Long-Lasting Immune Responses 4 Years after GAD-Alum Treatment in Children with Type 1 Diabetes

A phase II clinical trial with glutamic acid decarboxylase (GAD) 65 formulated with aluminium hydroxide (GAD-alum) has shown efficacy in preserving residual insulin secretion in children and adolescents with recent-onset type 1 diabetes (T1D). We have performed a 4-year follow-up study of 59 of the original 70 patients to investigate long-term cellular and humoral immune responses after GAD-alum-treatment. Peripheral blood mononuclear cells (PBMC) were stimulated in vitro with GAD65. Frequencies of naïve, central and effector memory CD4+ and CD8+ T cells were measured, together with cytokine secretion, proliferation, gene expression and serum GAD65 autoantibody (GADA) levels. We here show that GAD-alum-treated patients display increased memory T-cell frequencies and prompt T-cell activation upon in vitro stimulation with GAD65, but not with control antigens, compared with placebo subjects. GAD65-induced T-cell activation was accompanied by secretion of T helper (Th) 1, Th2 and T regulatory cytokines and by induction of T-cell inhibitory pathways. Moreover, post-treatment serum GADA titres remained persistently increased in the GAD-alum arm, but did not inhibit GAD65 enzymatic activity. In conclusion, memory T- and B-cell responses persist 4 years after GAD-alum-treatment. In parallel to a GAD65-induced T-cell activation, our results show induction of T-cell inhibitory pathways important for regulating the GAD65 immunity.


Introduction
Type 1 diabetes (T1D) is an autoimmune disease caused by auto-reactive immune cells which destroy insulin-producing bcells, eventually leading to complete insulin deficiency [1]. Since auto-reactive T cells play a key role in the pathogenesis of T1D, they are considered an attractive therapeutic target for immunomodulatory strategies aiming at preventing or delaying disease onset [2,3]. Glutamic acid decarboxylase 65 (GAD 65 ) is one of the major autoantigens targeted by self-reactive T cells in T1D [4,5]. Despite recent setbacks in a phase II [6] and a phase III clinical trial (submitted manuscript) using GAD 65 formulated with aluminium hydroxide (GAD-alum), we and others have previously shown preservation of residual insulin secretion by GAD-alum treatment, in clinical phase II trials involving recent-onset T1D children [7] and LADA patients [8]. In addition to the clinical efficacy, we have reported that GAD-alum induced an early T helper 2 (Th2)-associated immune deviation in response to GAD 65 [9] along with the appearance of GAD 65 -specific CD4 + CD25 high -FOXP3 + cells [10]. The treatment also enhanced GAD 65 autoantibody (GADA) levels [7] with an increase in subclasses IgG3 and IgG4 and a reduction in IgG1 suggestive of Th2 deviation, while IA-2 autoantibodies remained unaffected [11]. Altogether, these data indicate that GAD-alum treatment induced transient Th2-deviated GAD 65 -specific T-and B-cell responses during the 30-month study period. We have performed a 4-year follow-up study including 59 of the original 70 patients to evaluate long-term efficacy and safety of GAD-alum intervention. No treatment-related adverse events were reported and fasting Cpeptide remained better preserved relative to placebo in patients with ,6 months T1D duration at baseline [12].
Generation of a memory cell pool is important in the acquisition of effective immune therapy, and is formed by clonal expansion and differentiation of antigen-specific lymphocytes that ultimately persist for a lifetime [13]. Thus, the analysis of antigen-specific memory responses may be useful to understand the duration and stability of GAD-alum-induced immune responses. The leukocyte common antigen isoforms CD45RA and CD45RO have long been used to identify human naïve and memory T cells [14]. Naïve cells also express high levels of the chemokine receptor CCR7, which is essential for lymphocyte migration to lymph nodes [15]. Memory T cells contain two subsets, CD45RA -CCR7+ central memory (T CM ) and CD45RA -CCR7-effector memory (T EM ) cells, characterized by distinct homing capacities and effector functions [15]. Upon re-stimulation, T EM show a low threshold for activation and produce cytokines with rapid kinetics. Antigen rechallenge also initiates a memory Th-controlled memory B-cell response that promotes robust antibody production and enhancement of the antigen-specific memory B-cell compartment [16].
The aim of this study was to evaluate the long-term antigenspecific memory T-and B-cell responses in T1D children treated with GAD-alum. We here show that treated patients display sustained GADA levels, increased memory T-cell frequencies and prompt T-cell activation upon in vitro stimulation with GAD 65 , 4 years after GAD-alum intervention. In parallel to a GAD 65 -induced T-cell activation, our results show induction of T-cell inhibitory pathways important for regulating the GAD 65 immunity.

Ethics Statement
This study was approved by the Research Ethics Committee at the Faculty of Health Sciences, Linköping University, Sweden. Written informed consent was obtained from all patients, and for those ,18 years old also their parents, in accordance with the Declaration of Helsinki.

Subjects
The design and characteristics of the trial have previously been described [7]. Briefly, 70 T1D children between 10 and 18 years of age with less than 18 months of disease duration were recruited at 8 Swedish paediatric centres. All participants had a fasting serum Cpeptide level above 0.1 nmol/l and detectable GADA at inclusion. Patients were randomized to subcutaneous injections of 20 mg GAD-alum (DiamydH, Diamyd Medical; n = 35) or placebo (alum only; n = 35) at day 0 and a booster injection 4 weeks later in a double blind setting. After 4 years, patients and their parents were asked whether they were willing to participate in a follow-up study. Fifty-nine patients agreed to participate, of whom 29 had been treated with GAD-alum and 30 had received placebo.

Isolation of PBMC
PBMC were isolated from sodium-heparinised venous fasting blood samples as described previously [9], and immediately stimulated in vitro for Luminex cytokine assay, PCR array and flow cytometry analyse. Remaining PBMC were cryopreserved in aliquots and used for T-cell enzyme-linked immunospot (ELISpot) and proliferation assays. It was not possible to perform all the different laboratory analyses on each study participant, due to the limited sample size. All laboratory work was performed in a blinded manner.

Gene expression analysis by quantitative Real-Time PCR array
Expression of 15 selected genes (Table 1) was analyzed using a customized Human Gene RT 2 profiler TM PCR array (SABiosciences). PBMC were cultured for 24 h in AIM-V medium with or without 5 mg/ml of GAD 65 , and total RNA was isolated according to the RNeasy 96 vacuum/spin protocol (Qiagen) and quantified by optical density (OD) measurements at 260 nm. The purity of the RNA was ensured with an OD 260/280 ratio above 1.8, and RNA integrity was confirmed using Agilent 2100 bioanalyzer (Agilent Technologies). Each RNA sample (0.12 mg) was transcribed into PCR template with the RT 2 First Strand Kit (SABiosciences). Templates were then combined with RT 2 SYBRH Green/ROX TM qPCR Master Mix, and aliquots of 25 ml were loaded into each well containing the pre-dispensed gene-specific primer sets. ABI Prism 7900HT was employed for sequence detection, and sequence detection systems (SDS) version 2.3 (Applied Biosystems) was used to calculate the threshold cycle (Ct) values. An evaluation of the quality controls provided the relative levels of genomic DNA contamination and inhibition of either the reverse transcription or the PCR itself.
Relative gene expression was calculated with the delta-delta Ct (DDCt) method, using the normalized DCt value of each sample, calculated by subtracting the average Ct value of two housekeeping genes (GAPDH and HPRT1) from the Ct value of the gene of interest. The spontaneous Ct value was thereafter subtracted from the Ct value of the GAD 65 -stimulated sample. To calculate the DDCt, the average DCt value of each gene in the placebo group was subtracted from the average Ct value of the corresponding gene in the GAD-alum group. The fold-change for each gene was calculated as 2 (2DDCt) .

Detection of antigen-specific T-cell responses by ELISpot
Detection of antigen-specific T-cell responses was performed with an accelerated co-cultured dendritic cell (acDC)-amplified ELISpot assay, as described [18]. Briefly, cryopreserved PBMC were thawed, washed twice in AIM-V medium and re-suspended at 10610 6 /ml in AIM-V medium containing 1000 U/ml GM-CSF and 500 U/ml IL-4 (both from R&D). Cells were seeded at 10 6 /100 ml/well in flat-bottom 96-well plates and stimulated with 10 mg/ml GAD 65 , 40 mg/ml TTX (Statens Serum Institut) or no antigen at 37uC in 5% CO 2 . After 24 h, 100 ml AIM-V medium containing 100 U/ml TNF-a, 10 ng/ml IL-1b, 1 mM prostaglandin E2 and 0.5 ng/ml IL-7 was added to each well and cultured for another 24 h. Following this 48 h stimulation, non-adherent cells were washed, re-suspended in fresh AIM-V medium, seeded in quadruplicates at 1610 5 cells/well and incubated for 6 h in 96well PVDF plates (Millipore) precoated with anti-IFN-c or anti-IL-4 Abs (U-CyTech). Secretion of IFN-c and IL-4 was visualized with a biotin-conjugated anti-IFN-c or -IL-4 Ab (U-CyTech), alkaline phosphatase-conjugated ExtrAvidin and Sigmafast 5bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT) tablets (both from Sigma), as described [19]. Spots were counted using a Bioreader 5000 Pro-SF (Bio-Sys). Means of quadruplicate wells were calculated and the results expressed as spot-forming cells (SFC)/10 6 PBMC after background subtraction. The cut-off for a positive response was set at 3 SD above the average basal reactivity [19].

Autoantibody and GADA IgG subclass analyses
Serum GADA and IA-2A titres were determined using a radiobinding assay employing 35 S-labeled recombinant human GAD 65 and IA-2, as previously described [11]. The GADA IgG1, 2, 3 and 4 subclasses were measured using a modification of the conventional GADA assay [11].

GAD 65 enzymatic activity assay
Recombinant human GAD 65 enzymatic activity was measured in the presence of patient serum by a 14 CO 2 -trapping method based on the enzymatic conversion of glutamate to GABA as previously described [11], and expressed as a percentage of the maximum GAD 65 enzymatic activity. As GADA-positive serum from Stiff person syndrome (SPS) patients has been shown to inhibit this reaction [20], serum from one SPS patient was included in each assay as a positive control for inhibition.
C-peptide C-peptide levels were measured in serum samples with a timeresolved fluoroimmunoassay (AutoDELFIA TM C-peptide kit, Wallac), described previously [7]. Stimulated C-peptide was measured during a mixed meal tolerance test (MMTT) in patients who had a maximal C-peptide response of .0.20 nmol/l at the 30-month follow-up, i.e. 21 GAD-alum-treated patients and 10 patients in the placebo group. Clinical effect of treatment was defined by changes in stimulated C-peptide measured as area under the curve (AUC) from baseline to 48 months.

Statistical analysis
As the immunological markers were not normally distributed, non-parametric tests corrected for ties were used. Unpaired analyses were performed using the Mann-Whitney U-test, and correlations were analysed with Spearman's rank correlation coefficient test. Differences within groups were calculated by Wilcoxon signed rank test. A probability level of ,0.05 was considered statistically significant. Calculations were performed using PASW statistics version 18 for Windows (SPSS Inc).

GAD-alum-treated patients display increased memory CD4+ T-cell frequencies after in vitro GAD 65 stimulation
To test whether their frequencies were altered after GAD-alumtreatment, naïve and memory CD4+ and CD8+ subsets were analyzed in resting (Fig. 1A) and GAD 65 -stimulated (Fig. 1B) PBMC. The frequency of naïve (CD45RA + CCR7 + ), T CM (CD45RA -CCR7 + ) and T EM (CD45RA -CCR7 -) CD4+ cells in resting cultures did not differ between the placebo and GAD-alum groups (Fig. 1C). When stimulated with GAD 65 , frequencies of T CM (p = 0.040) and T EM (p,0.001) increased whereas naïve cells decreased (p,0.001) in GAD-alum-treated patients, while the placebo group remained unaffected. The frequency of T CM (p = 0.041) and T EM (p = 0.006) was also significantly higher, and naïve cells lower (p,0.001) in the GAD-alum group compared to the placebo group.
Similarly, the frequency of naïve, T CM and T EM CD8+ cells in resting cultures did not differ between the placebo-and GADalum-treated patients (Fig. 1D). However, after GAD 65 -stimulation, the frequency of T CM (p = 0.010) and T EM (p,0.001) increased in GAD-alum-treated patients, whereas naïve CD8+ decreased (p,0.001). The proportion of T EM was also significantly higher (p = 0.004), and naïve cells lower (p,0.001) compared to the placebo group, which remained unaffected upon GAD 65stimulation.
Induction of a cell subset with higher SSC and FSC was evident upon GAD 65 -stimulation only in GAD-alum treated patients ( Fig. 1E-F). The majority of cells within this population were CD4+ with a memory phenotype.
In vitro stimulation with GAD 65 induces T-cell activation and proliferation in GAD-alum-treated patients Proliferative responses to GAD 65 were also significantly higher in the GAD-alum-treated patients compared to placebo (p = 0.011; Fig. 2B). In contrast, proliferative responses to the control antigens IA-2 853-872 (T1D-associated antigen), TTX (irrelevant control) and PHA (positive control) did not differ between GAD-alum-and placebo-treated patients.

GAD 65 -stimulation of PBMC induces cytokine secretion in GAD-alum-treated patients
Since memory T cells are capable of immediate effector cytokine production when stimulated in vitro, we sought to study the cytokine profile in PBMC supernatants after antigen challenge using a multiplex Luminex assay. When stimulated with GAD 65 , secretion of IL-1b, IL-2, IL-5, IL-10, IL-13, IL-17, IFN-c and TNF-a was higher in PBMC from GAD-alum-treated patients compared to the placebo group (Fig. 3A). In contrast, spontaneous as well as TTX and IA-2 853-872 -induced secretion were similar in the two groups. Although the secreted levels of IL-7 and IL-15 were below the detection limit in the Luminex assay and although TGF-b was not available for multiplex testing, mRNA expression of these cytokines was up-regulated in PBMC from GAD-alumtreated patients re-challenged with GAD 65 , as shown in Fig. 2A.
In order to search for immune surrogate markers of clinical efficacy, we analyzed the association between cytokine secretion and b-cell function, as measured by stimulated C-peptide. No statistical significant associations were observed between cytokine production, or any other immune marker included in this study, and stimulated C-peptide. Still, to graphically illustrate the cytokine profile in relation to clinical efficacy, GAD-alum-treated patients were divided in two subgroups; patients with a loss of Cpeptide AUC #60% (n = 5), and patients with a loss of AUC .60% (n = 16; Fig. 3B). The cytokine profile in patients with a loss of AUC #60% was characterized by a more pronounced GAD 65induced IL-5, IL-10, IL-13 and IL-2 secretion, whereas patients with a loss of AUC .60% had a more pronounced inflammatory profile characterized by IFN-c, IL-1b and IL-17 secretion.

Sustained high serum GADA titres in GAD-alum-treated patients
Since autoantibody determination may be useful in assessing the long-lasting immunological impact of autoantigen treatment, we next analysed serum GADA titres. Our results show higher GADA levels in GAD-alum-treated patients compared to placebo, 4 years after treatment (p = 0.034; Fig. 5A). The GADA titres were also higher compared to baseline levels in the GAD-alum-(p = 0.007) but not in the placebo-treated group. In addition, IA-2A levels were determined in order to confirm that the persistent humoral response was antigen-specific. No difference between the two groups was observed (not shown). Further, the GADA IgG 1-4 subclass distribution was determined, as Th1-and Th2-cell cytokine production influence IgG-subclass switching [21,22]. However, the subclass distribution did not differ between groups (Fig. 5B), nor did it differ compared to baseline (not shown).
High GADA titres are often found in SPS patients, raising concerns that therapies boosting GADA may have deleterious neurological effects. However, SPS is characterized by GADA which inhibit the GAD 65 enzymatic activity [20], which is not the case in T1D. Therefore, we investigated whether serum containing high GADA titres generated by GAD-alum treatment, was inhibitory. The enzymatic activity did not differ when rhGAD 65 was incubated with serum from GAD-alum-(median 90 %, range 42-100) and placebo-treated patients (median 91 %, range 65-100), but was significantly higher compared to GAD 65 incubated with control serum from a SPS patient (median 20 %, range 19-24; p,0.001; Fig. 5C).

Discussion
We have shown significant preservation of residual insulin secretion 4 years after GAD-alum treatment in T1D children and adolescents with ,6 months T1D duration at inclusion, compared to placebo [12]. In the present study, we aimed to characterize the long-term antigen-specific memory T-and B-cell responses. Detection of antigen-specific memory cells ex vivo is a great challenge due to low frequencies. A previous study has demonstrated that only one in 30,000 or less CD4+ T-cells in peripheral blood from patients with recent-onset T1D is GAD 65specific [23], and activation and in vitro amplification of the GAD 65 -specific T-cells is crucial for detection. Our results show induction of T-cell subsets with a predominant memory phenotype upon in vitro GAD 65 -stimulation in PBMC from GAD-alum-treated patients. This suggests clonal expansion of the memory T-cell compartment upon antigen re-challenge, in parallel to the observed proportional reduction in naïve T-cell percentage.
When stimulated in vitro, memory T cells display low activation thresholds, immediate cytokine production and vigorous proliferation [13]. Our results show that the T-cell activation markers CD69, CD25 and PD-1 were all up-regulated in PBMC from GAD-alum-treated patients, and that proliferative T-cell responses to GAD 65 together with GAD 65 -induced cytokine secretion were significantly higher compared to placebo, the latter confirming our previous findings [7,9,10]. Given that the cytokine-and proliferative responses elicited by various control antigens in vitro were similar between the two treatment groups, the effect of GAD-alum appears to be antigen-specific. This selective immune modulation might also be considered as an indication of safety, since it would be undesirable to non-specifically influence immune responses to unrelated antigens. Besides detecting cytokine secretion in PBMC, we quantified antigen-specific IFN-c and IL-4 T-cell responses using an acDC assay. PBMC were selected on the basis of availability and thereby a limiting factor for including additional cytokines. In the assay, antigen and DC-activating agents rapidly induce, pulse and mature DCs, thus lining up the sequential steps of T-cell activation within 48 h and amplifying antigen-specific responses. The utility of these acDC-based assays for immune monitoring of vaccination trials has been previously demonstrated [18,24]. The number of GAD 65 -induced IL-4 SFC, a cytokine difficult to detect with the Luminex assay, was significantly increased in the GAD-alum group compared to placebo. Further, IL-4 SFC correlated with IL-13 secretion, two Th2 cytokines with overlapping biological effects that share receptor components [25]. In parallel, the GAD 65 -induced IFN-c SFC were also increased in the GAD-alum-treated patients, and correlated with the GAD 65induced IFN-c secretion, supporting the reliability of the cytokine observations.
Proliferation of memory T cells can be driven not only by antigenic stimulation but also by cytokines. Here we show that gene expression of IL-7 and IL-15, two cytokines that are constitutively produced by a variety of cells and play an essential role for maintenance of both CD4+ and CD8+ T cells [13], was higher after GAD 65 -stimulation in the GAD-alum-treated group. In addition, IL-2, which is involved in long-term survival of antigen experienced CD4+ and regulatory T cells [26,27], was also induced by GAD 65 -stimulation. Receptors for IL-2, IL-7 and IL-15 transmit signals mainly through STAT5, which is a critical factor for inducing and maintaining the expression of FOXP3 [28], and of the anti-apoptotic molecule Bcl-2 [29]. Upregulation of the aforementioned cytokines and their receptors upon GAD 65 -stimulation, together with that of their associated signalling pathways and transcription factors suggests their involvement in the maintenance of a long-lasting GAD 65 -specific T-cell memory population. B-cell memory is characterized by persistent elevated specific antibody titres and generation of long-lived memory B cells [30]. Elevated GADA levels 4 years after GAD-alum-treatment, together with up-regulated PRDM1, a transcription factor essential for development of Ig-secreting cells and maintenance of long-lived plasma cells [31], suggests an induction of plasma cells continuously secreting GADA. PRDM1 is also expressed in effector and memory T cells [32,33], and appears to have a role in Th2 cells by repressing Th1 genes [34].
The outcome of a T-cell response is shaped by the balance between co-stimulatory and co-inhibitory signals, which are often simultaneously provided to T cells by their surrounding cells. PD-1 is a member of the CD28 superfamily of immunoreceptors that is up-regulated following TCR stimulation [35], and interaction with its ligand PD-L1 inhibits T-cell effector functions [36]. Upregulation of PD-1/PD-L1 in parallel to GAD 65 -induced T-cell activation and proliferation in the GAD-alum group demonstrates activation of co-inhibitory pathways important for regulating the immune balance. Reliable biomarkers associated with therapeutic success following vaccination with b-cell antigens are still lacking. We have previously shown that, although GAD-alum-treatment induced a GAD 65 -specific cell population characterised by a broad cytokine profile [7], the response was preceded by an early Th2 immune deviation [9]. The cytokine profile observed in patients with better preserved C-peptide after 4 years, even though not statistically assured, may suggest that a beneficial clinical response might be associated with a persistent Th2/Tregskewed GAD 65 -specific immune response. In a vaccination trial by Harrison and co-workers using intranasal insulin, immune responses were characterized by IFN-c ELISpot and autoantibody measurements [24]. In contrast to our findings, antigenspecific IFN-c and antibody responses decreased following treatment, suggesting that the therapeutic effect (or lack thereof) may be linked to different immunological mechanisms. The administration route and the use of alum adjuvant may be important factors in triggering these different mechanisms. Recently a phase II trial [6] and a European phase III trial (submitted manuscript) using GAD-alum have failed to reach their primary outcome. However, it cannot yet be excluded that treatment might be beneficial in certain patient subgroups. Thus, in order to make improvements in b-cell antigen treatment, alone or in combination with other therapies, it is of utmost importance to learn more about the immunological effects.
In conclusion, we here show persistent GAD 65 -specific cellularand humoral immune responses 4 years after GAD-alum intervention in T1D children. Prompt re-activation of GAD 65reactive T cells upon in vitro antigen challenge was accompanied by secretion of Th1, Th2 and Treg cytokines and by induction of coinhibitory pathways that may be of importance for regulating the GAD 65 immunity.