Serodiagnostic of celiac disease: Patient derived monoclonal anti-gliadin

Serodiagnostic of celiac disease: Patient derived monoclonal anti-gliadin antibody harnessed in a novel inhibition assay Øyvind Steinsbø 1, Siri Dørum 1, Knut E.A. Lundin 1,2, Ludvig M. Sollid 1. 1 Centre for Immune Regulation and Department of Immunology, University of Oslo and Oslo University Hospital Rikshospitalet, Oslo, Norway 2 Department of Gastroenterology, Oslo University Hospital Rikshospitalet, Oslo, Norway Corresponding author: Ludvig M. Sollid Postal address: Department of Immunology Oslo University Hospital - Rikshospitalet 0424 Oslo, Norway Telephone number: +47 23 07 35 00 Fax number: +47 23 07 35 10 E-mail address: l.m.sollid@medisin.uio.no Disclosures: A patent application covering gliadin-specific monoclonal antibodies has been submitted with Øyvind Steinsbø, Ludvig M. Sollid, Carole H. Dunand and Patrick C. Wilson as inventors. The remaining authors declare no competing financial interests. Author contributions: L.M.S. managed the project. K.E.A.L. provided biological material. Ø.S. and S.D. performed the experiments. Ø.S. and L.M.S analyzed the data and wrote the manuscript with input from co-authors. Running title: Serodiagnostic of celiac disease Total word count: 5794 1

Abstract Background & Aims Serum antibodies to deamidated gliadin peptides (DGP) are used in diagnostic work-up of celiac disease. Recombinant gliadin-specific human monoclonal antibodies (hmabs) have been established by cloning antibody genes of single IgA plasma cells of celiac lesions. Here, we have developed a novel serologic inhibition assay based on the principle of inhibiting by test serum antibodies the binding of one such gliadin-specific hmab (1002-1E03) to its preferred target peptide antigen. Methods Two peptides, a 34mer and a 26mer found in -gliadins and low-molecular weight glutenins, were identified as preferred targets from antibody pull-down of a digest of gliadin treated by transglutaminase 2. The peptides, which contained repeat sequence motifs, were tested together with the hmab 1002-1E03 in an amplified luminescent proximity homogeneous (AlphaLISA) inhibition assay. A constructed peptide with a sequence motif repeated three times was also tested. Serum samples from untreated celiac patients (n=106) and two control groups (198 blood donors, 151 Crohn s disease patients) were analyzed in the assay as well as in conventional commercial anti-tg2 IgA and anti-dgp IgG serologic assays. Results In the serologic inhibition assays, the 34mer peptide demonstrated the highest diagnostic performance. This assay showed higher specificity than anti-dgp IgG, higher sensitivity than anti-tg2 IgA and detected the majority of the anti-tg2 IgA negative celiac patients without significant increase in false discovery rate. Conclusions The novel anti-gliadin inhibition assay was highly specific, and was particularly useful for detecting celiac disease patients scoring negative for anti-tg2 IgA. 2

Keywords Celiac disease; serology; antibodies; gliadin 3

Celiac disease is an immune-mediated disorder caused by a harmful immune response to ingested gluten proteins from grains of wheat, barley and rye. The disease develops in genetically predisposed individuals expressing the HLA molecules DQ2.5, DQ2.2 or DQ8. 1 Establishing the diagnosis can be challenging due to a varied clinical presentation of the disease. Despite this, the prevalence of diagnosed patients has increased steadily over the last decades. 2, 3 Development of better serologic tests is a major reason for more efficient diagnostics, and disease-specific antibodies have become increasingly important in the diagnostic workup of the disorder. 4-7 Gluten is a heterogeneous mixture of proteins, and in wheat, gluten consists of many different gliadin and glutenin proteins. 8 Upon gastrointestinal digestion, these proteins are proteolyzed to shorter peptides by digestive enzymes. Both T cells 9 and antibodies 10, 11 of celiac disease patients demonstrate better reactivity to gluten peptides that have become deamidated. Evidence suggests that the enzyme transglutaminase 2 (TG2) is mediating this posttranslational modification by 12, 13 conversion of certain glutamine residues to glutamate. TG2 also is a target for autoantibodies in celiac disease. 14 Serum antibodies reactive with TG2 and with deamidated gliadin peptides (DGP) are used as diagnostic markers for the disease. In a large population study, initial combined screening for both anti-tg2 and anti-dgp antibodies was found to give the most sensitive and efficient diagnostic workup. 15 Several other studies support the view that this is the best serodiagnostic approach for celiac disease. 16, 17 Most official guidelines, however, are more cautious, and tend to recommend primary use of anti-tg2 IgA in order to minimize false discovery rate due to inferior specificity of the present anti-dgp assays. 4, 7 Existing anti-dgp assays are based on synthetic antigenic peptides harboring gliadin motifs observed to give the highest antibody reactivity in patients sera. The literature is consistent in reporting enhanced antibody reactivity to synthetic gliadin peptides with glutamate residues at positions corresponding to glutamine residues targeted by TG2. 10, 18-20 Although highly sensitive, a recent study questioned the diagnostic value of these assays as they showed low predictive values in disease 4

susceptible individuals. 21 Moreover, high false discovery rates have been reported in disease control groups of autoimmune liver diseases and inflammatory bowel disease, 22-24 and meta-analyses have generally described the anti-dgp assays to be less specific as compared to the anti-tg2 assays. 25 Low specificity was also a documented problem with former anti-tg2 test kits, 26 but the use of pure recombinant TG2 antigen in new assays have considerably reduced this problem. 27 We recently established gliadin-specific human monoclonal antibodies (hmabs), by cloning the antibody genes of single IgA plasma cells isolated from small intestinal lesions of patients with active celiac disease. 11 In total, 38 hmabs from eleven different patients were produced. We have identified the preferred targets of these hmabs from a proteolytic digest of gliadin treated by TG2 (Dørum et al., submitted). Most of the hmabs preferred binding to long, TG2-deamidated gliadin peptides that each harbor several copies of epitopes. Our aim in this study was to utilize one of these hmabs and its preferred target peptides in a novel serologic competition assay. The principle was to measure binding of gliadin-specific hmab to target peptide, where inhibition of this binding indicated presence of gliadin-specific antibodies in serum. 5

Materials and Methods Patients and controls Celiac disease patients with biopsy-confirmed diagnosis were enrolled at Oslo University Hospital Rikshospitalet (n = 106, mean age 38 years, range 17-72, 69 females and 37 males). The Marsh scores were the following: Marsh 3A: n=17; Marsh 3B n=40; Marsh 3C: n=49. All patients were either positive for HLA-DQ2.5, HLA-DQ2.2 and/or HLA-DQ8. Serum was sampled at the day of endoscopy. Exclusion criterion was documented gluten-free diet prior to the endoscopy. None of the enrolled patients had a previous diagnosis of celiac disease. Informed consent for participation was obtained from all patients. Blood donors were recruited from Oslo University Hospital Ullevål aiming to match the sex and age distribution of the celiac disease group (n = 198, mean age 41 years, range 20-60 years, 132 females and 66 males). Donors with known celiac disease were excluded from the study. Serum and blood samples were de-identified after collection according to the protocol of the ethical approval. For sera giving positive serologic test result(s), the respective blood samples were used for genotyping. Crohn s disease patients under anti-tnf treatment were recruited from Oslo University Hospital Rikshospitalet (n = 151, mean age 38 years, range 18-82 years, 66 females and 85 males). The protocols were approved by Regional Committees for Medical Research Ethics in South East Norway (ethical approvals 2013/1352, 2014/432 and S-97201). ELISA peptide reactivity of gliadin-specific hmabs Biotinylated synthetic gliadin peptides containing GSGSGS C-terminal spacer (500 nm) (GL Biochem) were used as antigens in streptavidin-coated ELISA plates (Nunc, 436014). Gliadin-specific hmabs were titrated in fourfold dilutions starting at 6.67 nm. Alkaline phosphatase conjugated anti-human IgG (Southern Biotech 9040-04) in 1:4000 dilution was used as detecting antibody, and visualized 6

with phosphatase substrate (Sigma S0942-200TAB) reactivity measured at 405 nm. PBS ph 7.4 was used as buffer, and the plates were washed three times with 0.05% Tween in PBS between each step. Serologic inhibition assay The serologic inhibition assay was developed on an amplified luminescent proximity homogeneous assay (Alpha)LISA platform (Perkin Elmer) with customized AlphaLISA acceptor beads (Perkin Elmer) conjugated with the gliadin-specific hmab (1002-1E03), according to manufacturer s recommendations. Three different target peptides were tested in separate assays: biotin- QPEQPFPEQPEQPEQPFPQPEQPFPWQPEQPFPQ (termed b- 34), biotin- QPEQPFPEQPEQPEQPFPQPEQPFPW (b- 26), and biotin-pqpeqpfpqpeqpfpqpeqpfpqp (b- QPEQPFP 3 ). Target peptides were dissolved in AlphaLISA buffer (0.1% Pluviol in PBS ph 7.4) to 2.5 or 1.25 nm (b- 34), 5 or 7.5 nm (b- 26), and 4 or 6 nm (b-qpeqpfp 3 ). Next, 10 l was transferred to each well and incubated with 5 l serum for 1 hour at room temperature (RT). AlphaLISA 1002-1E03 acceptor bead stock solution was diluted to 3.5 μg/ml in AlphaLISA buffer, and 15 l was added per well. The plate was incubated in 45 minutes at RT in dark. Alphascreen streptavidin donor bead (Perkin Elmer) solution was diluted to 7 μg/ml in AlphaLISA buffer, 15 l was added per well, and the plate was incubated in 45 minutes at RT in dark. AlphaLISA signal was measured with Envision Multilabel Plate Reader (Perkin Elmer). The serum samples were tested in duplicates and the mean value of the two logarithmic AlphaLISA signals were used. Together with a reference curve, generated from healthy control serum spiked with known concentrations of gliadin-specific hmabs (1002-1E01, 1002-1E03 and 1130-3B03), the mean value was used to estimate serum antibody reactivity of equivalent amounts of gliadin-specific hmabs (mg/l). Anti-TG2 IgA and anti-dgp IgG Celikey Varelisa ttg IgA (Phadia, 181 96) and QUANTA Lite Gliadin IgG II (INOVA, 704520) were used for serologic testing of anti-tg2 IgA and anti-dgp IgG. The serologic testing was performed by the routine service laboratory at the Department of Immunology at Oslo University Hospital Ullevål. 7

Statistical analyses The statistical analyses and graphs of serologic test results were generated with GraphPad Prism 6 (GraphPad Software Inc.). 8

Results Natural binding targets of a patient-derived gliadin-specific antibody In the inhibition assay we sought to establish, the pair of the gliadin-specific hmab and its target peptide would be integral parts that should be carefully chosen. Out of a panel of 38 hmabs available, we selected the gliadin-specific hmab 1002-1E03 that showed good reactivity and a strong preference for TG2-deamidated gliadin antigen. 11 In an analysis of the preferred target antigen of the hmab by antibody pull-down from a proteolytic digest of gliadin treated by TG2, 381 different peptides were identified by mass spectrometric sequencing (Dørum et al., submitted). Here we report an in-depth analysis of these peptides. Most of them were fragments of -gliadins, -gliadins and low-molecular weight glutenins, and the majority contained deamidated residues (85 % of peptides in post pull-down versus 27 % of peptides in pre pull-down sample, Fisher s exact test p <0.0001). The pull-down with this antibody preferentially resulted in long peptide fragments (28.3±0.4 residues post versus 21.4±0.4 residues pre pull-down, Student s t-test p <0.0001) (Figure 1A). This is similar to what was observed for the other gliadin-specific hmabs (Dørum et al., submitted). As was observed for other anti-gliadin hmabs, the hmab 1002-1E03 pulled down long fragments that shared identical or very similar motif sequences (Figure 1B, Table 1). The most common 3mer motifs (QQP and PQQ) were present in 378 of the 381 peptides in multiple copies (1785 and 1634 times, respectively). Similarly, the most common 4mer motif was present in 374 peptides in 1479 copies, and this was a single amino acid extension of the 3mer motif (PQQP). The same was observed for motifs of five, six and seven residues length, and more than 95 % (363 of 381 peptides) harbored one or more copies of the same 7mer motif (QPQQPFP). In comparison, this 7mer was found in less than 35% of all peptides identified in the gliadin fraction before pull-down (Figure 1B). The majority of peptides in the pull down were deamidated, but as the mass spectrometry search engine did not unambiguously report the exact deamidation sites, native sequences are listed in Table 1. Manual inspection of mass spectrometry fragment spectra revealed that this motif mainly 9

existed as QPEQPFP in the pull-down sample which is in accordance with the specificity of TG2 28, 29. From eight residues the frequency in the pull-down decreased steadily down to 29% of the 381 peptides containing the most common 15mer. Altogether, these analyses suggested that long gliadin, -gliadin, and low-molecular weight glutenin peptides, with TG2 deamidated glutamate residues and with several copies of the above given motifs as optimal targets for the hmab 1002-1E03. Based on this analysis, we synthetized two -gliadin derived fragments being abundant in the pull-down fraction and used these as target peptides in the serologic inhibition assay (biotin- QPEQPFPEQPEQPEQPFPQPEQPFPWQPEQPFPQ, b- 34 and biotin- QPEQPFPEQPEQPEQPFPQPEQPFPW, b- 26). Both peptides comprised several copies (three to six) of the 3mer to 7mer identified as the most common motifs. ELISA reactivity of hmab 1002-1E03 to synthetic gliadin peptides To evaluate the specificity of the hmab 1002-1E03 in light of the identified common motifs, the hmab 1002-1E03 was tested against a panel of synthetic gliadin peptides in ELISA together with eight other hmabs (Table 2). Interestingly, the reactivity patterns of the nine hmabs were not identical. One hmab, 1114-1G01, showed almost identical peptide reactivity pattern as hmab 1002-1E03. Four hmabs, 1002-1E01, 1130-2A02, 1130-3A02 and 1130-3B04, were reactive to all peptides harboring the QPQQPFP or QPEQPFP sequences. The hmab 1130-3B01 showed binding to all peptides comprising QPQQPFP, QPEQPFP, or PQPELPYPQP. The last two hmabs, 1130-3B03 and 1130-3G05, only recognized peptides harboring the sequence PQPELPYPQP. As to the reactivity of the hmab 1002-1E03, the sequence PEQ was always present in the peptides which were recognized. However, not all peptides containing this 3mer sequence were recognized by the hmab suggesting that PEQ was important but not sufficient for binding. As mentioned, the PQQ sequence was usually seen as part of the QPQQPFP motif sequence among the peptides in the pull-down. The hmab 1002-1E03 showed good reactivity to synthetic peptides containing the QPEQPFP sequence, but not to peptides with the corresponding QPQQPFP native sequence. Thus the hmab 1002-1E03 is likely reactive with the QPEQPFP sequence. The QPEQPFP sequence was represented four times in b- 34 and three 10

times in b- 26. A peptide comprising three copies of the QPEQPFP motif (biotin- PQPEQPFPQPEQPFPQPEQPFPQP, b-qpeqpfp 3 for short) was also synthesized and used as a third target peptide in the serologic inhibition assay. Serologic inhibition assay Serum and biotinylated target peptides were first mixed and incubated. Second, AlphaLISA acceptor beads conjugated with hmab 1002-1E03 were added to the mixture. The binding of the hmab to the biotinylated peptide was detected with Alphascreen streptavidin donor beads. The principle of the assay is that the biotinylated target peptide binds to streptavidin donor beads, which then are brought in proximity to the hmab conjugated acceptor beads if the antibody binds to the target peptide. Upon light activation, the donor beads release energy that can activate the acceptor beads to emit light signal provided proximity of the beads. Gliadin-reactive serum antibodies can compete with the hmab for binding to the peptide and hence reduce the signal. We tested sera of untreated celiac disease patients (n = 106) and compared these with sera of control subjects (198 healthy blood donors and 151 patients with Crohn s disease). Three different target peptides (b-qpeqpfp 3, b- 34, and the shorter variant b- 26) were tested in the assay. Two of them represent peptide fragments that are naturally occurring in the gliadin proteome (b- 34 and b- 26), and the antibody pull-down experiment suggested these peptides to be of primary targets for the hmab. In contrast, the third peptide (b-qpeqpfp 3 ) was designed to represent the sequence QPEQPFP three times. The peptide b- 34 demonstrated highest diagnostic performance of the three target peptides, evaluated by the area under the curve (AUC) of the receiver operating characteristic (ROC) curve analysis (Supplementary Table 1). The most striking difference, though, was the large variation in negative controls in the assay employing the b-qpeqpfp 3 peptide, whereas the two assays using the -peptides only gave small variations in the same control groups (Supplementary Figure 1). There was a modest positive effect of combining the test result of the three target peptides 11

(Supplementary Table 1), but only the results of the assay with the b- 34 target peptide (Figure 2A) were used in the further analyses (hereafter termed anti- 34 Ig assay). Each serum sample was tested in duplicates and thus giving two values of logarithmic AlphaLISA Signal. The mean of these two values was used in the analysis (Figure 2A). In order to establish test characteristics of the assay and to establish a reference curve, a serum from a non-celiac control was spiked with dilutions of known concentrations of gliadin-specific hmabs (equimolar concentrations of 1002-1E01, 1002-1E03 and 1130-3B03), and tested (Figure 2B). We found that the performance of the assay was critically dependent on the concentration of target peptide. Low concentrations gave better sensitivity in terms of lower detection limit, but also lower specificity due to reduced signal/noise-ratio (Supplementary Figure 2). The peptide concentrations used in in this study gave dynamic range from approximately 1 mg/l to 20-30 mg/l of the gliadin-specific hmabs used as reference. Signal/noise-ratio under these conditions spanned approximately two logarithmic units. The mean values of the logarithmic AlphaLISA Signal (Figure 2A) and the reference curve (Figure 2B) were used to extrapolate the antibody activity in the serum samples to equivalent concentrations (mg/l) of reference gliadin-specific hmabs (Figure 2C). Optimal cut-off estimated by Youden index (sensitivity (1 specificity)) of the extrapolated values (Figure 2C) was < 1.02 mg/l (sensitivity 0.9292, specificity 0.9513), suggesting that the detection limit of the test platform was the main limitation of the sensitivity in this assay. This suggests that further improvements in sensitivity could possibly be achieved by using even lower peptide concentrations, although this was difficult to obtain without loss of specificity as observed in preliminary test results. Conceivably, to use < 1 mg/l as cut-off would be inadequate, as 1 mg/l of hmabs in reference serum gave no significant decrease in AlphaLISA Signal compared to background (Figure 2B). In contrast, the reference curve gave a clear difference in signal between 1 mg/l and 2 mg/l (Figure 2B), and hence was cut-off < 2 mg/l selected for the anti- 34 Ig assay. 12

Serologic anti- 34 Ig assay compared to anti-tg2 IgA and anti-dgp IgG assays The serum samples from celiac disease patients and control subjects (healthy blood donors and patients with Crohn s disease) were also analyzed for anti-tg2 IgA (Varelisa ttg IgA, Phadia) and anti- DGP IgG (QUANTA Lite Gliadin IgG II, INOVA) (Figure 3A-B). The manufacturers operated with two different recommended cut-off values, namely < 5 or < 8 U/ml for anti-tg2 IgA, and < 20 or < 30 U/ml for anti-dgp IgG. These are termed high (< 8 and < 30 U/ml) and low (< 5 and < 20 U/ml) cut-off values in the following. The test results of anti-tg2 IgA and anti-dgp IgG were compared with the results of the novel serologic anti- 34 Ig assay (Figure 2C). In addition to the mentioned cut-off < 2 mg/l, we also included a second cut-off < 3 mg/l, corresponding to the high cut-off values of the two commercial assays. The ROC-estimated AUCs were high (0.9715 0.9806) for all three assays (no significant differences, Supplementary Table 1). In general, all three assays showed very good specificity, whereas the sensitivity was low for all assays (Table 3). Of particular note, many of the celiac disease patients scored negative in the anti-tg2 IgA assay (Figure 3), and anti-tg2 IgA demonstrated lower sensitivity than both the anti-dgp IgG and anti- 34 Ig assays (Table 3). We next evaluated the specificity of the assays for the two control groups separately. For blood donors, the high and low cut-off values gave identical specificity results for all three assays. In brief, five of the 198 blood donors were seropositive, and three of them scored positive in two or more assays (Table 4). One donor had anti-dgp IgG 49 U/ml, but was negative for anti- 34 Ig (0 mg/l) and anti-tg2 IgA (0.088 U/ml). The fifth donor scored 6.6 mg/l in the anti- 34 Ig assay, had anti-tg2 IgA value of 2.6 U/ml and was anti-dgp IgG negative (below 5 U/ml). Notably, all five were HLA-DQ2.5 positive (Table 4). Thus, many of the seropositive blood donors were highly suspicious of having untreated celiac disease, which obviously made the specificity evaluation difficult in the blood donor group. 13

In contrast, the Crohn s disease control group demonstrated clear differences between the three assays. In total seven Crohn s disease patients scored positive using low cut-off values, and none of them were positive in more than one assay (Table 4). Only three patients had antibody activities above the high cut-off values, all in the anti-dgp IgG assay. In summary, for both low and high cut-off values anti- 34 Ig (0.9868 1.0) and anti-tg2 IgA (0.9934 1.0) demonstrated superior specificity as compared to the anti-dgp IgG assay (0.9735 0.9801) in the Crohn s disease control group. The diagnostic odds ratio (DOR), estimating the odds of (true) positive test (sensitivity / (1-sensitivity) in subject from disease group, relative to odds of (false) positive test ((1-specificity)/specificity) in control subject, is a commonly used measurement of the effectiveness of an assay. 30 The DOR results were comparable for all three tests using both high and low cut-off (range 321-453), except for anti- DGP IgG (<30 U/ml) that demonstrated the poorest odds with 167. Although not significant, anti- 34 Ig (< 2 mg/l) showed highest DOR (Table 3). Joint analysis of anti-tg2 IgA and anti- assays increases detection rate Recently, several studies have claimed an important role of gliadin-specific antibodies as they allow detection of celiac disease patients who are anti-tg2 IgA negative. 16, 31-33 In keeping with this, we next evaluated whether complementary use of anti- 34 Ig assay provided better diagnostic performance than single testing of anti-tg2 IgA. Anti- 34 Ig assay detected 57% and 75% of the celiac disease patients with negative anti-tg2 IgA titers for high and low cut-off values, respectively (i.e. 86 patients with anti-tg2 IgA > 5 U/ml, against 101 patients with anti-tg2 IgA > 5 and/or anti- 34 Ig > 2mg/L). The specificity was modestly decreased, and the DOR was more than doubled (Table 3). Anti-DGP IgG showed similarly good effect (Table 3). Thus, the combined use of anti-tg2 IgA and the novel anti- 34 Ig assay considerably increased the diagnostic performance. 14

Discussion We here describe a novel serologic inhibition assay for detection of gliadin-specific antibodies to be used as a diagnostic tool in celiac disease. The assay is based on a recombinant hmab, generated from the antibody genes of a single intestinal IgA plasma cell of a celiac disease patient, together with a preferred target antigen identified by antibody pull-down of peptides from a TG2-treated gliadin digest. The combination of this gliadin peptide and the hmab was employed in an amplified luminescent proximity homogeneous (AlphaLISA) assay whose signal could be specifically inhibited by serum antibodies. The inhibition assay was more specific than a commercial anti-dgp IgG assay, more sensitive than a commercial anti-tg2 IgA assay and detected up to 75% of the patients who scored negative for anti-tg2 IgA in this study. Two main characteristics separate the novel inhibition assay from conventional anti-tg2 and anti- DGP assays. First, it takes advantage of the specificity of the hmab originating from the plasma cell of a celiac lesion as well as the sequence of the peptide fragment preferentially recognized by this hmab. Second, it monitors antibody activity of all antibody isotypes in serum. Although the diagnostic performance of the anti-dgp assays is markedly better than assays detecting antibodies to whole native gliadin, the anti-dgp assays in general are reported to be less specific than the anti-tg2 assays 25 and are claimed to have low predictive values in the serologic diagnostics of celiac disease. 21 The inhibition assay we present exhibits better specificity than the anti-dgp IgG assay in the material we tested. This could relate to the use of the disease-specific hmab or the different gluten antigens used in the two assays. The antigen in the anti-dgp assays typically consists of gluten motifs empirically giving the highest antibody reactivity in patient sera. Of note, several independent studies have reported strikingly similar results, pointing at peptide sequences highly similar to the 7mer motif QPEQPFP enriched by the 1002-1E03 hmab investigated in this study. 10, 34, 35 Our anti- 34 Ig assay employed the 1002-1E03 hmab together with one of this hmab s preferred targets, the deamidated 34mer -gliadin fragment which contained four copies of the QPEQPFP sequence. 15

The fact that our assay measures antibody activity of all isotypes in serum is particularly relevant for patients with IgA deficiency. IgA deficiency has a higher prevalence among celiac disease patients. 36 For this reason, screening of anti-gliadin antibodies of IgG isotype is often recommended as the second test to anti-tg2 IgA. 7 IgG and IgA of patients sera show reactivity to the same motifs of the gluten proteome, 10, 18, 19 which indicates that the gluten antibodies of the different isotypes have overlapping epitope specificities. In assays where only one isotype is measured, antibodies of other isotypes would compete and potentially reduce the signal. Measuring all isotypes, as done in the inhibition assay of this study, may thus be an advantage. There is a further potential for improvements of the inhibition assay. We have only tested hmab 1002-1E03, one of several gliadin-specific monoclonal antibodies available. Other combinations of antibodies and target peptides could potentially perform better than the combinations investigated in this study. The goal would be to find the best possible combination. The fine specificities of nine hmabs tested with synthetic peptides in ELISA were not all the same speaking to the potential for assay improvements. The difference in fine specificity revealed in ELISA with synthetic peptides was only partially observed in the analysis of fragments isolated by antibody pull-down and mass spectrometry sequencing from a TG2-treated enzymatic digest of gliadin (Dørum et al., submitted). For the hmab 1002-1E03, we tested two other peptides, b- 26 the b-qpeqpfp 3, in addition to the b- 34 peptide. Notably, the b-qpeqpfp 3 peptide gave large variation in reactivity of sera from the control groups, which was not observed for the two -gliadin derived peptide fragments. This suggests that the sequence of the peptides impacted on the performance of the assay. Further, if the nine hmabs we tested and which showed different fine specificities are representative for polyclonal anti-gliadin responses in patients, it is to be expected that different target peptides would give different performance in the type of inhibition assay we have developed. Sensitivity estimates can vary considerably, depending on the patient population investigated. 37 A major critique of many serologic studies has been the possibility of ascertainment bias, as most 16

patients are selected for biopsy because they have positive serology. 25 How the anti- 34 Ig assay will perform in testing populations with lower disease prevalence remains to be established. Several recent publications have reported a similar increased detection rate when testing for both anti-tg2- and anti-gliadin-antibodies. 16, 31-33 Such a joint analysis appeared to give improved performance also in our test material. Taken together this study demonstrates that hmabs deriving from celiac disease patients can be harnessed together with their natural target peptides in inhibitory serologic assay for diagnostic purpose. The assay takes advantage of the specificity introduced by both the hmab and the antigen, and the results give promise for this type of assay to have the requested sensitivity and specificity for celiac disease diagnosis. 17

Acknowledgments This work was supported by the United European Gastroenterology Research Prize and a Research Excellence Prize from the Oslo University Hospital to LMS, by a grant from the European Research Council (ERC-2010-Ad-268541), and by the Research Council of Norway through its Centre of Excellence funding scheme, project number 179573/V40. We thank Eli Taraldsrud, Laila Steinbakk, Liv Slettevoll Steen and Vidar Bosnes at Oslo University Hospital Ullevål for excellent collaboration and for their work on serologic testing of anti-dgp IgG and anti-tg2 IgA. We also thank Patrick C. Wilson and Carole H. Dunand for providing antibody plasmids of the gliadin-specific hmabs investigated in this study, and Shuo-Wang Qiao for critical reading of the manuscript. 18

Figure legends Figure 1. (A) Length of peptides identified by mass spectrometry in fractions of a TG2-treated digest of gliadin before (grey) and after (black) pull-down by the human monoclonal antibody 1002-1E03. The y-axis unit is relative peptide concentrations as determined by MaxQuant quantitation. (B) The number of peptides sharing identical sequence motifs, of 3 to 15 residues in length in pre (grey) and post pull-down (black) samples. Figure 2. AlphaLISA anti- 34 Ig assay. (A) Inhibition of 1002-1E03 binding to b- 34 peptide by sera of three test groups as analyzed in AlphaLISA. (B) Reference curve established by negative control serum spiked in with serial dilutions of known and equimolar concentrations of three gliadin-specific hmabs. (C) Activity of gliadin-specific serum antibodies expressed as concentration equivalents (mg/l) of reference gliadin-specific hmabs. Each cross represents one subject except for the seropositive blood donors HLA typed to be DQ2.5 that are marked with triangles. Grey stapled lines represent the cut-off values of reference hmab concentration equivalents of 2 mg/l or 3 mg/l. Figure 3. (A) Anti-TG2 IgA and (B) anti-dgp IgG levels of participants from the three test groups. Each cross represents one subject except for the seropositive blood donors HLA typed to be DQ2.5 that are marked with triangles. Supplementary Figure 1. Inhibition of AlphaLISA signals of all three assays with hmab 1002-1E03 and the target peptides (A) b-qpeqpfp 3, (B) b- 26 and (C) b- 34 by sera of the test groups untreated celiac disease patients (celiac disease) and controls (Crohn disease patients and healthy subjects). Each cross represents one subject except for the seropositive blood donors HLA typed to be DQ2.5 that are marked with triangles. Supplementary Figure 2. The target peptide concentration affects the sensitivity, dynamic range and signal/noise-ratios as shown for three different concentrations of b- 34. Mean values for all three concentrations are shown in grey. 19

References 1. Sollid LM, Markussen G, Ek J, et al. Evidence for a primary association of celiac disease to a particular HLA-DQ / heterodimer. J Exp Med 1989;169:345-50. 2. West J, Fleming KM, Tata LJ, et al. Incidence and prevalence of celiac disease and dermatitis herpetiformis in the UK over two decades: population-based study. Am J Gastroenterol 2014;109:757-68. 3. Murray JA, Van Dyke C, Plevak MF, et al. Trends in the identification and clinical features of celiac disease in a North American community, 1950-2001. Clin Gastroenterol Hepatol 2003;1:19-27. 4. Husby S, Koletzko S, Korponay-Szabo IR, et al. European Society for Pediatric Gastroenterology, Hepatology, and Nutrition guidelines for the diagnosis of coeliac disease. J Pediatr Gastroenterol Nutr 2012;54:136-60. 5. Ludvigsson JF, Bai JC, Biagi F, et al. Diagnosis and management of adult coeliac disease: guidelines from the British Society of Gastroenterology. Gut 2014;63:1210-28. 6. Bai JC, Fried M, Corazza GR, et al. World Gastroenterology Organisation global guidelines on celiac disease. J Clin Gastroenterol 2013;47:121-6. 7. Rubio-Tapia A, Hill ID, Kelly CP, et al. ACG Clinical Guidelines: Diagnosis and management of celiac disease. Am J Gastroenterol 2013;108:656-76. 8. Ribeiro M, Nunes-Miranda JD, Branlard G, et al. One hundred years of grain omics: identifying the glutens that feed the world. J Proteome Res 2013;12:4702-16. 9. Sjöström H, Lundin KE, Molberg O, et al. Identification of a gliadin T-cell epitope in coeliac disease: general importance of gliadin deamidation for intestinal T-cell recognition. Scand J Immunol 1998;48:111-5. 10. Osman AA, Gunnel T, Dietl A, et al. B cell epitopes of gliadin. Clin Exp Immunol 2000;121:248-54. 11. Steinsbø Ø, Henry Dunand CJ, Huang M, et al. Restricted VH/VL usage and limited mutations in gluten-specific IgA of coeliac disease lesion plasma cells. Nat Commun 2014;5:4041. 12. Molberg Ø, McAdam SN, Korner R, et al. Tissue transglutaminase selectively modifies gliadin peptides that are recognized by gut-derived T cells in celiac disease. Nat Med 1998;4:713-7. 13. van de Wal Y, Kooy Y, van Veelen P, et al. Selective deamidation by tissue transglutaminase strongly enhances gliadin-specific T cell reactivity. J Immunol 1998;161:1585-8. 14. Dieterich W, Ehnis T, Bauer M, et al. Identification of tissue transglutaminase as the autoantigen of celiac disease. Nat Med 1997;3:797-801. 15. Anderson RP, Henry MJ, Taylor R, et al. A novel serogenetic approach determines the community prevalence of celiac disease and informs improved diagnostic pathways. BMC Med 2013;11:188. 16. Wolf J, Hasenclever D, Petroff D, et al. Antibodies in the diagnosis of coeliac disease: a biopsy-controlled, international, multicentre study of 376 children with coeliac disease and 695 controls. PLoS One 2014;9:e97853. 17. Agardh D. Antibodies against synthetic deamidated gliadin peptides and tissue transglutaminase for the identification of childhood celiac disease. Clin Gastroenterol Hepatol 2007;5:1276-81. 18. Aleanzi M, Demonte AM, Esper C, et al. Celiac disease: antibody recognition against native and selectively deamidated gliadin peptides. Clin Chem 2001;47:2023-8. 19. Bateman EA, Ferry BL, Hall A, et al. IgA antibodies of coeliac disease patients recognise a dominant T cell epitope of A-gliadin. Gut 2004;53:1274-8. 20. Schwertz E, Kahlenberg F, Sack U, et al. Serologic assay based on gliadin-related nonapeptides as a highly sensitive and specific diagnostic aid in celiac disease. Clin Chem 2004;50:2370-5. 20

21. Vriezinga SL, Auricchio R, Bravi E, et al. Randomized feeding intervention in infants at high risk for celiac disease. N Engl J Med 2014;371:1304-15. 22. Shor DB, Orbach H, Boaz M, et al. Gastrointestinal-associated autoantibodies in different autoimmune diseases. Am J Clin Exp Immunol 2012;1:49-55. 23. Gatselis NK, Zachou K, Norman GL, et al. IgA antibodies against deamidated gliadin peptides in patients with chronic liver diseases. Clin Chim Acta 2012;413:1683-8. 24. Vojdani A. The characterization of the repertoire of wheat antigens and peptides involved in the humoral immune responses in patients with gluten sensitivity and Crohn's disease. ISRN Allergy 2011;2011:950104. 25. Lewis NR, Scott BB. Meta-analysis: deamidated gliadin peptide antibody and tissue transglutaminase antibody compared as screening tests for coeliac disease. Aliment Pharmacol Ther 2010;31:73-81. 26. Lewis NR, Scott BB. Systematic review: the use of serology to exclude or diagnose coeliac disease (a comparison of the endomysial and tissue transglutaminase antibody tests). Aliment Pharmacol Ther 2006;24:47-54. 27. Leffler DA, Schuppan D. Update on serologic testing in celiac disease. Am J Gastroenterol 2010;105:2520-4. 28. Vader LW, de Ru A, van der Wal Y, et al. Specificity of tissue transglutaminase explains cereal toxicity in celiac disease. J Exp Med 2002;195:643-9. 29. Fleckenstein B, Molberg O, Qiao SW, et al. Gliadin T cell epitope selection by tissue transglutaminase in celiac disease. Role of enzyme specificity and ph influence on the transamidation versus deamidation process. J Biol Chem 2002;277:34109-16. 30. Glas AS, Lijmer JG, Prins MH, et al. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol 2003;56:1129-35. 31. Mooney PD, Wong SH, Johnston AJ, et al. Increased detection of celiac disease with measurement of deamidated gliadin peptide antibody before endoscopy. Clin Gastroenterol Hepatol 2015. 32. Sugai E, Hwang HJ, Vazquez H, et al. New serology assays can detect gluten sensitivity among enteropathy patients seronegative for anti-tissue transglutaminase. Clin Chem 2010;56:661-5. 33. Basso D, Guariso G, Bozzato D, et al. New screening tests enrich anti-transglutaminase results and support a highly sensitive two-test based strategy for celiac disease diagnosis. Clin Chim Acta 2011;412:1662-7. 34. Vallejo-Diez S, Bernardo D, Moreno Mde L, et al. Detection of specific IgA antibodies against a novel deamidated 8-mer gliadin peptide in blood plasma samples from celiac patients. PLoS One 2013;8:e80982. 35. Ballew JT, Murray JA, Collin P, et al. Antibody biomarker discovery through in vitro directed evolution of consensus recognition epitopes. Proc Natl Acad Sci U S A 2013;110:19330-19335. 36. Chow MA, Lebwohl B, Reilly NR, et al. Immunoglobulin A deficiency in celiac disease. J Clin Gastroenterol 2012;46:850-4. 37. Sugai E, Moreno ML, Hwang HJ, et al. Celiac disease serology in patients with different pretest probabilities: is biopsy avoidable? World J Gastroenterol 2010;16:3144-52. 21

FIGURES AND FIGURE LEGENDS Figure 1. (A) Length of peptides identified by mass spectrometry in fractions of a TG2-treated digest of gliadin before (grey) and after (black) pull-down by the human monoclonal antibody 1002-1E03. The y-axis unit is relative peptide concentrations as determined by MaxQuant quantitation. (B) The number of peptides sharing identical sequence motifs, of 3 to 15 residues in length in pre (grey) and post pull-down (black) samples. 22

Figure 2. AlphaLISA anti- 34 Ig assay. (A) Inhibition of 1002-1E03 binding to b- 34 peptide by sera of three test groups as analyzed in AlphaLISA. (B) Reference curve established by negative control serum spiked in with serial dilutions of known and equimolar concentrations of three gliadin-specific hmabs. (C) Activity of gliadin-specific serum antibodies expressed as concentration equivalents (mg/l) of reference gliadin-specific hmabs. Each cross represents one subject except for the seropositive blood donors HLA typed to be DQ2.5 that are marked with triangles. Grey stapled lines represent the cut-off values of reference hmab concentration equivalents of 2 mg/l or 3 mg/l. 23

Figure 3. (A) Anti-TG2 IgA and (B) anti-dgp IgG levels of participants from the three test groups. Each cross represents one subject except for the seropositive blood donors HLA typed to be DQ2.5 that are marked with triangles. 24

Supplementary Figure 1. Inhibition of AlphaLISA signals of all three assays with hmab 1002-1E03 and the target peptides (A) b-qpeqpfp 3, (B) b- 26 and (C) b- 34 by sera of the test groups untreated celiac disease patients (celiac disease) and controls (Crohn disease patients and healthy subjects). Each cross represents one subject except for the seropositive blood donors HLA typed to be DQ2.5 that are marked with triangles. 25

Supplementary Figure 2. The target peptide concentration affects the sensitivity, dynamic range and signal/noise-ratios as shown for three different concentrations of b- 34. Mean values for all three concentrations are shown in grey. 26

TABLES Table 1. Most frequent motifs of all sequences of 3-15 residues length found in the 381 gliadin peptides pulled down by the hmab 1002-1E03. The motifs are given with native sequences even though most peptides in the pull down were deamidated. Motif Motif count Peptides with given motif after pull-down Length Sequence after pull-down Peptide count Frequency 15 QPQQPFPQQPQQPFP 127 111 29 % 14 PQQPFPQQPQQPFP 130 113 30 % 13 QPQQPFPQQPQQP 163 145 38 % 12 QPQQPFPQQPQQ 181 163 43 % 11 FPQQPQQPFPQ 188 176 46 % 10 PQQPQQPFPQ 312 254 67 % 9 QQPQQPFPQ 502 302 79 % 8 QPQQPFPQ 680 338 89 % 7 QPQQPFP 774 363 95 % 6 QPQQPF 792 364 96 % 5 QPQQP 1055 370 97 % 4 PQQP 1479 374 98 % 3 QQP/PQQ 1785/1634 378 99 % 27

Table 2. Recognition in ELISA of synthetic gliadin peptides by the nine hmabs. The hmabs showed good reactivity (+), weak reactivity (+/-) or no reactivity (-) to the peptides in titration experiments. Not all hmabs were tested for reactivity to all peptides (empty squares). Synthetic gliadin peptides (sequences) 1002-1E03 1114-1G01 1002-1E01 1130-2A02 1130-3A02 1130-3B04 1130-3B01 1130-3B03 1130-3G05 PLQPEQPFP + + + + + + + - - PLQPQQPFP - - + + + + +/- PQQPEQPFP + + + + + + + QLQPFPQ - - - - - - PQQPFPQ - - - - - - QQPFPQ - - - - - - QPFPQ - - - - - - QPFP - - - - - - LQLQPFPQPELPYPQPELPYPQPELPYPQPQPF +/- - +/- - - +/- + + + QLWEIPEQS + - - PEQQLQFEE + + - - - - QLQPFPQPELP - - QPEQP - - +/- PEQPFPE + - - EQPFPE - - - PQPELPYPQP - - - - - - + + + PQPEQPFPQPEQPFPQPEQPFPQP + + + + + + + + + 28

Table 3. Diagnostic performance of the serologic tests at high and low cut-off values and diagnostic performance of the combination of serologic tests Cut-off Sensitivity Specificity DOR Anti-TG2 IgA < 5 U/ml 0.8113 0.9885 370 < 8 U/ml 0.7358 0.9914 321 Anti-DGP IgG < 20 U/ml 0.8794 0.9799 355 < 30 U/ml 0.7736 0.9799 167 Anti- 34 Ig < 2 mg/l 0.8679 0.9857 453 < 3 mg/l 0.7830 0.9914 416 Low cut-off values a Sensitivity Specificity DOR Anti-TG2 IgA + anti- 34 Ig 0.9528 0.9799 984 Anti-TG2 IgA + anti-dgp IgG 0.9622 0.977 1081 High cut-off values b Anti-TG2 IgA + anti- 34 Ig 0.8868 0.9885 673 Anti-TG2 IgA + anti-dgp IgG 0.9245 0.9799 597 a Anti-TG2 IgA < 5 U/ml, anti-dgp IgG < 20 U/ml and anti- 34 Ig < 2 mg/l b Anti-TG2 IgA < 8 U/ml, anti-dgp IgG < 30 U/ml and anti- 34 Ig < 3 mg/l 29

Table 4. Seropositive control subjects. Category Participant Anti- 34 Anti- Antiid Ig TG2 IgA DGP IgG DQB1 DQA1 HLA Blood donor 1008 2244 8.42 9.8 31 *02, *02 *05, *05 DQ2.5 1008 2342 0.50 10.3 34 *02, *05 *01, *05 DQ2.5 1009 2383 6.61 2.6 5 *02, *06 *01, *05 DQ2.5 1008 2407 0 0.09 49 *02, *06 *01, *05 DQ2.5 1008 2430 3.34 37 67 *02, *02 *02, *05 DQ2.5 Crohn s disease patient a n.d.: not determined 20400-IBD 0.33 0.81 71 n.d. a n.d n.d. 30901-IBD 0 0.36 31 n.d. n.d n.d. 40606-IBD 0 7.2 5 n.d. n.d n.d. 40607-IBD 0 0.72 100 n.d. n.d n.d. 40400-IBD 1.67 0.20 20 n.d. n.d n.d. 40602-IBD 2.52 0.18 17 n.d. n.d n.d. 30