Autoantibody Response to CD38 in Caucasian Patients With Type 1 and Type 2 Diabetes: Immunological and Genetic Characterization

A total of 208 diabetic patients (94 type 1 and 114 type 2) were randomly selected among the outpatients attending the Diabetology Clinic at the San Giovanni Battista Hospital in Torino, Italy. Diagnosis of either type 1 or type 2 diabetes was made according to the American Diabetes Association and World Health Organization (WHO) criteria (19,20). Healthy control subjects with no family history of diabetes, normal fasting plasma glucose levels, and similar sex, age, and BMI distribution were selected at the local blood bank. All blood samples were obtained in the morning after overnight fasting. Informed consent was obtained from each participant, and the study was approved by the San Giovanni Battista Hospital Review Board and Ethics Committee.

Recombinant soluble CD38 (rsCD38) (received from H.C. Lee, University of Minnesota, Minneapolis, MN) was used as the target antigen in the enzymatic immunoassay (EIA). This protein was obtained from Pichia pastoris yeast cells and represents the unglycosylated extracellular domain of CD38 (21). EIA plates were incubated overnight at 4°C with rsCD38 previously dialyzed against phosphate-buffered saline (PBS). After three washings in PBS, wells were saturated with 5% nonfat dry milk in PBS and washed again in PTN (0.05% Tween 20 and 0.05% sodium azide in PBS). Sera of patients diluted 1:640 in PTN were added and incubated for 2 h at 37°C. Plates were subsequently washed with PTN, and an alkaline phosphatase-conjugated anti-human IgG monoclonal antibody (mAb) (Cappel; Organon Teknika, Durham, NC) was added for 2 h at 37°C. After washing in PTN, paranitrophenylphosphate chromogen (1 mg/ml in 10% diethanolamine, 0.5% mmol/l MgCl₂, and 0.1 mmol/l sodium azide, pH 9.8) was added, and absorbance was read at 405 nm after 45 min. The IB4 murine anti-human CD38 mAb (IgG_2a) (22) was used as positive control. Each sample was assayed in triplicate in at least two different experiments; therefore, the mean of six measurements was used for all subsequent calculations. A standard curve was determined including a highly reactive serum in each experiment; the optical density (OD) obtained with this serum at a 1:640 dilution was conventionally defined as equal to 10,000 AU/ml (arbitrary units), and the reactivity of the other samples was calculated in proportion (23) with the following formula:

where the negative control was obtained by incubation of the immobilized rsCD38 with secondary mAb alone.

The upper limit of the normal range was defined as the mean + 3 SD of the values obtained from the sera of healthy control subjects. Sera with anti-CD38 response >10,000 AU/ml were diluted up to 1:2,000 to determine exact reactivity levels. For determination of intra-assay variability, a pool of sera with reactivity near the cutoff values (6,000–7,000 AU/ml) was tested 20 times in a single experiment. For interassay variability, the same pool was tested on seven different occasions.

Preadsorption experiments were performed by preincubating highly reactive sera (≥8,000 AU/ml) overnight with saturating concentrations of rsCD38 (1 in 10 μl of each serum), an equal amount of an irrelevant protein (HIV gp120; Chiron, Emeryville, CA), or PBS alone before EIA testing. For Western blot analysis, rsCD38 was run on 10% SDS-PAGE minigels, loading 2 μg of protein in each lane. Migrated proteins were then transferred to polyvinilydene difluoride (PVDF) membrane with a semi-dry transfer apparatus (Hoefer Pharmacia Biotech, San Francisco, CA) in Tris-glycine buffer containing 20% methanol and 0.035% SDS at 0.8 mA/cm². Filters were saturated with 5% nonfat dry milk and 1% bovine serum albumin (BSA) in Tris-buffered saline with Tween (TBST)-20 (10 mmol/l Tris [pH 7.4], 100 mmol/l NaCl, and 0.1% Tween-20) and incubated with sera of patients diluted 1:1,000 in 5% milk and 1% BSA TBST. After washing, a biotin-conjugated anti-human IgG mAb was added, followed by horseradish peroxidase–conjugated streptavidin (Amersham Italia, Milano, Italy). The reaction was visualized using enhanced chemiluminiscence reagents (Amersham) and exposure to X-ray film.

Analysis of Ca²⁺ mobilization mediated by anti-CD38 aAbs was performed on the Jurkat T acute lymphoblastic leukemia cell line, cloned by limiting dilution, and positively selected with anti-mouse IgG-coated magnetic beads and the MACS system (Miltenyi Biotech, Bergisch-Gladbach, Germany) to obtain a population with a homogeneous CD3⁺CD38⁺ phenotype. Cells were cultured in RPMI-1640 medium containing penicillin, streptomycin, glutamine, and 10% fetal calf serum (FCS), subsequently referred to as complete medium. Cells were loaded with Fluo 3-AM (Sigma, St. Louis, MO), a Ca²⁺-sensitive fluorescent dye (24,25). Briefly, cells were washed twice in Hanks’ balanced salt solution, pH 7.0, with 5% FCS, resuspended in the same buffer at a 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 F-127 detergent (Sigma). Cells were then washed twice, resuspended at a concentration of 10⁶ cells/ml, and used within 1 h. After calibration of the basal levels, 50 μl serum was added to a suspension of Jurkat cells (10⁷ cells/ml in 650 μl of complete medium). Effects on intracellular Ca²⁺ levels ([Ca²⁺]_i) were continuously analyzed at 37°C with a FACSort cytofluorimeter (Becton Dickinson, Milano, Italy) and Lysis II software. Cells were separated by size and side scatter to eliminate debris and dead cells. Controls included cells treated with the 4-bromo A23187 ionophore (Sigma) to ensure proper Fluo 3-AM loading and sera from healthy subjects and diabetic patients negative by anti-CD38 EIA. Modifications over time of [Ca²⁺]_i were followed by continuously recording fluorescence shifts of Fluo 3-AM during a 5-min time course and graphically represented as density plots. To explore potential antiagonistic properties of samples not featuring Ca²⁺ release (nonagonistic), competition experiments were performed by preincubating these sera with Jurkat cells, followed by the addition of suboptimal concentrations (<500 ng/ml) of the IB4 agonistic anti-CD38 mAb. CBT3G anti-CD3 mAb (IgG_2a) was added as control. In preadsorption experiments, sera were preincubated overnight with Sepharose-conjugated protein G (vol/vol) or with saturating amounts of rsCD38 (5 μg/150 μl serum) and subsequently used for determination of Ca²⁺ currents, as previously described.

Numerical presentation of data were obtained by extrapolating the Fluo 3-AM mean fluorescence intensity over the 5-min time-course from the density plots. The cutoff was arbitrarily set at the mean + 3 SD of the values obtained from the sera of diabetic patients negative by EIA.

BMI was calculated as the weight in kilograms divided by the square of the height in meters. Blood glucose control was assessed by means of fasting plasma glucose levels (measured by the glucose oxidase technique) and glycated hemoglobin (HbA_1c) (measured by high-performance liquid chromatography). Fasting plasma C-peptide and insulin levels were measured via radioimmunoassay (DPC, Los Angeles, CA, and Radim, Rome, Italy, respectively). Insulinemia was measured only in non–insulin-treated type 2 diabetic patients to avoid interference by anti-insulin aAbs (26). To obtain a parameter of the severity of the disease in type 2 diabetes, patients were assigned to only one of three therapy categories according to their current regimens (diet alone, oral hypoglycemic agents, or insulin), in a progressive manner paralleling the usual course of disease. Patients were considered as having hypertension either if they were on current therapy with antihypertensive agents or if they were presenting with blood pressure values >140/90 mmHg, according to the WHO guidelines (27).

A total of 84 type 2 diabetic subjects were screened for the Arg¹⁴⁰Trp mutation; 41 subjects (48.8%) had first-degree and five subjects (6%) had second-degree relatives affected with the disease, and 25 healthy individuals were included as control subjects. Genomic DNA was prepared from peripheral blood mononuclear cells according to the method of Laird et al. (28), and 50 ng was used per polymerase chain reaction (PCR). As the Arg¹⁴⁰Trp variant is located in CD38 exon three (14), this was amplified using primers specific for intron two (5′-GACATGCTAAATTGATCTCAG-3′) and intron three (5′-CAGCAGAAGTCACTCTGTTC-3′). Cycling parameters were as follows: initial denaturation at 95°C for 5 min, followed by 30 cycles at 94°C for 1 min, 48°C for 1 min, and 72°C for 1 min, with a final extension step at 72°C for 7 min. After ethanol precipitation, the amplified DNA was digested overnight with the restriction enzyme AciI (New England Biolabs, Beverly, MA) according to the manufacturer’s instructions. The enzyme’s 4-bp recognition sequence (G↓CGG) includes the Arg¹⁴⁰ codon (CGG) but does not cut the Trp variant (G↓TGG). Digests were separated by electrophoresis on a 2% agarose gel to distinguish the uncut 248-bp band from the 146- and 102-bp digested fragments.

Initially, 10 μg aliquots of genomic DNA, prepared as previously described, were digested overnight with 30 U of the PvuII endonuclease (Oncor Appligene, Rome, Italy). Digests were separated by electrophoresis on 0.7% agarose gels and transferred to Nylon Hybond-N membranes. The hybridization probe was a 308-bp HindIII-AccI CD38 cDNA fragment that detects a 12-kb PvuII fragment (CD38*A) or a 9-kb PvuII fragment (CD38*B), as previously reported (17).

A PCR-based screening method was subsequently devised to amplify the GC-rich genomic region containing the polymorphic PvuII site (17). The PCR (100 μl volume) consisted of 120 ng genomic DNA, 10× reaction buffer, 2.5 mmol/l each dNTP, 5 μmol/l forward primer (5′-CCGTCCTGGCGCGATGCGTCAAG-3′), 5 μmol/l reverse primer (5′-CCGTCCCTGAAGCGGTGAAGCGG-3′), and one unit Deep Vent polymerase (New England Biolabs). Cycling conditions were as follows: initial denaturation at 98°C for 8 min, followed by 33 cycles at 98°C for 1 min, 68°C for 1 min, and extension at 80°C for 1 min. A 5-min final extension at 80°C concluded the program. A 20-μl aliquot was digested overnight at 37°C with 10 U PvuII and analyzed by electrophoresis on a 2% gel.

All values are expressed as mean ± SD. Comparisons between proportions were made with the χ² test and Fisher’s exact test, when appropriate. Comparisons of the means between two groups were performed with the Student’s t test for normally distributed variables and the Mann-Whitney U test for nonnormal variables. Differences were considered statistically significant for P values <0.05.

The baseline characteristics of the 208 diabetic patients (94 type 1 and 114 type 2) selected in this study are shown in Tables 1 and 2. No exclusion criteria were used for the selection of patients in order to obtain a heterogeneous sample representative of a broad spectrum of clinical and metabolic phenotypes.

An immuno-enzymatic method using rsCD38 as the immobilized target molecule was developed to test for the presence of specific serum anti-CD38 aAbs. Figure 1 shows the OD reading obtained with two representative sera from diabetic patients, one of which was highly reactive (anti-CD38⁺) and the other displaying very low binding (anti-CD38⁻). Reactivity obtained with anti-CD38⁺ sera was evident starting from dilutions of 1:10 and persisted up to 1:5,120, showing dose-dependent binding. On the contrary, anti-CD38⁻ sera exhibited a binding to the target molecule that was weaker even at the lowest dilutions (1:10) and was almost totally lost at a dilution of 1:320. Furthermore, the difference between positive and negative samples was lost only at very high dilutions (≥1:10,240). Intra- and interassay variabilities were 4.30 and 11.63%, respectively. Similar results were obtained using plasma samples in place of sera or an anti-human polyvalent Ig in place of an anti-IgG as secondary mAb.

Preadsorption experiments were performed to confirm the specificity of the assay (Fig. 2); highly reactive sera (anti-CD38⁺) preincubated with saturating amounts of rsCD38 displayed a consistent decrease (93.3–95.5%) in EIA binding to the same rsCD38 target molecule as compared with the basal reactivity of sera preincubated with PBS alone or with the irrelevant protein HIV gp120. Moreover, when the immunoenzymatic assay was performed using wells coated with gp120, no reactivity was detected with any of the sera tested (data not shown).

Given the intrinsic lower specificity of EIA techniques as compared with the Western blot previously adopted (15,16), we also checked the anti-CD38⁺ samples against the same rsCD38 target protein in Western blot (Fig. 3) with the sera diluted 1:1,000, as previously described (16). The band highlighted by different anti-CD38⁺ samples (lanes 1–6) displayed a molecular weight of ∼30 kDa, corresponding to the rsCD38 used for EIA. On the contrary, no signal was detected by a representative anti-CD38⁻ serum sample (lane 7). Visualization of the molecular species corresponding to rsCD38 as the one recognized by the patient’s sera, together with the absence of signal obtained with the same samples on an irrelevant protein (data not shown), confirmed the specificity of the EIA procedure. Moreover, preadsorption experiments performed by preincubating the reactive sera with rsCD38 before Western blot abolished the signal (data not shown). Given the detection limit of a 1:5,120 dilution displayed by the EIA technique compared with the 1:1,000 dilution used for Western blot both in the present work and in previous ones (15,16), the EIA proved to be more sensitive in the identification of anti-CD38⁺ sera.

Between the two representative serum samples shown in Fig. 1, a whole range of intermediate reactivities was obtained with sera from other patients. Because the two curves were more widely separated using the 1:640 dilution (Fig. 1), this was chosen as the most informative for EIA testing. Moreover, to divide diabetic patients into anti-CD38⁺ and anti-CD38⁻ categories, cutoff levels were arbitrarily defined as the mean + 3 SD of the measures obtained from healthy control subjects. Thus, the values obtained were 6,015 AU/ml with the control subjects matched for type 1 and 7,021 AU/ml with the control subjects matched for type 2 diabetic subjects. Anti-CD38 aAb titers obtained for type 1 diabetic patients are shown in Fig. 4. Reactivities are expressed as arbitrary units per milliliter, using the highly positive serum shown in Fig. 1 as the reference standard. The distribution shows a continuous pattern spanning through all values; the majority of individuals (both diabetic and healthy) showed reactivities <1,250 AU/ml. Among the 94 type 1 diabetic subjects studied, 18 (19.15%) were above the cutoff value selected. On the contrary, only 2 of the 96 control subjects (2.08%) were characterized as anti-CD38⁺ (P = 0.0001); moreover, the reactivity ranked in the bottom half of the positive range. A similar distribution was observed for type 2 diabetic patients (Fig. 5); although the cutoff value obtained with the matched control subjects was higher, a similar number of positive subjects was selected (19 of 114 subjects, 16.67%; P = 0.64 for the comparison with anti-CD38⁺ type 1 diabetic subjects). Only 1 of the 112 control subjects included (0.89%) was positive (P = 0.00003) at low titers.

The anti-CD38 reactivity of some patients was also tested using serum samples obtained on two separate occasions, highlighting the persistence of similar titers, at least for the time intervals of ≤12 months analyzed (data not shown).

Next, the functional activity of anti-CD38 aAbs was evaluated. Anti-CD38 mAbs are generally identified as either agonistic or nonagonistic, depending on their binding to different portions of the CD38 protein (29) and the triggering of different signals. Therefore, the effect of anti-CD38 aAbs on Ca²⁺ release was analyzed in the human Jurkat T-cell line to characterize these serum immunoglobulins in a similar way. Different patterns of Ca²⁺ triggering by representative sera from patients are presented in Fig. 6. Fig. 6A shows an anti-CD38⁺ serum displaying agonistic activity, documented by a consistent shift in the [Ca²⁺]_i. Other anti-CD38⁺ sera did not have any effect on Ca²⁺ signaling (Fig. 6B) and thus were identified as nonagonistic. The Ca²⁺ mobilization pattern highlighted by nonagonistic anti-CD38⁺ sera (Fig. 6B) was similar to that evidenced by anti-CD38⁻ sera (Fig. 6C). Competition experiments between sera, such as the one presented in Fig. 6B, and the agonistic anti-CD38 mAb IB4 were performed to exclude the possibility that the lack of signaling was because of misidentification of the sera as anti-CD38⁺ by EIA. Whereas binding of the IB4 mAb alone was followed by massive Ca²⁺ mobilization (Fig. 6D), preincubation of target cells with the latter sera determined a significant decrease in the stimulatory efficiency of IB4 (Fig. 6E). In contrast, control anti-CD38⁻ sera did not interfere with the IB4-triggering effect (Fig. 6F). These results indicate that nonagonistic anti-CD38⁺ sera are not able to induce Ca²⁺ release as expected, but can inhibit the same effect mediated by the agonistic mAb IB4 through epitopic competition, thus indirectly confirming the specificity of such sera as that identified by EIA.

Further experiments were performed to confirm the specificity of the Ca²⁺-mobilizing effect as due to anti-CD38 aAbs. Removal of the Ig fraction from anti-CD38⁺ sera by means of sepharose–protein G pretreatment lead to a consistent decrease in Ca²⁺ mobilization (Fig. 6G), thus demonstrating the involvement of antibodies in the observed effect. Furthermore, these antibodies are specific for CD38, because anti-CD38⁺ sera preincubated with saturating amounts of rsCD38 triggered minimal Ca²⁺ currents (Fig. 6H). On the contrary, preincubation of anti-CD38⁺ sera with either PBS or gp120 did not have any influence on the agonistic activity (data not shown).

The relative prevalence of agonistic versus nonagonistic anti-CD38 aAbs was next evaluated by extrapolating the Fluo 3-AM mean fluorescence intensity from kinetics profiles, such as the ones depicted in Fig. 6. The arbitrary cutoff, defined as the mean + 3 SD of the Fluo 3 mean fluorescence intensity values obtained from anti-CD38⁻ diabetic sera, was 11.87. As shown in Fig. 7, 16 of the 28 sera considered (57.14%) displayed agonistic properties, whereas only 1 of the 21 anti-CD38⁻ sera (4.76%) was capable of triggering modest Ca²⁺ mobilization (P = 0.0001).

Anti-CD38⁺ and anti-CD38⁻ diabetic subjects were compared for relevant clinical and metabolic characteristics. Among type 1 diabetic subjects, no significant association of anti-CD38 aAbs was found with any of the parameters considered (Table 3). On the contrary, anti-CD38⁺ type 2 diabetic subjects displayed significantly higher fasting plasma C-peptide and insulin levels (Table 4), in line with the agonistic properties reported here and with the previously described insulin-secreting capacity of anti-CD38 aAbs (16). Higher BMI values and a lower frequency of insulin treatment was also noted among anti-CD38⁺ type 2 diabetic subjects, although statistical significance was not reached.

As CD38 displays genetic heterogeneity in terms of both pathological mutations and normal allelic variants, the potential influence of this genetic milieu in the Caucasian diabetic population was explored. The presence of the Arg¹⁴⁰Trp mutation described in Japanese type 2 diabetic individuals (14) was first evaluated using a PCR assay. Among the 84 Caucasian type 2 diabetic patients screened (46 of whom had first- or second-degree relatives affected), no individual carrying such a mutation was found. The only subject positive for the mutation was found among the 25 healthy control subjects; this subject repeatedly had normal fasting plasma glucose levels and no family history of diabetes. Presence of the Arg¹⁴⁰Trp mutation in a heterozygous state was also confirmed by genomic DNA sequence analysis (data not shown).

Because a CD38 genetic polymorphism was recently described in the healthy Caucasian population (17), we set out to determine whether any particular CD38 allele was preferentially associated with the diabetic condition. The sample of type 2 diabetic patients analyzed was larger than that of type 1 diabetic subjects, given the heavier influence of genetic factors on the former group (2). Results are summarized in Table 5 and 6; neither the distribution of the less frequent CD38*A allele nor that of the CD38*A/*A genotype differed significantly among type 1 diabetic patients, type 2 diabetic patients, and healthy control subjects. No correlation was observed between any CD38 allele and aAb status (data not shown). Furthermore, CD38*A homo- or heterozygotes did not display a stronger family history of diabetes.

Interactions between genes and environment are critical in the susceptibility to several human diseases; among these, diabetes is certainly one in which the “genetic versus acquired” debate proves to be extremely complex. The CD38 molecule was first suspected to be a mediator of insulin secretion in 1993 (11), in virtue of its multiple enzymatic activities, which are unique in leading to synthesis and degradation of cADPR (9). Since then, attention has shifted from physiology to pathology, as selected populations of diabetic patients (mainly type 2 and, to a lesser extent, type 1) have proved to carry a defective CD38 gene on the one side (14) and autoreactive antibodies targeting the molecule on the other (15,16).

The aim of this study was to elucidate immunological and genetic issues concerning human CD38 in diabetes among Caucasian individuals, who until now have been evaluated only in terms of anti-CD38 autoreactivity (16). A second related aim was the development of a more convenient screening test for anti-CD38 aAbs. Using an EIA technique, 19.15% of type 1 and 16.67% of type 2 diabetic patients displayed anti-CD38 autoreactive antibodies in their sera, as opposed to a negligible proportion (0.89–2.08%) of healthy control subjects. Prevalences obtained with the EIA technique are very similar to those reported by Ikehata et al. (15) in Japanese type 2 diabetic subjects (13.8%) and by Pupilli et al. (16) in Caucasian type 2 (9.7%) and type 1 (13.1%) diabetic patients, although the higher sensitivity intrinsic to the EIA technique as compared with the Western blot gave a slightly larger proportion of anti-CD38⁺ samples in our study. The assay also proved reliable in terms of reproducibility and specificity, as assessed by preadsorption experiments and by the conventional Western blot approach.

The EIA method presents some advantages over the Western blot assay—the latter exploits immobilized rsCD38 as the target molecule, but the protein is denatured by the electrophoresis in SDS. Consequently, only aAbs recognizing linear epitopes can bind, whereas this is not the case for aAbs against conformational epitopes. Apart from the higher sensitivity of EIA, the Western blot used in the two previous studies could therefore have led to an underestimation of the prevalence of anti-CD38 aAbs. A potential for further underestimation, unavoidable in both approaches, derives from the rsCD38 target molecules used in both the EIA (rsCD38 deprived of the putative sites of glycosylation) and the Western blot (CD38-maltose binding protein fusion protein) techniques (15,16). The glycosylation is important for mAb binding; indeed, glycosylation of both proteins is different from that of native human CD38, in which oligosaccharide residues account for ∼25% of the molecular weight (30). Moreover, the different molecular weight of CD38 expressed in human islet cells (48 vs. the conventional 45 kDa), possibly resulting from an alternative pattern of glycosylation, may lead to a different antigenic recognition pattern of the protein expressed in this tissue (R.M. and M. Volante, unpublished observations). Another pitfall shared by both approaches could derive from in vivo binding of aAbs to circulating soluble forms of CD38 (31,32); these molecular species could subsequently compete with the target proteins used in the assays. The EIA technique developed for this study might constitute a valuable routine test for the screening of anti-CD38 aAbs.

Although matching control subjects with type 2 diabetic patients was not optimal for age, this is unlikely to have determined major bias. Prevalence of various autoreactive antibodies increases up to fourfold in healthy nonagenarians as compared with middle-aged control subjects (33). Even if such an increase held true for anti-CD38 aAbs, the difference noted between diabetic patients and control subjects would have remained significant. Moreover, preliminary results obtained in our laboratory indicate that the prevalence of anti-CD38 aAbs in nondiabetic nonagenarians is not significantly higher than that in younger individuals.

The finding of such a high proportion of anti-CD38⁺ patients not only among type 1 but also among type 2 diabetic subjects is not surprising. These patients may belong to a distinct subgroup similar to the latent autoimmune diabetes in the adult (LADA) subgroup (34–36) for the autoimmune phenotype, but not for the clinical and metabolic characteristics. On the contrary, the association of anti-CD38 aAbs with higher fasting plasma C-peptide and insulin levels and higher BMI would suggest a phenotype opposite to that of LADA. The analysis of other conventional islet aAbs among anti-CD38⁺ and anti-CD38⁻ patients will further clarify this issue. Knowledge of whether these findings have a heuristic merit or a prognostic significance awaits results from ongoing evaluations of long-term complications of diabetes; further study may provide evidence in support of the use of anti-CD38 aAbs not only as a simple diagnostic marker of diabetes with autoimmune features, but also as a prognostic parameter for a particular course of the disease. Taken together, these results also suggest that the scenario of autoimmunity in type 2 diabetes might be more heterogeneous than originally presumed.

Genetic analysis did not highlight any significant association between the CD38*A allele recently identified in healthy Caucasian individuals (17) and subjects with diabetes or anti-CD38 aAbs. Also, the Arg¹⁴⁰Trp mutation described in Japanese type 2 diabetic patients (14) was not detected among our screened patients, but it was identified in one healthy control subject. This discrepancy can possibly be ascribed to the heterogeneity of the genetic background of diabetes in distant ethnic groups (37). The Arg¹⁴⁰Trp mutation is rare and probably not as critical among Caucasians. The potential interplay between genetic and autoimmune factors could have been because of a defective antigenic recognition of alternative forms of CD38. Alternatively, genetic variants could have exerted pathogenetic effects independently by altering CD38 enzymatic activities, as has been demonstrated for the Arg¹⁴⁰Trp mutation (14). This would have configured a pathological paradigm similar to that recently described for thrombotic thrombocytopenic purpura, in which the deficiency of the von Willebrand factor–cleaving protease responsible for the disease can be due to either genetic mutations or aAbs against the protease (38). However, according to the present data, the influence of the CD38 genetic background can be excluded bona fide, at least in the Caucasian population.

A further step of this work was the study of the functional properties of anti-CD38 aAbs in terms of agonistic or nonagonistic activity. To this end, analysis of the effects triggered on Ca²⁺ release was chosen, as it is critical for insulin secretion and many other cellular functions (39). Indeed, the majority (57.14%) of anti-CD38 aAbs were characterized as agonistic. This definition is relative, because it is the proportion of agonistic versus nonagonistic immunoglobulins of the polyclonal response toward the autoantigen that leads to the prevalence of stimulatory or blocking effects, as evidenced in the paradigmatic example of Graves’ disease and Hashimoto’s thyroiditis (40,41). Nonetheless, this point is relevant because the agonistic properties of anti-CD38 aAbs could determine an increase in insulin secretion, as directly demonstrated on purified human islet cells (16). In line with this hypothesis, higher plasma C-peptide and insulin levels were observed in anti-CD38⁺ type 2 diabetic patients; simultaneously, a trend toward a higher BMI and a lower frequency of insulin therapy was also noted. No significant differences in C-peptide levels were found between anti-CD38⁺ and anti-CD38⁻ type 1 diabetic patients, which is likely to be because of complete β-cell exhaustion. Therefore, the in vivo effect of anti-CD38 aAbs could be an enhanced insulin secretion from β-cells, possibly leading, in the long term, to insulin secretory failure from excess stimulation. Moreover, CD38 is also expressed in a number of extrapancreatic tissues at high epitopic densities. Thus, it is possible that the in vivo effects of circulating anti-CD38 aAbs may not be limited to β-cells. Among the CD38⁺ tissues, peripheral blood mononuclear cells (PBMCs) deserve special consideration with regard to the prevalent agonistic properties of anti-CD38 aAbs. In PBMCs, binding of agonistic anti-CD38 mAbs induces release of pro-inflammatory cytokines, i.e., interleukin (IL)-1, IL-6, and tumor necrosis factor-α (42), and similar effects could be triggered by anti-CD38 aAbs. Cytokine release could be an important contribution to the insulitis responsible for early presentation of type 1 diabetes and would be in line with the recently proposed models of type 2 diabetes as a disease of the innate immune system (43,44). Apart from these speculations, anti-CD38 aAbs constitute a new marker of β-cell autoimmunity in diabetes that, unlike islet cell antibodies, anti-GAD, and anti-IA-2 aAbs, is endowed with peculiar functional properties.

This work was supported in part by grants from Telethon, Rome, Italy; Special Project “Biotechnology” CNR/MURST, Rome, Italy; AIRC, Milan, Italy; and Special Project “AIDS,” ISS, Rome, Italy. Compagnia di Sanpaolo and Cariverona Foundations provided valuable financial contributions.

We thank Enza Ferrero, MD, for a critical reading of the manuscript and for constructive discussion. We also thank Dr. Gabriella Gruden and Dr. Alda Olivero for helpful suggestions. R.M., S.G., and E.L. are students at the Postgraduate School of Internal Medicine, University of Torino, Torino, Italy.

This work is dedicated to the memory of Teresa Sacco (1922–1994).