Human Anti-CD38 Autoantibodies Raise Intracellular Calcium and Stimulate Insulin Release in Human Pancreatic Islets

The clinical characteristics of the population screened for anti–CD38 autoimmunity and the methods for the detection of anti–CD38 antibodies have been reported previously (7). Briefly, recombinant CD38–maltose binding protein (MBP) fusion protein (rCD38-MBP) (68 kDa) was obtained from a human CD38 cDNA encoding amino acids 45–300 (8). Then, 10 μg of rCD38-MBP was separated on a 10% SDS polyacrylamide gel and transferred to a polyvinylidene fluoride membrane. The membranes were subsequently blocked with a solution of phosphate-buffered saline (PBS) containing 5% nonfat dry milk and 0.15% Tween-20 and incubated for 60 min. Sera from diabetic patients and control subjects were diluted (1:1,000) in the same buffer. After rinsing with PBS containing 0.15% Tween-20, the membranes were incubated for 60 min with a purified rabbit Ig specific for human Ig (G+M) labeled with horseradish-peroxidase. The signals were revealed by using an enhanced chemiluminescence detection system; the intensity of the bands was measured by densitometry (7). For the present studies, 46 sera were used: 34 from type 2 diabetic patients (13 men and 21 women, mean age 60 ± 9 years) and 12 from healthy subjects (5 men and 7 women, mean age 58 ± 10 years).

IB4 (an agonistic anti–CD38 mAb, IgG_2a), used as the intact molecule or the F(ab′)₂ fragment, was used as the reference agonistic mAb. IB6 (anti–CD38 mAb, IgG_2b) and OKT10 (anti–CD38 mAb, IgG₁), both nonagonistic mAbs, were used to compete with CD38 binding by autoantibodies. CB-gp120 (anti–HIV-1 gp120 mAb, IgG_2a) and CB-ICAM-1 (anti–CD54, IgG₁) were used as isotype controls. All mAbs used in the [Ca²⁺]_i mobilization experiments were purified by affinity chromatography and high-performance liquid chromatography on hydroxylapatite and prepared pyrogen-free in the laboratory (9).

Jurkat T-cells, a human T-cell acute lymphatic leukemia line cloned by limiting dilution or positively selected by MACS (Miltenyi Biotech GmbH, Bergisch-Gladbach, Germany) to obtain a population displaying a homogeneous CD3⁺CD38⁺CD31⁺ phenotype, were cultured in RPMI-1640 medium supplemented with penicillin, streptomycin, glutamine, and 10% fetal calf serum (FCS) (Seromed; Biochrom KG, Berlin), herein referred to as complete medium. Murine NIH-3T3 fibroblasts stably transfected with human CD38 gene and mock-transfected controls were obtained as described (10) and mantained in complete medium.

Isolation and culture of pancreatic islets from human donors were performed as previously described (7,11,12). Briefly, islets were prepared by collagenase digestion and density-gradient purification from the pancreas of human cadaver donors referred through the local organ-procurement organization (with the permission of the Institutional Ethics Committee of the University of Pisa School of Medicine). At the end of the isolation procedure, islets were resuspended in M199 tissue culture medium supplemented with 10% serum and antibiotics and kept at 37°C in a 5% CO₂ atmosphere.

Islets were dissociated into single cells according to previously reported procedures (13). Briefly, islets were gently resuspended in a calcium-free Krebs-Ringer-HEPES buffer with 1 mmol/l EGTA at 30°C. After 8 min, trypsin (16.5 μg/ml) and DNAase (2 μg/ml) were added, and a further 8- to 10-min incubation at 30°C was allowed. Islet cell dissociation was monitored by observing aliquots of the suspension under an inverted light microscope. When most cells were single or in small groups of two to three, the procedure was stopped by adding Krebs-Ringer bicarbonate (KRB) buffer containing 1.67 mmol/l calcium and 1% calf serum. At the end of the dissociation, cells were washed with PBS, resuspended in M199 culture medium, and eventually placed at 37°C in a CO₂ incubator.

Changes in [Ca²⁺]_i levels in Jurkat T-cells were monitored by flow cytometry after loading cells with fluo-3-acetoxymethyl ester (fluo-3-AM), a Ca²⁺-sensitive fluorescent dye, and 4-bromo A23187 ionophore. Fluo-3-AM, Hanks’ balanced salt solution (HBSS), Pluronic F-127, and DMSO were purchased from Sigma (St. Louis, MO). The loading procedure was adapted from the work of Deaglio et al. (14) and Morra et al. (15). Briefly, Jurkat T-cells were washed twice in HBSS, pH 7.0, 5% FCS, resuspended in the same medium at the concentration of 2 × 10⁶ cells/ml, and incubated for 30 min at 37°C with 5 μmol/l fluo-3-AM in the presence of 0.01% Pluronic F127. Cells were then washed twice and resuspended in test tubes at 10⁶ cells/ml and used within 1 h. After the base level was calibrated, sera under analysis (50 μl) were added to 650 μl of Jurkat cells (10⁷ cells/ml in complete medium).

[Ca²⁺]_i mobilization was analyzed according to previously published protocols (16). Briefly, dispersed cells were loaded with 2 μmol/l fluo-3-AM over 30 min of incubation in a HEPES-buffered Earle’s medium (124 mmol/l NaCl, 5.4 mmol/l KCl, 1.8 mmol/l CaCl₂, 0.8 mmol/l MgSO₄, 1.0 mmol/l NaH₂PO₄, 14.3 mmol/l NaHCO₃, 10 mmol/l HEPES, pH 7.3) containing 5.6 mmol/l d-glucose, 5 mg/ml bovine serum albumin (BSA) (Sigma), and 0.3 mg/ml DNAse (Boehringer Mannheim, Mannheim, Germany). Cells were then washed twice in HEPES-buffered Earle’s medium, resuspended in the same buffer, and analyzed after 15 min of incubation at 30°C. The gate for β-cell identification was drawn as a polygone on the dot-plot recorder, and for cell discrimination, both the fluo-3 mol/l signal and NADPH autofluorescence under ultraviolet excitation were used, as previously described (16). After we calibrated the base level signal, sera under analysis (50 μl) were added to 650 μl of cell suspension.

Dynamic changes in [Ca²⁺]_i were monitored at room temperature on a FACScalibur (Beckton Dickinson, San Jose, CA) flow cytometer by continuously plotting the shift in the fluo-3-AM fluorescence over a 540-s time period (intensity-over-time density plot). Mean fluorescence intensity (MFI) was calculated by Cellquest software. Standard controls included incubation with sera from patients negative for anti–CD38 autoantibodies and the A23187 ionophore (in Jurkat T-cells) or 16 mmol/l glucose (in dispersed islet cells). To confirm the specificity of signals obtained with anti–CD38-positive sera, before the [Ca²⁺]_i-mobilizing experiments, β-cells were preincubated with a mixture of IB6 and OKT10 (nonagonistic anti–CD38 mAbs, both used at 10 μg/ml) used as a blocking tool.

To differentiate agonistic activity mediated by CD38 signaling after autoantibody ligation from artifacts related to the presence of undefined serum components or contaminants, sera were preincubated (1 h at room temperature) with murine NIH-3T3 cells transfected with the human CD38 gene and expressing high surface levels of the molecule. Cells were then centrifuged and the supernatant was collected and used in experiments of [Ca²⁺]_i mobilization in Jurkat T-cells. Sera preincubated with mock-transfected NIH-3T3 were used as controls. The same sera were separately incubated (v/v) with Staphylococcus aureus protein A covalently linked to Sepharose CL-4B (Pharmacia, Uppsala, Sweden). After 4 h at room temperature, the mixture was centrifuged and the supernatant was collected and used in [Ca²⁺]_i mobilization assays.

CD38 blocking experiments were performed in an attempt to identify the serum component(s) responsible for [Ca²⁺]_i mobilization. Jurkat T-cells were loaded with fluo-3-AM and incubated (10 min at room temperature) with a mixture of IB6 and OKT10 (nonagonistic anti–CD38 mAbs)—both at 10 μg/ml—used as a blocking tool. A mixture of CB-ICAM-1 and CB-gp120 (17)—both used at the same concentrations as above—was used as control. Sera were then assayed on these cells for their ability to mobilize [Ca²⁺]_i.

Within 7–12 days of isolation, the insulin secretory function of the islets in response to 3.3 mmol/l glucose was assessed by the batch incubation method, as previously reported (13,18). Groups of islets of comparable size (∼50/tube) were preincubated at 37°C for 45 min in KRB solution and supplemented with 3.3 mmol/l glucose and 0.5% BSA. The islets were then washed and incubated at 37°C for 45 min in KRB solution containing 3.3 mmol/l glucose, either with or without the addition of 10% anti–CD38-positive or anti–CD38-negative serum. For each test serum, two to four groups of islets were incubated separately. At the end of the incubation period, aliquots of the medium were collected for the measurement of insulin concentration.

To control for the inherent variability of human islet preparations and to also test the effect of sera on islets exposed to high glucose (16), in additional islet preparations, 14 of the 22 sera tested as above were reassayed by measuring islet insulin content using acid alcohol extraction (13,18); the insulin secretory response to 3.3 or 16.7 mmol/l glucose was expressed as percentage of total islet hormone content. At 3.3 mmol/l glucose, there was a positive relationship between insulin release expressed as insulin concentration in the medium and as percentage of total islet content (r = 0.59, P < 0.04).

Sera were considered to be positive for anti–CD38 autoantibodies if the optical density on the Western blot was equal to 3 SD above the mean of the previously reported control group (i.e., ≥1.9 units) (7). In the [Ca²⁺]_i mobilization experiments, basal MFI averaged 69 ± 1 units in a total of 32 incubations. An increment in MFI of ≥20% of the mean basal MFI (i.e., ≥83 MFI units) was taken to be a positive [Ca²⁺]_i response.

Data are given as means ± SE. Because of their nonnormal distribution, insulin concentrations are given as median and interquartile range and were log-transformed for use in parametric statistical analyses. Comparison of mean group values was performed by analysis of variance (ANOVA); simultaneous testing for group and treatment was performed by two-way ANOVA, with Bonferroni-Dunn post hoc testing for individual group comparisons. The χ² test was used as a measure of association between categorical variables.

On the Western blot, a band of ∼68 kDa was observed in the sera of some diabetic patients and normal subjects. The signals were abolished by preincubating anti–CD38-positive sera with human recombinant CD38, whereas preincubation with LacZ(β-galactosidase)-MBP did not affect reactivity with the CD38 target. T16, an anti–human CD38 mAb, yielded a pattern similar to that obtained with anti–CD38-positive human sera. While we were immunoblotting sera containing putative anti–CD38 autoantibodies against recombinant CD38 from Pichia pastoris, a band of apparent molecular mass of ∼30 kDa (corresponding to the soluble CD38) was revealed; also evident was a second band of higher molecular weight, as previously reported (19). No signals were produced by sera lacking anti–CD38 activity. The signals (Figs. 1A and D) were abolished by preincubating anti–CD38-positive sera with recombinant CD38 from P. pastoris (Fig. 1B). Preincubation of anti–CD38-positive sera with S. aureus protein A covalently linked to Sepharose CL-4B also abolished the autoradiographic signals (Fig. 1C), indicating that the observed anti–CD38 activity was due to IgG.

We first tested the ability of sera (n = 46) to induce changes in [Ca²⁺]_i in Jurkat T-cells. The results of the experiments are summarized in Table 1. Of 19 anti–CD38-positive sera, 11 (58%) showed significant (≥20% of basal) increases in [Ca²⁺]_i (Fig. 2), whereas no [Ca²⁺]_i-mobilizing activity was found in 27 anti–CD38-negative sera (χ² = 20.5, P < 0.0001 for the association between CD38 positivity and [Ca²⁺]_i-mobilizing activity).

When preincubated with S. aureus protein A to remove IgG, anti–CD38-positive sera with [Ca²⁺]_i-mobilizing activity showed a 70 ± 5% reduction in activity (4 of 4 sera) (Fig. 3), indicating that the native [Ca²⁺]_i-mobilizing activity was likely due to IgG. In contrast, no change was observed when the same treatment was applied to sera without [Ca²⁺]_i-mobilizing activity, whether they were positive or negative for anti–CD38 autoantibodies.

To rule out unknown effects of serum IgGs other than those reacting with CD38, sera were incubated with CD38-transfected NIH-3T3 fibroblasts (CD38/NIH-3T3). Preincubation of anti–CD38-positive sera with CD38/NIH-3T3 fibroblasts abolished [Ca²⁺]_i mobilization (Fig. 3C), whereas no effect was observed with the preincubation with mock-transfected (CD38-negative) NIH-3T3 cells, indicating specific absorption of the anti–CD38 autoantibodies.

To verify that blocking CD38 counteracted native [Ca²⁺]_i-mobilizing activity, Jurkat T-cells were first preincubated with saturating concentrations (50 mg/ml) of IB6 and OKT10 (two nonagonist anti–CD38 mAbs) then used in [Ca²⁺]_i mobilization assays. In these experiments, anti–CD38-positive sera eliciting a [Ca²⁺]_i response in untreated cells lost their [Ca²⁺]_i-mobilizing activity after preincubation with either IB6 or OKT10, suggesting transmembrane signaling mediated by the CD38 molecule (Fig. 3D). In fact, CB-gp120 and CB-ICAM-1 mAbs, used as controls, were ineffective in blocking serum-induced [Ca²⁺]_i-mobilization.

Incubation of dispersed islet cells with 16 mmol/l glucose caused a prompt and sustained threefold increase in [Ca²⁺]_i (Fig. 4), demonstrating that the cell preparation protocol yielded functional β-cells. Incubation of a separate sample from the same cell preparation with an anti-CD38-positive serum gave a less intense (approximately twofold) but clearly detectable increase in [Ca²⁺]_i (Fig. 4). This signal disappeared when cells were preincubated with a mixture of IB6 and OKT10 (nonagonistic anti-CD38 mAbs). An anti–CD38-negative serum gave no signal.

A total of 14 sera, 11 of which were positive for anti-CD38 autoantibodies, were tested in islet cells; 5 sera positive for anti–CD38 displayed significant [Ca²⁺]_i-mobilizing activity (Table 1).

We first tested whether CD38 positivity or [Ca²⁺]_i-mobilizing activity in Jurkat T-cells was associated with stimulation of insulin release in isolated human islets. In islets cultured with 3.3 mmol/l glucose, 22 sera were tested, of which 6 were anti–CD38-positive and had [Ca²⁺]_i-mobilizing activity, 6 were anti–CD38-positive but devoid of [Ca²⁺]_i-mobilizing activity, and 10 had neither CD38 positivity nor [Ca²⁺]_i-mobilizing activity (no anti–CD38-negative sera were found to have detectable [Ca²⁺]_i-mobilizing activity). The results show that the six anti–CD38-positive sera with [Ca²⁺]_i-mobilizing activity induced significantly (P = 0.002) greater insulin release than control incubations, whereas sera without [Ca²⁺]_i-mobilizing activity, whether CD38 positive or negative, did not (Fig. 5).

Finally, we tested whether sera exhibiting [Ca²⁺]_i-mobilizing activity in dispersed human islet cells were also capable of stimulating insulin release in cultured human islets. In nine control incubations with 3.3 mmol/l glucose and five control incubations with 16.7 mmol/l glucose, insulin in the medium averaged 1.17 ± 0.07 and 2.30 ± 0.35%, respectively, of total islet insulin content. When incubated with sera positive for anti–CD38 and [Ca²⁺]_i-mobilizing activity (n = 5), cultured islets showed an increase in insulin release at both 3.3 and 16.7 mmol/l glucose, whereas sera positive for anti–CD38 but not [Ca²⁺]_i-mobilizing activity (n = 6) and sera negative for both anti–CD38 and [Ca²⁺]_i-mobilizing activity (n = 6) did not stimulate insulin release at either glucose concentration (P < 0.0001 for the effect of glucose and P < 0.002 for the effect of group by two-way ANOVA) (Fig. 6).

In a total of 14 sera, [Ca²⁺]_i-mobilizing activity in dispersed human islet cells was well correlated both with [Ca²⁺]_i-mobilizing activity in Jurkat T-cells and with insulin release at 3.3 mmol/l glucose in cultured islets (Fig. 7).

The present work confirms that human sera reacting with recombinant CD38 during immunoblotting contain IgG autoantibodies that recognize recombinant CD38 from P. pastoris (i.e., the soluble CD38). The new finding is that sera containing anti–CD38 reactive autoantibodies consistently stimulate [Ca²⁺]_i in Jurkat T-cells, a human T-cell clone displaying a homogeneous CD3⁺CD38⁺CD31⁺ phenotype. In the 46 human sera tested, the association between anti-CD38 reactivity and [Ca²⁺]_i-mobilizing activity, though incomplete (∼50%), was highly significant statistically. The [Ca²⁺]_i-mobilizing activity was largely lost when anti-CD38-positive sera were either depleted of their IgG or preincubated with murine fibroblasts transfected with human CD38 or when CD38 was specifically blocked by nonagonist anti–CD38 mAbs. Thus, the native anti–CD38 IgG autoantibodies not only bind to the soluble protein but also deliver an intracellular signal. To our knowledge, these results provide the first evidence that autoantibodies with agonist properties on the CD38 effector system—i.e., [Ca²⁺]_i—occur in nature. These findings also support the receptor nature of the CD38 molecule and its ability to activate outside/inside signaling (20). Furthermore, they indirectly confirm the existence of one or more ligands (14), whose interaction with functional domains of the CD38 molecule is mimicked by the natural autoantibodies. Recent work in the MC3T3.E1 cell line (21) has demonstrated the presence of CD38 on the inner nuclear membrane, where it triggers nucleoplasmic Ca²⁺ release through the intranuclear cyclization of NAD to cADPR. Taken together, these observations support the notion that CD38 modulates Ca²⁺ homeostasis in both cytosolic and nucleoplasmic cellular compartments. The biological effects of CD38 activation depend on the cell type as well as the intracellular localization of the Ca²⁺ transients evoked by cADPR binding. Thus, through nucleoplasmic Ca²⁺ waves, CD38 may transduce metabolic stress (signaled by the amount of NAD generated by metabolic activity) to highly Ca²⁺-sensitive nuclear processes, such as gene transcription and apoptosis, as seen in T-cells (15).

The relevance of CD38 agonism to the modulation of [Ca²⁺]_i in β-cells has been debated (22). For example, it has been difficult to reproduce both cADPR increases by glucose in rat pancreatic islets (23) and Ca²⁺ release by cADPR in rat islets, ob/ob mouse β-cells, and RINm5F or INS-1 cell lines (24,25,26). On the other hand, Okamoto et al. (5) have produced in vitro and experimental evidence that stimulation of CD38 activity is involved in insulin stimulus-secretion coupling. According to their results, cADPR accumulates in response to glucose stimulation in rat islets (27) and cADPR releases Ca²⁺ in rat islet microsomes (28). To address this issue directly, we measured [Ca²⁺]_i in human islet cells. In our study, half of the anti–CD38-positive sera (and none of the anti–CD38-negative sera) produced a clear-cut increment in [Ca²⁺]_i. This response, which was approximately one-third the magnitude of that elicited by high glucose, was specific, because it could be blocked by nonagonistic mAbs. Moreover, in the 14 human sera that could be tested in both cell systems, there was a very good correspondence between the ability of sera to generate a Ca²⁺ signal in Jurkat T-cells and in human islet cells. The only precedent for our finding that we are aware of is the study by Juntti-Berggren et al. (29), in which sera from four patients with type 1 diabetes (two of whom had the disease diagnosed after the age of 40 years) were found to increase the activity of l-type Ca²⁺ channels in mouse β-cells, RINm5F cells, and in a cell line (GH₃) derived from a pituitary tumor. Interestingly, the [Ca²⁺]_i-mobilizing activity of these sera was lost after depletion of immunoglobulins and could not be identified with anti-GAD autoantibodies. In that study, the insulin secretory response was not measured, but prolonged (48-h) incubation of RINm5F cells with these sera was associated with enhanced DNA fragmentation.

With regard to the physiological consequences of raising [Ca²⁺]_i by CD38 agonism in β-cells, our previous work in cultured human islets (7) demonstrated that human anti-CD38-positive sera generally stimulate insulin release. We now report that the insulin-stimulating activity of anti-CD38-positive sera is largely restricted to those displaying [Ca²⁺]_i-mobilizing activity in Jurkat T-cells (Fig. 5) or human islet cells (Fig. 6). In fact, the observed stimulation of insulin release was roughly proportional to the [Ca²⁺]_i-mobilizing activity (Fig. 7). Therefore, our findings are a strong indication that the ability of antibodies raised against CD38 to stimulate insulin release is mediated by intracellular Ca²⁺ mobilization.

The overall role that native CD38 autoimmunity may play in the β-cell dysfunction of human diabetes remains to be defined. It is possible that chronic stimulation of insulin release may lead to β-cell exhaustion, a phenomenon long suspected to underlie the progression of normal glucose tolerance to clinical hyperglycemia in obese subjects (30). Alternatively, stimulating anti–CD38 antibodies may sustain insulin release by β-cells that are failing under the pressure of genetic programming or environmental insults. Finally, the biologically relevant consequences of autoimmune CD38 activation may be the longer-term effects mediated by nucleoplasmic [Ca²⁺]_i mobilization—i.e., gene expression and apoptosis (21,29)—rather than the acute insulin release triggered by cytoplasmic [Ca²⁺]_i transients. To complicate the matter, in animal models of diabetes, CD38 expression is reduced and cADPR generation is downregulated (5). Consequently, the impact of CD38 autoimmunity may be different in islets of diabetic and nondiabetic subjects.

This work was supported through a grant from the Yamanouchi European Foundation.