Nature Chemical Biology 1, 93-97 (2005)

Nature Chemical Biology1, 93-97 (2005)
doi: 10.1038/nchembio713

Homomultimeric complexes of CD22 in B cells revealed by protein-glycan cross-linking

Shoufa Han1,2, Brian E Collins1,2, Per Bengtson1 andJames C Paulson1

CD22 is a negative regulator of B-cell receptor signaling, an activity mediated by recruitment of SH2 domain−containing phosphatase 1 through a phosphorylated immunoreceptor tyrosine inhibitory motif in its cytoplasmic domain1. As in other members of the sialic acid−binding immunoglobulin-like lectin, or siglec, family, the extracellular N-terminal immunoglobulin domain of CD22 binds to glycan ligands containing sialic acid, which are highly expressed on B-cell glycoproteins2. B-cell glycoproteins bind to CD22 in cis and 'mask' the ligand-binding domain3, modulating its activity as a regulator of B-cell signaling4, 5, 6. To assess cell-surface cis ligand interactions, we developed a new method for in situ photoaffinity cross-linking of glycan ligands to CD22. Notably, CD45, surfaceIgM (sIgM) and other glycoproteins that bind to CD22 in vitro7, 8 do not appear to be important cis ligands of CD22 in situ. Instead, CD22 seems to recognize glycans of neighboring CD22 molecules as cis ligands, forming homomultimeric complexes.

Activation of B cells after antigen binding to the B-cell receptor (BCR) is a primary immune response to pathogens, resulting in proliferation and production of antigen-specific antibodies. BCR activation is highly regulated by coreceptors like CD22, which set a signaling threshold that prevents an aberrant immune response and autoimmune disease1. Although it is clear that cis ligands modulate the activity of CD22 (refs. 4−6), their identity has not been unequivocally established. CD22 binds specifically to the terminal sequence N-acetylneuraminic acid (2-6) galactose (NeuAc-(2-6)-Gal) present on many B-cell glycoproteins2. Although glycan-mediated binding of CD22 to CD45 and sIgM has been demonstrated in vitro, in situ interactions of CD22 with these glycoproteins detected with protein cross-linking reagents were found to be glycan independent9. Because the bulky glycan contacting CD22 could leave a gap between the proteins too large to be bridged by a bifunctional protein cross-linking reagent, we reasoned that cross-linking to the glycan ligand itself would provide a more direct approach to identifying the glycoprotein(s) comprising the cis ligands of CD22.

To achieve covalent cross-linking between CD22 and its glycan ligands in situ, we adopted a strategy for metabolic installment of a photo-cross-linker into cell-surface glycans (Scheme 1). The strategy relies on the ability of mammalian cells to take up sialic acid analogs and incorporate them through the normal biosynthetic pathway into cell-surface glycoproteins10, 11. We selected the aryl-azide (AAz) moiety as a preferred functional group for photo-cross-linking12 and installed AAz at the C-9 position of N-acetylneuraminic acid (NeuAc, Compound 1) to give 9-AAz-NeuAc (Compound 4; Scheme 2), as CD22 shows enhanced binding affinity to NeuAc modified to contain aryl substituents at C-9 (refs. 5,13).

Scheme 1:

Strategy for metabolic incorporation of 9-AAz-NeuAc into cell-surface glycoproteins.

Full scheme and legend (34K) Figures, schemes & tables index

Scheme 2:

Synthesis of 9-AAz-NeuAc (Compound 4).

Full scheme and legend (6K) Figures, schemes & tables index

To determine whether cells could use 9-AAz-NeuAc in place of NeuAc, we used the B-cell line BJAB K20 (K20), which is deficient in UDP-GlcNAc-2-epimerase, a key step in sialic acid biosynthesis14. K20 cells are thus deficient in sialic acid, unless the medium is supplemented with NeuAc, N-acetylmannosamine (ManNAc) or substituted NeuAc derivatives. Supplementing the medium with 3 mM NeuAc, 9-AAz-NeuAc or ManNAc (data not shown) restores cell-surface sialic acids and CD22 ligands on K20 to levels similar to those of wild-type BJAB K88 (K88), as detected with labeled Sambucus nigra agglutinin (SNA) or CD22-Fc chimera (CD22-Fc), respectively (Fig. 1a). Notably, SNA, which is specific for the NeuAc-(2-6)-Gal linkage, seems unaffected by the 9-AAz substituent. CD22-Fc showed slightly enhanced binding to 9-AAz-NeuAc cells, as anticipated, because of the aryl substituent at this position. We verified the presence of the AAz moiety on the cell surface by the Staudinger-Bertozzi ligation15, which results in covalent attachment of biotin to the cell surface for subsequent detection with labeled streptavidin (Scheme 3 and Fig. 1b).

Figure 1:Introduction of a carbohydrate photoaffinity cross-linker into cell-surface glycoprotein ligands in situ.

(a) K88 cells (red bold), K20 cells (gray shaded) and K20 cells cultured with 3 mM NeuAc (black thin) or 9-AAz-NeuAc (blue bold), stained with CD22-Fc chimera and FITC-labeled anti−human IgG (Fc specific) or FITC-labeled SNA to detect incorporated NeuAc or 9-AAz-NeuAc. (b) K20 cells cultured with 3 mM NeuAc (black thin), 9-AAz-NeuAc (blue bold) or no added sugar (gray shaded), then reacted with triphenylphosphine-biotin; incorporated biotin was detected with FITC-streptavidin.

Full figure and legend (14K) Figures, schemes & tables index

Scheme 3:

Azide reactivity on the cell surface determined by the Staudinger-Bertozzi ligation (incorporation of a biotin group).

Full scheme and legend (14K) Figures, schemes & tables index

To assess the cross-linking of 9-AAz-NeuAc to CD22, we used a biotinylated sialoside, 9-AAz-NeuAc-(2-6)-Gal-(1-4)-GlcNAc--spacer-biotin (Compound 6; Scheme 4). Incubation with CD22-Fc resulted in cross-linking of the 9-AAz-NeuAc−containing sialoside to CD22-Fc after UV irradiation (Fig. 2a). We observed no cross-linking in analogous experiments with a variety of control proteins, indicating that the reaction of 9-AAz-NeuAc-sialoside with CD22 is not a result of nonspecific cross-linking (data not shown).

Figure 2:Cross-linking of 9-AAz-NeuAc−containing glycans to CD22-Fc chimera in vitro, or endogenous B-cell membrane−bound CD22 in situ.

(a) Western blot analysis to detect UV-induced cross-linking of 9-AAz-NeuAc-(2-6)-Gal-(1-4)-GlcNAc--Lc-Lc-biotin to CD22-Fc in vitro. (b) In situ cross-linking of CD22. Western blot analysis of CD22 from K20 cells cultured with or without exogenous NeuAc or 9-AAz-NeuAc for 18 h and exposed to UV light to induce cross-linking. (c) Western blot analysis of CD22 from K20 cells cultured with or without 9-AAz-NeuAc or NeuAc for the indicated time, and then exposed to UV light.

Full figure and legend (13K) Figures, schemes & tables index

Scheme 4:

Chemoenzymatic synthesis of 9-AAz-NeuAc-(2-6)-Gal-(1-4)-GlcNAc--Lc-Lc-biotin (Compound 6).

Full scheme and legend (7K) Figures, schemes & tables index

To determine whether cell-surface glycoproteins containing 9-AAz-NeuAc would cross-link to CD22 in situ, we harvested K20 cells grown in medium supplemented with NeuAc or 9-AAz-NeuAc and irradiated them with UV light to induce cross-linking. Most of the CD22 from cells cultured with 9-AAz-NeuAc yielded a diffuse, high-molecular-weight band of >250 kDa in western blots, indicative of extensive cross-linking (Fig. 2b), whereas CD22 from cells cultured with NeuAc gave a discrete band at the expected molecular weight of 120 kDa. Culturing cells with 9-AAz-NeuAc for as little as 6 h resulted in detectable CD22 cross-linking, with maximal cross-linking achieved after 14−18 h of culture (Fig. 2c and Supplementary Fig. 1).

To identify the candidate cis ligands of CD22, we evaluated the ability of CD22-Fc to precipitate them from cell lysates in vitro. Although neither SNA nor CD22-Fc precipitated glycoproteins from lysates of K20 cultured in the absence of sialic acids, both precipitated numerous glycoproteins from lysates of K20 cultured with ManNAc, NeuAc or 9-AAz-NeuAc at levels similar to that from wild-type K88 (Fig. 3a and Supplementary Fig. 2). SNA precipitated lower-molecular-weight proteins more efficiently than CD22-Fc, except in K20 cultured with 9-AAz-NeuAc, presumably because of the enhanced affinity this sialic acid provided for CD22. Otherwise, CD22 seemed to recognize the same pattern of glycoproteins expressed on K88 or K20 cells cultured with ManNAc, NeuAc or 9-AAz-NeuAc. Both CD22-Fc and SNA also precipitated specific proteins previously implicated as cis ligands of CD22, including CD45 (refs. 7,16,17), CD19 (ref. 18) and the plasma membrane calcium ATPase PMCA19 in a sialic acid−dependent manner (Fig. 3b andSupplementary Fig. 3). Amounts precipitated from K88 and K20 cultured with sialic acid precursors were comparable (with ManNAc) or slightly enhanced (with 9-AAz-NeuAc) in three separate experiments.

Figure 3:Differential recognition of B-cell glycoproteins by CD22 in vitro and in situ.

(a) Streptavidin-HRP immunoblotting for proteins 'immunoprecipitated' with CD22-Fc or SNA-agarose in a sialic acid−dependent manner from K88 cells or K20 cells cultured exogenous NeuAc, 9-AAz-NeuAc, or ManNAc, then surface biotinylated. (b) Immunoblotting for CD45 and CD19 in total lysate or CD22-Fc immunoprecipitants from cells treated as in a. (c) Immunoblotting for CD22, CD45, IgM and PMCA after CD22 immunoprecipitation of cross-linked lysates from K20 cells with or without NeuAc or 9-AAz-NeuAc and UV light to induce cross-linking. (d) Differential recognition of glycoproteins by CD22 in vitro as compared with in situ, possibly driven by microdomain localization or spatial juxtaposition.

Full figure and legend (40K) Figures, schemes & tables index

To assess whether these same proteins are cis ligands of CD22 in situ, we immunoprecipitated lysates of UV-irradiated cells cultured with NeuAc or 9-AAz-NeuAc with anti-CD22 and sequentially probed the blots with antibodies to CD22, CD45, PMCA, IgM and CD19. In contrast to CD22, which is extensively cross-linked in cells cultured with 9-AAz-NeuAc, negligible cross-linking of the other glycoproteins was observed, and none coprecipitated with CD22 (Fig. 3c and Supplementary Fig. 4). Because CD22 is estimated to be on the surface of B cells at levels equal to or 10-fold higher than CD19 and IgM, respectively20, 21, these glycoproteins should be detected in the anti-CD22−immunoprecipitated pellet even if only a small fraction of them are cross-linked to CD22. CD45, however, is expressed at levels 10-fold higher than CD22 (ref. 22), and a small fraction cross-linked to CD22 could be missed. To address this possibility, we used anti-CD45 for immunoprecipitation. Only the resulting supernatant contained the cross-linked CD22, indicating no significant cross-linking of CD45 to CD22 (Supplementary Fig. 4). Thus, despite the fact that these glycoproteins carry glycan ligands recognized by CD22 (Fig. 3b)8, none of them seem to represent significant cis ligands of CD22 in resting B cells.

Masking of CD22 to multivalent sialoside probes by cis ligands has been reported to be partially reversed after BCR activation3. In activated cells, we observed no significant change in the amount or size of the CD22 cross-linked complex, nor did we find any cross-linking of CD45, PMCA or sIgM to CD22 (Supplementary Fig. 5). This is perhaps not surprising, because in separate experiments evaluating the binding of multivalent sialoside probes, we observed no 'unmasking' of CD22 after activation of K88 or K20 cultured in the presence of sialic acids (data not shown).

Taken together, the results suggest that many, if not most, B-cell glycoproteins carry glycans recognized by CD22-Fc in vitro (Fig. 3a,b), but they do not seem to be significant cis ligands in situ (Fig. 3c). It follows therefore that the extracellular ligand-binding domain of CD22 has access to only a limited subset of glycoprotein glycans when anchored to the B-cell membrane. This could result from colocalization of glycoproteins to the same microdomain as CD22, from a requirement that the glycoprotein glycans be physically juxtaposed with the CD22 binding site (the same distance from the membrane), or both (Fig. 3d).

One glycoprotein that fits both criteria is CD22 itself. Indeed, there are four potential N-linked glycosylation sites in the N-terminal immunoglobulin domain of CD22 (ref. 23) that have been suggested as potential cis ligands9, 24. Moreover, the high molecular weight of cross-linked CD22 (>400 kDa) is consistent with recognition of glycans on adjacent CD22 molecules, leading to the formation of CD22 'multimers'. Indeed, with low sialic acid incorporation, discrete bands consistent with the formation of dimers, trimers and tetramers of CD22 could be detected (Fig. 2c).

To test this possibility, we prepared expression vectors producing CD22 fusion proteins with C-terminal (cytoplasmic) Flag-tag or Myc-tag (Supplementary Fig. 6) and transfected them simultaneously into K20, then cultured them with either NeuAc or 9-AAz-NeuAc. Coprecipitation of CD22-Myc with anti-Flag after irradiation with UV light showed protein-glycan cross-linking between CD22-Flag and CD22-Myc (Fig. 4). Notably, only the most highly cross-linked species of CD22-Myc coprecipitated with anti-Flag, whereas the non-cross-linked CD22-Myc (120 kDa) remained in the supernatant. Immunoprecipitation with anti-Myc gave similar results (Supplementary Fig. 6). The high degree of cross-linking between the Myc-tag and Flag-tag constructs, coupled with the virtual absence of cross-linking with other major glycoproteins containing glycans recognized by CD22 (Fig. 3c), suggested that glycans of neighboring CD22 molecules are the predominant cis ligands of CD22.

Figure 4:CD22 forms homomultimeric complexes with glycans on neighboring CD22 molecules.

K20 cells transfected with CD22-Flag, CD22-Myc or both, cultured in the presence of exogenous NeuAc or 9-AAz-NeuAc and exposed to UV light to induce cross-linking. Western blot analysis for Myc- and Flag-tags. Lysates were immunoprecipitated with anti-Flag M2 agarose.

Full figure and legend (10K) Figures, schemes & tables index

Interactions of cis ligands with CD22 have been variously proposed to enhance or suppress its function as a regulator of BCR signaling4, 5, 6, 9, 25; consequently, a consensus view of the role of these interactions on CD22 function has not emerged. We showed here that CD22 exhibits highly restricted recognition of the glycans of neighboring CD22 molecules in situ, to the exclusion of many other B-cell glycoproteins recognized in vitro. Although we have not excluded protein-glycan cross-linking of other glycoproteins to CD22, mass spectrometric analysis of the cross-linked band detected only CD22 (data not shown). We also note that protein-protein interactions documented for CD22 (refs. 9,19) would not be detected unless the interactions simultaneously involved binding to glycans. Nonetheless, the selectivity of CD22 for cis ligands emphasizes the need for a better understanding of the membrane microdomain distribution of CD22 and other glycoproteins involved in BCR regulation. Significantly, CD22 contains a cytoplasmic motif that binds to a component of the scaffolding of clathrin-coated pits26. Additionally, we observed colocalization of approximately 75% of CD22 with clathrin-coated pits in murine B cells, consistent with microdomain localization being an important factor in CD22 selectivity (data not shown). Although homomultimerization of CD22 through protein-glycan interactions may simply be a consequence of its density in microdomains like clathrin-coated pits, multimerization through protein-glycan interactions may also serve to modulate the activity of CD22 as do protein-protein interactions that modulate the activity of other regulators of lymphocyte signaling27.

In addition to binding in cis, CD22 binds ligands in trans on neighboring cells, causing redistribution to sites of cell contact and dampening of B-cell activation28, 29. It will be of interest to apply the 9AAz-NeuAc cross-linking approach to identifying trans ligands of CD22. Most members of the sialic acid−binding immunoglobulin-like lectin (siglec) family similarly interact with both cis and trans ligands30. To the extent that 9-AAz-NeuAc or related analogs are recognized by siglecs13 and are incorporated into cells bearing their ligands, the approach described here may be of general utility in identifying the in situ ligands of other members of the siglec family.

Top of page

Methods

Antibodies.

Antibodies and dilutions for western blots were as follows: anti-CD22 (1:500, Santa Cruz Biotechnology, H-221), anti-CD45 (1:500, Santa Cruz Biotechnology, 35-Z6), anti-PMCA (1:1,000, Affinity BioReagents, 5F10), rabbit anti-CD19 (1:5,000, Cell Signaling Technology), goat anti-IgM (1:1,000, Pierce Biotechnology), mouse anti-Flag (1:5,000, Sigma, M2), rabbit anti-Myc (1:10,000, Abcam), horseradish peroxidase (HRP)−conjugated goat anti−human IgG (Fc specific, 1:5,000, Sigma), HRP-conjugated streptavidin (1:20,000, Jackson Immunoresearch Labs), HRP-conjugated goat anti-mouse (1:10,000, Upstate Biotechnology), HRP-conjugated donkey anti-rabbit (1:10,000, Amersham), and HRP-conjugated rabbit anti-goat (1:30,000 Jackson Immunoresearch Labs). For immunoprecipitations, we used anti-CD22 (Santa Cruz Biotechnology, M20) and anti-CD45 (Santa Cruz Biotechnology, 35-Z6).

In vitro binding of lectins to 9-(p-azidobenzoylamino)-NeuAc-(2-6)-Gal-(1-4)-GlcNAc--Lc-Lc-biotin (6).

9-AAz-NeuAc (Compound 4) and Compound 6 were synthesized and characterized as described in Supplementary Methods online. Compound Compound 6 (40 l, 4 mM) was incubated at room temperature (22 °C) in the dark for 30 min with agitation with each of the following lectins (40 g): CD22-Fc (produced as described in Supplementary Methods online), SNA-II−HRP (1 mg ml-1, E.Y. Laboratories), MAA-HRP (1 mg ml-1, E. Y. Laboratories) and PSL-HRP (0.3 mg ml-1, E. Y. Laboratories). Aliquots (40 l) of each in a 12-well tissue culture plate were exposed to UV light (254 nm) for 20 min at a distance of 1.5 cm, and the remainder was kept in the dark. All samples were ultrafiltered in the dark (Centricon, molecular weight cutoff 10,000 Da) with PBS wash (three times) to remove unbound Compound 6 and then characterized by western blot analysis.

Cell culture and UV cross-linking.

K20 cells were cultured in RPMI/Nutridoma SP (Roche) and 10 mM HEPES. Medium was further supplemented with 3 mM 9-AAz-NeuAc, NeuAc or ManNAc for 18 h. K88 cells were cultured in RPMI/10% FCS and 10 mM HEPES. Cells were washed with PBS to remove unincorporated NeuAc or derivatives to yield a single cell suspension and were used for cross-linking or flow cytometry. For cross-linking, cells were divided into aliquots (1−2 106 cells per ml in PBS) in either 12- or 6-well tissue-culture plates, cross-linked by exposure to a handheld 254-nM UV source 1.5 cm from the cells for 20 min with agitation, then washed and lysed at 1 107 cells per ml in 50 mM Tris (pH 8.0), 150 mM NaCl, 5 mM EDTA and 1% Nonidet P-40 with a protease inhibitor cocktail (Calbiochem).

Lectin staining and flow cytometry.

K20 or K88 cells (1 105 cells per 100 l) were stained with 1 g of fluorescein isothiocyanate (FITC)-labeled SNA (Vector Labs) for 30 min on ice before analysis by flow cytometry. Alternatively, 8 g of CD22-Fc chimera was precomplexed with 4 g of FITC-labeled anti−human IgG (Fc-specific) in 20 l for 15 min, and 10 l of the complex was added to K20 cells as above. Cell-surface azide reactivity was detected by means of the Staudinger-Bertozzi ligation with a triphenylphosphine-biotin probe (Compound 5). In short, cells were resuspended in 400 l of PBS containing 5 mg ml-1 of BSA and 1 mMCompound 5 and allowed to react for 1 h on ice. Cells were then washed with PBS/BSA twice, and incorporated biotin was detected with FITC-conjugated streptavidin (Jackson Immunoresearch Labs, 1 g of streptavidin-FITC per 100 l).

Western blotting and immunoprecipitations.

Precleared lysates (1 ml) were incubated overnight at 4 °C with end-over-end rotation with 5 g of anti-CD22 or anti-Myc, followed by two additional hours with protein A−Sepharose. Alternatively, 25 l of anti-Flag M2 affinity agarose was added to lysates and incubated overnight. Pellets were washed three times with lysis buffer, eluted with 2 Bis-Tris sample buffer (Invitrogen), resolved on 10% NuPAGE gels, transferred to nitrocellulose and blocked with 10% milk in TBS with 0.05% Tween-20 (TTBS). Blots were probed with the indicated antibodies diluted in 10% milk/TTBS and visualized with chemiluminescence. For reprobing, blots were stripped by incubation in 10 mM Tris and 150 mM NaCl (pH 2.3) at 60 °C for 30 min, washed extensively with TTBS, and blocked again as above.