Dendritic cells and follicular dendritic cells express a novel ligand for CD38 which influences their maturation and antibody responses

Abstract

CD38 is a cell surface molecule with ADP-ribosyl cyclase activity, which is predominantly expressed on lymphoid and myeloid cells. CD38 has a significant role in B-cell function as some anti-CD38 antibodies can deliver potent growth and differentiation signals, but the ligand that delivers this signal in mice is unknown. We used a chimeric protein of mouse CD38 and human immunogobulin G (IgG) (CD38-Ig) to identify a novel ligand for murine CD38 (CD38L) on networks of follicular dendritic cells (FDCs) as well as dendritic cells (DCs) in the spleen. Flow-cytometry found that all DC subsets expressed cytoplasmic CD38L but only fresh ex vivo CD11c⁺ CD11b⁻ DCs had cell surface CD38L. Anti-CD38 antibody blocked the binding of CD38-Ig to CD38L, confirming the specificity of detection. CD38-Ig immuno-precipitated ligands of 66 and 130 kDa. Functional studies found that CD38-Ig along with anti-CD40 and anti-major histocompatibility complex (MHC) class II antibody provided maturation signals to DCs in vitro. When CD38-Ig was administered in vivo with antigen, IgG2a responses were significantly reduced, suggesting that B and T cells expressing CD38 may modulate the isotype of antibodies produced through interaction with CD38L on DCs. CD38-Ig also expanded FDC networks when administered in vivo. In conclusion, this study has identified a novel ligand for CD38 which has a role in functional interactions between lymphocytes and DCs or FDCs.

Introduction

CD38 is an ADP-ribosyl cyclase enzyme that regulates the activation and growth of myeloid and lymphoid cells. Early studies found that mouse CD38 was a B-cell coreceptor capable of modulating signals through the antigen receptor.¹ These studies also showed that CD38 signalling and coreceptor activity were regulated by conformational changes induced in the extracellular domain upon ligand binding.² Other studies found that ligation of CD38 on B cells induced germline immunoglobulin G1 (IgG1) transcription³ and proliferation of somatically mutated B cells,⁴ but inhibited the growth of immature B cells in the bone marrow.⁵ Cross-linking of CD38 molecules on human thymocytes induced expression of the interleukin-2 receptor and HLA-DR⁶ and cross-linking of CD38 on T cells initiated the secretion of cytokines.⁷ CD38 was also shown to influence intracellular transduction of signals in natural killer cells.⁸ CD38 expression does, however, differ between mice and humans. CD38 is expressed on human germinal centre B cells and plasma cells⁹ but not on these cells in mice.¹⁰ This does raise the possibility that CD38 functional involvement in postgerminal centre B-cell differentiation differs between humans and mice.

Studies in mice with a null mutation for CD38 (CD38^−/−) found that, whilst CD38 was not essential for hematopoiesis or lymphopoiesis, it did have a significant role in antibody responses.¹¹ CD38^−/− mice showed marked deficiencies in antibody responses to T-cell-dependent protein antigens, whilst responses to at least one T-cell-independent type 2 polysaccharide antigen were augmented.¹¹ These data suggested that CD38 played an important role in vivo, in regulating humoral immune responses, but the mechanism by which CD38 mediated changes in antibody function was not identified.

Studies have suggested that human CD38 mediates adhesion of lymphoid cells to the endothelium via homotypic adhesion and that human CD31 is the ligand for CD38.^12–14 These studies were based on the observations that CD38-mediated binding of lymphocytes to human umbilical vein endothelial cells (HUVECs) was blocked by antibody specific for CD31.^15,16 Given that CD31 is expressed on natural killer cells, platelets, neutrophils and monocytes, it was difficult to envisage how these cells modulated the changes to antibody responses noted in the CD38^−/− mice, suggesting that CD31 may not be the only ligand for CD38. Moreover, CD38 was recently shown to regulate trafficking of dendritic cells (DCs) in mice,¹⁷ and human plasmcytoid DCs were shown to induce differentiation of CD38+ plasma cells,¹⁸ suggesting that other ligand(s) may exist in mice and humans. In the light of these observations, in our studies we used a fusion protein of mouse CD38 and the Fc portion of human IgG1, to identify the ligand for murine CD38 (CD38L). The ligand was expressed on DCs and follicular dendritic cells (FDCs) and had a role in humoral immune responses, DC maturation and expansion of FDC networks.

Materials and methods

Construction of a soluble mouse CD38-human IgG1 Fc chimeric protein

The primers used to amplify the extracellular domain of mouse CD38¹⁹ were 5′ (AGG CCG CGC TCA CTC CTG GTG GTG GTG TGG) and 3′ (TCA TCA CGT ATT AAG TCT ACA CGA TGG GTG CTC). A cDNA library of spleen cells was used as a template for PCR amplification. Each primer contained restriction sites that allowed subcloning into the vector pCR^TM3-Uni, which contained the sequence encoding the CH2CH3 (Fc) domains of human IgG1. The N-terminus of extracellular CD38 was fused to the Ig tail. The resulting plasmid construct for CD38-Ig was transfected into J558L cells as previously described.²⁰ Soluble CD38-Ig protein was purified from culture supernatants using Gamma-bind plus, recombinant Protein G columns (Pharmacia Biotech, Uppsala, Sweden) in phosphate-buffered saline (PBS), and eluted with 0·1 m glycerine buffer. This protein was shown to have < 2 EU/ml endotoxin using the E-toxate assay (Sigma, St Louis, MO).

Mice

C57BL/6J mice were maintained under specific pathogen-free conditions at either the Sir William Dunn School of Pathology (Oxford, UK) or the Mater Medical Research Institute (Brisbane, Australia) research facilities.

Identification of cells expressing CD38L using immunohistochemistry

Spleens were collected from naïve mice or mice given sheep red cells and keyhole limpet Hemocyanin (KLH) 1, 3, 5, 7 and 9 days after immunization. Cryostat sections were fixed with acetone and pretreated with a monoclonal antibody (2.4.G2) specific for mouse Fc receptors, known to block the binding of immunoglobulin Fc to receptors.²¹ CD38L was then detected by labelling tissue with either 10 µg/ml CD38-Ig or 10 µg/ml human IgG1 as a control, followed by biotinylated antihuman IgG-Fc (Jackson, ImmunoResearch Labs, Pennsylvania, PA) and Streptavidin-Texas Red (Vector Labs, California, USA). Sections were then labelled with FITC-labelled antibodies to detect MHC class II, CD3 (T cells), B220 (B cells), 3D6 (marginal zone macrophages), CD86 (DCs), F4/80 (red pulp macrophages) and FDC-M1 (FDCs).

Isolation of splenic DCs

DCs were enriched from spleens of naïve C57BL/6J mice and CD11c cells sorted as required.^22–24

Flow cytometry

Purified DCs were labelled with either CD38-Ig or human IgG1 followed by biotinylated antihuman Ig and Streptavidin-Quantum Red (Sigma). Cells were routinely also simultaneously labelled with either CD4-FITC and CD8-PE, or CD31-FITC (Becton Dickinson Biosciences, California, USA). The expression of CD38L was investigated on the surface of fresh DCs immediately after isolation or after overnight culture. The cells were treated with either CD38-Ig or human IgG1 (Serotec, Oxford, UK) in PBS containing 2% bovine serum albumin (BSA) and 1% rat serum. Bound CD38-Ig or human IgG1 was detected with biotinylated anti-human IgG and Quantum Red-Streptavidin. To determine whether CD38L was expressed in the cytoplasm of DCs, this labelling procedure was used on cells that were permeabilized with 0·3% saponin in PBS with 2% BSA and 1% rat serum,²⁵ and the saponin was included in all steps of the procedure, to keep the cell membrane permeabilized. Following cell labelling, cells were fixed with 2% paraformaldehyde and the cells were analysed using FACSCalibur (Becton Dickinson).

To identify subpopulations of DCs expressing CD38L, DCs were labelled with different combinations of CD8-PE, CD4-FITC, CD11c-PE, CD11b-FITC (Pharmingen, CA, USA), B220-FITC (Cedarlane Laboratories, Ontario, Canada) or CD3-FITC (Southern Biotechnology Associates, AL, USA) followed by cytoplasmic or surface labelling of CD38L. The cells were fixed in 2% paraformaldehyde prior to analysis.

Blocking of CD38 detection

To confirm that CD38-Ig binding to CD38L was mediated by the CD38 component, DCs were incubated with 10 µg/ml CD38-Ig, which had been preincubated with 10-fold excess of rat anti-CD38 antibody (NIMR-5; Pharmingen). To control for the non-specific effects of rat immunoglobulin on CD38-Ig binding, the same quantity of CD38-Ig was incubated with 10-fold excess of rat immunoglobulin. The human IgG1 (10 µg/ml) was also incubated with 10-fold excess of rat immunoglobulin. These DC preparations were then labelled with biotinylated anti-human Ig Fc and Quantum Red-Streptavidin for flow cytometry.

Detection of the CD38L molecule by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE)

Approximately 5 × 10⁶ naïve DCs or DCs stimulated overnight with lipopolysaccharide (LPS) were washed in PBS thrice and lysed in a cell lysis buffer [50 mm Tris-HCl, 150 mm NaCl, 2 mm ethylenediaminetetraacetic acid (EDTA), 34 µg/ml phenylmethylsulphonyl fluoride (PMSF), 1% NP₄O and Complete® protease inhibitor from Boehringer, Mannheim, Germany] at 4° for 30 min, and cell lysates were then incubated with Protein A on Sepharose beads to deplete non-specifically adherent proteins (Pharmacia). The suspension was spun and the supernatant taken for immunoprecipitation. CD38-Ig or human Ig, irreversibly bound to recombinant Protein A on Sepharose beads, was then added to the lysate and mixed for 5 hr at 4° and the beads were then washed 5 times with lysis buffer. The immunoprecipitated protein was eluted directly into a non-reducing sample buffer at room temperature for 15 min, the beads were removed following centrifugation, and the eluted sample was then boiled and run on a 4–20% SDS gradient gel (Bio-Rad, CA, USA). The eluted proteins were visualized using a silver staining kit (Bio-Rad). Every batch of CD38-Ig or human Ig, irreversibly bound to Protein A on Sepharose beads, was always tested for ‘leaching’ by mixing with non-reducing sample buffer for 15 min and running the eluent on SDS-PAGE gel.

In vivo administration of soluble CD38-Ig

Groups of four mice were immunized with 50 µg of soluble di-nitrophenyl (DNP)-KLH (Calbiochem-Novobiochem, Darmstadt, Germany) and were then injected with either CD38-Ig or human IgG (Binding Site, Birmingham, UK) at 100 µg/mouse/day for 4 days starting at the day of immunization. The mice were bled after 14 days and anti-DNP and anti-KLH responses measured by enzyme-linked immunosorbant assay (ELISA). These experiments were repeated 3 times.

Study of FDC networks using tissue sections

Spleen and lymph nodes were collected from mice 14 days after they had been given DNP-KLH and CD38-Ig or human Ig for cryostat sections. Sections were fixed with either 2% paraformaldehyde or cold acetone. The sections were then stained with either FDC-M1 (FDC and tingible body macrophages), B220, CD3 or M115.4 (anti-class II) and peroxidase-anti-rat-Ig or biotinylated peanut agglutinin (PNA) followed by streptavidin-peroxidase. The number of FDC networks per ×20 field stained by FDC-M1 was counted in spleen sections of groups CD38-Ig and control Ig-treated mice. Between 11 and 15 fields were counted per section per mouse and a t-test was used to determine the statistical significance of differences in mean number of networks per field.

Maturation of DCs

To determine if CD38-Ig affected maturation of DCs, the cells were treated with either 10 µg/ml CD38-Ig or human IgG1 (Binding Site Ltd, Birmingham, UK) alone or in combination with anti-CD40 (FGK-45) and anti-class II (M115.4), or with LPS as described in the results section. The concentration of reagents used was determined by titration to give maximum maturation of DCs within 18 hr. The cells were cultured for 18–24 hr and labelled with antibodies to detect class I (28.8.6S, kindly provided by Dr P. Fairchild, Sir William Dunn School of Pathology, Oxford, UK), class II (Cedarlane, Laboratories), CD80 and CD86 (kindly provided by Dr Elizabeth Adams, Sir William Dunn School of Pathology, Oxford, UK). As a comparison, DCs were prepared by an adherence technique.²⁶

Statistics

Error bars shown represent mean + standard error of mean. P-values were calculated using the Mann–Whitney non-parametric t-test, with a one-sided tail, based on pooled data from several experiments.

Results

Characterization of CD38-Ig

A chimeric protein of murine CD38 and the Fc portion of human IgG (CD38-Ig) was purified on Protein G columns following secretion from transfected eukaryotic cells. CD38-Ig, studied by Western blotting by labelling with anti-human Ig, was found to be predominantly a dimer, but some monomeric protein was seen under non-reducing conditions (Fig. 1a). The valency of the protein construct was confirmed by demonstrating a single band with an apparent molecular weight of 55 kDa after reduction. Furthermore, CD38-Ig was shown to be glycosylated as N-endoglycosidase reduced the apparent molecular weight of the protein by approximately 5 kDa (Fig. 1b). Also, two batches of approximately 5 µg of purified CD38-Ig run under non-reducing conditions and stained with Coomassie blue had no detectable contamination by other proteins (Fig. 1c). We also found that CD38-Ig was poorly detected by anti-CD38 in Western blots even under non-reducing conditions.

Figure 1.

Characterization of CD38-Ig. CD38-Ig was isolated using Protein G, run on SDS-PAGE and blotted onto a membrane before detection of proteins expressing human IgG-Fc. (a) CD38-Ig under reducing and non-reducing conditions. (b) CD38-Ig was run under reducing conditions following control treatment, treatment with N-endoglycosidase or no treatment. Several batches of CD38-Ig were tested and were found to have the same molecular weight. (c) Two batches of CD38-Ig were run on SDS-PAGE and the gel was stained with Coomassie blue to show purity.

Localization of the ligand for CD38 in mice using immunocytochemistry

CD38-Ig was used to localize CD38L within the spleen (Fig. 2) and lymph node (data not shown as results were similar). Cohorts of mice were given sheep red cells to initiate an immune response and the spleens and lymph nodes frozen after 2, 5, 7, and 9 days for tissue sections. Tissue sections from these mice and from naïve mice were labelled with CD38-Ig or human IgG1 and an appropriate detection system to identify cells expressing the ligand for CD38 (CD38L). In naïve spleens, CD38-Ιg bound to few isolated, irregularly shaped cells in the red pulp of the spleen, which also co-expressed MHC class II (Figs 2a, b and 2c). The morphology and co-expression of MHC class II suggested that these cells were DCs. During the first 7 days of the primary immune response, CD38L expression appeared on clusters of irregular cells that expressed MHC class II (Figs 2d, e and f) and CD8 (data not shown) within the T-cell areas, further suggesting that they were DCs. In addition, CD38L was also detected on the FDC networks, which coexpressed FDC-M1, a marker specific for these cells (Figs 2g, h and i). CD38L was not detected on T cells, B cells or any macrophage subpopulation. There was some very weak staining of the vascular endothelium. The absence of any binding of soluble CD38-Ig to cells expressing CD38 ruled out homotypic adhesion. At all time points, control human IgG only bound to very occasional cells in the red pulp (Figs 2k and l), but these cells never expressed MHC class II (Figs 2j and 2l). Moreover, whereas CD38-Ig binding was seen in the cytoplasm of cells expressing MHC class II (Figs 2a–2f), the human ‘background’ binding was at the cell surface (Figs 2j–2l).

Figure 2.

CD38L is expressed on DCs and FDCs. Frozen tissue sections of spleens from naïve (a–c) and immunized (d–i) mice were labelled with CD38-Ig and an anti-human Ig-Texas red detection system to identify cells expressing CD38L, which appear red (b, e, h). These sections were also labelled to detect cells expressing MHC class II (a, d) or follicular dendritic cells (g), which appear green. Cells co-expressing both molecules appear orange or yellow (c, f, i). CD38L was identified on isolated DCs in naïve tissue (a–c), in clusters of DCs in immunized tissue (d–f), and on FDC networks (g–i). Human Ig bound very occasional cells in the red pulp (k) which did not coexpress MHC class II (j, l). The labelling was repeated seven times with different reagents. The magnification of cells labelled by immunofluorescence is × 1000.

Flow cytometric analysis of CD38L expression by DCs

Because studies on spleen tissue had suggested that DCs expressed CD38L, isolated splenic DCs were examined by flow cytometry. CD38L was detected in the cytoplasm of CD4⁺, CD8⁺ and CD4^– CD8^– DCs (Fig. 3a). Whereas CD38L was detected in the cytoplasm of all freshly isolated DCs at approximately equal levels (Fig. 3a), approximately 47·4 ± 3·7% (n = 3) of fresh ex vivo CD11c⁺ CD11b^– DCs expressed CD38L on their surface (Fig. 3b). However, as only approximately 5% of DCs expressed CD31 on their surface or cytoplasm (Fig. 3c), CD31 expression was not required for CD38 binding to the majority of DCs. To confirm that the CD38-Ig was binding its corresponding ligand, the CD38-Ig was pretreated with an excess of a known agonistic anti-CD38 antibody in an attempt to block specific CD38 binding. This pretreatment significantly reduced the binding of CD38-Ig to DCs (Fig. 3d), indicating that the CD38 component was binding specifically to its natural ligand. Attempts to up-regulate the expression of CD38L on DCs by LPS, or signals such as anti-CD40, were inconsistent (data not shown).

Figure 3.

Flow cytometry profiles of cytoplasmic and surface CD38L and CD31 expression on DCs and confirmation of the specificity of CD38L detection. DCs were labelled to detect CD4, CD8, CD11c or CD11b and either surface or cytoplasmic CD38L. (a) CD4⁺, CD8⁺ and CD4^– CD8^– DCs expressed CD38L in their cytoplasm. (b) CD11c⁺ CD11b^– DCs expressed surface CD38L. (c) CD31 on surface and cytoplasm of DCs. (d) CD38L expression in the cytoplasm of DCs was confirmed by blocking the binding site of the soluble CD38-Ig with an agonistic anti-CD38 antibody. In the flow cytometry profiles, the dark lines represent labelling with the relevant antibody or soluble CD38-Ig, and the thin line labelling with either human IgG1 control antibody or an appropriate isotype control antibody. The profiles shown are examples of data from at least four experiments, except the blocking study, which was repeated twice and gave the same results.

Characterization of CD38L by SDS-PAGE

To determine the molecular weight of CD38L, CD38-Ig or control Ig was irreversibly bound to Protein A and used to immunoprecipitate CD38L from freshly isolated spleen DCs and DCs treated with LPS. The irreversible binding of CD38-Ig to protein A was always confirmed by SDS-PAGE. CD38-Ig consistently immunoprecipitated proteins of 66 and 130 kDa from lysates of fresh (data not shown) or LPS-treated DCs (Fig. 4). A very faint band of 50 kDa was also seen in preparations of fresh ex vivo DCs.

Figure 4.

Molecular weight of CD38L. Lysates of DCs were mixed with CD38-Ig or human Ig irreversibly bound to recombinant Protein G, and the immunoprecipitate was run on SDS-PAGE gels. This was followed by silver staining of the gel. The immunoprecipitation was repeated 3 times and gave the same bands.

CD38 has a role in DC maturation

To identify a function for CD38L on DCs, ex vivo splenic DCs were treated for 20 hr with soluble CD38-Ig alone or in combination with anti-CD40 and anti-MHC class II antibodies. Human IgG1 was used to determine whether the Fc portion of CD38-Ig contributed to any effects of CD38-Ig. LPS was used to induce maximal maturation of DCs and these were compared to DCs prepared by the adherence (mature DCs). In general, LPS treatment of fresh DCs increased the expression of MHC class I and CD86 to levels comparable with those of mature DCs but did not significantly affect CD80 expression (Fig. 5).

Figure 5.

Histogram profiles of DC maturation following treatment with CD38-Ig and/or costimulatory signals. Splenic DCs were treated overnight with soluble CD38-Ig, human Ig, anti-class II + anti-CD40, CD38-Ig + anti-class II + anti-CD40, CD38-Ig + anti-CD40 or LPS, and phenotypic changes compared to the phenotype of DC isolated by adherence to plastic, which are mature DCs. The DCs were then labelled for class I, MHC class II, CD80 and CD86 to determine the effects of CD38 on DC maturation. The figure shows representative flow cytometry profiles from replicate experiments. In the flow cytometry profiles, the dark lines represent labelling with the relevant antibody and the thin line labelling with an appropriate isotype control antibody. The numbers in parentheses, for (experiment 1) and [experiment 2], show geometric mean fluorescence intensity data from duplicate experiments. The numbers in bold represent significant differences from treatment of DCs with human Ig.

In comparison to treatment of DCs with human Ig, treatment of these cells with anti-class II and anti-CD40 alone induced a limited increase in class I expression but did not up-regulate the expression of other markers. However, treatment of DCs with CD38-Ig alone increased class I expression slightly, and in combination with anti-CD40, or anti-CD40 and anti-class II, increased class I expression to levels comparable to those induced by LPS treatment or seen in mature DCs. MHC class II levels (detected by a different monoclonal antibody) on the treated cells in general were not notably affected by treatment. Moreover, CD80 levels were slightly reduced by all treatments, especially CD38-Ig in combination with anti-CD40 and class II, when compared to treatment with human IgG1 or mature DCs. In contrast, treatment of DCs with CD38-Ig in combination with anti-CD40 and class II increased CD86 expression to levels equivalent to those found for LPS treatment or mature splenic DCs prepared by adherence.

A role for CD38 in humoral immune responses and germinal centre development

To determine whether CD38–CD38L interactions affected antibody responses, cohorts of mice were given human IgG1 or CD38-Ig for 4 days with soluble DNP-KLH on the first day. The mice were bled after 14 days and the spleen sections stained with various markers. A measurement of anti-DNP and anti-KLH responses in serum showed that IgG1 responses were not affected, whereas IgG2a anti-DNP and anti-KLH responses were significantly reduced (P < 0·03; Fig. 6).

Figure 6.

CD38 modulates the isotype of immune responses. Cohorts of mice were given DNP-KLH and either CD38-Ig or human Ig for 4 days. Anti-DNP titres were then assessed after 14 days to determine the effects of CD38 on IgG1 and IgG2a antibody responses. The error bars shown are the standard error of the mean and the data shown are an example of data from three experiments with at least five mice in each cohort. P-values were calculated using the Mann–Whitney non-parametric t-test, with a one-sided tail, based on pooled data from several experiments.

To determine whether CD38-Ig treatment induced any morphological changes in the spleen, sections were analysed by immuno-histochemistry. Staining with antibodies specific for MHC class II, B220 or CD3 did not show any obvious change in the splenic architecture or distribution of cells within the white pulp (data not shown). However, frozen sections labelled with peanut agglutinin, which binds to germinal centre cells or FDC-M1, a monoclonal antibody specific for follicular dendritic cells (FDCs), showed that the number of FDC networks as labelled by FDC-M1 increased by 2·8-fold following CD38-Ig treatment from an average of 0·69 networks per ×20 field in control mice to 2·0 networks per ×20 field (t-test, P < 0·001). The networks also appeared to be significantly larger following treatment with CD38-Ig (Fig. 7a) compared to human IgG1 treatment (Fig. 7b).

Figure 7.

The effects of soluble CD38 on FDC networks. Cohorts of mice were given DNP-KLH and either (a) CD38-Ig or (b) human Ig daily for 4 days. Frozen spleen sections were labelled with FDC-M1 to examine FDC networks and every mouse given the treatment had notably larger networks. Sections were examined from three experiments with four to seven mice per group.

Discussion

CD38 is widely expressed on haemopoeitic cells in humans and mice. Monoclonal antibodies to CD38 have been shown in vitro to have important immunoregulatory effects including proliferation,²⁵ protection of B cells from apoptosis²⁷ and inhibition of B-cell lymphopoiesis.⁵ CD38 is also thought to function as an adhesion molecule via CD31,^12–14 but it is unclear if CD31 represents the only, or the major, functional ligand for CD38. Moreover, mice with a null mutation for CD38 exhibited marked deficiencies in antibody responses to T-cell-dependent protein antigens and augmented antibody responses to at least one T-cell-independent type 2 polysaccharide antigen.¹¹ These data suggested that CD38 played an important role in vivo in regulating humoral immune responses. However, it was difficult to explain how CD31, which is expressed on monocytes, natural killer (NK) cells, neutrophils and platelets, could modulate B-cell responses. We constructed a chimeric CD38-Ig protein to enable us to define the location of CD38L in mice. CD38L was detected on DCs in the red pulp and T-cell areas of the spleen and lymph nodes, on FDC networks after immunization, and on occasional endothelial cells in the endothelium. The ligand was not detected on T cells or B cells in tissue or by flow cytometry (data not shown). Immunoprecipitation studies using CD38-Ig found that CD38 had two ligands on DCs of 66 and 130 kDa. The 130-kDa ligand was probably CD31, as inhibition studies have shown that the ligand for CD38 in humans is CD31, which has the same molecular weight. However, CD31 expression was not required for CD38 binding, as CD38L was detected in the cytoplasm of all DCs but only a small proportion of these DCs expressed CD31, and FDCs do not express CD31. The 66-kDa protein appeared to be a novel ligand for CD38 expressed on DCs and perhaps also on FDCs; however, we were unable to isolate sufficient numbers of FDCs for biochemical studies. Its expression on both DCs and FDCs would explain how T and B cells may receive critical CD38-mediated signals which modulate immune responses as described using agonistic antibodies or CD38^−/− mice.

Interactions between DCs and T cells have been extensively investigated. We have previously shown that DCs can take up and retain soluble antigen and present this antigen to B cells to initiate antibody responses,²³ and recently DCs were shown to present antigen to marginal zone B cells.²⁸ However, except for CD40–CD40L,²⁹ few signals that occur during DC–B-cell interactions have been identified. As anti-CD38 signals initiate proliferation of B cells²⁵ and DCs also initiate B-cell proliferation,^29,30 this study suggests that CD38L on DCs could initiate proliferation of B cells. Moreover, the administration of soluble CD38-Ig in vivo significantly reduced IgG2a responses, whilst having little effect on IgG1 antibody responses. It has been previously shown that B cells regulate the capacity of DCs to promote interleukin (IL)-4 secretion, by down-regulating their secretion of IL-12, thereby favouring the induction of a non-polarized immune response.³¹ In the light of this observation, it seems possible that CD38 on B cells may down-regulate IL-12 secretion via CD38L on DCs, and this would explain why CD38^−/− mice exhibited marked deficiencies in antibody responses to T-cell-dependent protein antigens,¹¹ as IL-12 is essential for IgM³² as well as interferon-γ-mediated IgG2a responses.³³ Further studies on DC function are required but were beyond the scope of this work.

Studies on DCs generally focus on the effects of DCs on lymphocyte function.^{23,28–30,32,34–44} However, it is known that T cells affect the development and function of DCs.⁴⁵ As CD4⁺ T cells, CD8⁺ T cells and B cells express CD38, it is highly likely that these cells also affect DC development and function. To test this hypothesis, DCs were cultured with soluble CD38-Ig, anti-MHC class II or anti-CD40 antibody, representing signals which can be provided by T and/or B cells, in order to determine their effects on DCs. These studies found that signals to MHC class II, CD40 and CD38L on DCs up-regulated expression of MHC class I and CD86 but not MHC class II and CD80. MHC class I and CD86 expression in combination has been shown to be crucial for cytotoxic T-lymphocyte (CTL) development,⁴⁶ suggesting that CD38 on T cells may also have the potential to modulate DCs to a phenotype that promotes CTL development, a suggestion open to further experimental evaluation.

Finally, CD38L was also identified on FDCs 3–5 days after the initiation of an immune response, and when soluble CD38 was administered during this time the size and numbers of FDC networks were significantly increased. As B cells express CD38 and are an integral part of germinal centres, we hypothesize that, when B cells interact with CD38L on FDCs, this interaction stimulates an expansion of FDC networks. CD38L is the first molecule known to expand FDC networks, which has implications for germinal centre and antibody memory development. Since germinal centres are important for the development of B-cell memory,²⁰ an expansion and enhancement of this environment could affect the development of antibody memory.

In conclusion, this study has identified a novel ligand for CD38 on DCs and FDCs which influences antibody responses, DC maturation and the development of FDC networks. Characterization of this ligand provides insight into interactions between B or T cells expressing CD38 and DCs, and how this interaction affects primary and possibly secondary immune responses.