ATPase4A Autoreactivity and ItsAssociation With AutoimmunePhenotypes in the Type 1 DiabetesGenetics Consortium StudyDiabetes Care 2015;38(Suppl. 2):S29–S36| DOI: 10.2337/dcs15-2006

Autoantibodies targeting the H+/K+-ATPase proton pump of the gastric parietalcell (parietal cell antibodies [PCA]) are diagnostic of atrophic body gastritis (ABG)leading to pernicious anemia (PA). PCA, ABG, and PA occur in increased frequencyin patients with type 1 diabetes and their relatives and are considered “minor”components of forms of autoimmune polyglandular syndrome (APS). A custom-ized radioimmunoprecipitation assay was applied to 6,749 samples from the Type 1Diabetes Genetics Consortium to measure ATP4A autoreactivity. Autoantibodyprevalence was correlated with variants in HLA class II, PTPN22, and CTLA4 genes.With an ATP4A radioimmunoprecipitation assay, PCA were detected in sera from20.9% of affected individuals. PCA prevalence increased with age and was greaterin females (25.3%) than males (16.5%) and among Hispanics (36.3%) and blacks(26.2%) compared with non-Hispanic whites (20.8%) and Asians (16.7%). PCA andother organ-specific autoantibodies GAD65, IA-2, thyroid peroxidase (TPO), 21-hydroxylase (21-OH), and transglutaminase (TG) clustered within families withheritability estimates from71 to 95%. PCA clusteredwith TPO, 21-OH, and persistentGAD65 autoantibodies but not with celiac (TG) or IA-2 autoantibodies. PCA-positivesubjects showed an increased frequency of DRB1*0404, DPB1*0201, and PTPN22R620W (rs2476601-T) and a decreased frequency of DRB1*0101, DPB1*0301, andCTLA4 CT60 (rs3087243-T). Genetic variants accounted for 4–5% of the heritablerisk for PCA. The same alleles were associated with other autoantibody phenotypesin a consistent pattern. Whereas most of the heritable risk for PCA and other anti-bodies reflects genetic effects that are tissue specific, parietal cell autoimmunityis a major pathogenetic contributor in APS2.

Autoantibodies targeting the H+/K+-ATPase proton pump on the plasmamembraneof the gastric parietal cell (parietal cell antibodies [PCA]) are a specific and sensitivediagnostic of atrophic body gastritis (ABG). Chronic ABGmay lead topernicious anemia(PA) as a result of impairedproductionof intrinsic factor and presence of intrinsic factorautoantibodies, which hinder cobalamin uptake in the duodenum leading to cobaltdeficiency. PCA and PA are common autoimmune conditions, affecting 2% and 0.15–1% of the general population, respectively. PCA prevalence increases with age, up to12%, yet the true incidences of ABG and PA are uncertain, as ABG is generally asymp-tomatic and cobalt deficiency and its clinical manifestations are slow to develop. ABG isalso associated with gastric malignant lesions in ;10% of individuals (1). The identifi-cation of demographic, genetic, and immunological risk factors associated with ABG isvital to its diagnosis and early intervention to prevent the serious clinical sequela of PA.

1Barbara Davis Center for Childhood Diabetes,University of Colorado Denver, Aurora, CO2Digestive and Liver Disease Unit, University “LaSapienza,” Sant’Andrea Hospital, Rome, Italy

Corresponding author: Janet M. Wenzlau, janet.wenzlau@ucdenver.edu.

Received 23 March 2015 and accepted 4 July2015.

†Deceased.

This publication is based on the presentationsfrom the Type 1 Diabetes Genetics Consortium(T1DGC) Autoantibody Workshop, which washeld on 7 June 2011 in Bethesda, MD. The pub-lication of this supplement wasmade possible byresources provided by the T1DGC, a collaborativeclinical study sponsored by the National Instituteof Diabetes and Digestive and Kidney Diseases,the National Institute of Allergy and InfectiousDiseases, the National Human Genome ResearchInstitute, the Eunice Kennedy Shriver NationalInstitute of Child Health and Human Develop-ment, and JDRF and supported by grant U01DK062418.

© 2015 by the American Diabetes Association.Readersmayuse this article as longas thework isproperly cited, the use is educational and not forprofit, and the work is not altered.

Janet M. Wenzlau,1 Pamela R. Fain,1

Thomas J. Gardner,1 Lisa M. Frisch,1

Bruno Annibale,2 and John C. Hutton1†

Diabetes Care Volume 38, Supplement 2, October 2015 S29 T1DGCAUTO

ANTIB

ODYWORKSH

OP

Familial aggregation of PCA and PA isconsidered well established, althoughbased on limited data in families of pa-tients with PA (2,3). Parietal cell autoim-munity, in common with many otherautoimmune disorders, may be deter-mined by an interplay of genetic andenvironmental risk factors with involve-ment of HLA class I and/or HLA class IIgenes. Most of the evidence support-ing a genetic basis for PCA comes fromstudies documenting the frequency ofPCA and PA in patients and their rela-tives with other autoimmune diseases,including type 1 diabetes, autoim-mune thyroid disease, and adrenalautoimmunity (1–5).The gastric proton pump H+/K+-ATPase

is a heterodimer composed of an eighttransmembrane–spanning catalytic Asubunit (ATP4A) and a single transmem-brane highly glycosylated B subunit(ATP4B), both of which are antigenictargets for T and B cells (6,7). A mousemodel of autoimmune gastritis, inducedvia thymectomy, mimics the human dis-ease including submucosal infiltration,parietal cell destruction, and autoanti-bodies targeting both A and B H+/K+-ATPase subunits. Transgenic expressionof the B subunit protects mice fromdeveloping autoimmune gastritis, imply-ing tolerance induction within the thy-mus. Moreover, the initial immuneresponse may target the B subunit, asthe other autoantigen reactivates areavoided in the transgenic mouse (8). De-tection of PCA has relied on immu-nocytochemistry (9) and ELISA, usingbiochemically purified pig stomach en-zyme as the target antigen (10) with di-agnostic confirmation by biopsy. Mostautoantibodies target epitopes in thecarbohydrate moiety of the B subunit,whereas in our study we use the ATPaseA subunit protein as the antigen probe.We recently developed a radioimmuno-precipitation assay (RIA) to defined epitopesin ATP4A using in vitro translated humanATP4A cDNA to generate a “biochemical”assay of high sensitivity (95%) and spec-ificity (100%) to measure antibodies inindividuals with histologically confirmedABG (11). In the 2012 InternationalAutoantibody Study Program, this novelassay detected autoantibodies in 30%of type 1 diabetes samples and 4% ofcontrols.In the current study, the new radio-

immunoprecipitation assay was used to

characterize ATP4A autoreactivity in6,749 subjects as part of the Type 1 Di-abetes Genetics Consortium (T1DGC)Autoantibody Workshop. The primaryobjective was to characterize pheno-typic and genetic differences betweenATP4A sero-positive and sero-negativeindividuals with type 1 diabetes. Wedemonstrate that observed differencesare consistent with a genetic basis. Thisresult likely reflects predisposing andprotective alleles primarily in gene(s)with relatively specific effects on gastricautoimmunity. Furthermore, there arelikely strong contributions by HLA classII and non-HLA variants with pleiotropiceffects on gastric, thyroid, and adrenalautoimmunity.

RESEARCH DESIGN AND METHODS

Radioimmunoprecipitation AssayA full-length gastric ATPase4A cDNAclone was generated by reverse tran-scription of human stomach RNA(iScript; Bio-Rad, Hercules, CA), andsequences encoding antigenic epitopeswere subcloned into the pCMVTNT vec-tor for coupled in vitro transcription/translation (cat. no. L1170; PromegaTNT kit) to produce [35S]-methionine(PerkinElmer)–labeled antigen probes.Unincorporated [35S]-labeled aminoacids were removed by G-25 columns(cat. no. 17-0853-02; NAP-5, GE Health-Care), and the labeled probe was quanti-fied by trichloroacetic acid precipitation.

Serum samples (2.5 mL) were incu-bated overnight at 48C with the antigen(20,000 counts per minute [CPM]) in 60 mLof 20 mmol/L Tris-HCl, pH 7.4, containing150 mmol/L NaCl, 0.1% BSA, 0.15%Tween-20, and 0.1% sodium azide in96-well plates. Immune complexeswere captured by addition of 25 mL ofa 50% Protein A Sepharose slurry (cat.no. 17-5280; 4 Fast Flow, GE Health-Care), transferred to 96-well filtrationplates (cat. no. MSDVN6B50; Millipore,Billerica, MA) and washed sequentiallywith 150 mL assay buffer at 48C using anautomated, vacuum-operated 96-wellplatewasher (Tecan,Mannedorf, Switzer-land). After air-drying, 25 mL of scintil-lation fluid (MicroScint-20; PerkinElmer,Waltham, MA) was added to eachwell and radioactivity monitored on aWallac 1450 MicroBeta TriLux b-counter(PerkinElmer).

Duplicate positive standards (a poolof eight ATPase4A high-titer sera samples)

and individual negative control sera wereincluded in each assay set for calibrationand results expressed as a binding index(sample CPM – negative control CPM)/(positive control CPM – negative controlCPM). The upper limit of normal (index0.020) was established as the 95th percen-tile from receiver operator characteristiccurves generated from a panel of 220type 1 diabetes autoantibody–negativefirst-degree relatives of subjects withtype 1 diabetes attending the BarbaraDavis Center (median age 6.32 years[range 1–52]; 75% Hispanic, 25% whiteCaucasian/non-Hispanic). These samplesserved as healthy controls for subsequentassays.

SeraSera from 6,749 individuals with type 1diabetes were provided by the T1DGCCoordinating Center. These sampleswere from 3,384 males (50.14%) and3,365 females (49.86%). A total of5,223 subjects were non-Hispanicwhites (NHWs) (77.39%), with the ma-jority from 2,444 type 1 diabetes–affected sibling pair (ASP) families aswell as a small subset of 299 singletons(case subjects with type 1 diabetes withno parental sample); the average age ofthe NHW subjects was 22.5 years. Therewere 450 Hispanic (Mexican) ancestrysamples (6.67% of the total collection)from 62 ASP families as well as 324 sin-gletons, with an average age of 14.0years. There were 671 African ancestryparticipants (9.94% of the total collec-tion) from 30 multiplex families withan additional 609 singletons, with an av-erage age of 14.8 years. There were 365Asian ancestry participants (5.41% ofthe total) from 32 ASP families with anadditional 299 singletons, with an aver-age age of 17.4 years. There were 21mixed-ethnicity (0.31% of total) partici-pants, all singletons, and 19 subjects of“other ethnicity” (0.28% of the total)from nine ASP families as well as 10 sin-gletons, with an average age of 11 years.

GenotypesThe primary genetic risk for type 1 diabe-tes resides in the MHC, with specific riskand protective antigens defined by theHLA class I and HLA class II genes. HLAloci, HLA-A, -B, -C, -DRB1, -DQA1, -DQB1,and -DPA1, were genotyped at 4-digit res-olution in all T1DGC ASP and trio (two pa-rents and one type 1 diabetes–affectedchild) families, as well as in collections of

S30 Parietal Cell Autoantibodies in Type 1 Diabetes Diabetes Care Volume 38, Supplement 2, October 2015

case subjects with type 1 diabetes andcontrol subjects as previously described(12). Single nucleotide polymorphism(SNP) markers for the PTPN22 and CTLA4genes were genotyped in a subset ofT1DGC subjects as part of the T1DGCRapid Response Project (13), including2,520 of 6,749 subjects tested for ATP4Aautoantibodies.

Statistical AnalysisThe association of each autoantibodyphenotype with age, age of onset oftype 1 diabetes, duration of type 1 di-abetes, sex, HLA genotypes, CTLA4 SNPs,and PTPN22 SNPs was tested using lo-gistic regression methods with robustvariance estimators, stratified by ethnicity,and clustered by pedigree to accommo-date the correlation between affectedsiblings. Initial screening for genetic as-sociations was restricted to 38 common(minor allele frequency [MAF] .3%)HLA alleles in 5,219 NHWs. Similarmethods were used to test for associa-tion with 43 common PTPN22 andCTLA4 SNPs (MAF .10%) in 2,003NHWs.HLA class II haplotypes were recon-

structed using MENDEL software to as-sign the parental origin of HLA-DRB1,-DQA1, and -DQB1 alleles. Inmost cases,parental origin of the alleles at all threeHLA class II loci were assigned unambig-uously from family data, including bothparents and unaffected siblings. Haplo-types with unphased alleles were in-ferred from linkage disequilibriumpatterns.Estimates of the prevalence of ATP4A

autoantibodies in subjects of Hispanic(Mexican), African, and Asian ancestrywere standardized to the age and sexdistribution of NHWs. Regression coef-ficients, SEs, significance levels (Waldtest), and standardized prevalence esti-mates were calculated using STATA (ver-sion 11.2; StataCorp, College Station,TX). Statistical significance of genetic as-sociation tests was based on a P, 0.001threshold, based on a Bonferroni correc-tion allowing for 50 independent tests, aconservative criterion given the stronglinkage disequilibrium between HLA al-leles and SNP markers for both theCTLA4 and PTPN22 genes. Univariateand bivariate heritability analyses,allowing for phenotype-specific covari-ates, were conducted on 2,444 multi-plex families (NHWs only), estimating

heritability (h2) of each phenotype andgenetic correlation (rg) between pheno-types, and significance levels using theliability threshold model for discretetraits (14) as implemented in SOLAR(15,16).

RESULTS

Demographic AssociationsAmong6,749T1DGCsubjects, autoantibod-ies to the ATP4A antigen were detectedin 1,408 (20.9%), varying significantly byage, sex, and ethnicity (Table 1). Theprevalence of ATP4A autoreactivity in-creased from 13.6% in subjects #10years of age to 32.6% in those 40$ years(P , 10210). Prevalence of ATP4A auto-antibodies differed by sex (16.5% inmales vs. 25.3% in females; odds ratio[OR] 1.7; P , 10210) after adjustmentfor age and ethnicity. ATP4A auto-antibodies differed by ethnicity, withprevalence of 20.8% in NHWs comparedwith 36.3% of those of Hispanic/Mexicanancestry (P, 1027), 26.2% of African an-cestry (P = 0.05), and 16.7% of Asianancestry (P = 0.10) after adjustment forage and sex. There was no evidencethat the prevalence of ATP4A autoanti-bodies differed by disease duration afterallowance for the effects of age (P =0.212).

Risks to SiblingsIn analyses of 2,444 NHW type 1 diabe-tes ASP families, ATP4A autoantibodieswere detected in 46.5% of siblings ofATP4A autoantibody–positive probandscompared with 20.8% of all siblings (Fig.1). Anti-ATP4A sero-positive probandsand their siblings were more likely tohave other autoantibodies. There wasan increase in islet (anti-GAD65) autoan-tibodies (43.6% of all affected siblings,55.7% of ATP4A sero-positive pro-bands, and 53.3% of siblings of ATP4Asero-positive probands), thyroid (anti–thyroid peroxidase [TPO]) autoantibod-ies (27.4% of all affected siblings, 52.1%of ATP4A sero-positive probands, and39.6% of siblings of ATP4A sero-positiveprobands), and adrenal (anti–21-hydroxylase [21-OH]) autoantibodies(1.8% of all affected siblings, 4.5% ofATP4A sero-positive probands, and2.9% of siblings of ATP4A sero-positiveprobands). Those anti-ATP4A probandsand their siblings were less likely to haveother islet (anti–IA-2) autoantibodies(46.5% of all affected siblings, 40.6% of

ATP4A sero-positive probands, and 42.9%of siblings of ATP4A sero-positive pro-bands) or thyroid (anti-transglutaminase[TG]) autoantibodies (7.4% of all affectedsiblings, 6.8% of ATP4A sero-positiveprobands, and 5.4% of siblings of ATP4Asero-positive probands).

Univariate (Heritability) and Bivariate(Genetic Correlation) EstimatesFamily data for NHW subjects were usedto estimate heritability assuming theliability threshold model (14). The liabil-ity threshold model assumes that an in-dividual’s “liability” for developingATP4A autoantibodies is the sum oftheir inherited risk and environmentalexposures, with the expression ofautoantibodies occurring in individualswith liabilities exceeding a threshold de-fined by the prevalence in the studypopulation (20.8% for ATP4 autoanti-bodies in NHW subjects with type 1 di-abetes). To the extent that geneticdifferences contribute to differences inliability, the mean liability of siblings ofATP4A autoantibody–positive probandswill be higher than the mean liability inthe study population, and thus a higherfrequency of siblings will exceed thethreshold for ATP4A autoantibody–positivity (43.6% vs. 20.8%). Taken to-gether with the expected familialresemblance (genetic correlation) be-tween siblings (siblings share one-halfof their genome, on average, or 0.5),the shift in mean liability provides anestimate of heritability (h2).

Heritability analyses were performedfor all autoantibody phenotypes (ATP4A,IA-2, GAD65, TPO, 21-OH, and TG), withadjustment for phenotype-specific co-variates. In this sample of participantswith type 1 diabetes, ATP4A autoanti-body positivity showed the highestheritability, with genetic variation ac-counting for 95% of differences in risk(P , 10246) (Fig. 2). Similarly, the fiveother autoantibody phenotypes ex-hibited high heritability estimates,ranging from 71% for autoantibodiesto GAD65 (P, 10230) to 82% for auto-antibodies to IA-2 (P , 10243), withautoantibodies to TPO, 21-OH, and TGhaving heritability estimates of 81%(P , 10233), 79% (P , 1024), and 76%(P , 10213), respectively.

Bivariate analyses (estimating singletrait heritability and cross-trait geneticcorrelation) were performed in 2,444

care.diabetesjournals.org Wenzlau and Associates S31

T1DGC NHW families (as above) to de-termine whether latent genetic factorsshared by different autoantibody phe-notypes account for a significant frac-tion of the observed heritability. Theresults supported extensive shared ge-netic factors, particularly for ATP4A,TPO, and 21-OH autoantibodies. Thegenetic correlation estimates were0.496 for TPO and ATP4A (P , 10257),0.353 for TPO and 21-OH (P, 0.01), and0.369 for ATP4A and 21-OH (P , 0.001)(Fig. 2). Significant genetic correlationswere also observed for GAD65 autoanti-bodies with both TPO (0.359; P, 1027)and ATP4A autoantibodies (0.375; P ,10219). There was no significant geneticcorrelation estimated between GAD65with 21-OH autoantibodies (0.07; P =0.88). The genetic correlation between

the islet autoantibodies GAD65 and IA-2was relatively modest (0.12; P = 0.05).There was no evidence of a genetic as-sociation significantly different fromzero for IA-2 or TG autoantibodies withany other autoantibody phenotype.

HLA Class II Association WithAutoantibody PhenotypesOf 38 common HLA class I (HLA-A, -B,and -C) and HLA class II (HLA-DR, -DQ,and -DP) alleles tested for associationwith ATP4A autoantibodies in NHWs,four HLA class II alleles exhibited in-creased risk for ATP4A autoantibodies,while four HLA class II alleles had de-creased risk. There was no evidence ofassociation of ATP4A autoantibodieswith any HLA class I alleles. The HLA classII alleles associated with increased riskare DRB1*0404 (OR 1.53; P, 33 1024),DQA1*0301 (OR 1.40; P , 3 3 1027),DQB1*0302 (OR 1.38; P , 1.5 3 1026),and DPB1*0201 (OR 1.40; P , 1.6 31026). The class II alleles associated withdecreased risk are DRB1*0101 (OR 0.62;P , 1.7 3 1024), DQA1*0101 (OR0.63; P , 2.4 3 1025), DQB1*0501 (OR0.63; P , 3.2 3 1025), and DPB1*0301(OR 0.62; P , 2.1 3 1028). The ATP4Aassociations with DPB1*0201 andDPB1*0301 are independent of theATP4A associations with HLA-DR andHLA-DQ alleles; however, the effectsof the HLA-DR and HLA-DQ alleles can-not be distinguished owing to stronglinkage disequilibrium among the high-risk alleles DRB1*0404, DQA1*0301,and DQB1*0302 and low-risk allelesDRB1*0101, DQA1*0101, and DQB1*0501.In addition to the results observed inNHWs, the decreased risk associatedwith DPB1*0301 is also significant in sub-jects of Hispanic/Mexican ancestry (OR0.39; P , 3.1 3 1025). Further, the in-creased risk associated with DRB1*0404attained nominal significance in subjects

of African ancestry (OR 2.04; P , 0.05).Overall, the effects of the HLA class IIalleles appear similar across ethnicities,except for DPB1*0201, which is hetero-geneous across ethnic groups (Fig. 3).

The HLA class II alleles associatedwithATP4A autoreactivity are also associatedwith other autoantibody phenotypes(Fig. 4). The pattern of associations isconsistent with the pattern of familialclustering (Fig. 1) and strength of thegenetic correlation between phenotypes(Fig. 2), particularly for HLA-DRB1*0404and HLA-DPB1*0201. DRB1*0404 isassociated with ATP4A autoantibodies(OR 1.53; P, 3.03 1024), TPO autoan-tibodies (OR 1.84; P, 1.23 1029), and21-OH autoantibodies (OR 3.55; P ,1.0 3 1029). DRB1*0404 is not associ-ated, however, with GAD65 autoanti-bodies (OR 1.04; P = 0.67), IA-2autoantibodies (OR 0.99; P = 0.87), orTG autoantibodies (OR 0.93; P = 0.68)(Fig. 4).

HLA-DPB1*0201 is associated withATP4A autoantibodies (OR 1.40; P ,1.6 3 1026), TPO autoantibodies (OR1.32; P, 3.03 1025), and GAD65 auto-antibodies (OR 1.41; P , 9.0 3 1029)but not with IA-2 autoantibodies (OR1.00; P = 0.97). In addition, DPB1*0201shows negative association (protection)with TG autoantibodies (OR 0.73; P ,0.009) but not with 21-OH autoantibod-ies (OR 0.65; P , 0.08). In contrast, thelow-risk HLA-DRB1*0101 allele is nega-tively associated with many autoanti-body phenotypes: ATP4 autoantibodies(OR 0.62; P, 1.73 1024), GAD65 auto-antibodies (OR 0.58; P , 8.0 3 1029),TPO autoantibodies (OR 0.54; P, 2.031027), 21-OH autoantibodies (OR 0.07;P , 0.01), and TG autoantibodies(OR 0.38; P , 2.0 3 1025) but not IA-2autoantibodies (OR 0.85; P , 0.07).The effects of the low-risk HLA-DPB1*0301 allele appear specific to

Table 1—Demographic characteristics of the T1DGC data set and prevalence of ATP4A autoantibodies standardized to the ageand sex distribution of NHWs

Ethnicity Subjects (n) Multiplex sibships (n) Median (range) age, years % Female Standardized prevalence (95% CI)

NHW 5,223 2,444 19 (1–76) 48.9 0.208 (0.198, 0.220)

Hispanic 450 62 13 (2–60) 49.3 0.363 (0.301, 0.434)

Black 671 30 13 (1–64) 56.2 0.262 (0.220, 0.309)

Asian 365 32 16 (2–56) 52.3 0.167 (0.126, 0.218)

Mixed 21 0 13 (5–37) 76.2 0.171 (0.032, 0.506)

Other 19 9 10 (3–19) 15.8 0.387 (0.101, 1.000)

Figure 1—Genetic associations betweenautoantibody phenotypes shown as an in-creased prevalence of ATP4A autoantibodiesin siblings of ATP4A sero-positive pro-bands (far left); an increased prevalenceof GAD65, IA-2, and TPO autoantibodies;and decreased prevalence of 21-OH andTG autoantibodies in ATP4A sero-positivesubjects (gray) and their siblings (black)compared with all T1DGC subjects (lightgray). Analyses were restricted to NHWsubjects.

S32 Parietal Cell Autoantibodies in Type 1 Diabetes Diabetes Care Volume 38, Supplement 2, October 2015

ATP4A autoantibodies (OR 0.62; P, 2.131028) and TPO autoantibodies (OR 0.77;P , 4.0 3 1024), with an opposite effectfor GAD65 autoantibodies (OR 1.30; P ,5.0 3 1025). The DPB1*0301 allele hasno effect on IA-2 autoantibodies (OR0.94; P = 0.30), 21-OH autoantibodies(OR 1.10; P = 0.63), or TG autoantibodies(OR 1.13; P = 0.27).

CTLA4 and PTPN22 AssociationsGenotypes for SNPs in the PTPN22 re-gion on human chromosome 1p13.2(24 SNPs with MAF .0.05) and in theCTLA4 region on chromosome 2q33.3(19 common SNPs) were available in asubset of 2,003 NHWs. Of the PTPN22region SNPs, only the rs2476601 variant

in the coding region of PTPN22 (R620W),known to be associated with type 1diabetes risk, is significantly associatedwith ATP4A autoantibodies (OR 1.40;P , 0.001) (Fig. 5). The association ofrs2476601 with TPO autoantibodies isrelatively weak (OR 1.31; P , 0.01).There is no evidence of associationof PTPN22 rs2476601 with GAD65autoantibodies (OR 1.13; P = 0.18), 21-OH autoantibodies (OR 1.17; P = 0.57),TG autoantibodies (OR 1.27; P = 0.16),or IA-2 autoantibodies (OR 0.93; P = 0.45).

In contrast, four of the CTLA4 re-gion SNPs, rs3087243, rs15231811,rs11571293, and rs11571316, are nega-tively associated with ATP4A autoanti-bodies. CTLA4 SNP rs3087243 (CT60)has the strongest association withATP4A autoantibodies (OR 0.74; P ,3.03 1024). Results of conditional anal-yses suggest that the ATP4A associa-tions with rs15231811, rs11571293,and rs11571316 are most likely due tolinkage disequilibrium with rs3087243.Although none of the other five autoan-tibody phenotypes show significant as-sociation with rs3087243, there aresuggestive associations for 21-OH auto-antibodies (OR 0.61; P, 0.08), but giventhe low effect size (OR 0.62), wide con-fidence limits, and relatively smallsample size (38 21-OH autoantibody–positive NHWs with SNP genotypes),caution must be taken in this suggestiveinterpretation of effect. The effect sizeof CTLA4 SNP rs3087243 is not farfrom neutrality for IA-2 autoantibodies

(OR 1.02; P = 0.71), GAD65 autoanti-bodies (OR 0.94; P = 0.39), TPO auto-antibodies (OR 0.99; P = 0.84), and TGautoantibodies (OR 1.11; P = 0.39).

CONCLUSIONS

Gastric autoimmunity, defined as pres-ence of autoantibodies targeting theH+/K+-ATPase proton pump (PCA) ofthe gastric parietal cell, is understudied,lacking data or consensus about rela-tionships with age, sex, ethnicity, andkey genetic markers. The vast majorityof previous studies of gastric autoimmu-nity used indirect immunofluorescencefor detecting PCA in a relatively smallnumber of samples from patients withPA or other autoimmune diseases (17).Our results demonstrate that the prev-alence of ATP4A autoantibodies in-creases with age, is higher in femalesthan in males, and is higher in subjectsof Hispanic/Mexican and African ances-try and lower in those of Asian ancestrythan in NHWs (Table 1). ATP4A autoanti-bodies are also more common amongthose with HLA-DRB1*0404-DQA1*0301-DQB1*0302, HLA-DPB1*0201, orthe PTPN22 coding variant, R620W(rs2476601-T). ATP4A autoantibodies areless common among those with HLA-DRB1*0101-DQA1*0101-DQB1*0501,DPB1*0301, or the CTLA4 variantrs3087243-T (CT60) (Figs. 3–5). Except forDPB1*0201, the HLA effects are consistentacross ethnicity (Fig. 3). Although we didnot confirm previous reports of an as-sociation of PCA with DQA1*0501-DQB1*0301 (18) in those of NHW (OR1.02; P = 0.91) or Asian (OR 1.07; P =0.91) ancestry, there is suggestive evi-dence of association with DQA1*0501-DQB1*0301 in those of Hispanic/Mexican(OR 2.13; P = 0.038) and African (OR 1.91;P = 0.023) ancestry (data not shown).

There are few, if any, reports of ethnicdifferences in the frequency of PCA insubjects with type 1 diabetes of appre-ciable size. An early study observed sim-ilar frequencies in blacks and whites(19), whereas we observed an increasedfrequency in blacks (26.2%) and His-panics (36.3%) compared with NHWs(20.8%). The decreased frequency ofATP4A autoantibody positivity seen inAsians (16.7%) is consistent with studiesshowing that ABG and PA are extremelyrare among Asians in general (20), butthe increased frequencies of ATP4A au-toantibody positivity seen in blacks and

Figure 2—Heritabilities (h2) for ATP4A auto-antibodies and five other autoantibody phe-notypes (IA-2, GAD65, TPO, 21-OH, and TG)and genetic correlations (rg) between differ-ent autoantibody phenotypes in NHW sub-jects from T1DGC.

Figure 3—Ethnicity-specific ORs and 95% CIs for HLA alleles identified as positively (blue) ornegatively (yellow) associated with ATP4A autoantibodies in NHWs. A: HLA-DRB1 alleles asso-ciated with ATP4A autoantibodies in NHW subjects. B: HLA-DPB1 alleles associated with ATP4Aautoantibodies in NHW subjects.

care.diabetesjournals.org Wenzlau and Associates S33

Hispanics conflict with studies showingthat ABG and PA are also less common inthese groups. However, previous stud-ies of PA focused on elderly subjects.Differences seen in the relatively youngT1DGC cohort could relate to atypicalclinical and immunological expressionof PA in Hispanics and blacks, includingaccelerated onset of PA (21).

The T1DGC Autoantibody Workshop,providing data collected on NHW fami-lies with two or more type 1 diabetes–affected siblings, is the first systematicinvestigation of the heritability of ATP4Aautoreactivity and other autoantibodiesas well as bivariate heritability (geneticcorrelation) of ATP4A with other auto-antibodies (Figs. 1 and 2). In addition,these data permitted evaluation of theinfluence of genetic variation of the HLAclass I, HLA class II, CTLA4, and PTPN22genes on the autoantibodies. The re-sults demonstrate that the differencebetween ATP4A sero-positive (20.8%)and sero-negative (79.2%) subjects isprimarily genetic (;95% after allow-ance for age and sex), as is also thecase for each of five other organ-specificautoantibodies targeting the GAD65and IA-2 islet autoantigens, thy-roid TPO autoantigen, adrenal 21-OHautoantigen, and celiac-associated TGautoantigen (h2 . 70%). Although asignificant fraction of the heritabilityof ATP4A autoreactivity is shared (thegenetic correlation) with TPO (24.6%),21-OH (13.6%), and persistent GAD65(14.1%) autoantibodies, most of theheritability is explained by genetic influ-ences that do not relate to other auto-antibody phenotypes and may be morespecific to parietal cell autoimmunity(.75%).

The combined effects of class II HLA(DRB1*0404, DRB1*0101, DPB1*0201,

DPB1*0301), PTPN22 (rs2476601-T),and CTLA4 (rs3087243-T) variants ac-count for ;4–5% of PCA heritability,much of which is shared with one ormore autoantibody phenotypes in a pat-tern consistent with observed geneticcorrelations (Figs. 2, 4, and 5). A recentgenome-wide association study of PCAin participants with type 1 diabetes whowere unselected for family historyidentified a strong negative association(OR 0.34) with a marker tagging ABOblood type O (22,23). These resultscould relate to early reports of an in-creased frequency of blood type A inpatients with PA (24). Since no otherautoantibody phenotype or autoim-mune disease has been found to beassociated with ABO, the observed asso-ciation with PCA may contribute tothe parietal cell–specific component ofPCA heritability. That same study didnot detect an association with ei-ther rs2476601 (PTPN22) or rs3087243(CTLA4), even using a less stringent level(P, 0.01) of statistical significance. Thiscould be due to a lower frequency ofPCA (;10% vs. ;20% in the T1DGC Au-toantibodyWorkshop samples), younger-aged subjects with type 1 diabetes (13years vs. 22 years in T1DGC), or geneticdifferences between T1DGC ASP familiesversus unselected cases.

The genes and variants identified byscreening for association with ATP4Aautoantibodies may be involved in low-ering the threshold for developing au-toimmunity or in determining thespecific organs and tissues targeted bythe autoimmune response. Most auto-immune diseases have been found to beassociated with one or more HLA genes,particularly the HLA class II genes HLA-DRB1 and HLA-DQB1, and to a lesserextent HLA-DPB1 (25–27). The HLA classII genes encode the HLA-DR, -DQ, and-DP proteins residing on the surface ofantigen-presenting cells. Antigenbound to the groove of an HLA class IImolecule enables recognition by theT-cell receptor/CD3 complex, with addi-tional stimuli required for T-cell activa-tion and proliferation provided bygenes regulating the T-cell response,such as CTLA4 and PTPN22. DifferentHLA class II alleles encode differentpeptide-binding structures, thus im-pacting the binding affinity for differenttissue-specific autoantigens. Oftenviewed as primary determinants of

Figure 4—Autoantibody-specific ORs for HLA alleles identified as positively (blue) or negatively(yellow) associated with ATP4A autoantibodies in NHWs. A: HLA-DRB1 alleles associated withATP4A autoantibodies in NHW subjects. B: HLA-DPB1 alleles associated with ATP4A autoanti-bodies in NHW subjects.

Figure 5—Autoantibody-specific ORs for theCTLA4 CT60 variant (rs3087243-T) identifiedas negatively (yellow) associated withATP4A autoantibodies and the PTPN22R620W variant (rs2476601-T) identified aspositively (blue) associated with ATP4Aautoantibodies in NHW subjects.

S34 Parietal Cell Autoantibodies in Type 1 Diabetes Diabetes Care Volume 38, Supplement 2, October 2015

tissue specificity in autoimmunity(25,26,28), the association of the sameHLA alleles with multiple organ-specificautoimmune phenotypes could bedue to linkage disequilibrium betweendifferent HLA class II genes. Experi-mental data suggest that the HLA-DQb chain may be most critical forbinding immunogenic peptides derivedfrom pancreas in autoimmune diabe-tes (29–31), whereas the HLA-DRbchain may be most critical for bindingimmunogenic peptides derived fromthyroid in autoimmune thyroid dis-ease (32,33). Thus, t ight l inkagedisequilibrium between the genesencoding HLA-DR and -DQ could ex-plain the increased risk for both thy-roid and pancreatic autoimmunityassociated with specific DR-DQ hap-lotypes such as DRB1*0404-DQ8.Nevertheless, commonality in HLAassociations for a spectrum of autoim-mune diseases suggests a more generaleffect on susceptibility and resistanceto autoimmunity, particularly for thehigh-risk diabetes haplotypes, DR3-DQ2and DR4-DQ8 (34). Specific residues inboth DRb (35,36) and DPb (37,38)binding domains have been implicatedin multicellular autoimmunity, possiblyby influencing T-cell receptor docking,autoantigen recognition by CD4+T-helper cells, or both (39).The non-HLA genes, CTLA4 and

PTPN22, influence the T-cell responseto antigen presentation, and thus theirrole in multicellular autoimmunity ismore obvious than for the multiplegenes and variants in the HLA complex.The CTLA4 gene is expressed on acti-vated CD4+ and CD8+ T cells, wherethe gene product downregulates T-cellactivation, proliferation, and differen-tiation by interacting with costimulatorymolecules on the surface of antigen-presenting cells. Screening 19 commonpolymorphisms in CTLA4 identified an as-sociation of ATP4A autoantibodieswith the CTLA4 rs3087243 SNP. Subse-quent analyses showed little or no ev-idence for association with otherautoantibody phenotypes. The sameCTLA4 variant rs3087243 (CT60) is as-sociated with several autoimmune dis-eases, including autoimmune thyroiddisease, type 1 diabetes, Addison dis-ease, and celiac disease (40). Initialstructural and functional studies sug-gest that the rs3087243 variant may

modify the expression of alternativemRNA splice variants of CTLA4 in unsti-mulated CD4+ T cells (41), though a sub-sequent study failed to confirm theseresults (42). Similar to CTLA4, the asso-ciation of ATP4A with the PTPN22rs2476601 (R620W) SNP was identifiedby screening 24 commonpolymorphisms.In contrast, a trend in association wasseen consistently for multiple autoanti-body phenotypes. This variant is also as-sociated with multiple autoimmunediseases, and functional studies suggestthat the autoimmunity-predisposingvariant decreases T-cell receptor–mediated signaling and subsequentT-cell activation (43).

The T1DGC Autoantibody Workshopand the current analyses demonstratethat the pathogenesis of parietal cell, thy-roid, and adrenal autoimmunity in thosewith type 1 diabetes involves commongenetic pathways, perhaps also sharedwith persistent GAD65 autoimmunity,but not with persistent IA-2 or celiac au-toimmunity. The results are consistentwith many studies showing an increasedfrequency of PCA and other organ-specific autoantibodies in families withtype 1 diabetes, including reported asso-ciations between PCA and persistentGAD65 antibody positivity, but not per-sistent IA-2 antibody positivity (1,17),possibly reflecting a broader tissue distri-bution for GAD65 than for IA-2, andin line with studies showing that GAD65autoantibodies are less specific markersof pancreatic autoimmunity than areIA-2 autoantibodies (44,45). The co-morbidity of type 1 diabetes and otherchronic inflammatory diseases, par-ticularly Addison disease, chroniclymphocytic thyroiditis, and gastricautoimmunity, was one of the firstclues pointing to an autoimmune basisfor type 1 diabetes, ultimately distin-guishing type 1 and type 2 diabetes, aswell as autoimmune and nonautoim-mune forms of Addison disease andABG (2,46). Family studies showing overtor latent autoimmunity also occurred inpatients’ relatives and led to the recog-nition of autoimmune polyglandularsyndromes and to the notion that multi-ple autoimmune diseases can arise fromthe same inherited immunologicaldefect (18).

Two distinct genetic syndromes,APS1 and APS2, were described basedon autosomal recessive inheritance of

candidiasis-associated hypoparathy-roidism and polyendocrine failure insome families (APS1) and autosomaldominant or complex HLA-associatedinheritance of polyendocrine failurewithout candidiasis or hypothyroidismin others (APS2) (47,48). Whereas adre-nal insufficiency with polyendocrinefailure was initially considered the car-dinal manifestation of both APS1 andAPS2, more recent data show a muchwider range of familial autoimmune dis-ease associations, encompassing bothorgan-specific and systemic disorderssuch as systemic lupus erythematosusand rheumatoid arthritis (49). In con-trast to adrenal autoimmunity, gastricautoimmunity and PCA have beenconsidered a “minor” component ofvirtually all validated (APS1 andAPS2) and proposed (APS3 and APS4)(50) APS subtypes. However, results ofthe T1DGC ASP family study indicatethat PCA is a major component of agenetically distinct autoimmune clus-ter that includes both thyroid andadrenal autoimmunity and may wellrepresent a pathogenetic subtype oftype 1 diabetes. Our results point tothe important role of family studiesin dissecting the complex genetics ofautoimmunity targeting multiple cellsand tissues.

Funding. This research uses resources providedby the T1DGC, a collaborative clinical studysponsored by the National Institute of Diabetesand Digestive and Kidney Diseases (NIDDK), theNational Institute of Allergy and Infectious Dis-eases, the National Human Genome ResearchInstitute, the Eunice Kennedy Shriver NationalInstitute of Child Health and Human Develop-ment, and JDRF and supported by U01DK062418. The study was also supported inpart by grants from the BDC-JDRF Autoimmu-nity Prevention Center (4-2007-1056); theNational Institutes of Health (NIH) Diabetesand Endocrine Research Center (NIDDK P30DK57516); and the NIH (DK32083 andAI50864).Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. J.M.W. researcheddata and wrote the manuscript. P.R.F. wrotethe manuscript and was responsible for thestatistical analyses. T.J.G., L.M.F., and B.A.researched data. J.C.H. conceived the study,researched data, contributed to discussion,and wrote, reviewed, and edited the manu-script. J.C.H. is the guarantor of thiswork and, assuch, had full access to all the data in the studyand takes responsibility for the integrity of thedata and the accuracy of the data analysis.

care.diabetesjournals.org Wenzlau and Associates S35

References1. De Block CEM, De Leeuw IH, Van Gaal LF.Autoimmune gastritis in type 1 diabetes: a clin-ically oriented review. J Clin Endocrinol Metab2008;93:363–3712. Whittingham S, Mackay IR. Autoimmunegastritis: historical antecedents, outstandingdiscoveries, and unresolved problems. Int RevImmunol 2005;24:1–293. Banka S, Ryan K, Thomson W, Newman WG.Pernicious anemia - genetic insights. Autoim-mun Rev 2011;10:455–4594. Baxter AG, Jordan MA, Silveira PA, WilsonWE, Van Driel IR. Genetic control of susceptibil-ity to autoimmune gastritis. Int Rev Immunol2005;24:55–625. Johnston C, Millward BA, Leslie RD, Pyke DA,Bottazzo GF. Are thyrogastric autoantibodiesassociated with an increased susceptibility todeveloping type 1 (insulin-dependent) diabe-tes? A study in identical twins. Autoimmunity1990;6:195–2016. Karlsson F, Burman P, Loof L, Mardh S. Majorparietal cell antigen in autoimmune gastritiswith pernicious anemia is the acidproducingH+,K+adenosine triphosphatase of the stomach.J Clin Invest 1988;81:4754797. Toh BH, Sentry JW, Alderuccio F. The causa-tive H+/K+ ATPase antigen in the pathogenesisof autoimmune gastritis (Abstract). ImmunolToday 2000;21:3483548. Alderuccio F, Toh BH, Tan SS, Gleeson PA, vanDriel IR. An autoimmune disease with multiplemolecular targets abrogated by the transgenicexpression of a single autoantigen in the thy-mus. J Exp Med 1993;178:419–4269. Fisher JM, Taylor KB. A comparison of auto-immune phenomena in pernicious anemia andchronic atrophic gastritis. N Engl J Med 1965;272:499–50310. Chuang JS, Callaghan JM, Gleeson PA, TohBH. Diagnostic ELISA for parietal cell autoantibodyusing tomato lectin-purified gastric H+/K(+)-AT-Pase (proton pump). Autoimmunity 1992;12:1–711. Wenzlau JM, Gardner TJ, Frisch LM,Davidson HW, Hutton JC. Development ofa novel autoantibody assay for autoimmunegastritis in type 1 diabetic individuals. DiabetesMetab Res Rev 2001;8:887–89012. Mychaleckyj JC, Noble JA, Moonsamy PV,et al.; T1DGC. HLA genotyping in the interna-tional Type 1 Diabetes Genetics Consortium.Clin Trials 2010;7(Suppl.):S75–S8713. Brown WM, Pierce JJ, Hilner JE, Perdue LH,Lohman K, Lu L, de Bakker PIW, Irenze K, ZiaugraL, Mirel DB. Overview of the Rapid Responsedata. Genes Immun 2009;10:S5–S1514. Falconer D. The inheritance of liability to cer-taindiseases, estimated fromthe incidenceamongrelatives. Ann Hum Genet 1965;29:51–7615. Duggirala R, Williams JT, Williams-BlangeroS, Blangero J. A variance component approachto dichotomous trait linkage analysis using athreshold model. Genet Epidemiol 1997;14:987–99216. Almasy L, Dyer TD, Blangero J. Bivariatequantitative trait linkage analysis: pleiotropyversus co-incident linkages. Genet Epidemiol1997;14:953–958

17. de Graaff L, Smit J, Radder J. Prevalence andclinical significance of organspecific autoanti-bodies in type 1 diabetes mellitus. Neth J Med2007;65:235–24718. De Block CEM, De Leeuw IH, VertommenJJF, et al.; Belgian Diabetes Registry. Beta-cell,thyroid, gastric, adrenal and coeliac autoimmu-nity and HLA-DQ types in type 1 diabetes. ClinExp Immunol 2001;126:236–24119. Maclaren NK, RileyWJ. Thyroid, gastric, andadrenal autoimmunities associatedwith insulin-dependent diabetes mellitus. Diabetes Care1985;8(Suppl. 1):34–3820. Irvine WJ, McFadzean AJ, Todd D, Tso SC,Yeung RT. Pernicious anaemia in the Chinese:a clinical and immunological study. Clin Exp Im-munol 1969;4:375–38621. Carmel R. Reassessment of the relativeprevalences of antibodies to gastric parietalcell and to intrinsic factor in patients with per-nicious anaemia: influence of patient age andrace. Clin Exp Immunol 1992;89:74–7722. Plagnol V, Howson JMM,SmythDJ, et al.; Type1DiabetesGenetics Consortium.Genome-wide as-sociation analysis of autoantibody positivity in type1 diabetes cases. PLoS Genet 2011;7:e100221623. Aird I, Bentall HH, Bingham J, et al. An as-sociation between blood group A and perni-cious anaemia: a collective series froma number of centres. BMJ 1956;2:723–72424. Callender ST, Denborough MA, Sneath J.Blood groups and other inherited charactersin pernicious anaemia. Br J Haematol 1957;3:107–11425. Trowsdale J. The MHC, disease and selec-tion. Immunol Lett 2011;137:1–826. Goris A, Liston A. The immunogenetic archi-tecture of autoimmune disease. Cold SpringHarb Perspect Biol 2012;4:a00726027. Varney MD, Valdes AM, Carlson JA, et al.;Type 1 Diabetes Genetics Consortium. HLADPA1, DPB1 alleles and haplotypes contributeto the risk associated with type 1 diabetes: anal-ysis of the Type 1 Diabetes Genetics Consortiumfamilies. Diabetes 2010;59:2055–206228. Michels AW, Gottlieb PA. Autoimmune poly-glandular syndromes. Nat Rev Endocrinol 2010;6:270–27729. Kwok WW, Domeier ME, Johnson ML,Nepom GT, Koelle DM. HLA-DQB1 codon 57 iscritical for peptide binding and recognition. JExp Med 1996;183:1253–125830. Yu B, Gauthier L , Hausmann DHF,Wucherpfennig KW. Binding of conserved isletpeptides by human and murine MHC class IImolecules associated with susceptibility to typeI diabetes. Eur J Immunol 2000;30:2497–250631. Lee KH, Wucherpfennig KW, Wiley DC.Structure of a human insulin peptide-HLA-DQ8complex and susceptibility to type 1 diabetes.Nat Immunol 2001;2:501–50732. Sawai Y, DeGroot LJ. Binding of human thy-rotropin receptor peptides to a Graves’ disease-predisposing human leukocyte antigen class IImolecule. J Clin Endocrinol Metab 2000;85:1176–117933. HuberA,Menconi F, Corathers S, JacobsonEM,Tomer Y. Joint genetic susceptibility to type 1 di-abetes and autoimmune thyroiditis: from epidemi-ology tomechanisms. Endocr Rev2008;29:697–725

34. Fernando MMA, Stevens CR, Walsh EC,et al. Defining the role of the MHC in autoim-munity: a review and pooled analysis. PLoSGenet 2008;4:e100002435. Hammer J, Gallazzi F, Bono E, et al. Peptidebinding specificity of HLA-DR4molecules: corre-lation with rheumatoid arthritis association.J Exp Med 1995;181:1847–185536. Nepom BS, Nepom GT, Coleman M, KwokWW. Critical contribution of beta chain residue57 in peptide binding ability of both HLA-DR and-DQ molecules. Proc Natl Acad Sci USA 1996;93:7202–720637. Nicholson I, Varney M, Kanaan C, et al. Al-loresponses to HLA-DP detected in the primaryMLR: correlation with a single amino acid differ-ence. Hum Immunol 1997;55:163–16938. D ıaz G, Catalfamo M, Coiras MT, et al.HLA-DPbeta residue 69 plays a crucial role inallorecognition. Tissue Antigens 1998;52:27–3639. Gough SC, Simmonds MJ. The HLA regionand autoimmune disease: associations andmechanisms of action. Curr Genomics 2007;8:453–46540. Gough SC, Walker LS, Sansom DM. CTLA4gene polymorphism and autoimmunity. Immu-nol Rev 2005;204:102–11541. Ueda H, Howson JM, Esposito L, et al. Asso-ciation of the T-cell regulatory gene CTLA4 withsusceptibility to autoimmune disease. Nature2003;423:506–51142. Mayans S, Lackovic K, Nyholm C, et al. CT60genotype does not affect CTLA-4 isoform ex-pression despite association to T1D and AITDin northern Sweden. BMC Med Genet 2007;8:343. Stanford SM, Mustelin TM, Bottini N. Lym-phoid tyrosine phosphatase and autoimmunity:human genetics rediscovers tyrosine phospha-tases. Semin Immunopathol 2010;32:127–13644. Tuomi T, Bjorses P, Falorni A, et al. Anti-bodies to glutamic acid decarboxylase and insulin-dependent diabetes in patientswith autoimmunepolyendocrine syndrome type I. J Clin EndocrinolMetab 1996;81:1488–149445. Soderbergh A, Myhre AG, Ekwall O, et al.Prevalence and clinical associations of 10 de-fined autoantibodies in autoimmune polyendo-crine syndrome type I. J Clin Endocrinol Metab2004;89:557–56246. Betterle C, Lazzarotto F, Presotto F. Auto-immune polyglandular syndrome type 2: the tipof an iceberg? Clin Exp Immunol 2004;137:225–23347. Wirfalt A. Genetic heterogeneity in autoim-mune polyglandular failure. Acta Med Scand1981;210:7–1348. Neufeld M, Maclaren NK, Blizzard RM. Twotypes of autoimmune Addison’s disease associ-ated with different polyglandular autoimmune(PGA) syndromes. Medicine (Baltimore) 1981;60:355–36249. Weetman AP. Diseases associated with thy-roid autoimmunity: explanations for the ex-panding spectrum. Clin Endocrinol (Oxf) 2011;74:411–41850. Betterle C, Zanchetta R. Update on autoim-mune polyendocrine syndromes (APS). Acta Bio-med 2003;74:9–33

S36 Parietal Cell Autoantibodies in Type 1 Diabetes Diabetes Care Volume 38, Supplement 2, October 2015