CD22 (Siglec-2) is a B-cell membrane-bound lectin that recognizes glycan ligands containing α2,6-linked sialic acid (α2,6Sia) and negatively regulates signaling through the B-cell Ag receptor (BCR). Although CD22 has been investigated extensively, its precise function remains unclear due to acting multiple phases. Here, we demonstrate that CD22 is efficiently activated in trans by complexes of Ag and soluble IgM (sIgM) due to the presence of glycan ligands on sIgM. This result strongly suggests sIgM as a natural trans ligand for CD22. Also, CD22 appears to serve as a receptor for sIgM, which induces a negative feedback loop for B-cell activation similar to the Fc receptor for IgG (FcγRIIB).
CD22 is a 140 kDa glycoprotein on the surface of B cells that negatively regulates signaling through the B-cell Ag receptor (BCR) 1–3. There are six tyrosine residues within the cytoplasmic portion of CD22, four of which are located within ITIMs 4. These tyrosine residues are phosphorylated upon BCR cross-linking, leading to recruitment of SHP-1 4, 5. SHP-1 subsequently dephosphorylates the BCR-proximal signaling molecules, resulting in downmodulation of BCR signaling. Consistent with this, B cells from CD22-deficient mice are hyperactive 6–9. The extracellular portion of CD22 is composed of seven immunoglobulin (Ig)-like domains, the most distal of which is a V-set Ig-like domain that recognizes α2,6-linked sialic acid (α2,6Sia)-containing glycoconjugates 3, 10. α2,6Sia is common at the terminal of N-linked glycans and is abundantly expressed on various kinds of cells, including erythrocytes, monocytes, B cells, and T cells. α2,6Sia also exists on soluble plasma proteins such as serum-soluble IgM (sIgM) 11. CD22 is a member of the sialic acid-binding Ig-like lectin (Siglec) family, and is also referred to as Siglec-2. CD22 appears to interact with various ligands in cis and in trans to modulate B-cell activity 10.
Potential CD22 ligands, including IgM, CD45, and CD22 itself, have been identified 12. Among them, only CD22 has been identified as a natural glycan ligand for CD22 in cis13. Furthermore, CD22 regulates BCR signaling induced by Ags expressed on other cells in an α2,6Sia-dependent manner 14. It has recently been reported that sialylated multivalent Ags engage CD22 in trans and inhibit B-cell activation 15. Thus, various interactions between CD22 and its ligands have been shown. However, the overall interactions and the subsequent effects on B-cell activation are not fully understood. In this study, we further evaluated the role of CD22 ligand binding in trans in B-cell activation and propose a novel model of CD22 function.
Results and discussion
sIgM binds to CD22 on the cells in a β-galactoside α2,6-sialyltransferase I-dependent manner
Since sIgM has been shown to bind to recombinant CD22 fusion protein (CD22-Fc) 11, we tested whether sIgM binds to CD22-expressing cells. The mouse myeloma line J558L fails to express the CD22 glycan ligand α2,6Sia at the terminal of N-glycan due to a lack of β-galactoside α2,6-sialyltransferase I (ST6GalI) expression. Introduction of a ST6GalI expression vector can restore α2,6Sia on cell-surface glycoproteins and we showed previously that the soluble CD22 fusion protein (CD22-Fc) bound to J558L cells expressing ST6GalI (J558L/ST6) but not to J558L cells 16. Since sIgM has been identified as a potential glycan ligand for CD22 11, we examined whether sIgM binds to these cells. (4-Hydroxy-3-nitrophenyl) acetyl (NP)-specific sIgM bound to Ag NP-PE (Ag/sIgM) was able to bind to CD22-expresssing (J558L/CD22) cells but not to CD22-deficient (J558L) cells (Fig. 1A). Double staining with anti-CD22 mAb is shown in Supporting Information Fig. 1. This binding was not prevented by the presence of FCS containing α2,6Sia, suggesting that CD22 selectively binds to sIgM. CD22 lectin activity is masked on the cells harboring α2,6Sia-containing glycan on the cell surface, since CD22 is heavily glycosylated and interacts with neighboring CD22 via glycan ligands 13. Therefore, we tested whether Ag/sIgM binds to CD22 on J558L/CD22/ST6 cells that express the CD22 glycan ligands. As shown in Fig. 1A, sIgM did not bind to CD22 on J558L/CD22/ST6 cells. Furthermore, we examined their interaction by using spleen B cells treated with or without sialidase (Fig. 1B). sIgM did not interact with spleen B cells from wild-type C57BL/6 mice (Fig. 1A). However, Ag/sIgM bound to sialidase-treated cells, suggesting that sIgM can potentially interact with CD22 on B cells, but endogenous α2,6Sia prevents this interaction. The formation of multimeric CD22 complexes via in cis glycan ligands, probably on CD22 13, may prevent inappropriate interactions between CD22 and molecules harboring α2,6Sia, such as sIgM, in the serum.
Ag/sIgM complex efficiently activates CD22 in trans
While sIgM seems to bind to CD22 on B cells, it cannot bind to CD22 on α2,6Sia-harboring cells. We asked whether the complex of Ag and Ag-specific sIgM (Ag/sIgM) can induce CD22 activation as is the case for synthetic α2,6sialylated Ag 15. Since most B cells from QM mice are NP specific 17, we conjugated NP to non-NP-specific sIgM (NP-sIgM) as an Ag/sIgM and treated with or without sialidase (Supporting Information Fig. 2). We stimulated spleen follicular B cells from QM mice with sialidase-treated Ag/sIgM (α2,6Sia-deficient Ag/sIgM) or untreated Ag/sIgM. Sialidase-treated Ag/sIgM induced augmented BCR signaling, including ERK activation and Ca2+ mobilization, compared with that induced by untreated Ag/sIgM (Fig. 2A and B). In contrast, in B cells from CD22−/− QM mice, Ag/sIgM induced a similar level of BCR signaling to that induced by sialidase-treated Ag/sIgM. In particular, Ag/sIgM induced less Ca2+ mobilization in B cells from WT QM mice than NP-BSA did, whereas Ag/sIgM induced stronger Ca2+ mobilization in CD22−/− QM mouse B cells than NP-BSA did. Furthermore, we stimulated a mouse B lymphoma line, K46μvCD22, which harbored NP-specific BCR with sialidase-treated or untreated Ag/sIgM. As a control, K46μvCD72 which expresses another inhibitory coreceptor, CD72 18, instead of CD22, was used. CD22-expressing cells (K46μvCD22) yielded the results similar to those obtained in QM B cells, whereas the non-CD22-expressing cells (K46μvCD72) exhibited similar results to CD22−/− QM B cells (Fig. 2C and D). Both ERK activation and Ca2+ mobilization were attenuated in K46μvCD22 cells by Ag/sIgM compared with sialidase-treated Ag/sIgM. In contrast, Ag/sIgM and sialidase-treated Ag/sIgM induced a similar level of the BCR signaling in control cells (K46μvCD72). These results imply that CD22 activation is augmented by glycan ligand on sIgM.
Next, we examined whether Ag/sIgM regulates CD22 activation in trans. Ag/sIgM induced CD22 phosphorylation and subsequent recruitment of SHP-1 more efficiently than sialidase-treated Ag/sIgM (Fig. 3A). Furthermore, CD22 preferentially coprecipitated with Ag/sIgM but not sialidase-treated Ag/sIgM, suggesting that CD22 physically binds to sIgM in an α2,6Sia-dependent manner (Fig. 3B). Membrane IgM (mIgM) also coprecipitated with Ag/sIgM regardless of sialidase-treatment. This interaction is probably mediated by Ag. Immunoprecipitation of SHP-1/SHIP-1 revealed that CD22 appears to be a major phospho-protein upon BCR cross-linking by Ag/sIgM (Supporting Information Fig. 3A). Moreover, FcγRIIB, an inhibitory Fc receptor for IgG on B cells seems not to be activated by sIgM because its recruitment of SHIP-1 did not increase by Ag/sIgM as was the case for NP-BSA (Supporting Information Fig. 3B). These results strongly suggest that Ag/sIgM induces a negative feedback loop for BCR signaling via CD22 in trans in a glycan ligand-dependent manner, most likely by coligation of CD22 with BCR (Fig. 3C).
CD22 on B cells cannot bind to sIgM with different Ag specificities and sIgM cannot bind to CD22 on α2,6Sia-expressing cells (Fig. 1). However, when the BCR bears the same Ag specificity as sIgM, the interaction of the BCR with Ag/sIgM may bring CD22 and sIgM into proximity, resulting in the coligation of BCR and CD22 via Ag/sIgM. Thus, Ag/sIgM complexes induce CD22 activation and trigger a negative feedback loop for B-cell activation, as is the case for the FcγRIIB 19–21. These molecular mechanisms prevent autoimmunity and excess immunity depending on the quality and quantity of Ags, i.e. size and valency. When excess amounts of Abs exist, Ags are heavily covered by Abs to induce complement activation, resulting in clearance of Ags by phagocytes without B-cell activation. However, under some circumstances Ag/sIgM complexes that induce either immunity or tolerance are generated. If a relatively large Ag can induce a conformational change in sIgM, complement is activated by the C3d(g)/IgM/Ag interaction. This results in the induction of positive feedback for B-cell activation via the complement receptor CR2/CD21, which is associated with the positive regulatory molecule CD19 22. Small Ags that do not evoke a conformational change in sIgM do not activate complement, instead Ag/sIgM complexes may induce negative feedback for B-cell activation via CD22 as shown in Fig. 3C.
Recently, FcμR on B cells has been identified 23, 24. However, this receptor is undetectable on freshly isolated spleen B cells and its expression is upregulated by BCR stimulation or special culture conditions. Therefore, it is unlikely that this receptor might be involved in sIgM-mediated downmodulation of BCR signaling on normal spleen B cells.
Several lines of evidence support our model. B-cell activation by Ag displayed on a target cell is depressed if the target coexpresses α2,6Sia-containing glycoconjugates 14, 25. Furthermore, it has recently been reported that sialylated multivalent Ags engage CD22 in trans and inhibit B-cell activation 15. Since α2,6-sialylation is largely a feature of higher eukaryotes, this interaction of CD22 may serve to dampen the B-cell response to self-Ags. In addition, sIgM has been identified as a potential CD22 ligand in trans in an α2,6Sia-dependent manner 11. Therefore, Ag/sIgM complexes may act as α2,6Sia-multivalent Ags and induce CD22-mediated negative regulation of BCR signaling in order to prevent B-cell activation. Indeed, sIgM-deficient mice 26 as well as CD22-defficient mice 27 exhibited autoimmunity, suggesting that sIgM prevents autoimmunity. Therefore, sIgM contributes to not only the clearance of Ags, but also to CD22-mediated suppression of B-cell activation to maintain tolerance. CD22 as a receptor for IgM appears to induce negative regulation of B-cell activation.
We demonstrate that CD22 is activated efficiently by Ag/sIgM and negatively regulates BCR signaling in a glycan ligand-dependent manner. Our data strongly suggest that CD22 serves as a receptor for sIgM in a glycan ligand-dependent manner in trans. Together with sIgM as a natural glycan ligand in trans, CD22 regulates a negative feedback loop for B-cell activation and may contribute to B-cell tolerance.
Materials and methods
The retrovirus vectors pMx-CD22 and pMx-ST6GalI have been described previously 16, 28.
Cells and mice
The mouse myeloma lines J558L, and NP-specific BCR-reconstituted J558L, J558Lμm3, and NP-specific BCR-reconstituted mouse B lymphoma line K46μv were described previously 16, 28, 29. To obtain retrovirus, plasmids were transfected with Plat-E cells 30 by a method of calcium phosphate precipitation. Cells were infected with the retrovirus expressing mouse CD22 and/or ST6GalI. Spleen CD23+ B cells from QM mice and CD22−/− QM mice 9, 17 were purified as described previously 31. Mice including WT C57BL/6 mice were maintained under specific pathogen-free conditions according to the guidelines set forth by the animal committee of Tokyo Medical and Dental University.
Cells were cultured as described previously 18. Cells were stimulated with NP-conjugated BSA, or alternatively NP-conjugated sIgM (NP-sIgM) or sialidase (Roche Applied Science)-treated NP-sIgM.
Immunoprecipitation and Western blotting
Cell lysates were immunoprecipitated with rabbit anti-mouse CD22 Ab 32, anti-SHP-1 Ab, anti-SHIP-1 (these two Abs were from Santa Cruz Biotechnology), anti-FcγRII/III mAb 2.4G2 (BD Biosciences) or NP-specific IgG Ab from QM mice together with protein G-Sepharose (Amersham Pharmacia Biotech). Total cell lysates or immunoprecipitates were separated on SDS-PAGE and transferred to membranes. Membranes were incubated with peroxidase-conjugated anti-phospho-tyrosine mAb 4G10 (Upstate Biotechnology), rabbit anti-SHP-1 Ab (Santa Cruz Biotechnology), rabbit anti-phospho-ERK Ab (New England Biolabs), goat anti-mouse IgM Ab (Southern Biotech), anti-β-tubulin mAb TUB 2.1 (Seikagaku Kogyo) or rabbit anti-CD22 Ab followed by appropriate peroxidase-conjugated Abs, anti-rabbit IgG Ab (New England Biolabs), anti-goat IgG (Southern Biotech) or anti-mouse IgG Ab (Amersham Pharmacia Biotech). Proteins were then visualized by a Chemi-Lumi One system (Nacalai Tesque).
Cells were incubated with biotin-labeled CD22-Fc 16 or anti-mouse CD22 mAb Cy34.1 (BD Biosciences), followed by reaction with FITC-labeled streptavidin (Dako). Alternatively, cells were stained with NP-specific IgM, B1-8 33 and NP-conjugated phycoerythrin (NP-PE) or Alexa647-conjugated sIgM (non-NP specific). Cells were then analyzed by flow cytometry using a CyAn ADP (Beckman Coulter).
Measurement of intracellular calcium mobilization
Cells were incubated in culture medium containing 5 μM Fluo-4/AM (Molecular Probes) for 30 min. Cells were stimulated with Ag and analyzed by flow cytometry using a CyAn ADP (Beckman Coulter).
The authors thank K. Mizuno, T. Asano, A. Ogawa, and A. Yoshino for technical assistance. This work was supported in part by grants from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.
Conflict of interest: The authors declare no financial or commercial conflict of interest.