644 Diabetes Care Volume 38, April 2015

Simke Demeester,1 Bart Keymeulen,1 Preexisting Insulin Autoantibodies Leonard Kaufman,1 Annelien Van Dalem,1 Eric V. Balti,1 Ursule Van de Velde,1 Predict Efficacy of Otelixizumab in Patrick Goubert,1 Katrijn Verhaeghen,1 Howard W. Davidson,2 Janet M. Wenzlau,2 Preserving Residual b-Cell Ilse Weets,1 Daniel G. Pipeleers,1 and Function in Recent-Onset Type 1 Frans K. Gorus1 Diabetes Diabetes Care 2015;38:644–651 | DOI: 10.2337/dc14-1575

OBJECTIVE Immune intervention trials in recent-onset would benefit from biomarkers associated with good therapeutic response. In the previously reported randomized placebo-controlled anti-CD3 study (otelixizumab; GlaxoSmithKline), we tested the hypothesis that specific diabetes autoantibodies might serve this purpose.

RESEARCH DESIGN AND METHODS In the included patients (n = 40 otelixizumab, n =40placebo),b-cell function was assessed as area under the curve (AUC) C-peptide release during a hyperglycemic glucose clamp at baseline (median duration of insulin treatment: 6 days) and every 6 months until 18 months after randomization. (Auto) against insulin (I[A]A), GAD (GADA), IA-2 (IA-2A), and ZnT8 (ZnT8A) were determined on stored EMERGING TECHNOLOGIES AND THERAPEUTICS sera by liquid-phase radiobinding assay.

RESULTS At baseline, only better preserved AUC C-peptide release and higher levels of IAA were associated with better preservation of b-cell function and lower insulin needs under anti-CD3 treatment. In multivariate analysis, IAA (P = 0.022) or the interaction of IAA and C-peptide (P = 0.013) independently predicted outcome 1Diabetes Research Center and University Hospi- together with treatment. During follow-up, good responders to anti-CD3 treat- tal Brussels (UZ Brussel), Vrije Universiteit Brus- + b ‡ sel, Brussels, Belgium ment (i.e., IAA participants with relatively preserved -cell function [ 25% of 2Barbara Davis Center for Childhood Diabetes, healthy control subjects]) experienced a less pronounced insulin-induced rise in University of Colorado at Denver, Aurora, CO I(A)A and lower insulin needs. GADA, IA-2A, and ZnT8A levels were not influenced Corresponding author: Ilse Weets, ilse.weets@ by anti-CD3 treatment, and their changes showed no relation to functional uzbrussel.be. outcome. Received 27 June 2014 and accepted 14 Decem- ber 2014. CONCLUSIONS This article contains Supplementary Data online There is important specificity of IAA among other diabetes autoantibodies to at http://care.diabetesjournals.org/lookup/ predict good therapeutic response of recent-onset type 1 diabetic patients to suppl/doi:10.2337/dc14-1575/-/DC1. anti-CD3 treatment. If confirmed, future immune intervention trials in type 1 © 2015 by the American Diabetes Association. Readers may use this article as long as the work diabetes should consider both relatively preserved functional b-cell mass and is properly cited, the use is educational and not presence of IAA as inclusion criteria. for profit, and the work is not altered. care.diabetesjournals.org Demeester and Associates 645

Type 1 diabetes is a chronic –mediated terms of preservation of functional 6 months, and positivity for Epstein- autoimmune disease ultimately leading to a b-cell mass, determined as area under Barr virus IgG. Patients received an infu- major loss of insulin-secreting b-cells, hy- the curve (AUC) of second-phase glucose- sion of ChAglyCD3 (otelixizumab, n =40) perglycemia due to insulinopenia, anddif stimulated C-peptide release during a hy- or placebo (n = 40), administered during not well controlleddlife-threatening perglycemic clamp in addition to already 2–4 h on 6 consecutive days (64 mg cumu- complications (1). Humanized nonmito- established factors (4,5), and might thus lative dose in the first 4 patients; 48 mg genic Fc-mutated monoclonal anti-CD3 serve as independent predictors of clinical cumulative dose in the following 36 pa- antibodiesdhOKT3g1(Ala-Ala) (teplizu- outcome. In addition, we investigated tients). Treatment was randomized accord- mab; Macrogenics) (2,3) and ChAglyCD3 whether treatment with anti-CD3 influ- ing to trial center (four in Belgium, one in (otelixizumab) (4,5)dcould slow dis- enced the natural history of diabetes anti- Germany), age (,15 or $15 years) and ease progression by targeting activated body patterns after diagnosis (i.e., the presence or absence of ICA (4). The initial T lymphocytes in recent-onset type 1 declining trend of GADA, IA-2A, and protocol was planned for an 18-month diabetic patients, but preservation of ZnT8A and the insulin treatment–induced study. Efficacy and safety data were re- functional b-cell mass was transient rise in insulin antibodies [IA]) (11–13). ported previously (4). Written informed and largely confined to individuals consent was obtained from each patient. with relatively intact C-peptide secre- RESEARCH DESIGN AND METHODS The primary end point was second-phase tion and young age (,27 years) at di- Patient Selection and Treatment AUC C-peptide release (min 60–140) during agnosis (2–5). Likewise, the efficacy of Eighty recent-onset type 1 diabetic pa- the hyperglycemic clamp test (4). Be- several other immune interventions in tients were included in a randomized cause the number of participants who recent-onset diabetes was highest in phase 2 placebo-controlled trial (4) (trial underwent hyperglycemic clamp testing participants with younger age at inclu- number NCT00627146) (Supplementary during a 48-month follow-up was insuf- sion, shorter disease duration, or higher Fig 1). They were selected according to ficient to use clamp-derived C-peptide residual insulin-producing capacity at the following criteria: age 12–39 years, release as a primary end point (n = 36) the start of treatment (1,6). positivity for ICA and/or GADA, random (5), we limited the current study on Future trials, particularly if planned at plasma C-peptide level $0.2 nmol/L at a autoantibody predictor ability to 18 the preclinical stage, would benefitfrom glycemia of 10.0–13.9 mmol/L, treat- months of follow-up. The anti-CD3–treated biomarkers that identify responders to a ment with insulin for #4 weeks before group and the placebo group did not dif- given intervention. This would avoid ex- enrollment, polyuria for ,6months, fer statistically in clinical or laboratory posing nonresponders needlessly to im- ,10% weight loss during the previous characteristics (Table 1) or in the munomodulators with potentially harmful adverse effects (1,7,8). Diabe- tes autoantibodies are obvious candi- Table 1—Characteristics of patients at screening and start of the treatment dates in this respect because (changes Placebo group ChAglyCD3 in) status or levels have been Variable (n =40) (n =40) P associated with clinical outcome in islet At screening or pancreas transplantation protocols Men 26 (65) 25 (63) 0.816 and in the oral arm of the DPT-1 trial Age (years)* 26 (7) 27 (7) 0.596 (9,10). Taking advantage of the data Time of insulin treatment (days) 6 (2–12) 7 (3–14) 0.432 and sample base from the previously re- IAA+ 15 (38) 10 (25) 0.228 fi GADA+ 37 (93) 34 (85) 0.288 ported rst randomized placebo- + controlled anti-CD3 study originally IA-2A 21 (53) 23 (58) 0.653 ZnT8A+ 19 (48) 20 (50) 0.823 designed to test the safety and b-cell $1 autoAb+ 39 (98) 38 (95) 0.556 preserving effects of otelixizumab in $2 autoAb+ 28 (70) 27 (68) 0.809 recent-onset type 1 diabetes (4), we IAA levels in IAA+ patients wanted to test the hypothesis that spe- (% tracer binding) 1.3 (0.7–1.9) 0.9 (0.7–1.3) 0.276 cific autoantibody profiles at diagnosis GADA levels in GADA+ patients fi (WHO units/mL) 532 (267–3,551) 279 (110–1,585) 0.088 might predict the ef cacy of a short + course (6 days) of anti-CD3 treatment. IA-2A levels in IA-2A patients (WHO units/mL) 271 (93–1,407) 620 (95–1,852) 0.411 In the original study, only the presence ZnT8A levels in ZnT8A+ patients of islet cell antibodies (ICA) and/or GADA (index) 0.17 (0.08–0.38) 0.32 (0.10–0.54) 0.275 positivity were examined as potential pre- At start of treatment dictive autoantibody markers (4). We there- Time since diagnosis (days) 25 (19–27) 21 (18–24) 0.113 fore measured autoantibodies against Time of insulin treatment (days) 20 (13–25) 20 (16–24) 0.769 2 2 insulin (IAA), GAD (GADA), insulinoma- Insulin dose (units z day 1 z kg 1)* 0.38 (0.20) 0.46 (0.27) 0.201 associated protein-2 (IA-2A), and zinc HbA1c (%)* 8.1 (1.5) 8.6 (1.5) 0.059 transporter8(ZnT8A)atclinicalonsetin HbA1c (mmol/mol)* 65 (16) 71 (16) C-peptide release under glucose participants in this study (4). We investi- clamping (nmol z L21 z min21) gated whether autoantibody levels could AUC min 60–140* 0.61 (0.29) 0.54 (0.30) 0.160 help identify individuals who benefited Data are mean (SD)*, n (%), or median (IQR). most from otelixizumab treatment in 646 IAA Predict Otelixizumab Efficacy Diabetes Care Volume 38, April 2015

expression of susceptible or protective cold or labeled insulin was removed in .250 nondiabetic control subjects HLA haplotypes (data not shown) (4). by a washing step, and the radioactivity and amounted to $0.60% tracer binding of the polyethylene glycol precipitate for I(A)A, $23.0 WHO units/mL for Patient Follow-up was measured, averaged, and ex- GADA, $1.4 WHO units/mL for IA-2A, Patients received intensive insulin ther- pressed as percentage added tracer $0.039 index for ZnT8A, and $0.48 aU apy for the entire period, and at least bound (24,000 cpm/tube) after correc- for the I(A)A microassay. The (auto)anti- once every 3 months, doses were ad- tion for displacement. The latter was bodies were determined by technicians justed according to metabolic needs. considered complete if the counts blinded to the participants’ identification Outpatient study visits occurred every could be lowered to the level of IAA- and treatment regimen. 3 months at the trial center or at the negative control samples (typically local hospital. The parameters recorded 200–250 cpm), and this was the case Statistical Analysis included type and dose of insulin, body for all analyzed samples. Data were stored at the Belgian Diabe- weight, home blood glucose levels, con- Strongly IA-positive samples were di- tes Registry. Statistical differences be- comitant medication, and adverse events luted until their binding signal fell into tween groups were assessed by the (4). HbA levels were determined every 3 1c the linear range of the assay, and the Mann-Whitney U test (unpaired) and months, and if the patient agreed, a glu- results were multiplied by the dilution Wilcoxon test (paired) for continuous cose clamp was performed at baseline factor. This explains why some sera had data and a x2 test for categorical data. and every 6 months thereafter (4). The levels exceeding 100% tracer binding. The change in mean clamp-derived C-peptide concentration was determined This assay performed best in the Fourth C-peptide release versus baseline was by time-resolved immunofluorescence Immunology of Diabetes Workshop on selected as the end point (4). To evalu- assay before and at 60, 90, 120, and 140 Insulin Autoantibodies (16). Like all as- ate the effect of potential prognostic min after the start of the hypergly- says for IAA or IA, this method is unable factors on this end point, forward step- cemic phase of the clamp. The induced to distinguish between antibodies wise multiple regression analysis was C-peptide release was expressed as AUC against endogenous or exogenous insu- used. In the original anti-CD3 protocol, per minute or in multivariate analysis lin (17). GADA, IA-2A, and ZnT8A levels the predefined prognostic factors were as mean C-peptide (i.e., the average of were determined as described before age, sex, body weight, BMI, duration of the C-peptide concentrations at 60, 90, (18) and expressed as World Health Or- insulin therapy at screening, status of 120, and 140 min). In patients who were ganization (WHO) units/mL by compari- ICA and GADA (positive vs. negative), C-peptide negative at a previous visit, the sonwiththeWHOstandardserum and HLA-DQ (protective vs. nonprotec- clamp test was not done and a zero value (GADA, IA-2A) (19) or as a relative index tive) at screening, as well as HbA level was filled in for subsequent time points. 1c by comparison with an in-house positive and clamp-derived C-peptide release C-peptide negativity was defined as AUC standard serum (ZnT8A) (20). (above or below the median [P50] of C-peptide #0.03 nmol z L21 z min21 dur- In all but three samples at screening, all 80 included patients) at start of the ing the hyperglycemic clamp (4). IAA could also be measured by a modified treatment. In the present multivariate Diabetes Autoantibodies microassay (21). Briefly, serum samples analysis, we included only the variables Autoantibodies were determined at were incubated with 125I-labeled human that were significant in previous analy- screening (i.e., after a median [in- insulin in the presence or absence of an ses (i.e., initial clamp-derived C-peptide terquartile range {IQR}] duration of in- excess unlabeled insulin (Actrapid; Novo release, which was the only significant sulin treatment of 6 [3–14] days) and at Nordisk; final concentration: 2.24 3 1025 predictor of C-peptide preservation at 6, 12, and 18 months after the start of mol/L), and immune complexes were pre- month 18 in the original multivariate the anti-CD3 treatment by liquid-phase cipitated with a mixture of 20% glycine analysis) (4) and age (predictor of insulin radiobinding assays. IAA and IA were pretreated protein A-Sepharose and needs at month 48) (5), together with determined by a modification (14) of 8% ethanolamine pretreated protein the newly determined -specific the Palmer assay (15) using acid char- G-Sepharose. The differences in counts autoantibody levels at screening (IAA, coal extraction of serum to deplete IAA were expressed as arbitrary units (aU) GADA, IA-2A, and ZnT8A). Values below and IA from bound endogenous or ex- by comparison with an in-house positive the limit of detection (0.3% tracer bind- ogenous insulin. Briefly, sera were in- standard serum. ing for IAA) were assigned a value of cubated with acid charcoal to disrupt cDNAs for the preparation of radioli- 0.2%. The parameters with a P value of circulating complexes of antibodies gands by in vitro transcription-translation #0.05 in univariate analysis were fur- with endogenous or exogenous insulin encoded the intracellular portion of IA-2 ther tested in multiple linear regression and to absorb unbound insulin. After (gift of M. Christie, King’s College School analysis performed in a stepwise fash- neutralization and centrifugation, the of Medicine and Dentistry, London, U.K.), ion. Statistical analysis was performed insulin-free supernatants were incu- full-length 65kDa GAD (gift of Ǻ.Lernmark separately in the placebo-treated and bated in duplicate with radioactive- when at the University of Washington, the ChAglyCD3-treated group as well labeled human recombinant insulin Seattle, WA), and the fused Arg325 and as in the entire study group. Statistical in the presence and absence of Trp325 carboxy-termini of ZnT8A (gift of tests were performed two-sided at the excess cold insulin (final concentration J.C. Hutton, Barbara Davis Center for Child- 5% level of significance with SPSS 20.0 1.76 3 1026 mol/L). Next, the formed hood Diabetes, Aurora, CO). Cutoff values software (IBM SPSS Statistics, Armonk, immune complexes were precipitated for antibody positivity were determined as NY) and the multivariate analyses with using polyethylene glycol. Unbound the 99th percentile of autoantibody levels SAS 9.2 software (SAS Institute, Inc., Cary, care.diabetesjournals.org Demeester and Associates 647

NC), and the figures were generated with Table 2—Forward multiple linear regression with the change in C-peptide GraphPad Prism 5.00 software (GraphPad concentration over 18 months as a dependent variable Software, Inc., San Diego, CA). No adjust- Change in mean C-peptide concentration (start-month 18) ments were made for multiplicity. Multivariate Multivariate RESULTS Univariate model 1 model 2 Baseline Characteristics Independent variables PPb P b Otelixizumab-treated and placebo- Placebo-treated patients treated new-onset patients had similar Age* 0.519 NM NM baseline characteristics. The groups did C-peptide (P50)† 0.033 0.050 20.354 NM not differ in age and sex distribution, du- GADA* 0.395 NM NM ration of insulin treatment, prevalence ZnT8A* 0.988 NM NM (alone or in combination) and levels (in IA-2A* 0.586 NM NM 2 case of positivity) of autoantibodies at IAA* 0.024 0.024 0.397 NM IAA 3 C-peptide (P50)† 0.016 d 0.016 20.422 screening, or in HbA level, C-peptide re- 1c ChAglyCD3-treated patients lease during hyperglycemic clamp, time Age* 0.383 NM NM since diagnosis, or in duration and dosage C-peptide (P50)† 0.830 NM NM of insulin therapy at the start of the trial GADA* 0.807 NM NM (Table 1). The median (IQR) time since ZnT8A* 0.333 NM NM start insulin treatment of the entire group IA-2A* 0.359 NM NM was 6 (3–14) days; hence, most samples IAA* 0.894 NM NM 3 † were taken within 2 weeks of insulin treat- IAA C-peptide (P50) 0.677 NM NM ment. Moreover, IAA levels at screening All patients fi Age* 0.519 NM NM were not signi cantly different from the C-peptide (P50)† 0.033 NM NM levels at the start of treatment within 4 GADA* 0.899 NM NM weeks of diagnosis (P = 0.214 by Wilcoxon ZnT8A* 0.710 NM NM test) and can thus safely be assumed to IA-2A* 0.207 NM NM represent true IAA (14). IAA* 0.009 0.022 20.264 NM Treatment‡ 0.005 0.020 0.268 0.023 0.260 IAA Levels at Diagnosis Predict a IAA 3 C-peptide (P50)† 0.004 d 0.013 20.286 Better Response to Anti-CD3 b, standardized coefficient; NM, not selected for entry in the multivariate model. Boldface text Treatment indicates P , 0.05. *Variable at screening. †Variable at start of the treatment, AUC 60–140 min , We examined by multiple linear regres- P50 (0) or $P50 (1) of the entire group. ‡Placebo-treated (0), ChAglyCD3-treated (1). sion whether baseline IAA levels could serve as independent predictor(s) of the response to anti-CD3 therapy apart administration (model 1, Table 2). A sig- Figure 1A shows the significantly less from residual b-cell function or age at nificant interaction between higher IAA rapid decline in AUC C-peptide release, diagnosis (4,5). Higher baseline IAA lev- levels and C-peptide release was ob- and Fig. 1B shows the associated els were thus found to independently served (model 2, Table 2). Age was not less rapid increase in insulin needs in- predict better preservation of functional retained as an independent predictor of duced by anti-CD3 treatment in patients b-cell mass at month 18, expressed as a clinical outcome at 18 months, which with IAA $P50 but not in patients with lesser change in clamp-derived C-peptide agrees with our previous analysis at IAA ,P50 (Fig. 1C and D). Likewise, release from baseline and reflected by a this time point (4). Results were similar anti-CD3 treatment preserved clamp- negative standardized coefficient b when I(A)A levels at the start of the trial derived C-peptide release and the asso- (Table 2). In the placebo group, we con- were used instead of IAA levels at ciated lower insulin needs, compared firmed (4) that clamp-derived baseline screening (not shown). with the placebo group, in individuals C-peptide levels above the median inde- Figure 1 illustrates that the results of with AUC C-peptide release at inclusion pendently predicted better outcome. In themultivariateanalysisshouldbeinter- $P50 of the entire group (Fig. 1E and F) addition, higher IAA levels at screening preted in the sense that in the placebo- but not in those with initial values came out as the strongest independent treated group, patients with higher IAA ,P50 (Fig. 1G and H), as already shown predictor, whereas none of the tested levels (above or equal to the P50 of the in the original study report (4). The variables appeared predictive in the entire patient group, corresponding to C-peptide–preserving effect of anti- ChAglyCD3-treated group (Table 2). Sim- $0.50% tracer binding) or with higher CD3 treatment was largely confined ilar results, albeit with slightly higher clamp-derived AUC C-peptide release at to individuals with both AUC C- P values, were obtained when IAA levels baseline ($P50 of the whole group cor- peptide and IAA levels $P50 of the were determined with an I(A)A microassay responding to $25% of healthy control whole patient group, hereby confirm- (not shown) (21). When all participants subjects) (4) experienced a rapid loss of ing the interaction observed in were considered together in multivari- functional b-cell mass versus baseline, multivariate analysis (Fig. 2). IAA+ and ate analysis, IAA levels at screening re- which was preferentially slowed by IAA2 anti-CD3–treated patients did not mained the only independent predictor anti-CD3 treatment to the same level differ in thedevelopment or in the levels of better outcome, apart from anti-CD3 as observed in the other subgroups. of anti-otelixizumab antibodies as 648 IAA Predict Otelixizumab Efficacy Diabetes Care Volume 38, April 2015

Figure 1—Efficacy tests for AUC C-peptide release (panels A,C,E,andG) and insulin dose (panels B, D, F,andH) during 18-month follow-up in placebo-treated (C) and ChAglyCD3-treated (▲) patients stratified according to IAA levels at screening above (panels A and B) or below (panels C and D) the median of all 80 patients or for clamp-derived AUC C-peptide release above (panels E and F) or below (panels G and H) the median value of the entire group. Values represent means 6 SE. The small numbers next to the whiskers indicate the number of observations at the respective times. Significant P values for differences in change in AUC C-peptide from baseline between anti-CD3 and placebo groups are shown in italic above the relevant times.

determined by two different methods independent predictor of the efficacy existing guidelines for C-peptide (maxi- (not shown) (22). of anti-CD3 (otelixizumab) treatment mally stimulated C-peptide $0.2 nmol/L to preserve residual b-cell mass for 18 during a mixed-meal tolerance test) (23), Influence of Anti-CD3 Treatment on months. Its predictive value significantly above which good responders to anti-CD3 Diabetes (Auto)Antibody Levels interacts with a relatively preserved treatment might by identified and thereby During Follow-up b-cell mass at baseline, previously rec- further paving the way to personalized The natural history of GADA, IA-2A and ognized as a predictor of good func- immune interventions. Anti-CD3 treat- ZnT8A autoantibody levels after diabetes tional outcome (4). IAA+ new-onset ment did not influence the natural history onset was not meaningfully altered by patients with relatively preserved of other autoantibodies after diagnosis otelixizumab treatment. The levels b-cell function are therefore proposed but was associated with a lesser rise in tended overall to progressively decrease as candidates of choice for future immune I(A)A levels selectively in patients who re- with time after diagnosis, in line with pre- intervention trials in type 1 diabetes. If sponded well to anti-CD3 therapy and re- vious observations (12,13), and this de- confirmed prospectively in sufficiently quired less insulin. cline occurred at a similar rate in both powered studies, specific cutoffs may be The strengths of the study include the treatment arms during the 18-month determined for IAA levels, similar to randomized, placebo-controlled nature study period (Supplementary Fig. 2). Nei- ther did both groups differ in autoanti- body kinetics after screening when stratified according to baseline residual C-peptide release during hyperglycemic clamp or age ($P50 or ,P50 of entire group, not shown), parameters previ- ously shown to predict clinical outcome (4,5). In contrast, the rise in insulin- induced IA was less pronounced under otelixizumab treatment during the 18- month study period, but only in initially IAA+ patients with relatively preserved functional b-cell mass at diagnosis $ ( P50 AUC C-peptide release at baseline) Figure 2—AUC C-peptide release during 18-month follow-up in placebo-treated (C) and ChA- (Supplementary Fig. 3). It was associated glyCD3-treated (▲) patients stratified according to clamp-derived C-peptide release and IAA with less insulin use (not shown). levels at screening. Panel A: Patients with clamp-derived AUC C-peptide and IAA above the respective median value of the whole group. Panel B: Patients with clamp-derived C-peptide 6 CONCLUSIONS and/or IAA below the respective median values of the whole group. Values represent means SE. The small numbers next to the whiskers indicate the number of observations at the re- This study identified elevated IAA levels spective times. Significant P values for differences in change in AUC C-peptide from baseline at screening of type 1 diabetes as an between anti-CD3 and placebo groups are shown in italic above the relevant times. care.diabetesjournals.org Demeester and Associates 649

of the otelixizumab trial (4) in new-onset with insulin autoimmunity playing a key role to cure autoimmune diabetes in NOD diabetes with a very short duration of in the disease pathogenesis (25,28,29). In mice (34). In contrast, a clinical trial disease (,4 weeks insulin treatment) support, almost half of the clonable T with , another humanized compared with other studies (1–7), the cells from pancreatic draining lymph no- anti-CD3 similar measurement of functional b-cell mass des of individuals with type 1 diabetes to otelixizumab, failed to identify IAA by hyperglycemic clampsdthe gold recognized an insulin epitope in a DR4- at baseline (within 8 weeks after diag- standard (24)dto derive objective restricted manner (30). nosis) as a predictor of good clinical outcome measures and stratification Anti-CD3 treatment is believed to pri- response (8). The reasons for the dis- criteria, the assessment of functional marily target pathogenic T cells, consid- crepant results may relate to the lower outcome as change versus baseline, as ered to be the main effectors of b-cell numbers of participants (22 responders well as the use of the slightly adapted destruction(31),inthecontextofa vs. 27 nonresponders) than in the cur- (14) original Palmer I(A)A assay (15) with primed and ongoing immune response, rent study. Moreover, our observations acid charcoal extraction to strip IAA or IA while preserving regulatory T cells (32). may have benefitted from the better from bound insulin before incubation Conceivably, IAA+ patients in a more ag- reproducibility and accuracy of hyper- with insulin tracer and with correction gressive disease phase may benefitmost glycemic clamp–derived measures of of each individual IAA result for nonspe- from otelixizumab treatment as a result functional b-cell mass compared with cific binding in the presence of an excess of diminished autoreactive T-cell activ- parameters derived from other stimula- cold insulin. Because most participants ity and/or a relative predominance of tion tests, as reflected by their superior were sampled for antibodies within 2 regulatory T cells. Anti-CD3 treatment predictive value for clinical outcome in weeks from start of insulin treatment also caused a transient partial depletion pretype 1 diabetes and their close cor- (median 6 days), the reported IAA levels in B lymphocytes (4). As suggested in the relation with implanted b-cell mass in represent true autoantibodies. The trial (33), B lymphocytes may islet transplantation or histopathologi- otelixizumab trial was originally powered contribute to the disease process by cal findings at clinical disease onset to detect differences in C-peptide release augmenting local immune response (26,35–37). but not to test the present hypothesis. through cytokines and/or antigen pre- We have shown that anti-CD3 treat- This retrospective study is limited by sentation to T lymphocytes. ment did not alter the natural history of the number of patients included in the Our results do not allow us to unam- progressively declining GADA, IA-2A, trial (n = 80) and the IAA prevalence, lead- biguously identify the nature and extent and ZnT8A levels during an 18-month ing to a relatively low number of partic- of the mechanisms underlying our period after diagnosis, whereas it selec- ipants in subgroups, particularly when observations. They have, however, ex- tively suppressed the insulin-induced stratifying for two variables in each treat- cluded that differences in the develop- rise in I(A)A in the subgroup of initially ment arm. Nevertheless, stratification ac- ment of anti-otelixizumab antibodies IAA+ patients with relatively preserved cording to initial IAA status and b-cell could be responsible for the observed C-peptide at diagnosis and better clini- function allowed disclosing differences differences in outcome between IAA+ cal outcome. Current I(A)A assays do not between subgroups, which were con- and IAA2 patients, as was already pre- allow us to establish to what extent this firmed by multivariate analysis in the viously shown not to be the case for lesser IA rise is due to a suppression of entire patient group. differences between patients with ini- insulin (auto)immunity, possibly via the How might a better efficacy of anti- tially high or low AUC C-peptide (22). mechanisms discussed above, or just CD3 in IAA+ patients with relatively pre- Finally, we consider it unlikely that IAA the consequence of a lower cumulative served AUC C-peptide tentatively be levels represent a surrogate for younger insulin dose (17). Nevertheless, our explained? First, individuals with rela- age because they outperformed age for follow-up results are in line with obser- tively preserved b-cell mass at diagnosis, the prediction of functional outcome in vations in several other immune inter- corresponding to $25% of healthy con- multivariate analysis; moreover, in this ventions that have been reported to trol subjects (4), have of course the po- study, IAA prevalence and levels did not selectively influence I(A)A levels in tential to lose more b-cells than those in differ according to age in either treat- insulin-treated patients in association whom most cells have already been de- ment arm (not shown). Our findings with better outcome. For example, this stroyed at inclusion. Next, the presence may thus help explain why age at diag- was the case for rituximab in recent- of IAA is associated with younger age at nosis has been associated with better onset type 1 diabetes (28) and for rapa- diagnosis of type 1 diabetes, higher prev- efficacy in several immune interven- mycin in long-standing disease (38). alence of multiple autoantibodies and tions in type 1 diabetes (1–7). In conclusion, our results have identi- HLA class II susceptibility genotypes, and In support of the ability of IAA, but fiedthepresenceofIAAshortlyafter more prominent insulitis (1,12,25,26). not of the other autoantibodies, to pre- diagnosis as a powerful predictor of Likewise, younger age at first autoanti- dict better responsiveness to immune good responsiveness to anti-CD3 treat- body appearance (in general IAA) and modulation, oral administration of insu- ment. This observation warrants the higher levels of IAA, but not of GADA or lin significantly delayed onset of type 1 measurement of IAA in future trials be- IA-2A, are associated with more rapid diabetes in first-degree relatives with fore intervention in recent-onset type 1 progression to diabetes in risk groups high IAA levels for as long as the treat- diabetes. If confirmed, the added value (27). This suggests that in general, ment continued (10). Likewise, preexist- of baseline IAA in combination with suf- IAA+ individuals experience a more se- ing IAA predicted efficacy of oral insulin ficient residual C-peptide at diagnosis vere underlying autoimmune reactivity, in combination with anti-CD3 treatment (4) can be specified with a higher power. 650 IAA Predict Otelixizumab Efficacy Diabetes Care Volume 38, April 2015

Itmayalsoleadtothedefinition of References Autoimmunity to insulin, beta cell dysfunction, cutoffs for IAA levels above which the 1. Gorus FK, Keymeulen B, Veld PA, Pipeleers and development of insulin-dependent diabe- – prediction of a successful metabolic out- DG. Predictors of progression to Type 1 diabe- tes mellitus. Diabetes 1986;35:139 142 16. Greenbaum CJ, Palmer JP, Kuglin B, Kolb H. come can be improved, thereby paving tes: preparing for immune interventions in the preclinical disease phase. Expert Rev Clin Immu- Insulin autoantibodies measured by radioimmu- the way to more personalized immune nol 2013;9:1173–1183 noassay methodology are more related to interventions. A meta-analysis of previ- 2. Herold KC, Hagopian W, Auger JA, et al. Anti- insulin-dependent diabetes mellitus than those ously completed effective immune in- CD3 monoclonal antibody in new-onset type 1 di- measured by enzyme-linked immunosorbent terventions with enrollment shortly abetes mellitus. N Engl J Med 2002;346:1692–1698 assay: results of the Fourth International Work- shop on the Standardization of Insulin Autoan- after diagnosis may already indicate 3. Sherry N, Hagopian W, Ludvigsson J, et al.; Proteg´ e´ Trial Investigators. Teplizumab for tibody Measurement. J Clin Endocrinol Metab the potential of IAA levels. treatment of type 1 diabetes (Proteg´ e´ study): 1992;74:1040–1044 1-year results from a randomised, placebo- 17. Franke B, Galloway TS, Wilkin TJ. Develop- controlled trial. Lancet 2011;378:487–497 ments in the prediction of type 1 diabetes mellitus, The authors gratefully 4. Keymeulen B, Vandemeulebroucke E, Ziegler with special reference to insulin autoantibodies. Acknowledgments. – acknowledge the several contributions to this AG, et al. Insulin needs after CD3-antibody ther- Diabetes Metab Res Rev 2005;21:395 415 work of their late mentor John C. Hutton, PhD, apy in new-onset type 1 diabetes. N Engl J Med 18. De Grijse J, Asanghanwa M, Nouthe B, et al.; and gratefully acknowledge the expert technical 2005;352:2598–2608 Belgian Diabetes Registry. Predictive power of assistance of coworkers at the central unit and 5. Keymeulen B, Walter M, Mathieu C, et al. screening for antibodies against insulinoma- the clinical biology reference laboratory of Four-year metabolic outcome of a randomised associated protein 2 beta (IA-2beta) and zinc fi the Belgian Diabetes Registry (V. Baeten, T. De controlled CD3-antibody trial in recent-onset transporter-8 to select rst-degree relatives of Mesmaeker, H. Dewinter, N. Diependaele, type 1 diabetic patients depends on their age type 1 diabetic patients with risk of rapid pro- S. Exterbille, T. Glorieux, C. Groven, T. Haulet, and baseline residual beta cell mass. Diabetolo- gression to clinical onset of the disease: impli- A. Ivens, D. Kesler, F. Lebleu, M. Van Molle, gia 2010;53:614–623 cations for prevention trials. Diabetologia 2010; – S. Vanderstraeten, and A. Walgrave from the 6. Skyler JS, Ricordi C. Stopping type 1 diabetes: 53:517 524 Department of Clinical Chemistry and Radio- attempts to prevent or cure type 1 diabetes in 19. Lernmark A, Kolb H, Mire-Sluis T. Towards a immunology, University Hospital Brussels-UZ man. Diabetes 2011;60:1–8 World Health Organization (WHO) approved Brussel, Brussels, Belgium, and E. Quartier, 7. Ludvigsson J; Linkoping¨ Diabetes Immune In- standard sample for islet cell antibodies, G. Schoonjans from the Diabetes Research tervention Study Group. The role of immuno- GAD65 and IA-2 autoantibodies. Diabetologia – Center, Vrije Universiteit Brussel). The authors modulation therapy in autoimmune diabetes. 1999;42:381 382 are grateful to Dr. G. Hale (when at Therapeutic J Diabetes Sci Tech 2009;3:320–330 20. Wenzlau JM, Juhl K, Yu L, et al. The cation fl Antibody Centre, Oxford, U.K.) for performing 8. Herold KC, Gitelman SE, Ehlers MR, et al.; ef ux transporter ZnT8 (Slc30A8) is a major the analyses on prevalence and titers of anti- AbATE Study Team. Teplizumab (anti-CD3 autoantigen in human type 1 diabetes. Proc otelixizumab antibodies and for providing the mAb) treatment preserves C-peptide responses Natl Acad Sci U S A 2007;104:17040–17045 results (22). The authors sincerely thank all par- in patients with new-onset type 1 diabetes in a 21. Williams AJ, Norcross AJ, Chandler KA, ticipating patients and all health care workers randomized controlled trial: metabolic and Bingley PJ. Non-specific binding to protein A who contributed to their recruitment and immunologic features at baseline identify a sub- Sepharose and protein G Sepharose in insulin follow-up in the otelixizumab trial (list provided group of responders. Diabetes 2013;62:3766– autoantibody assays may be reduced by pre- in Keymeulen et al. [4]). 3774 treatment with glycine or ethanolamine. Funding. The otelixizumab study was sup- 9. Piemonti L, Everly MJ, Maffi P, et al. Alloan- J Immunol Methods 2006;314:170–173 ported by the JDRF (grants 4-2001-434 and tibody and autoantibody monitoring predicts 22. Hale G, Rebello P, Al Bakir I, et al. Pharma- 4-2005-1327). The autoantibody measurements islet transplantation outcome in human type 1 cokinetics and antibody responses to the CD3 an- for the present report were performed in the diabetes. Diabetes 2013;62:1656–1664 tibody otelixizumab used in the treatment of type context of the EU-FP7 project 241883 and 10. Vehik K, Cuthbertson D, Ruhlig H, Schatz 1 diabetes. J Clin Pharmacol 2010;50:1238–1248 project G.0868.11 of the Research Foundation DA, Peakman M, Krischer JP; DPT-1 and TrialNet 23. Palmer JP. C-peptide in the natural history Flanders (FWO-Vlaanderen, Belgium). B.K. and Study Groups. Long-term outcome of individu- of type 1 diabetes. Diabetes Metab Res Rev I.W. are Senior Clinical Investigators for the als treated with oral insulin: diabetes preven- 2009;25:325–328 Research Foundation Flanders. K.V. is supported tion trial (DPT-1) oral insulin trial. Diabetes 24. DeFronzo RA, Tobin JD, Andres R. Glucose by EU-FP7 project 241883. H.W.D. and J.M.W. Care 2011;34:1585–1590 clamp technique: a method for quantifying in- received funding from National Institutes of Health 11. Ludvigsson J, Binder C, Mandrup-Poulsen T. sulin secretion and resistance. Am J Physiol (R01-DK-052068, R56-AI-094389) and JDRF (4- Insulin autoantibodies are associated with islet 1979;237:E214–E223 2007-1056). The sponsor of the study had no cell antibodies; their relation to insulin antibod- 25. Jasinski JM, Eisenbarth GS. Insulin as a pri- role in study design, data collection, data analysis, ies and B-cell function in diabetic children. Dia- mary autoantigen for type 1A diabetes. Clin Dev data interpretation, or writing of the report. betologia 1988;31:647–651 Immunol 2005;12:181–186 Duality of Interest. H.W.D. and J.M.W. are 12. Gorus FK. Diabetes registries and early bi- 26. Pipeleers D, Ling Z. Pancreatic beta cells in coinventors of U.S. patent No. 12/521,022 ological markers of insulin-dependent diabetes insulin-dependent diabetes. Diabetes Metab based on PCT/US2007/089125. No other poten- mellitus. Diabetes Metab Rev 1997;13:247–274 Rev 1992;8:209–227 tial conflicts of interest relevant to this article 13. Wenzlau JM, Walter M, Gardner TJ, et al. 27. Steck AK, Johnson K, Barriga KJ, et al. Age of were reported. Kinetics of the post-onset decline in zinc trans- islet autoantibody appearance and mean levels Author Contributions. S.D., B.K., D.G.P., and porter 8 autoantibodies in type 1 diabetic hu- of insulin, but not GAD or IA-2 autoantibodies, F.K.G. designed the study; searched the litera- man subjects. J Clin Endocrinol Metab 2010;95: predict age of diagnosis of type 1 diabetes: di- ture; collected, analyzed, and interpreted data; 4712–4719 abetes autoimmunity study in the young. Dia- and wrote the manuscript. L.K. and I.W. ana- 14. Gorus FK, Sodoyez JC, Pipeleers DG, betes Care 2011;34:1397–1399 lyzed and interpreted data and provided statis- Keymeulen B, Foriers A, Van Schravendijk CF. 28. Yu L, Herold K, Krause-Steinrauf H, et al.; Type tical and writing assistance. A.V.D., E.V.B., Detection of autoantibodies against islet amy- 1 Diabetes TrialNet Anti-CD20 Study Group. U.V.d.V., P.G., K.V., H.W.D., and J.M.W. partic- loid polypeptide in human serum. Lack of asso- Rituximab selectively suppresses specificislet ipated in data collection and provided writing ciation with type 1 (insulin-dependent) diabetes antibodies. Diabetes 2011;60:2560–2565 assistance. All authors approved the final ver- mellitus, or with conditions favouring amyloid 29. Atkinson MA. The pathogenesis and natural sion. F.K.G. is the guarantor of this work, and, as deposition in islets. The Belgian Diabetes Regis- history of type 1 diabetes. Cold Spring Harb Per- such, had full access to all the data in the study try. Diabetologia 1992;35:1080–1086 spect Med 2012;2:a007641 and takes responsibility for the integrity of the 15. Srikanta S, Ricker AT, McCulloch DK, 30. Kent SC, Chen Y, Bregoli L, et al. Expanded T data and the accuracy of the data analysis. Soeldner JS, Eisenbarth GS, Palmer JP. cells from pancreatic lymph nodes of type 1 care.diabetesjournals.org Demeester and Associates 651

diabetic subjects recognize an insulin epitope. depletion, and preservation of beta-cell function. 36. Vandemeulebroucke E, Keymeulen B, Nature 2005;435:224–228 N Engl J Med 2009;361:2143–2152 Decochez K, et al.; Belgian Diabetes Registry. Hy- 31. Martin S, Wolf-Eichbaum D, Duinkerken G, 34. Mamchak AA, Manenkova Y, Leconet W, et al. perglycaemic clamp test for diabetes risk assess- et al. Development of type 1 diabetes despite Preexisting autoantibodies predict efficacy of oral ment in IA-2-antibody-positive relatives of type 1 severe hereditary B-lymphocyte deficiency. insulin to cure autoimmune diabetes in combina- diabetic patients. Diabetologia 2010;53:36–44 N Engl J Med 2001;345:1036–1040 tion with anti-CD3. Diabetes 2012;61:1490–1499 37. Keymeulen B, Gillard P, Mathieu C, et al. Cor- 32. Chatenoud L, Warncke K, Ziegler AG. Clinical 35. Balti EV, Vandemeulebroucke E, Weets I, relation between beta cell mass and glycemic con- immunologic interventions for the treatment of et al. Hyperglycemic clamp and oral glucose tol- trol in type 1 diabetic recipients of islet cell graft. type 1 diabetes. Cold Spring Harb Perspect Med erance test for 3-year prediction of clinical on- Proc Natl Acad Sci U S A 2006;103:17444–17449 2012;2:a007716 set in persistently autoantibody-positive 38. Piemonti L, Maffi P, Monti L, et al. Beta cell 33. Pescovitz MD, Greenbaum CJ, Krause- offspring and siblings of type 1 diabetic pa- function during rapamycin monotherapy in Steinrauf H, et al.; Type 1 Diabetes TrialNet tients. J Clin Endocrinol Metab. 18 November long-term type 1 diabetes. Diabetologia 2011; Anti-CD20 Study Group. Rituximab, B-lymphocyte 2014 [Epub ahead of print] 54:433–439