New functions for the sialic acid-binding adhesion molecule CD22, a member of the growing family of Siglecs.

Abstract

CD22, a B-cell specific transmembrane protein, is a member of the family of Siglecs (sialic acid-binding immunoglobulin(Ig)-like lectins). The Siglecs are adhesion receptors that interact specifically with sialic acids. Sialic acids are abundantly expressed on cell surfaces and on secreted glycoproteins. More than 40 natural forms exist in nature and they can be attached in a variety of linkages to other sugars, thereby generating a considerable degree of molecular diversity and specificity [1]. On the cellular surface, sialic acids are a major carrier of negative charge and can reduce nonspecific cell adhesion through a charge repulsion effect. Although a range of pathogenic viruses, bacteria and protozoa use sialic acids on the cell surface for recognition, it is only recently that mammalian lectins have been shown to specifically bind sialic acids. The characterization of Siglecs as sialic acid-binding proteins began with independent studies on sialoadhesin (Sn), a receptor on macrophages [2] and CD22 on B cells [3]. When both cDNAs were cloned, a significant degree of sequence similarity of these two members of the Ig superfamily was observed [4]. Two other known mammalian members of the Ig superfamily, namely myelin-associated glycoprotein (MAG) and CD33, were found to share sequence similarity with Sn and were subsequently shown to be sialic acid-binding proteins [5, 6]. Most recently, several new members of the Siglec family have been discovered, most of them by sequencing of randomly selected human cDNAs. Apart from MAG (Siglec-4a), which is found exclusively in the nervous system, all Siglecs are expressed within the haemopoietic and the immune systems. These new members illustrate a common theme for the Siglecs: all are expressed in a highly tissue-specific manner, have a specificity for certain types of sialic acid in particular linkages and often carry one or several tyrosine-based inhibitory motifs (ITIMs) in their intracellular tails [7-14] (Fig. 1). The ITIM motifs are known to be involved in the negative regulation of the cellular signalling. These properties suggest that most Siglec family members may be involved in cell to cell interactions that result in intracellular inhibitory signals. The Siglec family. The diagram shows the domain organization and presence of tyrosine-based inhibitory motifs (ITIM) or the membrane-distal SLAM-like Tyr motifs (distal Y) (28) on the cytoplasmic tails of mammalian Siglecs. The greatest sequence similarity between Siglecs occurs in the V-set domain (red) and the adjacent 1–2 N-terminal domains (orange). The sialic acid binding site is contained in the V-set domain. Names of the various members, the expression pattern and sialic acid specificities are shown. Siglec-5–Siglec-9 have only been identified in humans so far. Sialoadhesin (Sn, Siglec-1, CD169) is expressed on macrophages in a regulated fashion, with high levels on distinct sets of macrophages, for example those present in the perifollicular zones of lymphoid tissues [15]. The Sn expression can be induced by soluble factors or under pathological conditions [16-19]. Sn has a specificity for sialic acids (Sia) in α2,3 linkage (e.g. NeuNAc-α2,3-Gal-β1,3GalNAc) which are expressed for example on the surface of myeloid cells and in the extracellular matrix. Certain pathogenic micro-organisms also express sialylated glycoconjugates on their surface. Therefore, it is thought that the Sn may function in host defence as well as acting as an accessory molecule to promote interactions of macrophages with the host microenvironment. CD22 (Siglec-2) is expressed in a B-cell specific fashion. It has a high specificity for α2,6 Sia (human CD22 the NeuNAc, mouse CD22 the NeuNGc-form) and carries three ITIM motifs on its intracellular tail. Potentially, CD22 can bind to target cells such as lymphocytes or cytokine-activated endothelial cells that express high levels of α2,6 Sia on the surface [5, 20, 21]. The role of CD22 in modulating B-cell signalling is quite well understood, owing to the characterization of CD22-deficient mice, as will be discussed below. CD33 (Siglec-3) was originally shown to have specificity for α2,3-linked Sia [6], although recent observations suggest that CD33 may prefer the α2,6 linkage [22]. CD33 has been extensively used as a marker of myeloid cell progenitors that is absent from pluripotential haemopoeitic stem cells. During myelopoiesis, CD33 is lost from neutrophils but retained on monocytes and some tissue macrophages [6]. MAG (Siglec 4a) and its close homologue SMP (Siglec 4b), which has only been identified in birds, are expressed on glial cells. MAG had been postulated as being important for the process of myelin formation. However, MAG-deficient mice develop essentially normal and only show some defects in myelin–axon interactions at older age (Fig. 1) [23, 24]. Identification and cloning of new human members of the Siglec family in the last 2 years demonstrate that the Siglecs comprise a large gene family. All of these new members (Siglec-5–Siglec-9) are most strongly related to CD33 (with 50–80% amino acid identity), thus comprise a subfamily. Most of them have so far only been characterized in terms of tissue distribution and sugar-binding specificity, but little is currently known concerning their functional properties. Apart from Siglec-8, all CD33-related Siglecs carry a membrane proximal ITIM-like motif, which can be tyrosine-phosphorylated, as shown in the cases of CD33 and Siglec-7, and recruit the tyrosine phosphatases SHP-1 and SHP-2 [25-27]. In addition to the ITIM, they have the well-conserved membrane-distal tyrosine based motif: TEYSE(I/V) that is similar to a motif in SLAM (signalling lymphocytes activation molecule) known to be important for the recruitment of SAP (SLAM-associated protein) [28]. SAP has been shown to prevent interactions with the SHP-2 tyrosine phosphatase and hence block signalling through this mechanism [29]. Considerably more work is needed to understand the role of both this motif and the ITIM motif in the signalling functions of CD33-related Siglecs. Siglec-5 was identified as a new Siglec expressed on myeloid cells which bound α2,3- and α2,6-linked sialic acids equally [7]. Studies using human leukaemic cell lines indicated that Siglec-5 is expressed at a later stage than CD33 during haemopoietic differentiation and is retained on both neutrophils and monocytes, unlike CD33 which is lost on mature neutrophils. Siglec-6 (also called OB-BP1 or CD33L1) was identified as an anonymous cDNA [30] as well as in screens for leptin-binding proteins [8] and was subsequently shown to bind sialic acid. It is expressed at high levels on placental cells and at lower levels on B cells. Siglec-6 has an unusual specificity for the sialyl Tn antigen (Neu5Acα2–6GalNAcα) [8]. Siglec-7 [9, 10] (also identified as p75/AIRM1 in screens for inhibitory natural killer (NK) cell receptors [11]) is expressed mainly by NK cells and monocytes. It carries a classical ITIM motif that can be tyrosine-phosphorylated and recruit SHP-1 in NK cells [11]. Siglec-7 is about 80% identical to Siglec-9 which shows the highest expression on monocytes, neutrophils and a minor population of NK-like cells with the phenotype: CD16+CD56−[13, 14]. Both Siglec-7 and Siglec-9 have a similar binding specificity for α2,3 Sia and α2,6 Sia. Finally, Siglec-8 was characterized as a new eosinophil-specific receptor. It preferentially binds to α2,3 Sia. Siglec-8 has a shorter cytoplasmic tail than the other members of the CD33-related Siglecs and lacks the two Tyr-based motifs [12] (Fig. 1). Recent studies indicate that Siglec-8 may also be expressed at very low levels on basophils [31]. Only recently have mammalian lectins been identified that mediate sialic acid-dependent interactions. An important class of such lectins are the selectins which mediate lymphocyte homing and leucocyte migration into tissues during inflammation. Like the Siglecs, physiological interactions mediated by the selectins are sialic acid dependent. The ligands for selectins contain sialylated, fucosylated and sulfated components (sialyl-lewisx derivatives). Recent structural findings indicate that selectins do not make direct, molecular contacts with sialic acid. Rather, specific interactions of selectins are to the fucose ring and may be supported by nonspecific electrostatic effects by the negatively charged sialic acid to positively charged amino acids of the selectin [32]. In contrast, Siglecs are true sialic acid-binding proteins that form direct molecular contacts with the chemical substituents of sialic acid, namely the carboxylate group, the glycerol side chain and the N-acetyl group [33]. The extracellular region of each Siglec is made up of an N-terminal V-set domain, followed by varying numbers of C2-set domains (Fig. 1). The N-terminal two domains are predicted to be linked by an unusual disulfide bond that is likely to have a major influence on their orientation. However, the significance of this disulfide bond for ligand recognition is unclear because truncation-deletion studies, site-directed mutagenesis experiments, x-ray crystallography and Nuclear Magnetic Resonance (NMR) analysis showed that the sialic acid-binding site is contained exclusively within the N-terminal V-set domain [33-36]. The recent structural determination of the isolated V-set domain of Sn in a complex with α2,3 sialyllactose has directly demonstrated the interaction of Sn to its ligand [33] (Fig. 2A). This structure has provided a template for the sialic acid recognition that is likely to be applicable also to other Siglecs. Structure of the N-terminal domain of sialoadhesin. (A) The V-set domain of Sn in complex with 3′ sialyllactose is shown. The 3′ sialyllactose makes interactions with residues positioned on the A, G and F strands. (B) Structure of the sialic acid binding site in Sn. Positions of key residues near the sialic acid residue in the crystal structure of the N-terminal domain of Sn complexed with 3′ sialylactose are shown. For details see [33]. By solving the crystal structure of Sn complexed to α2,3 sialyllactose it became clear that most of the molecular contacts are occurring with the sialic acid, rather than to the attached sugar units. As shown in Fig. 2(B) an arginine residue (Arg97 in Sn) which is highly conserved in the Siglec family forms a salt bridge with the carboxylate group of the sialic acid. Two well-conserved aromatic groups (both tryptophan for Sn) are involved in the hydrophobic interactions with the N-acetyl and glycerol side chains of the sialic acid [33]. Results of site directed mutagenesis studies carried out both with Sn and CD22 are in good agreement with the crystal structure. In particular, the crucial importance of the conserved arginine has been demonstrated by the finding that even a conservative substitution with lysine leads to an about 10-fold reduction in the binding of Sn and unmeasurable binding in the case of CD22 [34, 35]. Recent studies of Sn–sialoside interactions by NMR have provided evidence that the molecular interactions characterized in crystals also occur in solution and have demonstrated that the affinity of Sn for sialosides is very low, around 10−3 M [36]. The very low affinity of Sn and other Siglecs could be important in their ability to mediate cell–cell interactions in plasma, because, despite being abundant, poorly clustered glycans on plasma proteins would not be expected to compete efficiently with the highly clustered sialic acids present on cell surfaces. Because sialic acids are abundantly expressed on cell surfaces, it could be expected that Siglecs are involved in multiple cellular interactions. However, while all of the Siglecs studied so far can mediate sialic-acid dependent binding to cells as purified proteins, there are striking differences in their ability to do this when expressed on cell surfaces. This was first noted for CD33, which when expressed transiently on monkey COS-1 (COS) cells, required sialidase treatment to promote binding to cells carrying sialylated ligands [6]. When CD22 was expressed in COS or Chinese hamster ovary (CHO) cells, high levels of cellular binding were observed [3, 5]. However these cells lack ST6GalI, the sialyltransferase that creates ligands recognized by CD22. When ST6GalI was coexpressed with CD22 in CHO or COS cells, all CD22-dependent binding activity was lost. The binding could be restored by sialidase treatment of the transfected cells [20, 37]. These observations raise the question of whether CD22 on B lymphocytes can mediate cell–cell interactions because these cells usually express high levels of ST6GalI [38]. It was recently shown both for human peripheral blood lymphocytes as well as for mouse lymphocytes of peripheral organs that the Sia-binding activity of CD22 was largely undetectable. After activation of human B cells, CD22-dependent binding of α2,6 sialylated ligands could be demonstrated [39]. This unmasking effect may be owing to reduced cell-surface expression of 2,6 Sia resulting from either an increased activity of sialidase or a decreased expression of ST6GalI. The findings with CD33, CD22 and other Siglecs provide a possible explanation why Sn has evolved the large number of 17 Ig domains. This extended structure may position the sialic acid binding site away from the plasma membrane, in order to avoid the cis inhibitory sialic acids present on the same cellular surface (Fig. 3). Potential cis-inhibition of Siglec binding sites. Diagram illustrating how the extension of the sialic acid-binding site (N-terminal V-set domain) of Siglecs away from the plasma membrane could be important in the regulation of cell–cell interactions. According to this model, Siglecs such as Sn that extend the sialic acid-binding site a long distance from the plasma membrane are likely to be less prone to cis-inhibition by sialic acids present on the plasma membrane. In comparison, the binding site of CD33 is thought to be completely masked owing to its relatively small size. CD22 is one of a group of transmembrane proteins on the B-cell surface which modulates the B-cell antigen receptor (BCR) signal. CD22 is associated in low stochiometry with the BCR on the cellular surface [40, 41]. After BCR engagement, CD22 is tyrosine-phosphorylated on its cytoplasmic tail by the Src-like kinase Lyn [42, 43]. This leads to binding of the phosphatase SHP-1 to the three ITIM motifs of the CD22 tail and subsequent activation of the phosphatase [42]. Other signalling molecules such as Syk, PLCγ, PI3K, Grb-2 and Shc can also bind to the phophorylated CD22 tail [18, 44, 45]. The binding sites for these factors are partially distinct, partially overlapping with SHP-1. By generation of CD22-deficient mice, we and others have shown that CD22 has an inhibitory function on the BCR signal [46-49]. The Ca2+ flux of CD22−/– B cells was greatly enhanced and prolonged after the BCR stimulation, when compared to wild-type cells. Accordingly, B cells of CD22-deficient mice showed a (mildly) activated phenotype: a higher percentage of mature IgDhiIgMlo cells, two-fold higher expression of major histocompatibility complex (MHC) class II, two-fold higher secretion of IgM, higher responsiveness to lipopolysaccharide (LPS) and development of autoantibodies at older age [50]. CD22 seems to be important for controlling the signalling threshold, particularly at an early B-cell activation. Thus, it could control processes such as apoptosis induction and negative selection [46, 51]. However, when T-cell help and costimulation is provided, CD22−/– B cells respond normally, as shown by immune responses, germinal centre formation and distribution of Igs of all switched isotypes in CD22-deficient mice [46-49]. While the SHP-1-mediated inhibitory function of CD22 was evident in CD22-deficient mice, the role of the other, largely positively regulating, molecules binding to the cytoplasmic tail is unclear and still a matter of dispute. In primary mouse B cells, the SHP-1 is clearly the most prominent protein binding to the tyrosine-phosphorylated CD22 tail, while many of the other proteins which have been demonstrated to bind in vitro, are hardly detectable (Gerlach and Nitschke, unpublished). CD22 regulates the BCR-mediated signals in concert with the costimulatory complex CD19/CD21/CD81 [52]. It is still quite unclear by what mechanism CD22 associates to the BCR and what the role of the lectin domain is. Tooze et al. [53] have demonstrated three possible states on the B-cell surface, which result in different levels of MAP kinase activation: a forced coligation of CD22 with the BCR which gave the lowest signal; stimulation of the BCR alone resulting in an intermediate level of MAP kinase activation; and finally sequestering CD22 away from the BCR complex by pretreating with anti-CD22 coated beads, which gave the highest MAP kinase stimulation. These results suggest that the association of CD22 to the BCR could be a regulated process. The regulation could come from the sialic acid binding adhesion domain of CD22, which by binding to glycans on the cell surface could control the association to the BCR. Thus, the sialic acid expression level would determine the degree of signalling. The open question is: does a higher degree of sialylation sequester CD22 away from the BCR or facilitate the binding to surface Ig? The role of the lectin domain of CD22 for in vivo regulation of signalling can best be studied by the generation of a mouse line with a mutated adhesion domain of CD22. A ST6GalI-deficient mouse line, with no 2,6Sia expressed on the B-cell surface owing to the deficiency of the responsible sialyltransferase, didn't clarify the issue. These mice had a more severe immunodeficiency than CD22-deficient mice, suggesting a more general function ST6GalI in regulating B-cell responsiveness [38]. An unexpected consequence of the loss of the CD22 receptor from the surface of B cells was the specific reduction of mature (IgDhi) B cells in the bone marrow of CD22-deficient mice, despite a normal number of these cells in peripheral lymphoid organs. Mature, resting B cells are the product of a final B-cell maturation step which takes place in the spleen. These cells then recirculate through the blood and migrate back to tissues, amongst them bone marrow. The first evidence that the deficiency of this B-cell population in the bone marrow of CD22−/– mice was caused by a defective migration rather than by, e.g. a decreased life span of CD22−/– B cells was obtained by adoptive transfer studies. The bone marrow of RAG2−/– mice was two-fold higher repopulated with injected wild-type cells than with CD22-deficient B cells [46]. If the CD22 was directly involved in bone marrow homing of mature B cells, we expected that ligands for CD22 should be expressed in this tissue. This was demonstrated by using a recombinant CD22-Fc protein for the staining of tissue sections. Sialylated ligands for CD22 were expressed on endothelial cells of sinusoids in the bone marrow, but not on endothelial cells in other tissues examined [54]. In vivo injection of CD22-Fc protein into mice revealed that these ligands were accessible to the circulation. A single injection of CD22-Fc protein led to a 50% reduction in mature recirculating cells in the bone marrow, without affecting B-cell numbers in the spleen [54]. These results suggested that CD22 may act as a homing receptor for the bone marrow by allowing interaction of B cells with CD22 ligands on endothelial cells and subsequent transmigration into the parenchyma (Fig. 4). Expression of CD22 ligands is regulated by α2,6 sialyltransferases, especially ST6GalI. However, ST6GalI-deficient mice showed a normal composition of B cells in the bone marrow, including mature B cells [38]. This unexpected result could be owing to the expression of an alternative α2,6 sialyltransferase in bone marrow (BM) endothelial cells or to the use of compensatory mechanisms in ST6GalI-deficient mice. Diagram showing the potential role of CD22 in mediating homing of B-cell subsets to the bone marrow. B-cell development from preB to transitional B cells occurs in the haemopoetic tissue of the bone marrow. Transitional B cells leave the tissue and migrate via the blood stream into the periphery. This process is CD22-independent. Finally matured immunoglobulin (Ig)D+ B cells circulate back to the bone marrow. This IgD+ population contains two subsets: one carries a ‘lectin-competent’ CD22 and may engage sialylated ligands on bone marrow, thereby initiating transmigration into the haemopoetic spaces. The other subset B cells expressing CD22 whose binding site is masked by the endogenous α2,6-linked Sias. These would not be expected to interact with sialylated ligands on bone marrow endothelium and hence remain in the circulation. CD22 can mediate adhesion to endothelial target cells only if it is not masked by α2,6-linked sialic acids which are coexpressed on the B-cell surface (Fig. 3). Both human and mouse peripheral resting B cells express a large amount of α2,6-linked sialic acids on the surface which leads to the inhibition of the lectin function of CD22 in cis[38, 39]. We recently carried out a systematic analysis of the masking status of CD22 on the surface of B cells in various organs [55]. For this purpose we used N-acetyl α2,6 sialyllactose coupled to biotinylated polyacrylamide as a synthetic CD22 ligand. We found that this probe bound to a small fraction of peripheral IgD+ B cells (which are the B cells which have the highest level of CD22) and that the binding was CD22-specific, as B cells from CD22-deficient mice were not stained with the probe. While only 2–5% of the IgD+ B cells in the spleen or mesenteric lymph nodes bound this synthetic ligand, the fraction of IgD+ B cells with unmasked CD22 was increased in the bone marrow two- to five-fold [55]. This means that a new subset of mature B cells in the mouse with unmasked CD22 has been identified. This subset can potentially bind to CD22 ligands on bone marrow sinusoidal endothelium, leading to its selective homing and enrichment in this organ (Fig. 4). Unmasking of CD22 could be the result of either decreased activity of sialyl-transferases or activation of sialidases. The question remains what could be the biological relevance of a specific bone marrow homing mechanism for mature B cells? It is hard to envisage a specific function of the recirculating pool of resting B cells in the bone marrow. However, this population may contain an activated subset of B cells. It is relevant in this respect that unmasking of CD22 has been demonstrated to occur after B-cell activation in vitro[39]. Additionally, in humans the IgD+ B-cell pool of the bone marrow contains a subpopulation of memory B cells with a high number of somatic mutations in their rearranged Ig genes [56, 57]. Because CD22-deficient mice have a reduced number of both IgD+ and plasma cells in the bone marrow while these cells accumulate in the spleen [54], it is conceivable that both populations use the same CD22-dependent homing mechanism. Although murine plasma cells have lost CD22 expression, plasma cell precursors may still express this adhesion receptor on the surface. The bone marrow seems to be an optimal environment particularly for long-lived plasma cells. If a specific bone marrow homing mechanism exists for plasma cells, it may also involve adhesion receptors other than CD22. It has been shown that bone marrow sinusoidal endothelium constitutively expresses adhesion receptors such as VCAM-1 and E-selectin [58, 59], that are induced on endothelium in other organs by inflammation. The expression of these adhesion receptors has been shown to be involved in the homing of haematopoetic precursor cells to bone marrow [60] and may also contribute to the homing of mature B cells and plasma cells in addition to CD22 ligands. CD22 is the first Siglec for which a specific homing function has been identified. It will be interesting to see whether similar adhesion functions can be defined also for the new members of the Siglec family. We thank Drs A. May, Y. Jones and K. Drickamer for their help in preparing Figure 2. The work in the corresponding author’s laboratory was supported by grants from the Human Frontier Science Program and the Deutsche Forschungsgemeinschaft. Since completion of this manuscript another human CD33-related Siglec, Siglec-10, has been characterized [A1]. In addition, both this report [A1] and another study [A2] have documented the existence of an alternatively spliced longer form of Siglec-8 that contains the ITIM and SLAM-like Tyr motifs illustrated in Fig. 1 of this review. A1. Munday J, Kerr S, Ni J, et al. Identification, characterization and leucocyte expression of a novel human Siglec. Biochem J, in press. A2. Foussias G, Yousef GM, Diamandis EP. Molecular characterization of a siglec8 variant containing cytoplasmic tyrosine-based motifs, and mapping of the siglec8 gene. Biochem Biophys Res Commun 2000;278:775–81.