1690 Diabetes Volume 65, June 2016

Kerry A. McLaughlin,1 Carolyn C. Richardson,1,2 Aarthi Ravishankar,1 Cristina Brigatti,3 Daniela Liberati,4 Vito Lampasona,4 Lorenzo Piemonti,3 Diana Morgan,5 Richard G. Feltbower,5 and Michael R. Christie1,2

Identification of -7 as a Target of Autoantibodies in Type 1 Diabetes

Diabetes 2016;65:1690–1698 | DOI: 10.2337/db15-1058

The presence of autoantibodies to multiple-islet autoan- allowed the development of high-throughput autoanti- tigens confers high risk for the development of type 1 body assays for clinical diagnosis of type 1 diabetes and diabetes. Four major autoantigens are established (insulin, identification of individuals at risk for disease. Evidence glutamate decarboxylase, IA2, and zinc transporter-8), from both animal studies and human trials (3,4) indicates but the molecular identity of a fifth, a 38-kDa membrane that type 1 diabetes may be prevented in individuals at glycoprotein (Glima), is unknown. Glima antibodies have risk. Hence, a range of therapies to interfere with immune been detectable only by immunoprecipitation from ex- responses has proved to be effective in preventing disease tracts of radiolabeled islet or neuronal cells. We sought to development in animal models of diabetes (5) and in slow- fi identify Glima to enable ef cient assay of these autoan- ing the loss of b-cell function occurring in the months tibodies. Mouse brain and lung were shown to express after disease diagnosis in humans (6–8). There is now a Glima. Membrane glycoproteins from extracts of these focus on the development of procedures to interfere spe- organs were enriched by detergent phase separation, cifically in immune responses that cause type 1 diabetes, lectin affinity chromatography, and SDS-PAGE. were also immunoaffinity purified from brain extracts using requiring knowledge of the major targets of the autoim- autoantibodies from the sera of patients with diabetes mune response. There is no single common autoimmune fi before SDS-PAGE. Eluates from gel regions equivalent to target, and individuals differ in the antigen speci city of 38 kDa were analyzed by liquid chromatography–tandem the autoimmune responses that develop. Four major hu- fi mass spectrometry for identification. Three pro- moral autoantigens have been identi ed in type 1 diabetes teins were detected in samples from the brain and lung by defining the specificity of autoantibodies in the disease:

IMMUNOLOGY AND TRANSPLANTATION extracts, and in the immunoaffinity-purified sample, but insulin (9), glutamate decarboxylase (10), IA2 (11), and zinc not in the negative control. Only tetraspanin-7, a multipass transporter-8 (12). Autoantibodies to a fifth major hu- transmembrane glycoprotein with neuroendocrine ex- moral autoantigen, a 38-kDa glycosylated membrane pro- pression, had physical characteristics expected of Glima. tein (Glima), have been detected in 19–38% of patients Tetraspanin-7 was confirmed as an autoantigen by dem- with type 1 diabetes, with significantly higher prevalence onstrating binding to autoantibodies in type 1 diabetes. (up to 50%) in children (13–15). The molecular identity of We identify tetraspanin-7 as a target of autoimmunity in Glima has for many years proved elusive, hampering the diabetes, allowing its exploitation for diabetes prediction characterization of autoimmunity to the protein and the and immunotherapy. development of sensitive, specific autoantibody assays. Glima is expressed in pancreatic b-cell and neuronal cell Detection of circulating autoantibodies to pancreatic islets lines; is hydrophobic; is heavily N-glycosylated, having (1), and identification of their molecular targets (2), has affinity for the lectin wheat germ agglutinin; and has a

1Diabetes Research Group, Division of Diabetes & Nutritional Sciences, King’s Corresponding author: Michael R. Christie, [email protected]. College London, London, U.K. Received 31 July 2015 and accepted 1 March 2016. 2School of Life Sciences, University of Lincoln, Lincoln, U.K. This article contains Supplementary Data online at http://diabetes 3Diabetes Research Institute, Istituto di Ricovero e Cura a Carattere Scientifico, .diabetesjournals.org/lookup/suppl/doi:10.2337/db15-1058/-/DC1. San Raffaele Scientific Institute, Milan, Italy 4Division of Genetics and Cellular Biology, Istituto di Ricovero e Cura a Carattere © 2016 by the American Diabetes Association. Readers may use this article as Scientifico, San Raffaele Scientific Institute, Milan, Italy long as the work is properly cited, the use is educational and not for profit, and 5Division of Epidemiology & Biostatistics, School of Medicine, University of Leeds, the work is not altered. Leeds, U.K. diabetes.diabetesjournals.org McLaughlin and Associates 1691 core protein backbone of ;22 kDa (13–15). The aim of Glima for binding to Glima antibodies in serum from a this study was to take advantage of these known physical high-titer Glima antibody–positive patient. Mouse kidney, properties to prepare Glima-enriched extracts for the iden- brain, heart, liver, thyroid, muscle, salivary gland, thymus, tification of the autoantigen by mass spectrometry. pancreas, spleen, adrenal, pituitary, and lung were dis- sected, frozen in liquid nitrogen, and stored at 280°C RESEARCH DESIGN AND METHODS before extraction. Tissues were homogenized in homog- Patients enization buffer (10 mmol/L HEPES, pH 7.4, 0.25 mol/L Serum samples were obtained from 40 patients with type 1 sucrose, 10 mmol/L benzamidine), and membrane frac- diabetes (12–26 years of age) within 6 months of diagnosis tions sedimented by centrifugation at 15,000g for 30 min from clinics in West Yorkshire with informed consent for at 4°C. Supernatants were removed and pellets extracted in screening for high-titer Glima antibodies, from 94 additional 2% Triton X-100 extraction buffer for 2 h on ice. Extracts patients (12–63 years of age) for assay verification, and were centrifuged at 15,000g for 30 min at 4°C, and super- from 52 individuals without diabetes as negative control natants were collected. The protein concentrations of ex- subjects. Approval for the analysis of autoantibodies in sera tracts were determined using the Pierce BCA Protein Assay from these individuals was obtained from the Yorkshire and Kit (ThermoFisher Scientific, Loughborough, U.K.). the Humber–Bradford Leeds Research Ethics Committee. For the competition assay, wheat germ agglutinin agarose eluates from extracts of 35S-methionine–labeled GT1.7 cells Screen of Sera of Patients With Type 1 Diabetes for m Glima Antibodies were prepared as described above. Aliquots (20 L) of 3 5 Glima antibodies were detected by a modification of GT1.7 cell glycoproteins containing 3 10 cpm radiolabel- m immunoprecipitation assays previously described (13– ed proteins were incubated for 18 h at 4°C with 5 Lof m 15) using the neuronal mouse cell line GT1.7 as a source of serum from patient 029 alone, or with 10 Lofde- m antigen. Endogenous proteins in GT1.7 cells were labeled tergent extracts containing 100 gofextractedprotein by incubation in methionine-free DMEM medium contain- from each mouse tissue. Immune complexes were captured ing 4 MBq/mL 35S-methionine for 7 h at 37°C. Cells were on Protein A-Sepharose and processed for SDS-PAGE and washed with HEPES buffer (10 mmol/L HEPES, pH 7.4, autoradiography. 150 mmol/L NaCl, 10 mmol/L benzamidine) and stored at Partial Purification of Glima From Mouse Brain 2 80°C. Frozen cell pellets were extracted in HEPES buffer and Lung containing 2% Triton X-100 for 2 h on ice and insoluble Mouse brain and lung, shown to express Glima in the tissue material removed by centrifugation at 15,000g for 15 min screen (see RESULTS), were homogenized in ice-cold homog- at 4°C. Membrane glycoproteins were isolated by incubat- enization buffer in a Dounce homogenizer, and cell debris ing cell extracts with wheat germ agglutinin-agarose on ice was removed by centrifugation at 500g for 5 min at 4°C. A for 30 min and, after washing in HEPES buffer containing membrane fraction was prepared by centrifugation of the 0.5 mmol/L methionine, 100 mg/L BSA, and 0.5% Triton supernatant at 10,000g for 15 min at 4°C, and the pellet X-100, were eluted in the same buffer containing 0.5 mol/L was washed and extracted in 2% Triton X-114 extraction m N-acetyl glucosamine. Aliquots (20 L) of eluate containing buffer for 2 h at 4°C. Insoluble material was removed by 3 5 3 10 cpm radiolabeled protein were incubated with centrifugation at 10,000g for 15 min at 4°C, and a detergent m 5 L of test sera for 18 h at 4°C, and immune complexes phase was prepared by heat-induced phase separation as m were captured on 5 L of Protein A-Sepharose. Immuno- previously described (13). Fractions were added to wheat m precipitated proteins were eluted in 15 L of SDS-PAGE germ agglutinin-agarose at a ratio of 100 mL lectin-agarose Loading Buffer (Novex; Life Technologies, Paisley, U.K.) with to 5 mg total protein and incubated overnight at 4°C with heating at 90°C for 5 min and were subjected to SDS-PAGE gentle mixing. The beads were washed twice with HEPES on 12% polyacrylamide gels. After electrophoresis, gels were buffer containing 0.5% Triton X-100 and twice in NOG incubated in 40% v/v methanol, 2.5% v/v acetic acid, and buffer (1% N-octyl-glucopyroside in HEPES buffer). Wheat subsequently in Enlightning Autoradiographic Enhancer germ agglutinin–binding proteins were eluted in 0.5 mol/L (PerkinElmer, Coventry, U.K.), each for 30 min. Gels were N-acetyl-glucosamine in NOG buffer. fi fi dried and contacted with X-ray lm (BioMax MR lm; Eluates were concentrated using a Pierce SDS-PAGE Kodak, Watford, U.K.) for up to 2 weeks. After exposure, Sample Preparation Kit (Life Technologies), solubilized in fi X-ray lm was developed to detect radiolabeled proteins SDS-PAGE Loading Buffer (Novex; Life Technologies) for fi speci cally immunoprecipitated by sera from patients with 10 min at 60°C, and electrophoresed on 12% Bis-Tris gels type 1 diabetes, with bands detected in the 38,000 Mr re- in MOPS running buffer. Gels were stained with Brilliant gion, indicating positivity for Glima antibodies. Blue G-Colloidal Coomassie (Sigma-Aldrich, Poole, U.K.), Tissue Expression Screen and gel slices corresponding to the 38,000 Mr region were excised for mass spectrometry. To identify mouse organs containing the highest levels of Glima for use in antigen purification, competitive binding Immunoaffinity Purification With Glima Antibodies studies were performed using detergent extracts of organs For immunoaffinity purification, 250 mL of pooled sera as unlabeled competitors with 35S-methionine–labeled from three patients with high levels of Glima antibodies 1692 Tetraspanin-7 in Type 1 Diabetes Diabetes Volume 65, June 2016 was used with 250 mL of sera from three antibody-negative Scaffold (version 4.3.2; Proteome Software Inc., Port- individuals as a negative control. Sera were incubated with land, OR) was used to validate MS/MS-based peptide and Protein A-Sepharose (250 mL) for 1 h at room temperature protein identifications, which were assigned by Peptide with rolling and washed three times in borate buffer (100 Prophet algorithms (16,17) and accepted at .95.0% prob- mmol/L boric acid, pH 8.3). Antibodies were cross-linked ability. The UniProt database was manually searched for to Protein A-Sepharose with 20 mmol/L dimethyl pimelimi- the physical characteristics of the proteins identified, in- date in borate buffer for 1 h. Unreacted sites were blocked cluding molecular weight, tissue distribution, and glycosyl- with20mmol/Lethanolaminefor10minandwashedbe- ation. Of those identified, only one, tetraspanin-7 (Tspan7), fore use. matched the known properties of Glima (see RESULTS)and Triton X-114 detergent phase–purified amphiphilic pro- was characterized further. teins from mouse brains prepared as above were added to the Glima antibody–positive and Glima antibody–negative Immunohistochemistry beads and incubated overnight at 4°C with mixing. Beads Tspan7 localization in rodent tissues was performed by fi fi were washed with 0.5% Triton X-100 in HEPES buffer prior immunohistochemistry. Sections of formalin- xed, paraf n- to elution in 2% SDS at 90°C for 10 min. The eluate was embedded rat brain, pituitary gland, pancreas, adrenal concentrated to 20 mLusingtheSDS-PAGESample gland, lung, muscle, heart, liver, kidney, spleen, and thymus Preparation Kit and was subjected to SDS-PAGE and were dewaxed, and subjected to epitope retrieval in Colloidal Coomassie gel staining as above, and gel slices a microwave pressure cooker in 10 mmol/L citric acid (pH 6.0) and 0.05% Tween 20. Endogenous peroxidase in the 38,000 Mr region were excised for mass spectrometry. activity was inhibited with 0.3% H2O2,andnonspecific In-Gel Trypsin Digestion and Mass Spectrometry binding was blocked with 25% nonimmune swine serum in PBS. Primary antibody to Tspan7 (anti-TM4SF2, catalog of 38,000 Mr Proteins #HPA003140; Sigma-Aldrich) was applied at 1:1,000 di- Gel slices representing 38,000 Mr regions of all samples were processed using the Pierce In-Gel Tryptic Digestion lution and incubated overnight at 4°C. Antibody labeling Kit (ThermoFisher Scientific) according to the manufac- was detected with the Envision Kit (Dako, Ely, U.K.) turer instructions, and the trypsin-treated extracts were vac- according to the manufacturer instructions, and sections ’ uum dried and stored at 220°C prior to mass spectrometry. were counterstained in Mayer sHematoxylinSolution Samples were reconstituted in 30 mL of 50 mmol/L am- (Sigma-Aldrich) and visualized by microscopy. monium bicarbonate for 30 min at room temperature and Cloning and Expression of Recombinant Tspan7 centrifuged at 15,000g for 15 min to remove insoluble cDNA for the coding region of mouse Tspan7 was am- material. Samples were transferred to autosampler tubes, plified by RT-PCR from the mouse islet cell line Min6 m – and 10 L of each was analyzed by liquid chromatography using primers (59-GAATTCATGGCATCGAGGAGAATGG- tandem mass spectrometry (LC-MS/MS). Peptides were 39 and 59-AGATCTCACCATCTCATACTGATTGGC-39) that m resolved by reversed-phase chromatography on a 75- m introduce EcoR1 and BglII sites at the 59 and 39 ends, C18 EASY column using a linear gradient of acetonitrile in respectively, with the native stop codon removed to allow fl 0.1% formic acid at a ow rate of 300 nL/min over 50 min expression as a fusion protein with a COOH-terminal pu- fi on an EASY Nano LC system (ThermoFisher Scienti c). The rification tag. The PCR product was cloned into the pFLAG- eluate was ionized by electrospray ionization using an CTS expression vector for protein expression in Escherichia Orbitrap Velos Pro mass spectrometer (ThermoFisher Scien- coli BL21 cells after induction with isopropyl b-D-1- fi ti c) operating under Xcalibur (version 2.2; ThermoFisher thiogalactopyranoside. Expressed protein was extracted from fi Scienti c) and by precursor ions selected based on their cells with Hen Egg Lysozyme (1 mg/mL) in PBS containing intensity for sequencing by collision-induced fragmenta- 10 mmol/L benzamidine, 1 mmol/L phenylmethylsulfonyl tion. The tandem mass spectrometry (MS/MS) analyses fluoride for 30 min at room temperature, followed by fi were conducted using collision energy pro les that were incubation in Triton X-100 (0.1%) for 5 min and DNase chosen based on the charge/mass ratio and the charge (1 mg/mL) for 10 min. The lysate was centrifuged at state of the peptide. 10,000g for 10 min at 4°C, and the supernatant was Tandem mass spectra were processed into peak lists used in immunoprecipitation assays. using Proteome Discoverer (version 1.3; ThermoFisher Scientific). All MS/MS samples were analyzed using Tspan7 Binding to Autoantibodies in Type 1 Diabetes Mascot (version 2.2.06; Matrix Science, London, U.K.) Individual Glima antibody–positive and Glima antibody– searching the UniProt Mus musculus database, assuming negative human sera (15 mL) were incubated with Protein digestion with trypsin. Mascot was searched with a frag- A-Sepharose (15 mL), and the Ig captured was cross-linked ment ion mass tolerance of 0.80 Da and a parent ion tol- to beads with dimethyl pimelimidate (18). Bead-bound erance of 10.0 ppm, with oxidation of methionine and antibodies were incubated overnight at 4°C with Triton carbamidomethylation of cysteine as variable modifica- X-100 extracts of mouse brain or with lysates of E. coli tions. Each data set was analyzed with a reverse FASTA expressing recombinant mouse Tspan7. Beads were washed database acting as a decoy. three times in 0.5% Triton X-100 in HEPES buffer, and diabetes.diabetesjournals.org McLaughlin and Associates 1693 captured proteins were subjected to SDS-PAGE and indicative of Glima immunoreactivity in these tissues Western blotting using rabbit anti-Tspan7 antibody (Fig. 2). (anti-TM4SF, catalog #HPA003140; Sigma-Aldrich) at Identification of Glima Candidate Proteins by 1:250 dilution overnight at 4°C. Immunoprecipitated Mass Spectrometry Tspan7 was detected with goat anti-rabbit IgG-peroxidase Extracts enriched for glycosylated membrane proteins from (catalog #A0545; Sigma-Aldrich) and SuperSignal West Pico both brain and lung were prepared using Triton X-114 fi Chemiluminescent substrate (ThermoFisher Scienti c). phase separation of amphiphilic membrane proteins fol- Luminescent Immunoprecipitation Assay for Detection lowed by wheat germ agglutinin affinity purification. of Tspan7 Antibodies Proteins migrating at 38,000 Mr by SDS-PAGE were tryp- The coding region of human Tspan7 was cloned into the sinized and analyzed by LC-MS/MS. A total of 65 candi- pCMVTnT vector as a fusion with nanoluciferase (NanoLuc) datesinbrainand25inlungwereidentified, of which 20 (Promega,Southampton,U.K.)atthe39 end. The construct were common to both samples (Supplementary Table 1). was transfected into HEK 293 cells with ExpiFectamine Glycosylated membrane proteins immunoprecipitated from (ThermoFisher Scientific) for transient expression of an- brain extracts by antibodies in the high Glima antibody titer tigen. Transfected cells were extracted in either 2% patients’ serum pool were also subjected to SDS-PAGE and Triton X-114 or passive lysis buffer (Promega), and insol- LC-MS/MS analysis, and 3 of the 20 protein candidates uble material was removed by centrifugation. Triton common to brain and lung were present in the Glima X-114 extracts were subjected to phase separation as de- antibody–positive sample, but not in the negative con- scribed above and analyzed for fusion protein by Western trol(SupplementaryTable1).Thesewereasfollows:1) blotting with rabbit antibodies to NanoLuc (a gift from cytoplasmic actin-1 (Actb), a ubiquitous nonglycosylated Promega) or Tspan7 (anti-TM4SF). Luciferase expression cytoskeletal protein with a predicted molecular weight of in the cell extracts was quantified by luminometry using 42 kDa; 2) guanine nucleotide-binding protein G(i) sub- Nano-Glo assay reagent (Promega), and aliquots of pas- unit a-2 (Gnai2), a nonglycosylated membrane–associated sive lysis buffer extracts containing 106 light units of 40-kDa protein with a wide tissue distribution; and 3) antigen were incubated with 5 mL of serum at 4°C for Tspan7, a hydrophobic four-transmembrane domain pro- 16 h prior to the capture of antibody complexes on Pro- tein with a core molecular weight of 27.5 kDa, five putative tein A-Sepharose. Mouse brain extracts (150 mg protein) N-glycosylation sites, and a neuroendocrine distribution. or lysates of E. coli expressing Tspan7 (250 mgprotein) Tspan7 closely matched the known properties of Glima, were added to reactions as sources of Tspan7 to com- and additional studies were performed to compare proper- pete for antibody binding. Complexes were washed and ties and validate Tspan7 as the autoantigen. luciferase activity immunoprecipitation determined by Localization of Tspan7 in Rat Tissues luminometry with Nano-Glo assay reagent. The tissue distribution of Tspan7 was determined by immunohistochemistry for comparison with patterns of RESULTS Selection of Glima Antibody–Positive Sera To identify patients with high levels of Glima antibodies for immunoaffinity purification, sera from 40 patients with recent-onset type 1 diabetes were screened by immuno- precipitation using radiolabeled mouse GT1.7 cell extracts (Fig. 1). Intense diffuse 38,000 Mr bands, which are indic- ative of high levels of Glima antibodies, were detected for three patients (Fig. 1, patients 029, 037, and 110). These three sera were used for subsequent Glima characterization and purification. Weaker 38,000 Mr bands indicative of Glima antibody positivity were detected in an additional 11 patients (Fig. 1). Tissue Specificity of Glima Expression To identify large organs in which Glima is expressed at suitably high levels for antigen purification, competi- Figure 1—Autoradiogram showing a screen of serum samples from tive binding studies were performed in which detergent patients with type 1 diabetes for Glima antibodies by immunopre- cipitation of the 38,000 Mr protein from extracts of GT1.7 cells with extracts of normal mouse tissues acted as unlabeled detection by SDS-PAGE and autoradiography. A normal control competitors with radiolabeled Glima from GT1.7 cell serum (-ve) and a patient with type 1 diabetes previously deter- lysates for binding to antibodies in serum from patient mined to be positive for Glima antibodies (SL) were included in 029, who was strongly Glima antibody positive. Ex- the assay. The location of Glima on the autoradiograph is marked. An indication of whether the samples were determined to be neg- tracts of brain, pituitary, and lung reduced the in- ative (2) or positive (+) for Glima antibodies is shown under each fi tensity of radiolabeled 38,000 Mr protein, which is lane of the gure. 1694 Tetraspanin-7 in Type 1 Diabetes Diabetes Volume 65, June 2016

Glima expression in the competition experiments described rabbit polyclonal antibodies to both NanoLuc and Tspan7 above. Strong immunolabeling for Tspan7 was detected in detected diffuse 38,000 Mr bands (the expected size of the the rat brain, in particular the cerebral cortex, hippocam- nonglycosylated fusion protein) as the dominant immu- pus, cerebellum, striatum, and thalamus (Fig. 3A); in the noreactivity in cells transfected with the construct, with pancreatic islets (Fig. 3B); in the anterior pituitary (Fig. additional bands at approximately 80,000 Mr (Fig. 4C). 3C); and in epithelial cells lining the alveoli in the lung The 38,000 Mr protein partitioned into the detergent on (Fig. 3D). These observations agree with the Glima immuno- temperature-induced phase separation in Triton X-114. reactivity described above and previously (13,15). Weak Transfected cell extracts were used in immunoprecipitation Tspan7 immunolabeling was also found in cells of the ad- studies with normal control sera or with sera from Glima renal gland (Fig. 3E). No evidence of Tspan7 expression was antibody–positive and Glima antibody–negative patients found in the exocrine pancreas (Fig. 3B), muscle, heart, liver, with type 1 diabetes. All but one of the control subjects kidney, spleen, or thymus (data not shown). (control sample V015) (n = 52) had low levels of Tspan7 Immunoprecipitation of Tspan7 by Antibodies antibodies (Fig. 4D). Four patients with high levels of in Type 1 Diabetes Glima antibodies (Fig. 1) also immunoprecipitated high To demonstrate that Tspan7 is a target for autoantibodies luciferase activity in the Tspan7 antibody assay (Fig. 4D), fi in type 1 diabetes, extracts of mouse brain and lysates of and signi cantly higher levels of Tspan7 antibodies were – E. coli expressing recombinant mouse Tspan7 were subject found in Glima antibody positive patients than Glima – , to immunoprecipitation with Glima antibody–positive and antibody negative patients (P 0.0001; Mann-Whitney Glima antibody–negative sera followed by Western blot- U test). In competition assays, natural or recombinant ting with a rabbit antibody to Tspan7. A 38,000 M band Tspan7 in brain or E. coli extracts partially (control sample r – representing Tspan7 was selectively immunoprecipi- V015) or completely (Glima antibody positive patients tated from mouse brain detergent extract by three Glima with type 1 diabetes) blocked antibody binding to the antibody–positive sera, but not by control samples. The NanoLuc-Tspan7 construct (Fig. 4E). Control sample V015 bands detected comigrated with the brain lysate control did not bind Tspan7 from mouse brain extracts when (Fig. 4A). Antibodies in the sera of patients with type 1 tested in the Western blotting assay. A second set of 94 diabetes also specifically immunoprecipitated Tspan7 from patients with recent onset of type 1 diabetes was also bacterial lysates containing recombinant protein (Fig. 4B). tested in the Tspan7 antibody assay. Using a cutoff of ; mean 63 SDs of control subjects (omitting the outlier), Here, the protein migrated at 22,000 Mr, which is con- sistent with a lack of glycosylation in bacteria (Fig. 4B). The 40 (43%) were positive for Tspan7 antibodies (Fig. 4D). fi results con rm Tspan7 as a target of autoantibodies in DISCUSSION type 1 diabetes. Autoantibodies to Glima in type 1 diabetes were first Analysis of Tspan7 Antibodies by Luminescence reported in 1996 (13), but its molecular identity has since Immunoprecipitation Assay then remained unknown. We used mass spectrometry of Patients screened for Glima antibodies were analyzed for Glima-enriched fractions of brain and lung to search for Tspan antibodies by immunoprecipitation of recombinant likely candidates for Glima. LC-MS/MS analysis identified NanoLuc-tagged human Tspan7. Western blotting with 65 proteins in 38,000 Mr gel samples of amphiphilic mem- brane glycoproteins from brain and 25 proteins from lung, of which 20 were common to both (Supplementary Table 1). Additional LC-MS/MS analysis of immunoaffinity- purified proteins isolated from brain extracts using Protein A-Sepharose–coupled Igs from Glima antibody–positive sera (with similar preparations from Glima antibody– negative sera as a negative control) further narrowed down the potential candidates for the testing of auto- antigenicity. Only six proteins detected in the brain or lung extracts were also present in the immunoaffinity- purified sample but absent in the negative control. Of these, five (Actb, Gnai2, Sfxn5 [sideroflexin-5], Kctd12 [potassium channel tetramerization domain 12], and Tuba1b [tubulin a-1B chain]) had a considerably higher . Figure 2—Autoradiograph of tissue expression screen demonstrat- predicted molecular weight ( 36 kDa) than expected for ing competition for Glima antibody binding to Glima antibody– the nonglycosylated Glima protein (;22 kDa) (15). Fur- positive (Glima Ab +ve) serum sample from patient 029 by proteins thermore, Actb, Gnai2, and Tuba1a are ubiquitous cyto- in Triton X-100 detergent (100 mg) extracts of normal mouse tis- sues. Serum from an individual without diabetes was included as a plasmic proteins lacking the amphiphilic characteristics negative control (-ve control). Reduced intensity of the 38,000 Mr expected of Glima. Sfxn5 and Kctd12 were not detected band is indicative of Glima immunoreactivity in that tissue. in the lung extract. Tspan7 was, consequently, the most diabetes.diabetesjournals.org McLaughlin and Associates 1695

Figure 3—Immunohistochemical analysis of Tspan7 expression in rat tissues. Sections of formalin-fixed, paraffin-embedded rat tissues were labeled with rabbit anti-serum to Tspan7 and labeling detected by a peroxidase/3,3-diaminobenzidine–based system, with positive labeling detected as brown staining under the microscope. Representative images of the labeling of tissue sections of brain (A), pancreas (B), pituitary (C), lung (D), and adrenal gland (E) are shown.

promising candidate for Glima identified in the LS-MS/MS being detected in regions of the adult mouse brain and analyses. lung (20). In the pancreas, Tspan7 is found specifically in Tspan7 is a member of the tetraspanin family, members the islets of Langerhans (21). Functionally, tetraspanin of which share structural characteristics of four trans- family members are involved in mediating signal trans- membrane domains, with one short (EC1) and one long duction events and have been noted to regulate cell de- (EC2) extracellular loop (19). Four of the putative N- velopment, activation, growth, and motility through the glycosylation sites are contained within the EC2 domain trafficking of other transmembrane proteins (22). Both with a further site located in EC1. The presence of multiple the extracellular and intracellular domains are able to in- transmembrane domains and N-glycosylation sites within teract with other proteins, and a number of Tspan7 is consistent with the hydrophobic properties and bind integrins, thereby forming links with the actin cyto- heavy N-glycosylation previously reported for Glima (13,15). skeleton (23). Mutations in the Tspan7 are associated There are four amino acid differences between mouse with X-linked mental retardation and neuropsychiatric dis- and human Tspan7, all of which are located in the long eases, potentially as a result of impaired ability of the actin extracellular loop. The tissue distribution of Tspan7 ex- to drive neurite outgrowth (24). The ability of pression has not been widely investigated, but analysis tetraspanins to form complexes with other membrane and of the transcriptional activity of the Tspan7 gene has cytosolic proteins may explain the copurification of multiple shown restricted tissue distribution with high levels protein fragments identified in the LC-MS/MS analysis. 1696 Tetraspanin-7 in Type 1 Diabetes Diabetes Volume 65, June 2016

Figure 4—A: Tspan7 labeling of Western blots of mouse brain proteins immunoprecipitated by antibodies in sera from Glima antibody– negative (-ve) and Glima antibody–positive (+ve) patients with recent onset of type 1 diabetes. The migration of molecular weight markers 23 (10 3 Mr) on the gel and the localization of Tspan7-specific bands at 38,000 Mr are marked. B: Tspan7 labeling of Western blots of proteins from lysates of Tspan7 expressing E. coli immunoprecipitated by antibodies in sera from Glima antibody–negative and Glima antibody–positive patients with recent onset of type 1 diabetes. The localization of Tspan7-specific bands at ;22,000 Mr is marked; the migration of molecular weight markers is as in A. In both panels A and B, IgG heavy chains (50,000 Mr), light chains (25,000 Mr), and cross- linked Ig (>100,000 Mr) from all serum samples were also detected on the blot as a consequence of cross-reactivity with the peroxidase- conjugated anti-rabbit detection antibody. C: Tspan7 was expressed as a fusion protein with NanoLuc, and Triton X-114 extracts of cells were subject to heat-induced phase separation. Detergent and aqueous phases were subject to SDS-PAGE and Western blotting with 23 antibodies to NanoLuc or Tspan7. The migration of molecular weight markers are shown (10 3 Mr). D: Detergent extracts of NanoLuc- tagged Tspan7 were immunoprecipitated with normal control sera (Controls) (n = 30), sera from Glima antibody (Ab)–positive patients with type 1 diabetes (T1D) (n = 15), and the sera of Glima antibody–negative patients with type 1 diabetes and luciferase activity associated with each immunoprecipitate determined by luminometry. Data are plotted as luciferase activity immunoprecipitated in kilo light units (kLU), and sample codes for control or individuals with diabetes with high levels of antibodies are shown. E: Samples from control individuals or Glima antibody–positive patients with type 1 diabetes were tested for competitive binding by natural or recombinant Tspan7 in the LIPS by performing the immunoprecipitations in the absence (black bars) or presence of 150 mg of mouse brain extract (white bars) or 250 mgof lysates of E. coli expressing recombinant Tspan7 (hatched bars). Assays were performed in triplicate. The addition of brain and E. coli lysate significantly blocked antibody binding for all samples (P < 0.0001; ANOVA with Dunnett correction for multiple comparisons), with the exception of control sample CH.

Proteins immunoprecipitated by Glima antibody–positive in E. coli. A luminescence-based immunoprecipitation system sera from brain extracts, or from bacterial extracts con- (LIPS) (25) using NanoLuc-tagged human Tspan7 expressed taining recombinant Tspan7, also bound rabbit antibodies in mammalian cells showed that patients with high levels to Tspan7 by Western blotting, confirming Tspan7 as the of Glima autoantibodies were also strongly positive in the target of the antibodies. Glima autoantibodies bound both anti-Tspan7 LIPS. The relationship between Glima and the glycosylated natural 38,000 Mr Tspan7 in brain and the Tspan autoreactivity is imperfect, which may in part be 22,000 Mr nonglycosylated form of the protein expressed the consequence of difficulties in ascertaining whether or diabetes.diabetesjournals.org McLaughlin and Associates 1697 not diffuse Glima bands are present on autoradiographs of Prior Presentation. Parts of this study were presented in abstract form at immunoprecipitation reactions (Fig. 1). Individuals without the 75th Scientific Sessions of the American Diabetes Association, Boston, MA, diabetes had low levels of Tspan7 antibodies, with the 5–9 June 2015. exception of one strongly positive control. Antibodies in References this control serum bound poorly to natural Tspan7 from 1. Bottazzo GF, Doniach D. Islet-cell antibodies (ICA) in diabetes mellitus mouse brain extracts, suggesting that antibodies in this (evidence of an autoantigen common to all cells in the islet of Langerhans). Ric particular sample may bind epitopes not displayed on the Clin Lab 1978;8:29–38 natural protein. The SDS-PAGE gel migration of the fu- 2. Weenink SM, Christie MR. Autoantibodies in diabetes. In Autoantibodies and sion protein indicated that the majority of recombinant Autoimmunity: Molecular Mechanisms in Health and Disease. Pollard M, Ed. luciferase-tagged proteins were not subject to the heavy Weinheim, Germany, VCH Verlag GmbH & Co, 2005, p. 321–349 glycosylation found on the natural protein, which is in- 3. Skyler JS. Primary and secondary prevention of type 1 diabetes. Diabet Med – dicative of incorrect membrane insertion, protein folding, 2013;30:161 169 or intracellular targeting of the fusion protein required 4. Michels A, Zhang L, Khadra A, Kushner JA, Redondo MJ, Pietropaolo M. Prediction and prevention of type 1 diabetes: update on success of prediction and for appropriate posttranslational modification. Incorrect struggles at prevention. Pediatr Diabetes 2015;16:465–484 folding or lack of glycosylation may reveal antibody epi- 5. Luo X, Herold KC, Miller SD. Immunotherapy of type 1 diabetes: where are topes not normally displayed on the natural Tspan7. Fur- we and where should we be going? Immunity 2010;32:488–499 ther optimization of Tspan7 expression should permit the 6. Keymeulen B, Vandemeulebroucke E, Ziegler AG, et al. Insulin needs after development of high-throughput assays for the detection CD3-antibody therapy in new-onset type 1 diabetes. N Engl J Med 2005;352: of diabetes-associated Tspan7 autoantibodies with high 2598–2608 sensitivity and specificity. 7. Pescovitz MD, Greenbaum CJ, Krause-Steinrauf H, et al.; Type 1 Diabetes Autoimmunity to major autoantigens in type 1 diabetes TrialNet Anti-CD20 Study Group. Rituximab, B-lymphocyte depletion, and pre- appears within the first 5 years of life in at-risk children servation of beta-cell function. N Engl J Med 2009;361:2143–2152 (26), with individual immune responses developing se- 8. Feutren G, Papoz L, Assan R, et al. Cyclosporin increases the rate and quentially rather than simultaneously (27). Autoimmunity length of remissions in insulin-dependent diabetes of recent onset. Results of a – in the disease is therefore progressive, with the order of multicentre double-blind trial. Lancet 1986;2:119 124 9. Palmer JP, Asplin CM, Clemons P, et al. Insulin antibodies in insulin- appearance of autoimmune responses to individual anti- – fi dependent diabetics before insulin treatment. Science 1983;222:1337 1339 gens differing between individuals and diversi cation of 10. Baekkeskov S, Aanstoot HJ, Christgau S, et al. Identification of the 64K the immune response being essential for disease progres- autoantigen in insulin-dependent diabetes as the GABA-synthesizing enzyme sion; therefore, disease rarely develops in individuals in glutamic acid decarboxylase. Nature 1990;347:151–156 whom autoimmunity develops only to single autoantigens 11. Payton MA, Hawkes CJ, Christie MR. Relationship of the 37,000- and (28). The optimum strategy currently adopted for assessing 40,000-M(r) tryptic fragments of islet antigens in insulin-dependent diabetes to diseaseriskistoscreenindividualsforthepresenceofauto- the protein tyrosine phosphatase-like molecule IA-2 (ICA512). J Clin Invest 1995; antibodies to multiple islet autoantigens. The inclusion of 96:1506–1511 Tspan7 antibodies in the screen may improve the sensitivity 12. Wenzlau JM, Juhl K, Yu L, et al. The cation efflux transporter ZnT8 and specificity of disease prediction inlargepopulations,and (Slc30A8) is a major autoantigen in human type 1 diabetes. Proc Natl Acad Sci will provide a fuller description of the major autoimmune U S A 2007;104:17040–17045 fi responses that are developing in that individual, which is 13. Aanstoot HJ, Kang SM, Kim J, et al. Identi cation and characterization of necessary for guiding the selection of autoantigen-specific glima 38, a glycosylated islet cell membrane antigen, which together with GAD65 and IA2 marks the early phases of autoimmune response in type 1 diabetes. immunotherapeutic agents to prevent the disease. J Clin Invest 1996;97:2772–2783 14. Winnock F, Christie MR, Batstra MR, et al.; Belgian Diabetes Registry. Autoantibodies to a 38-kDa glycosylated islet cell membrane-associated antigen Acknowledgments. The authors thank Raymond Chung and Malcolm in (pre)type 1 diabetes: association with IA-2 and islet cell autoantibodies. Di- Ward of King’s College London Proteomics Facility for LC-MS/MS analyses. abetes Care 2001;24:1181–1186 Funding. This study was funded by a research grant from Diabetes UK (grant 15. Roll U, Turck CW, Gitelman SE, et al. Peptide mapping and characterisation 11/0004297) and by a Society for Endocrinology Early Career Award to K.A.M. of glycation patterns of the glima 38 antigen recognised by autoantibodies in type C.C.R. was supported by a PhD Studentship from King’s College London Grad- I diabetic patients. Diabetologia 2000;43:598–608 uate School. Research by C.B., D.L., V.L., and L.P. was conducted within the 16. Keller A, Nesvizhskii AI, Kolker E, Aebersold R. Empirical statistical model to framework of the Italian Ministry of Research project “Ivascomar project, Clus- estimate the accuracy of peptide identifications made by MS/MS and database ter Tecnologico Nazionale Scienze della Vita ALISEI.” search. Anal Chem 2002;74:5383–5392 Duality of Interest. No potential conflicts of interest relevant to this article 17. Nesvizhskii AI, Keller A, Kolker E, Aebersold R. A statistical model for iden- were reported. tifying proteins by tandem mass spectrometry. Anal Chem 2003;75:4646–4658 Author Contributions. K.A.M. designed the study, researched and analyzed 18. Harlow E, Lane D. Antibodies: A Laboratory Manual. Cold Spring Harbor, NY, data, and wrote the manuscript. C.C.R., A.R., C.B., D.L., V.L., L.P., D.M., and R.G.F. Cold Spring Harbor Laboratory Press, 1988 researched and analyzed data. M.R.C. designed the study, researched and analyzed 19. Hemler ME. Tetraspanin functions and associated microdomains. Nat Rev data, and contributed to the writing of the manuscript. All authors reviewed and edited Mol Cell Biol 2005;6:801–811 the manuscript and approved the final version for submission. M.R.C. is the guarantor 20. Lizio M, Harshbarger J, Shimoji H, et al.; FANTOM consortium. Gateways to of this work and, as such, had full access to all the data in the study and takes the FANTOM5 promoter level mammalian expression atlas. Genome Biol 2015; responsibility for the integrity of the data and the accuracy of the data analysis. 16:22 1698 Tetraspanin-7 in Type 1 Diabetes Diabetes Volume 65, June 2016

21. Hald J, Galbo T, Rescan C, et al. Pancreatic islet and progenitor cell surface profiles using LIPS (luciferase immunoprecipitation systems). Biochem Biophys markers with cell sorting potential. Diabetologia 2012;55:154–165 Res Commun 2007;352:889–895 22. Maecker HT, Todd SC, Levy S. The tetraspanin superfamily: molecular fa- 26. Ziegler AG, Bonifacio E; BABYDIAB-BABYDIET Study Group. Age-related islet cilitators. FASEB J 1997;11:428–442 autoantibody incidence in offspring of patients with type 1 diabetes. Diabetologia 23. Berditchevski F. Complexes of tetraspanins with integrins: more than meets 2012;55:1937–1943 the eye. J Cell Sci 2001;114:4143–4151 27. Barker JM, Barriga KJ, Yu L, et al.; Diabetes Autoimmunity Study in the 24. Zemni R, Bienvenu T, Vinet MC, et al. A new gene involved in X-linked Young. Prediction of autoantibody positivity and progression to type 1 diabetes: mental retardation identified by analysis of an X;2 balanced translocation. Nat Diabetes Autoimmunity Study in the Young (DAISY). J Clin Endocrinol Metab Genet 2000;24:167–170 2004;89:3896–3902 25. Burbelo PD, Ching KH, Mattson TL, Light JS, Bishop LR, Kovacs JA. Rapid 28. Ziegler AG, Rewers M, Simell O, et al. Seroconversion to multiple islet autoan- antibody quantification and generation of whole proteome antibody response tibodies and risk of progression to diabetes in children. JAMA 2013;309:2473–2479