showed similar titers in the ELISA assay, while control sera showed only background reading. PCA-positive sera did not react by ELISA with membrane preparations from pig liver or kidney. These ELISA results confirm the immunofluorescence data that the autoantibodies in the PCA-positive sera are tissue- specific and react with gastric parietal cell membranes.

Identification of Parietal Cell Autoantigens-Immunoblot analysis of the porcine gastric membrane preparation under nonreducing conditions showed that 24 out of 34 PCA-positive sera specifically reacted with a broad band of 60-90 kDa. A typical example is shown in Fig. 2A (lune a). In addition, 3 of the 34 PCA-positive sera reacted with an extensive 60-120- kDa smear (Fig. 2 A , lane b) , 1 of the sera reacted with a 92- kDa sharp band, and 6 sera did not react. The 20 PCA- negative sera showed no reactivity by immunoblotting with any of these components (Fig. 2 A , lane c ) . An additional nonspecific band of approximately 300 kDa and an occasional 45-kDa band were detected with the positive sera as well as with control sera. Reactivity of PCA-positive sera by immu- noblotting was observed with gastric membranes but not with porcine liver or kidney membranes (Fig. 2 A , lanes d and e). As the most common reactivity of the PCA-positive sera by immunoblotting was with the broad 60-90-kDa band, the remaining immunoblotting experiments in this study were performed exclusively with these sera unless otherwise spec- ified.

The reactivity of PCA-positive serum with gastric mucosal membranes of rat, rabbit, dog, and pig were compared by immunoblotting (Fig. 3). The patterns obtained for each were very similar in that a typical broad band was detected, al- though some variation was observed in the apparent M, of these diffuse bands.

As immunoblotting requires conservation of the antigenic determinants after SDS-PAGE and therefore some antigens may not be detected, we carried out immunoprecipitation experiments using solubilized iodinated porcine gastric mu- cosal membranes. In contrast to the variations observed by immunoblotting, the analysis of immunoprecipitates by SDS- PAGE under nonreducing conditions showed that all PCA- positive sera specifically immunoprecipitated a 60-90-kDa antigen and 100-120-kDa antigen. Fig. 4 (lane a) shows a

typical example with one of the positive sera. Reactivity with either of the two components alone was not observed. The components were present in porcine gastric mucosa but not in porcine liver or kidney preparations (Fig. 4). None of the 20 control sera reacted with the two antigens. Additional nonspecific 300- and 45-kDa components were observed in the immunoprecipitates of both PCA-positive and PCA-neg- ative sera.

Analysis of immunoprecipitates under reducing conditions showed that the size of the 100-120-kDa component slightly increased, which is indicative of intra-chain disulfide bonds, whereas the relative mobility of the 60-90-kDa antigen did not change. The reduced sample also resolved an additional 92-kDa sharp band, which was possibly obscured by the 100- 120-kDa antigen in the nonreduced sample (Fig. 4, lane c).

Characterization of Parietal Cell Antigens-As previous work suggested that parietal cell autoantigens include lipids (7), we extracted lipids and glycolipids from gastric mem- branes with chloroform/methanol and n-butanol-saturated H,O (29). Dot-blot assay showed that a PCA-positive serum reacted exclusively with the residue and not with the lipid fractions. Also, the 60-90-kDa band, demonstrated by im- munoblotting, was detected in the lipid extracted residue (Fig. 2B).

Treatment of gastric membranes with Pronase resulted in a complete loss of reactivity with PCA-positive sera on im- munoblotting (data not shown). Furthermore, reduction of the membrane preparation (100 mM dithiothreitol) prior to immunoblotting completely abolished PCA reactivity with the 60-90-kDa antigen (data not shown), indicating that intact disulfide bonds are essential for antibody binding.

Marked heterogeneity of the 60-90 and 100-120-kDa anti- gens on SDS-PAGE suggests extensive glycosylation. To de- termine if carbohydrate is required for antibody binding, we carried out preliminary experiments involving oxidation of the antigens with sodium periodate. Sodium periodate treat- ment of the gastric membrane preparation resulted in a dra- matic loss of PCA binding as detected by ELISA (Fig. 5). This was further investigated by assessing the reactivity of the 60-90-kDa antigen in immunoblotting experiments. so- dium periodate treatment of the nitrocellulose transfer re- sulted in complete loss of antibody binding to the 60-90-kDa component (Fig. 6, lanes a and b) . A control to assess the extent of protein degradation by this oxidation procedure was the reaction of the non-glycosylated protein BSA with an anti-BSA serum. No apparent loss of antibody binding to

I -

FIG. 1. ELISA of anti-gastric membrane activity of human sera. A Microplates were coated with porcine gastric mucosal membranes (750 ng/ E 0 . 6 - well). Solid lines show typical PCA activ- r

ity in three sera of patients having PCA b c a detected by immunofluorescence. Dotted lines show PCA activity in three control sera. The color reaction was developed with H,02/0-phenylenediamine and ab- %J2nm

sorbance at 492 nm measured. All assays 0 .2 - were carried out in duplicate.

0 . 8 -

b

0.4 - e

*""" *" 1 -a

0 :="+z-*&>-e-$ "" A"=* "" "" "-3 """ " - - *"""""" -"")-

I I I 1 I I I I I

50 100 200 400 800 1600 3200 8400 12800

R ~ ~ l p r o c o l o f s e r u m dllullon

Gastric Parietal Cell Autoantigens 18771

94- 68-

45-

45-

a ~ c a e

a b FIG. 2. A, immunoblot analysis of porcine membranes with human

sera. Microsomal fractions of gastric mucosa (lanes a-c), liver (lane d ) , or kidney (lane e ) were solubilized in nonreducing SDS-sample buffer. The proteins, separated in 10% polyacrylamide slab gel, were electrophoretically transferred to nitrocellulose membranes and in- cubated with PCA-positive sera (lanes a, b, d, and e ) or a PCA- negative serum (lane c) followed by horseradish periodase-conjugated rabbit anti-human immunoglobulin. Lanes a and b show two types of reactivities observed with PCA positive sera. S = stomach; L = liver; K = kidney. B, immunoblot analysis of delipidated porcine gastric membranes. Lane a, total membranes; lane b, residue after lipid extraction.

M, X ~ O - ~ I : c

U

45-

a b c a

FIG. 3. Immunoblot analysis of gastric membrane prepara- tions from different species with PCA-positive serum. Micro- somal fractions of gastric mucosa from rat (lane a), rabbit (lane b) , dog (lane c) and pig (lane d ) were solubilized in nonreducing SDS- sample buffer, and the proteins separated in a 10% polyacrylamide slab gel and electroblotted onto nitrocellulose membranes. The mem- branes were incubated with a PCA-positive serum followed by a horseradish periodase-conjugated rabbit anti-human immuno$lobulin as described under “Experimental Procedures.”

BSA was observed after sodium periodate treatment (Fig. 6, lanes d and e ) . Moreover, the nonspecific 300-kDa band was detected in both treated and untreated samples (Fig. 6, lanes a-c). Preservation of these proteins during periodate oxida- tion procedure suggests little or no protein degradation. How- ever, degradation cannot be excluded, as sodium periodate is known to affect certain amino acids, such as tyrosine and

- D l 1 + D l 1

S e r a + - Source S S S S L K

+ - + +

MJO-~ e: P I

92-

66-

a b c d e f FIG. 4. Immunoprecipitation of ‘251-labeled porcine mem-

branes with PCA-positive sera. Solubilized, iodinated microsomal fractions of gastric mucosa (lanes a-d), liver (lane e ) , or kidney (lane f ) were reacted with a typical PCA-positive serum (lanes a c, e, and f ) or with a PCA-negative serum (lanes b, and d ) as described under “Experimental Procedures.” Immunoprecipitates were solubilized in nonreducing (lanes a and b) or reducing (lanes c-f) SDS-sample buffer. 20,000 cpm of each sample were analyzed on a 10% polyacryl- amide slab gel. The dried gel was exposed to Kodak X-Omat AR film for 3 days a t -70 “C with an intensifying screen. DTT, dithiothreitol; S, stomach; L, liver; K, kidney.

tryptophan (30). The susceptibility of the 100-120-kDa com- ponent to sodium periodate oxidation was not assessed.

The use of endoglycosidases provides a more specific means of modifying carbohydrates compared with chemical treat- ment. N-glycanase specifically hydrolyzes N-linked oligosac- charides from glycoproteins (27). Treatment of gastric mu- cosal membranes with N-glycanase followed by immunoblot- ting resulted in the complete loss of antibody binding to the 60-90-kDa antigen after 3 h of N-glycanase treatment (Fig. 7, lane b). Control digestions of ribonuclease B and fetuin resulted in the reduction of the apparent Mr of fetuin and ribonuclease B to the expected size of the deglycosylated proteins (data not shown), indicating that the enzyme prep- aration was active and free of contaminating proteases. Five different PCA sera have been tested and all showed no reac- tivity with N-glycanase-treated gastric membranes.

The extent of N-glycosylation of the components was also assessed. Solubilized ’2sI-labeled gastric membranes were im- munoprecipitated with a PCA-positive serum and the 60-90- and 100-120-kDa antigens individually electroeluted from preparative SDS gels. About 70% of the iodinated 60-90-kDa antigen and 10-15% of the 100-120-kDa band were recovered from the gel. The 12sI-labeled electroeluted antigens were individually digested with N-glycanase for various time pe- riods to 16 h. The broad 60-90-kDa antigen showed a drastic reduction in M,, and after the 16 h N-glycanase digestion it had collapsed into a sharp band of about 34 kDa (Fig. 8, lane e). A control incubation of the lZ5I-labeled 60-90-kDa antigen without enzyme showed no change in electrophoretic mobility. The recovery of the ’2sI-labeled 100-120-kDa component was low; however, preliminary experiments suggest this compo- nent is insensitive to N-glycanase treatment over a 16-h period.

The extent of sialylation of the 60-90- and 100-120-kDa antigens was assessed by treatment with neuraminidase. Pre- treatment of gastric membranes with neuraminidase had no effect on immunoblotting with two different PCA-positive sera (data not shown). In addition, neuraminidase treatment

18772 Gastric Parietal Cell Autoantigens

FIG. 5. Effect of periodate oxida- tion of porcine gastric membranes on reactivity wi th PCA-positive serum analyzed by ELISA. Untreated porcine gastric membranes (750 ng/well) or gastric membranes treated with 20 mM sodium periodate were used to coat ELISA plates. Serial dilutions of PCA- positive serum were added (50 p1 in du- plicate), and the antibody binding was estimated by measuring absorbance a t 492 nm as described under "Experimen- tal Procedures."

S BSA "

Ne104 + + - + + - T i m e ( h ) 2 3 3 2 3 3 -

M ~ x ~ o - ~ I 94-

68- ll 45-

0.8

A b

E 0.6 r b

0 . 4

0.2

0

a b c d e f FIG. 6. Immunoblot analysis of periodate-treated porcine

gastric membranes. Gastric membranes (lanes a-c) or bovine al- bumin (lanes d-f) were separated by 10% SDS-PAGE, electroblotted onto a nitrocellulose membrane, and incubated with 20 mM sodium periodate in the dark for 2 h (lanes a and d) , 3 h (lanes b and e ) or 3 h in the incubation buffer alone (lanes c and f ) . The membranes were probed with PCA-positive serum (lanes a-c) or with rabbit anti- albumin antiserum (lanes d-f), followed by peroxidase-conjugated anti-immunoglobulin. S, stomach.

of immunoprecipitates of solubilized lZ5I-labeled gastric mem- branes had no effect on the mobility of either the 60-90- or 100-120-kDa antigens by SDS-PAGE. As the neuraminidase was shown to be active with sialylated glycoprotein standards, neither antigen therefore appears to be sialylated to a detect- able level.

DISCUSSION

Our studies provide the first detailed biochemical charac- terization of parietal cell antigens recognized by PCA. We have identified three antigens in porcine gastric mucosa which react with PCA-positive sera. Two of these antigens are broad bands of 60-90 and 100-120 kDa. All 34 PCA-positive sera detected both antigens by immunoprecipitation. The third antigen, a sharp band of about 92 kDa was only detected in immunoprecipitates analyzed under reducing conditions; this

Df3 Untreated Sodium wriodate-treated

N-glycanase - + - Time(h) 0 3 3

' "

45-

a b c FIG. 7. Immunoblot analysis of N-glycanase treated porcine

gastric mucosal membranes. Lane a, untreated membranes; lane b, gastric membrane samples (10 mg/ml) incubated with N-glycanase (4.6 units/ml) for 3 h at 37 "C; lane c, 3 h of incubation without the enzyme. The samples were separated by SDS-PAGE (7.5%) and electroblotted onto a nitrocellulose membrane. The nitrocellulose was probed with PCA-positive serum followed by peroxidase-conjugated anti-human immunoglobulin.

antigen is probably obscured by the 100-120-kDa component under nonreducing conditions. In contrast, considerable var- iation was observed with immunoblotting. Most positive sera reacted only with a 60-90-kDa antigen, which is probably identical to the immunoprecipitated 60-90-kDa component. Three sera reacted with a smear from 60 to >120 kDa which may include the other two components present in the immu- noprecipitates. Six sera, despite being positive by immunoflu- orescence and immunoprecipitation, did not react by immu- noblotting. These results suggest that PCA-positive sera rec- ognize several different epitopes and that these epitopes differ in conformational sensitivity.

The 60-90-kDa molecule is probably identical to the 65- 75-kDa molecule previously reported from our laboratory (15) and by Karlsson et al. (16) since the antigens are broad bands of similar M , and are only reactive with the autoantibodies in the nonreduced state. However, in the latter study, the major parietal cell antigen detected by immunoblotting was a 92- kDa antigen. In contrast, this component was detected by

Gastric Parietal Cell Autoantigens 18773 N-glycanase - + + + + -

Time ( h ) 0 0.5 1 4 16 16

~ , x 1 0 - ~

92 - 66 - 45 -

c +

31 - a b c d e f

FIG. 8. N-glycanase treatment of isolated porcine "%la- beled 60-90-kDa antigen. The 12511-labeled-60-90-kDa antigen was purified by immunoprecipitation followed by electroelution from pre- parative SDS-PAGE, as described under "Experimental Procedures." Lane a, isolated untreated '251-60-90-kDa antigen; lane b, 30-min digestion; lune c, 1-h digestion; lane d, 4-h digestion; and lune e, 16-h digestion of 1ZsII-60-90-kDa antigen with N-glycanase; lane f, 16-h mock digestion without enzyme. Digests were separated on 10% polyacrylamide gel and autoradiographed for 5 days a t -70 "C using Kodak X-Omat AR film and intensifying screen.

only one of our 34 positive sera by immunoblotting. I t is likely that this apparent discrepancy in results reflects differences in reactivity between the positive sera by immunoblotting and/or differences in the preparation of antigen for analysis by SDS-PAGE. For instance, it has recently been shown that the components in antigen preparations, visible by Coomassie Blue staining or by reactivity with antibodies may differ depending upon whether the antigen preparations were boiled in SDS-sample buffer prior to analysis by SDS-PAGE (31). However, it is difficult to directly compare results, as Karlsson et al. (16) used a pool of 10 human sera and, in addition, did not carry out immunoprecipitation experiments.

The observation that both 100-120- and 60-90-kDa mole- cules were immunoprecipitated by all PCA-positive sera sug- gests that the two molecules either (a ) are physically associ- ated, ( b ) share epitopes, or (c) are recognized by two distinct subsets of antibodies each reactive with only one of the two components. At present we are unable to distinguish between these possibilities.

A PCA-positive serum which reacted with the 60-90-kDa band of pig gastric mucosa by immunoblotting showed similar reactivity with dog, rat, and rabbit gastric mucosal mem- branes, although the apparent molecular masses of the rabbit and rat antigens were somewhat lower for this antigen. The size difference between species may reflect differences in glycosylation. The conservation of the 60-90-kDa antigen is consistent with immunofluorescence studies on the species distribution of PCA reactivity (13).

The size heterogeneity of the 60-90- and 100-120-kDa antigens suggests that both are extensively glycosylated. How- ever, both components were insensitive to neuraminidase treatment, indicating the absence of sialic acid residues. This is not surprising as labile sialic acid residues would not survive the highly acidic gastric environment. N-Glycanase treatment of the 60-90-kDa antigen collapsed the molecule to a discrete 34-kDa band indicating that the 60-90-kDa antigen consists of a 34-kDa core protein heavily glycosylated with N-linked oligosaccharides. On the other hand, preliminary experiments suggest that the 100-120-kDa antigen is resistant to N-gly- canase digestion and therefore does not appear to carry N-

glycan chains. The size heterogeneity of the 100-120-kDa antigen, may, however, be due to 0-glycosylation.

Antibody binding to the 60-90-kDa antigen was eliminated on removal of the N-linked oligosaccharides, indicating that these carbohydrates are required for antibody binding. In addition, reduction of sulfhydryl groups of the 60-90-kDa antigen or Pronase treatment also abrogated reactivity with the antibodies. Taken together, these results suggest that the autoepitopes include both carbohydrate and protein moieties. Alternatively, the carbohydrate may play an indirect role by influencing the conformation of the polypeptide chain. Car- bohydrate antigens have been implicated in other autoim- mune conditions. For example, sera from patients with insu- lin-dependent diabetes mellitus bind to the carbohydrate of a glycolipid present in pancreatic islet cells (32). In addition, IgM in patients with paraproteinemia with associated periph- eral neuropathy reacts with a carbohydrate determinant shared by myelin-associated glycoprotein and a ganglioside of human sciatic nerve (33). Carbohydrate blood group antigens on erythrocyte membranes are also implicated in autoimmune hemolytic anemia (34).

A major function of the parietal cell is hydrochloric acid secretion, mediated by the H'/K' ATPase (H' proton pump). The H' pump from porcine stomachs has a major component of 95-100 kDa. Karlsson et al. (16) presented indirect evidence that the 92-kDa parietal cell antigen is the H'/K' ATPase itself. The 92-kDa antigen detected in this study is probably the same protein. However, the function of the 60-90- and 100-120-kDa antigens is not known. The conservation of the 60-90-kDa antigen across species suggests that it has an important physiological function. Possible candidates are the molecules of the K' C1' ion channel associated with the pump (9). Alternatively, the 60-90- and 100-120-kDa antigens may be components of the pump itself. Indeed, a broad band of periodate-positive staining material of 75-80 kDa copurifies with the H'/K' ATPase (35). Further studies will be directed at investigation of the relationship of these two antigens with the H+ pump and their localization in the resting and stimu- lated parietal cell. The relationship of these molecules to parietal cell surface autoantigen (36-38) and to the gastrin receptor autoantigen (39) is also under investigation. Char- acterization of these parietal cell antigens should contribute toward knowledge of the structure and function of these molecules and of their role in the pathogenesis of autoimmune gastritis.

Acknowledgments-We thank Berta Ungar and Catherine Turn- bull for samples of human sera.

REFERENCES 1. Strickland, R. G., and Mackay, I. R. (1973) Am. J. Dig. Dis. 18,

2. Whittingham, S., and Mackay, I. R. (1985) in The Immune Diseme (Rose, N. R., and Mackay, I. R., eds) pp. 243-266, Academic Press, New York

3. Kaye, M. D., Whorwell, P. J., and Wright, R. (1983) Clin. Im- munol. Immunopathol. 28,431-440

4. Taylor, K. B., Roitt, I. M., Doniach, D., Couchman, K. G., and Shapland, C. (1962) Rr. Med. J. ii, 1347-1354

5. Wright, R., Whitehead, R., Wangel, A. G., Salem, S. N., and Schiller, K. F. R. (1966) Lancet i, 618-621

6. Tanaka, N., and Glass, G. B. J. (1970) Gastroenterology, 58,482- 494

7. Ward, H. A., and Nairn, R. C. (1972) Clin. Exp. Immunol. 10, 435-457

8. Hoedemaeker, P. J., and Ito, S. (1970) Lab. Inuest. 22,184-188 9. Sachs, G., Tache, Y., Debas, H. T., Jacobson, E. D., and Freston,

10. Smolka, A., Helander, H. F., and Sachs, G. (1983) Am. J. Physiol.

426-440

J. W. (1987) Am. J. Med. 83,307-328

245, G589-G596

18774 Gastric Parietal Cell Autoantigens 11. 12.

13.

14.

15.

16.

17.

18.

19. 20.

21.

22. 23.

24.

Helander, H. F. (1981) Znt. Rev. Cytol. 7 0 , 217-289 Ito, S. (1981) in Physiology of the Gastrointestinal Tract (Johnson,

L. R., ed) pp. 517-550, Raven Press, New York Nairn, R. C., and De Boer, W. G. R. M. (1966) Nature 210,960-

962 Loveridge, N., Bitensky, L., Chayen, J., Hausman, T. U., Fisher,

J. M., Tailor, K. B., Gardner, J. D., Bottazzo, G. F., and Doniach, D. (1980) Clin. Exp. Zmmunol. 4 1 , 264-270

Dow, C. A., de Aizpurua, H. J., Pedersen, J. S., Ungar, B., and Toh, B. H. (1985) Clin. Exp. Zmmunol. 6 2 , 732-737

Karlsson, F. A., Burman, P., Loof, L., and Mardh, S. (1988) J. Clin. Invest. 8 1,415-479

Ungar, B., Stocks, A. E., Martin, F. I. R., Whittingham, S., and Mackay, I. R. (1968) Lancet ii, 415-417

Nairn, R. C. (1976) Fluorescent Protein Tracing, 4th Ed, pp. 369- 377, Churchill-Livingstone, Edinborough

Bradford, M. M. (1977) Anal. Biochem. 72 , 248-254 Greenwood, F. C., Hunter, W. M., and Glover, J. S. (1963)

Goding, J. W. (1986) Monoclonal Antibodies: Principles and Prac-

Laemmli, U. K. (1970) Nature 227, 680-685 Stearne, P. A., Van Driel, I. R., Grego, B., Simpson, R. J., and

Towbin, H., Staehlin, T., and Gordon, J. (1979) Proc. Natl. Acad.

Biochem. J. 8 9 , 114-123

tice pp. 165-171, Academic Press, New York

Goding, J. W. (1985) J. Zmmunol. 134,443-448

Sci. U. S. A. 76, 4350-4354

26. 27.

28.

29.

30. 31.

32.

33.

34.

35.

36.

37.

38.

Finne, J., and Krusius, T. (1982) Methods Enzymol. 83,269-277 Plummer, T. H. Jr., Elder, J. H., Alexander, S., Phelan, A. W.,

and Tarention, A. L. (1984) J. Biol. Chem. 2 5 9 , 10700-10704 Ada, G. L., French, E. L., and Lind, P. E. (1961) J. Gen. Microbiol.

Makita, A,, and Taniguchi, N. (1985) in New Comprehensive Biochemistry (Wiegandt, H., ed) Vol. 10, pp. 1-100, Elsevier Science Publishers B. V., Amsterdam

24,409-425

Clamp, J. R., and Hough, L. (1965) Biochem. J. 9 4 , 17-24 Mercier, F., Reggio, H., Devilliers, G., Bataille, D., and Mangeat,

P. (1989) Biol. Cell 6 5 , 7-20 Nayak, R. C., Omar, M. A. K., Rabizadeh, A., Srikanta, S., and

Eisenbarth, G. S. (1985) Diabetes 34,617-619 Ilyas, A. A., Quarles, R. H., MacIntosh, T. D., Dobersen, M. J.,

Trapp, B. D., Dalakas, M. C., and Brady, R. 0. (1984) Proc. Natl. Acad. Sci. U. S. A. 8 1 , 1225-1229

Profsky, B. (1985) in The Autoimmune Diseases (Rose, N. R., and Mackay, I. R., eds) pp. 469-491, Academic Press, New York

Nandi, J., Zhou, M-A,, and Ray, T. K. (1987) Biochemistry 26,

De Aizpurua, H. J., Toh, B. H., and Ungar, B. (1983) Clin. Exp.

De Aizpuma, H. J., Ungar, B., and Toh, B. H. (1982) Clin. Exp.

De Aizpurua, J. H., Cosgrove, L. J., Ungar, B., and Toh, B.-H.

4264-4272

Immunol. 52,341-349

Zmmunol. 54,405-410

(1983) New End. J. Med. 309.625-629 25. Hawkes, R., Niday, E., and Gordon, J. (1982) Anal. Biochem. 39. de'Aizpurua, H. i., Ungar, B., and Toh, B.-H. (1985) New Engl.

119,142-147 J. Med. 313,479-483