Phenotype and Functional Characteristics of Islet-Infiltrating B-Cells Suggest the Existence of Immune Regulatory Mechanisms in Islet Milieu

Breeding pairs of NOD, NOR (13), and BALB/c mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and maintained by brother-sister mating under specific pathogen-free conditions, following the European Laws and Directives concerning the protection of animals for experimentation. In our facility, 85% female and 37% male NOD mice developed spontaneous autoimmune diabetes at 32 weeks of age. (NOD×NOR)F1 mice were generated by crossing NOD with NOR mice. Most (NOD×NOR)F1 mice developed a nondestructive but severe form of insulitis, although only ∼19% of (NOD× NOR)F1 female mice developed overt diabetes. We used the group of (NOD× NOR)F1 mice as a resistance-to-disease-progression model to study the characteristics of islet-infiltrating B-cells in nondestructive insulitis.

Pancreatic cryosections from NOD female mice were air-dried for 30 min and incubated at 4°C with anti-CD45R/B220 (biotin RA3-6B2; BD Biosciences) and then with streptavidin–fluorescein isothiocyanate (FITC) (BD Biosciences) at 4°C for 45 min. After washing, the pancreata sections were stained for 2 h at 4°C with a rabbit anti-Ki67 antibody (NCL-ki67p; Novocastra, Newcastle upon Tyne, U.K.), washed again, and then incubated with the secondary antibody (goat anti-rabbit TRITC; Southern Biotechnology Associates, Birmingham, AL) at 4°C for 45 min.

For immunofluorescence staining, islet-infiltrating mononuclear cells were obtained as previously described elsewhere (14). Briefly, after their isolation by enzymatic digestion of the pancreas (Collagenase CLS-4; Worthington Biochemical, Lakewood, NJ), pancreatic islets were handpicked and mechanically disrupted to obtain islet single-cell suspension. Spleens were disrupted into single-cell suspensions and erythrocytes lysed in 0.87% ammonium chloride. B-cell phenotypic analysis was performed by staining the previous samples at 4°C for 20 min with a combination of the following monoclonal antibodies (mAbs): anti-CD45R/B220 (FITC RA3-6B2), anti-CD19 (PE 1D3), anti-CD21 (FITC 7G6), anti-CD5 (FITC 53-7.3), anti-IgA (biotin C10-1), anti-IgD (FITC 11-26c.2a), anti-IgG1 (biotin A85-1), anti-IgG2a2b (biotin R2-40, which also recognizes NOD IgG2c), IgG2b (biotin R12-3), anti-IgG3 (biotin R40-82), anti-IgM (biotin R6-60.2), anti-CD24 (FITC M1/69), anti-CD86 (biotin GL1), anti-CD80 (biotin 16-10A1), anti-CD40 (biotin 3/23), anti-CD44 (biotin IM7), anti-CD69 (FITC H1.2F3), anti–I-Ak (biotin 10-3.6), and anti–H-2kd (biotin SF1-1.1), all of which were purchased from BD Biosciences, and anti-CD11b (biotin M1/70), purchased from Southern Biotechnology Associates. Biotin-conjugated mAbs were subsequently stained with streptavidin-PerCP (BD Biosciences) and analyzed by FACScan Cell Analyzer (BD Biosciences) using CellQuest software. According to their forward and side characteristics, live cells (previously determined by propidium iodide) were gated and analyzed for the expression of the surface molecules of interest. When the fluorescence histogram or any label clearly showed a single peak in the B-cell population, we measured the median fluorescence intensity because this parameter is more constant when analyzing small samples. When the histogram showed a continuous shoulder or a bimodal distribution, we measured the percentage of positive cells. To discard any possible effect of collagenase on membrane cell markers, samples of splenic cells were overdigested with this enzyme at 37°C for 45 min (for islet isolation, we usually perform pancreatic digestion with collagenase at 37°C for 20 min). Samples were then stained and analyzed by flow cytometry, and the results were compared with those obtained from undigested samples.

After enzymatic isolation, pancreatic islets from 12-week female NOD or (NOD×NOR)F1 mice were cultured in 96-well tissue culture plates (20 islets/well) at 37°C in 5% CO₂ in complete culture medium (RPMI-1640 media containing 10% heat-inactivated fetal bovine serum [Life Technologies, Grand Island, NY], 100 units/ml penicillin, 100 μg/ml streptomycin, 2 mmol/l glutamine, 1 mmol/l sodium pyruvate, and 25 μmol/l β-mercaptoethanol). Likewise, 2.5 × 10⁵/well splenocytes were cultured under the same conditions. Activating stimuli consisted of 10 μg/ml lipopolysaccharide (LPS) (Sigma-Aldrich, St. Louis, MO) or 10 μg/ml anti-CD40 (clone 3/23; BD Biosciences) plus 10 units/ml rIL-4 (R&D Systems, Minneapolis, MN). Then cultured mononuclear cells, from either spleen or islet infiltrate, were harvested at 0, 24, and 48 h and analyzed by three-color flow cytometry, as previously described.

Splenic and islet-infiltrating mononuclear cells from 12-week NOD and (NOD×NOR)F1 female mice (right after islet isolation) were cultured at 37°C in 5% CO₂ for 5 h in complete culture medium in 5 ng/ml phorbol myristic acid and 500 ng/ml Ionomycin (both from Sigma-Aldrich) and 10 μmol/l monensin (eBiosience, San Diego, CA). Then cells were stained with anti-CD19 mAb (Phycoethrin 1D3; BD Biosciences), fixed in 4% formaldehyde, permeabilized in 0.1% saponin, and stained in permeabilization buffer with the following FITC-conjugated mAbs: anti–γ-interferon (IFN-γ) (XMG1.2; eBioscience), anti–interleukin (IL)-4 (BVD6-24G2; eBioscience), and anti–IL-10 (JES5-16E3; BD Pharmingen). Stained cells were then analyzed by flow cytometry.

Splenic and islet-infiltrating mononuclear cell suspensions were stained with anti-CD3 (FITC 17A2; BD Biosciences) and anti-CD19 (PE 1D3; BD Biosciences) and then sorted using the FACS Vantage cytometer (BD Biosciences). In all the assays, purity of T-cell–and B-cell–sorted populations was always >98%. Purified T-cells (5 × 10⁶/ml) were labeled with an equal volume of 1.25 μmol/l carboxy-fluorescein diacetate succinimidyl ester (CFSE) (5-[and 6]-CFSE; Molecular Probes, Eugene, OR) for 10 min at room temperature and then quenched with the same volume with heat-inactivated fetal bovine serum for 10 min. Then T-cells were washed twice with PBS and resuspended in culture medium. T- and B-cells were cocultured (200 μl supplemented culture medium) in plate-bound 10 μg/ml anti-CD3 mAb (145-2C11; BD Pharmingen) or soluble 5 μg/ml anti-CD3 alone or supplemented with islet extracts obtained by freeze-thaw cycles (10 islets/well) and/or 10 μg/ml anti-CD40 mAb (3/23; BD Pharmingen). For a positive control of T-cell proliferation, purified T-cells were cultured in a plate-bound anti-CD3 mAb or cocultured with unfractionated islet-infiltrating APCs plus islet extract. Unfractionated islet-infiltrating APCs were obtained by selecting negative anti-CD3 mononuclear cell populations by cell sorting. After 3 days of culture, cells were harvested and analyzed by flow cytometry.

All data are shown as mean ± SE unless otherwise indicated. Statistical analyses were performed using the Mann-Whitney U test (SPSS, Chicago, IL, and GraphPad Prism, Graphpad Software, San Diego, CA).

The main objective of the present study was to characterize the islet-infiltrating B-cell population in NOD and (NOD×NOR)F1 mice. Having used peripheral B-cells as reference cells, we previously characterized this population resident in the spleen. Splenic B-cells from 12-week female NOD, NOR, and BALB/c mice (used as controls) were stained with a combination of surface markers indicative of activation, maturation, and B-cell lineage and were subsequently analyzed by flow cytometry.

A significant increase in the IgM^highIgD^low/neg CD21⁺ B-cell population was detected in the spleen of NOD mice compared with Balb/c and NOR mice (Fig. 1A), thus confirming previous studies indicating that the marginal zone B-cell population is increased in the spleens of NOD mice (15,16).

No significant differences were found in the expression of CD21, CD40, CD45R, or CD86 molecules in splenic B-cells of the three mouse strains. However, flow cytometric analysis showed that splenic B-cells from NOD and NOR mice expressed higher levels of CD24, CD44, and major histocompatibility complex (MHC) class I molecules (H-2k^d) than those from Balb/c mice (Fig. 1B and C). CD24 and CD44 levels were significantly higher in NOD than in NOR mice, suggesting a higher B-cell activation in NOD mice. Interestingly, the expression levels of the B-cell coreceptor molecule CD19 were 50% higher in NOD and NOR splenic B-cells than in those of Balb/c mice. Furthermore, the percentage of splenic B-cells CD5⁺ (CD11b⁻) in NOD and NOR mice was higher than in Balb/c mice (24–30 vs. 7%). As these data suggest that splenic B-cells are activated similarly in NOD and NOR mice, this must reflect a consequence of their background and not a function of diabetes progression.

Next we studied the fate of B-cells during type 1 diabetes development. We assessed the percentage of B-cells in the spleens of NOR and NOD mice at different stages of the disease (at weeks 5, 7, 9, and 12), in adult resistant mice (at 32 weeks), and in recently diabetic mice. The islet infiltration analysis was pursued in NOD and (NOD×NOR)F1 mice, since NOR mice only develop mild peri-insulitis.

The percentage of splenic B-cells remained quite constant during the progression of the disease in NOD mice, with values between 50 and 60% in most cases (Fig. 2A). Similar results were found in NOR mice. However, the percentage of islet-infiltrating B-cells tended to increase during this period (Fig. 2B), especially in NOD and (NOD×NOR)F1 female mice, and the highest values (almost 50%) were detected at 12 weeks. This may be crucial in the disease development, since it occurs just before the onset of diabetes. After that age, the percentage of islet-infiltrating B-cells decreases with age. To elucidate whether B-cells accumulate into the pancreatic islets as a consequence of their recruitment or their proliferation in situ, we double-stained B-cells to analyze the expression of the cell proliferation marker ki67. However, we did not find B-cell proliferation in the pancreatic islets (Fig. 3). Therefore, these data indicate that B-cell recruitment into pancreatic islets is strongly influenced by gender traits and suggest that this may be essential to determine the progression of insulitis but not disease development.

To further characterize B-cell infiltrate in islets, we carried out new surface marker analysis on the cell population of 12-week NOD and (NOD×NOR)F1 female mice. To analyze B-cells in their native state, we performed all the stainings as quickly as possible after islet isolation. We considered the possible effects of collagenase digestion used for islet isolation protocol on the surface antigens, which is why we overdigested splenic B-cells with collagenase and then analyzed each surface marker to evaluate their integrity. The results indicated that only CD21, CD44, and CD45R/B220 molecules were affected by collagenase digestion; we therefore did not use them in the phenotype study of islet-infiltrating B-cells (data not shown).

The immunoglobulin isotype analysis of islet-infiltrating B-cells indicated that ∼90% of these cells were IgM IgD double-positive B-cells (Fig. 4), whereas ∼3% were positive for IgG1. Only a few B-cells expressed other isotypes (e.g., IgG2, IgG3, and IgA), and their percentage could not to be accurately determined (≤1%) (data not shown).

Cytometric analysis of islet-infiltrating B-cells also showed a slight but significant increase in the percentage of CD69⁺ (11 vs. 5% in splenic B-cells) and CD86⁺ cells (7 vs. 2.5% in the spleen) (Fig. 5A), but no differences were found in the expression of CD80 molecules (both islet-infiltrating and splenic B-cells were CD80-negative [data not shown]). Moreover, the percentage of CD5⁺ B-cells was similar to that found in the spleen. Interestingly, we found that CD19 expression decreased by >60% in the whole infiltrating B-cell population compared with that from the spleen (Fig. 5B and C), showing a possible hyporesponsive status of these cells, given the relevance of CD19 in regulating B-cell activation. CD19 downmodulation was more pronounced during the progression of the disease (Fig. 6). The decrease in CD19 expression coexisted with a significant increase in MHC class I and II molecule expression (Fig. 5D). The population of islet-infiltrating B-cells also showed a decrease in CD24 surface molecules (Fig. 5D) and mIgD expression but not in mIgM expression (Fig. 4). No significant differences were found in the expression levels of CD40 between splenic and islet-infiltrating B-cells. Similar results were observed in the phenotype analysis of (NOD×NOR)F1 mice. In general, these results suggest that islet-infiltrating B-cells are not naive cells, as they had undergone upregulation of the antigen presentation machinery, possibly as a result of the encounter with their specific antigen. In that case, islet-infiltrating B-cells may present the captured autoantigen in the absence of costimulatory molecules, and therefore they may not be able to provide adequate stimulatory signals to activate T-cells. On the contrary, it is also possible that this B-cell phenotype may be secondary to the encounter with activated T-cells, i.e., by bystander help, or to the proinflammatory status present at the islet milieu. As a result, despite their antigenic specificity, the behavior of islet-infiltrating B-cells may be impaired, and thus these cells are unable to respond normally to antigen-specific stimulation.

As the secretion of certain cytokines by islet-infiltrating B-cells may act in situ on other mononuclear cells, we analyzed the possible differential expression of IFN-γ, IL-10, and IL-4 cytokines. Cytokine staining showed that the percentage of B-cells producing each type of cytokine was similar in both the islet infiltrate and spleen of NOD female mice (3% IFN-γ⁺, 1% IL-10⁺, and almost undetectable for IL-4⁺) (Fig. 7). Similar results were found in 12-week female (NOD×NOR)F1 mice (data not shown). Thus, no differences were found in the cytokine profile of islet-infiltrating B-cells in either NOD or (NOD×NOR)F1 mice.

Although the phenotype expressed by islet-infiltrating B-cells does not correspond to that classically defined for anergic B-cells (Fig. 5), in view of CD19 downmodulation we considered it important to explore this hypothesis. Therefore, islet-infiltrating B-cells from 12-week NOD female mice were cultured with different stimulating agents. LPS was used to study the response to T-cell–independent polyclonal stimulation, whereas anti-CD40 plus IL-4 were used to simulate a T-cell–dependent response. Islet-infiltrating B-cells did not show proliferation arrest, which is typical of anergic cells, but reached higher levels of antigen-presenting MHC class I and II molecules, CD86 (but not CD80 [data not shown]), and activation markers (e.g., CD69 and CD44) than splenic B-cells (Fig. 8). This difference was more obvious when cultured in anti-CD40 plus IL-4 (or anti-CD40 alone [data not shown]). Interestingly, the expression of CD19 in islet-infiltrating B-cells cultured in LPS or anti-CD40 plus IL-4 was always lower than in splenic B-cells. This different reaction of islet-infiltrating B-cells to in vitro stimulating agents clearly indicates that they are antigen experienced. Note that changes in the expression levels of some surface molecules (i.e., CD19, H-2Kd, CD86) were detected even when islet-infiltrating B-cells were cultured in complete culture medium alone out of the islet milieu, thus supporting the idea that they are somehow silenced in the infiltrate environment. However, this silenced status is easily reversible, becoming potentially antigen-presenting cells, particularly under T-cell–dependent stimulus.

To establish whether islet-infiltrating B-cells could act as APCs for islet-infiltrating T-cells, purified T- and B-cells (>98% purity) from spleen or islet infiltration were cocultured under different stimulating conditions, after which they were harvested and analyzed by cytometry. T-cell proliferation was analyzed by CFSE cytometric assay because it allowed us to study the proliferation of small cell populations. The results showed that islet-infiltrating B-cells cannot induce polyclonal proliferation of T-cells by themselves (Fig. 9), even when anti-CD40 mAb or islet extract is added to the culture. Hence, B-cells cannot present the specific antigen or provide a correct costimulatory signal to T-cells. We conducted further experiments culturing T- and B-cells in plate-bound anti-CD3 mAb, soluble anti-CD3, or soluble anti-CD3 plus anti-CD40 or islet extracts. The results indicated that neither splenic nor islet-infiltrating B-cells inhibit plate-bound anti-CD3–induced T-cell proliferation; however, both B-cell subsets act as accessory cells to promote similar splenic and islet-infiltrating T-cell proliferation in the presence of soluble anti-CD3. No significant differences were found when we conducted these same proliferative studies adding anti-CD40 mAb or islet extract to the culture. Similar results were obtained when we analyzed (NOD×NOR)F1 female mice (data not shown). Therefore, under in vitro conditions, islet-infiltrating B-cells from both NOD and (NOD×NOR)F1 mice cannot restrain T-cell proliferation. On the contrary, they can only provide adequate signals to induce polyclonal T-cell proliferation in the presence of the mitogenic soluble anti-CD3 mAb.

During the last decade, many studies in NOD mice demonstrated that B-cells are essential in type 1 diabetes progression (3). How, when, and where they provide this diabetogenic activity remains unclear. B-cells may induce diabetogenesis by presenting autoantigens to β-cell CTLs and/or by producing pathogenic autoantibodies in the spleen, regional lymph nodes, and/or inside pancreatic islets at any stage of the disease. The aim of this study was to phenotypically and functionally characterize the B-cell population in spleen and islet infiltrate to determine its possible role in the disease.

Our studies revealed that the splenic marginal-zone B-cell population is increased in the spleens of NOD but not NOR mice. These results support recent studies by Rolf et al. (16) indicating that this trait maps to diabetes susceptibility Idd9 and Idd11 loci in NOD mice (which are not present in NOR mice) and that the increase in the marginal-zone B-cell population is related to type 1 diabetes development.

Signs of B-cell activation have been observed in the spleen of 1-month-old NOD mice (17). Some authors have described that splenic B-cells highly express CD19, CD21, CD40, and MCH class II molecules, which may suggest a role in the activation of autoreactive T-cells (18). In our studies, we confirmed splenic B-cell activation by an increase in cell surface expression of CD19, CD44, CD24, and H-2Kd molecules in this cell population in both NOD and NOR compared with Balb/c mice. Moreover, the percentage of splenic CD5⁺ B-cells in NOD and NOR mice was clearly higher than in Balb/c mice. Since they are Mac-1 negative, this CD5⁺ B-cell population corresponds to B-2 cells, in which, most likely, CD5 has been selectively induced to provide inhibitory signals after recognition of autoantigen (19,20). These so-called “induced CD5⁺ B-cells” are mainly found in the mantle zone of germinal centers (21). Although CD5⁺ B-cells have been related to autoimmunity in humans (22) and in mice (23), their role in autoimmune diseases has not been demonstrated. Interestingly, these CD5⁺ B-cells are found both in NOD and NOR mice, thus indicating that their presence is a consequence of the autoreactive background of both mouse strains and not of diabetes progression.

It has been shown that systemic (24) or peritoneal-selective (25) depletion of the B-cell population in wild-type NOD mice, or their absence in NOD.Igμ^null mice (26–28), results in type 1 diabetes and/or insulitis resistance. Hence, one of the most interesting findings of our study shows that during the pre-diabetic period, B-cells accumulate into the islets influenced by sex traits in both NOD and (NOD×NOR)F1 mice. The fact that the incidence of diabetes is higher in NOD and (NOD×NOR)F1 female mice than male strongly suggests that B-cell recruitment to the islets is one of the key features, though not critical, in the progression of the disease to overt diabetes.

Compared with splenic, the phenotype of most islet-infiltrating B-cells is characterized by downmodulation of CD19, a low expression of IgD (but not IgM), low levels of CD24, an upregulation of MHC class I and II molecules, and similar levels of CD40. CD19 is a component of the stimulatory coreceptor of B-cells that enhances B-cell receptor signal and is essential for the response of mature B-cells to antigenic challenge (particularly to thymus-dependent antigens). Therefore, the formation of germinal centers, the production of high-affinity antibodies, and the development of memory B-cells are impaired in CD19-deficient mice (29). On the contrary, CD19 overexpression induces autoantibody production in systemic sclerosis patients (30) and in CD19 transgenic mice (31) not genetically predisposed to autoimmunity, demonstrating the role of this molecule in the mechanisms of tolerance. Moreover, signaling pathways via CD19 may affect antigen processing and may enhance or inhibit the endocytic pathways, depending on the signal intensity (32). Apparently, our results suggest that most islet-infiltrating B-cells are downregulated and cannot act as regular APCs and activate islet-infiltrating T-cells.

B-cells that recognize autoantigens are anergyzed as a mechanism of immune tolerance. The anergic B-cell phenotype downmodulates IgM, CD21, CD20, CD22, CD23, CD40, and l-selectin molecules and upregulates CD24, CD44, and MHC class II molecules (33). Though we did not analyze all of these markers (some of which were affected by collagenase digestion), this does not seem to be the case of islet-infiltrating B-cells, which showed lower CD24 levels and similar IgM and CD40 levels compared with splenic B-cells. Moreover, they developed a strong APC phenotype after being cultured in LPS or anti-CD40. Note that, following the in vitro stimulation, the upregulation of MHC, CD44, CD69, and CD86 molecules in islet-infiltrating B-cells was even higher than in splenic B-cells, strongly suggesting that they are antigen experienced. In fact, several authors found that islet-infiltrating B-cells were significantly more responsive to anti-IgMF(ab′)₂ or IL-4 stimulations than splenic B-cells in NOD mice (34). Why, then, do islet-infiltrating B-cells downmodulate CD19 molecules? The possible effect of collagenase on CD19 was ruled out by digesting splenic B-cells with this enzyme for twice the period of time than is usual in islet isolation. Interestingly, after islet isolation, islet-infiltrating B-cells that leave the islets increase CD19 levels even when there is no stimulation, thus strongly suggesting that soluble factors, such as cytokines, released in the islet environment are responsible for the islet-infiltrating B-cell phenotype. However, after being challenged with LPS or anti-CD40, islet-infiltrating B-cells still show lower levels of CD19 than splenic B-cells. Physiological downmodulation of CD19 has been described during plasmatic cell development (35), but apparently this does not occur in islet-infiltrating B-cells. It has been described that the CD19 molecule is downmodulated by low-affinity FcγRII (CD32) stimulation (36). In addition, FcγRII acts as an inhibitory receptor in B-cells, thus preventing their activation. Therefore, this mechanism may also be involved in the downmodulation of CD19 in islet-infiltrating B-cells.

It has been shown that a significant number of B-1 cells accumulate in the islet infiltrate and play a critical role in the development of insulitis and diabetes (25). However, we found that most islet-infiltrating CD5⁺ B-cells were CD11b negative, thus indicating that they were not B-1 cells. Moreover, there were no differences in the percentage of CD5⁺ B-cells between the islet infiltrate and the spleen. Our results are consistent with a recently published report (38) suggesting that the B-1 cell population is not present in the islet infiltrate.

Only a few islet-infiltrating B-cells showed signs of activation (11% were CD69⁺ and 7% CD86⁺). These results do not agree with those of Hussain et al. (34), wherein 80% of islet-infiltrating B-cells were CD69⁺ and CD86⁺. We performed all staining procedures immediately after islet isolation, whereas in the study by Hussain et al., phenotypic analysis was carried out following overnight culture of islet infiltrates. We have already described a significant increase in CD19, CD44, CD69, CD86, and MHC class I molecules after a short-period culture of islet-infiltrating B-cells. Most likely, the observed differences are caused by the different methodologies used.

Islet-infiltrating B-cells have been assigned the ability to activate islet-infiltrating T-cells (37). Our findings indicate the opposite. In our studies, islet-infiltrating B-cells did not induce the proliferation of their counterpart T-cells by themselves, even when they were activated with anti-CD40, or in the presence of islet extract. However, this B-cell population fully induced T-cell activation when acting as accessory cells. These results strongly suggest that islet-infiltrating B-cells do not present the appropriate antigens to islet-infiltrating T-cells. In this regard, we have recently reported that only a few islet-infiltrating B-cells may be islet-antigen restricted (38), supporting the idea that islet-infiltrating B-cells do not act as APCs in situ.

In conclusion, islet-infiltrating B-cells accumulate into the islets influenced by sex traits in NOD and (NOD × NOR)F1 mice. Phenotype and functional characteristics of islet-infiltrating B-cells suggest that in the proinflammatory islet milieu there are also immune regulatory mediators that may prevent their activation. Our results strongly suggest that B-cells do not induce islet-infiltrating T-cell activation in situ but may have other unknown roles in the development of the disease.

This work was supported by a grant from the Fondo de Investigaciones Sanitarias of the Spanish National Institute of Health (FIS 00/0447 and FIS 01/1128). M.C.P. was supported by a predoctoral CIRIT fellowship (2001FI 0002) from the Catalan Government. A.A. was supported by a BEFI predoctoral fellowship (01/9065) from the Instituto Carlos III of Spanish National Institute of Health. R.M.A. and X.P. were supported by a Juvenile Diabetes Research Foundation research grant (1-2002-724). M.V.-P. is a researcher of the Fondo de Investigaciones Sanitarias of Spanish National Institute of Health. J.V. is an associate professor of the Serra-Hunter Programme from the Catalan Government.

We thank M.A. Fernandez for technical assistance and Miss D. Cullell-Young for editorial assistance.