The pioneering studies of Alan F. Williams and colleagues in the 1980s1 laid the foundations for our current understanding of the roles of the immunoglobulin superfamily (IgSF) molecules as recognition molecules in the immune system. The immunoglobulin domain is a highly versatile fold that can be used to bind an almost infinite array of molecular structures, as illustrated by antibodies and T-cell receptors. The 'immunoglobulin-type' (I-type) lectins are a discrete subset of the IgSF that exploit the remarkable structural diversity of glycans in their recognition functions2,3. The Siglecs (sialic-acid-binding immunoglobulin-like lectins) are the best characterized I-type lectins2,3,4,5,6,7,8,9. They are type 1 membrane proteins displaying an amino-terminal V-set immunoglobulin domain that binds sialic acid and variable numbers (16 in the case of sialoadhesin) of C2-set immunoglobulin domains. They are categorized into two subsets on the basis of their sequence similarity and evolutionary conservation. Sialoadhesin (also known as Siglec-1 and CD169), CD22 (also known as Siglec-2), myelin-associated glycoprotein (MAG; also known as Siglec-4) and the recently discovered Siglec-15 (T. Angata, personal communication) are quite distantly related (25–30% sequence identity) and have clear orthologues in all mammalian species examined. In comparison, the CD33-related Siglecs share 50–99% identity but seem to be evolving rapidly by multiple processes, including gene duplication, exon shuffling, exon loss and gene conversion10 (Box 1). This has resulted in important differences in the repertoires of CD33-related Siglecs among mammalian species. Initial analyses of the genomes of fish, amphibians and birds indicate that, whereas typical CD33-related Siglecs are absent, a clear orthologue of MAG is present in all three taxa11. In humans, there are nine CD33-related Siglecs and one Siglec-like protein, whereas in mice there are five CD33-related Siglecs. So, it is difficult to assign orthologues, which has required the use of different numbering systems for the human and mouse CD33-related Siglecs (Fig. 1).

Figure 1: Siglec-family proteins in humans and rodents.
figure 1

Sialic-acid-binding immunoglobulin-like lectins (Siglecs) are type 1 membrane proteins containing an amino-terminal V-set immunoglobulin domain that mediates sialic-acid recognition and varying numbers of C2-set immunoglobulin domains. Siglecs can be divided into two groups based on sequence similarity and evolutionary conservation. The CD33-related Siglecs differ in composition between species, share high sequence similarity in their extracellular regions and frequently contain conserved tyrosine–based signalling motifs in their intracellular domains. By contrast, there are orthologues, in all mammals examined, of sialoadhesin, CD22, myelin-associated glycoprotein (MAG) and Siglec-15 and they exhibit lower sequence similarity. Siglec-14 and Siglec-H have been shown to associate with DAP12 through a charged residue in their transmembrane domains. Electron microscopy has shown that purified sialoadhesin is a tadpole-shaped molecule of 40 nm in which the carboxy-terminal immunoglobulin-like domains are arranged into a globular shape128. The cell-expression patterns are indicated on the basis of antibody labelling, except for Siglec-14 and Siglec-15, which have not yet been determined. The brackets indicate low levels of expression. Siglec-13 is present in baboons and chimpanzees and is specifically deleted in humans. Siglec-12 in humans has lost the ability to bind sialic acids and is, hence, designated as Siglec-XII (not shown). B, B cells; Ba, basophils; cDCs, conventional dendritic cells; Eo, eosinophils; GRB2, growth-factor-receptor-bound protein 2; ITIM, immunoreceptor tyrosine-based inhibitory motif; Mac, macrophages; Mo, monocytes; MyP, myeloid progenitors; N, neutrophils; ND, not determined; NK, natural killer cells; OligoD, oligodendrocytes; pDCs, plasmacytoid dendritic cells; Schw, Schwann cells; Troph, trophoblasts.

With the exception of MAG and Siglec-6, Siglec expression has been found mainly in the haematopoietic and immune systems (Fig. 1). Some are expressed in a highly cell-type-restricted pattern. For example, sialoadhesin is a macrophage-specific adhesion molecule and CD22 is a well-characterized B-cell inhibitory receptor that regulates multiple B-cell functions, including cellular activation thresholds and survival. In general, the CD33-related Siglecs show more complex expression patterns in the innate immune system in both humans and mice (Fig. 1). With the exception of resting T cells, most cell types in the human and mouse immune systems express at least one Siglec and others express several. Each Siglec has a unique specificity for sialylated ligands, making it more probable that each protein mediates a distinct, partially overlapping function. CD22 and most CD33-related Siglecs have one or more cytosolic immunoreceptor tyrosine-based inhibitory motifs (ITIMs). Classically, receptors with ITIMs function as inhibitory receptors and suppress activation signals that emanate from receptors associated with immunoreceptor tyrosine-based activation motifs (ITAMs) through the recruitment of tyrosine and inositol phosphatases12. By contrast, mouse CD33 and Siglec-H, and human Siglec-14 and Siglec-15 lack ITIMs. Two of these non-ITIM-containing Siglecs (Siglec-H and Siglec-14) have been shown to interact with DAP12 (also known as KARAP)13,14, an ITAM-containing adaptor that triggers both activating and inhibitory signalling15. Siglec-13, which is present in other primates but is specifically deleted in humans9, also seems to recruit DAP12 (A.V., unpublished observations).

Taken together, it is clear that Siglecs in the immune system have the potential to mediate both cell–cell interactions and signalling functions. However, defining their precise functions and determining which ligands are biologically relevant pose an important challenge. This is beginning to be tackled using a combination of experimental approaches, including the production of genetically manipulated mice, biochemical analyses of ligand recognition and dissection of signalling pathways. In particular, several recent studies using mice that lack CD22, CD22 ligands or both, as well as mice expressing mutant forms of CD22 that cannot bind sialylated glycans, have begun to shed light on the complex factors involved. These have also provided a conceptual framework for understanding how the less well-characterized CD33-related Siglecs may contribute to regulation of leukocyte functions, as revealed in a recent study of Siglec-F-deficient mice107. Sialoadhesin (recognized by the antibody MOMA-1) is well known as a macrophage-specific marker and adhesion molecule but its biological functions have remained enigmatic. However, several recent studies of sialoadhesin-deficient mice have shown an unexpected role of this receptor in modulating immune and inflammatory responses. New data are available on the endocytic functions of Siglecs and their interactions with various sialylated pathogens that could be important in both host defence and pathogenicity. Finally, there is emerging evidence that CD33-related Siglecs have undergone significant changes during human evolution. In this Review, we discuss how these recent advances have significantly furthered our understanding of the roles of Siglecs in the immune system and wherever possible we attempt to relate these functions to glycan recognition and physiology.

Sialic-acid recognition by Siglecs

The mammalian glycome contains numerous sialylated glycans that can be potentially recognized as ligands by Siglecs (Box 2). It is assumed that this recognition is important for modulating the functions of Siglecs as regulators of adhesion, cell signalling and endocytosis5,6,7,8,9. In general, Siglecs show low affinity (a Kd of 0.1–3 mM) for the sialic acid N-acetylneuraminic acid (Neu5Ac) α2–3 and α2–6 linkages to galactose (Neu5Acα2–3Gal and Neu5Acα2–6Gal) (Fig. 2) that are commonly found as terminal sequences on glycans of glycoproteins and glycolipids of most mammalian cells16,17, and Siglecs have an overlapping specificity for such sialosides (sialic-acid-containing glycans). However, when examined for their ability to recognize a diverse set of natural sialoside structures found in mammalian species, each Siglec shows a characteristic specificity profile (Fig. 3). CD22 is unique in having a strong preference for Neu5Acα2–6Gal and Neu5Gcα2–6Gal structures18,19. Siglec-7 and Siglec-11 prefer sialosides with the Neu5Ac(α2–8)Neu5Ac structure20,21.

Figure 2: Sialic acids and their interactions with Siglecs.
figure 2

a | Sialic acid in mammals is commonly modified at the R and R′ positions with the substituents indicated. There are many other kinds of sialic acids in vertebrates, but apart from the commonly found forms shown here, Siglec (sialic-acid-binding immunoglobulin-like lectin) interactions with these have not been studied systematically. b | Two common sialoside sequences recognized as low-affinity ligands by many Siglecs are shown. c | Structure of the V-set domain of sialoadhesin (gold and grey) complexed with Neu5Acα2–3Gal (shown with turquoise carbons). Side chains of amino acids that make direct contact with the bound ligand are shown with purple carbons. The C–C´ loop that is implicated in fine specificity of Siglec binding is shown in orange. d | Structure of the V-set domain of Siglec-7 (gold and grey) complexed with Neu5Ac (shown with turquoise carbons). Side chains of amino acids that make direct contact with the bound ligand are shown with purple carbons. The C–C´ loop that is implicated in fine specificity of Siglec binding is shown in orange. Gal, galactose; Neu5Ac, N-acetylneuraminic acid; Neu5Gc, N-glycolylneuraminic acid. Images in parts c and d are reproduced with permission from Ref. 28 © (2003) American Society for Biochemistry and Molecular Biology.

Figure 3: Relative affinities of Siglecs with selected sialoside sequences.
figure 3

A comparison of the sialoside specificities, obtained from glycan array analysis and competitive binding experiments, for several sialic-acid-binding immunoglobulin-like lectins (Siglecs) against selected sialoside sequences found in mammalian glycoproteins and glycolipids17,22,23,24 (see also the Functional Glycomics web site). Despite the general low affinity of Siglecs towards the common mammalian sialoside structures containing the N-acetylneuraminic acid (Neu5Ac) α2–6 and α2–3 linkages, examination of the binding of each individual Siglec to a range of common sialoside structures shows that there is a unique sialoside specificity profile that is characteristic to each Siglec. +++, strong binding; ++, moderate binding; +, low binding; +/−, detectable binding; 0, no detectable binding; GlcNAc, N-acetylglucosamine; ND, not determined; Neu5Gc, N-glycolylneuraminic acid.

Several Siglecs (such as Siglec-8 and Siglec-9) have differential specificity for sialosides that contain both sialic acid and sulphate, with the position of the sulphate being an important determinant of specificity22,23,24,25. In addition, Siglecs have different specificities for the many sialic acid species found in nature (Fig. 2). Of particular interest is the evolutionary loss of N-glycolylneuraminic acid (Neu5Gc) in humans, as Neu5Gc is the preferred ligand for at least some Siglecs in the closely related great apes26.

The ligand-bound crystal structures of sialoadhesin and Siglec-7 offer insights into the structural basis for Siglec specificity27,28,29,30,31. In both cases, sialic acid is bound to a shallow pocket in the N-terminal immunoglobulin domain of the Siglec, with the negatively charged carboxyl group of sialic acid forming an important salt bridge with a conserved essential arginine (Fig. 2). As shown in domain-swapping experiments20, the highly variable C–C´ loop (residues 70–75 in Siglec-7) of the V-set domain is a key determinant of Siglec specificity. In the crystal structure of Siglec-7 complexed with the Neu5Acα2–8Neu5Ac-containing ganglioside known as GT1b, the C–C´ loop undergoes a marked conformational change30. Although binding between natural sialosides and Siglecs is relatively weak, increased binding affinity has been seen for sialoadhesin, MAG and CD22 using sialic acid analogues that have aromatic substituents at the C-9 position of the monosaccharide29. These have been exploited for the creation of high-affinity ligand-based probes of Siglec function29,31,32,33,34,35.

Cis and trans ligands. The local concentration of sialic acids on surfaces of immune cells is very high; for example, on B cells it has been estimated to exceed 100 mM36. As a consequence of this, and the fact that Siglecs usually recognize sialoside sequences that are commonly found on mammalian cells, Siglec binding sites are typically 'masked' by cis interactions with other glycan ligands expressed on the same cell37,38,39,40. This indicates that interactions with cis ligands may dominate over interactions with trans ligands in modulating the biological activities of Siglecs7,9 (Fig. 4). One exception to this rule is sialoadhesin, which owing to its extended structure is thought to project its sialic-acid-binding site away from the plasma membrane, which reduces its cis interactions41. Despite the likely importance of cis-ligand interactions in Siglec function, they do not necessarily prevent the binding of ligands in trans. CD22 on B cells can redistribute to the site of contact with another cell expressing CD22 ligands36. Moreover, high-affinity synthetic sialoside probes can out-compete cis ligands for binding to CD22 on native B cells34. These results show that cis ligands downregulate, but do not preclude, binding of ligands in trans, and that equilibrium-based binding of Siglecs to trans ligands can occur dynamically in the presence of cis ligands. The biological significance of CD22 trans-interactions was explored by Lanoue et al. who showed that B-cell activation in response to antigen-presenting cells is suppressed if antigen and the CD22 ligand are expressed on the same cell42. This could be important for preventing autoantibody production by self-reactive B cells.

Figure 4: Cis and trans interactions of Siglecs.
figure 4

a | Most sialic-acid-binding immunoglobulin-like lectins (Siglecs) are masked at the cell surface owing to cis interactions with abundantly expressed sialic acids. Following exposure of cells to sialidase, which cleaves the cis-interacting Siglec ligands, or in some cases following cellular activation, Siglecs become unmasked, which allows them to make interactions with ligands in trans. b | Even when Siglecs are masked by cis interactions, trans interactions might occur during an encounter with another cell or a pathogen expressing higher affinity ligands that can out-compete the cis interactions.

Siglec-7 is a natural killer (NK)-cell inhibitory receptor43,44, which is masked by cis ligands that may include the preferred Neu5Acα2–8Neu5Ac-containing glycan45. In NK-cell cytotoxicity experiments with target cells engineered to overexpress this glycan structure, a small Siglec-7-dependent reduction in NK-cell cytotoxicity was observed and this was increased when NK cells were pre-treated with sialidase, which cleaves the cis-interacting ligands from the cell surface46. Such trans-interactions could contribute to the suppression of NK-cell cytotoxicity in tissues, such as the nervous system, in which the Neu5Acα2–8Neu5Ac-containing glycans are abundantly expressed.

In common with many glycan-binding proteins, high-avidity interactions of Siglecs require clustering of both receptors and ligands. For membrane receptors, clustering can be achieved by lateral diffusion in the plasma membrane or by the formation of multimeric complexes. For the glycan ligands, clustering can occur by presentation on suitable carriers, either protein or lipid, and/or by clustering on another cell surface. The resulting 'counter-receptors' have the potential to be recognized by glycan-binding proteins with high avidity and selectivity47. Although information on the nature of physiologically important ligands and counter-receptors of the Siglec family remains fragmentary, some insights have been obtained for CD22. B cells from mice deficient in the sialyltransferase ST6GAL1, which transfers sialic acids to galactose in α2–6 linkages to produce the preferred ligands for CD22, were devoid of cis ligands, which indicates that this sequence is required for binding by CD22 (Ref. 48). A CD22–immunoglobulin Fc fusion protein has been shown to bind to CD45, serum IgM and many other B-cell (and T-cell) glycoproteins in vitro, and this implicates CD45 and IgM as candidate cis glycoprotein counter-receptors for CD22 on B cells49,50,51. However, using B cells that are metabolically labelled with a sialic-acid analogue containing a photo-activatable crosslinker, Han et al. found that CD45 and IgM are not preferentially recognized as cis ligands on the surface of the intact cell33. Instead, CD22 interacted preferentially in cis with glycans of neighbouring CD22 molecules. This is consistent with earlier crosslinking experiments52,53, which also showed the formation of homomultimeric CD22 complexes involving protein–protein interactions. The basis for this selectivity may stem from the preferential concentration of CD22 in clathrin domains35,54,55. These studies underscore the potential difficulty in establishing the in situ ligands of Siglecs in their native context.

Pathogen ligands and innate immunity. Over 20 pathogenic microorganisms are known to have evolved the capacity to synthesize or capture sialic acids from their hosts and incorporate these into their own glycoconjugates. Examples include well-known bacterial pathogens, such as Neisseria meningitidis, Haemophilus influenzae, group B Streptococcus, Campylobacter jejuni, and several strains of Escherichia coli that cause infant meningitis, septicaemia, respiratory infections, diarrhoea and various other conditions56. Sialylation of glycoconjugates in these pathogens seems to be crucial for their survival in the mammalian host, possibly serving as molecular mimics of host cell surfaces to avoid immune attack57. It is also widely assumed that negatively charged sialic acids could serve to reduce pathogen interactions with the host by electrostatic repulsion, and/or by inhibiting the alternative pathway of complement activation. So far, comparatively little attention has been paid to the possibility that pathogen sialic acids are important for actually promoting infection or mediating an innate immune response by attachment to Siglecs. One clear example of this is porcine reproductive and respiratory syndrome virus (PRRSV), which is a sialylated enveloped virus that infects pig alveolar macrophages in a sialic-acid and sialoadhesin-dependent manner58,59. Sialoadhesin and several CD33-related Siglecs can interact with sialic acids on N. meningitidis, C. jejuni, group B Streptococcus and Trypanosoma cruzi60,61,62,63 (Table 1). Siglec-dependent uptake of these pathogens could potentially benefit the host by promoting pathogen destruction and antigen presentation. It has also been suggested, but not proved, that sialylated pathogens modulate leukocyte activation through ITIM-mediated signalling of CD33-related Siglecs, thereby benefiting the pathogen by dampening inflammatory and immune responses8,60,62,63.

Table 1 Interactions between pathogens and Siglecs

Sialoadhesin as a modulator of immune responses

Sialoadhesin was discovered as a sialic-acid-dependent macrophage adhesion molecule64. It is one of the largest members of the IgSF with an extracellular region made up of 17 immunoglobulin domains65, a feature that is well conserved in mammals. Unlike most Siglecs, sialoadhesin lacks tyrosine-based signalling motifs and its cytoplasmic tail is poorly conserved, which suggests a primary role as a binding partner in cell–cell interactions, rather than in cell signalling.

There is increasing evidence to support a contribution for sialoadhesin in the pro-inflammatory functions of macrophages. Sialoadhesin is constitutively expressed on subpopulations of tissue-resident macrophages66,67, and is rapidly upregulated by inflammatory macrophages. In proliferative glomerulonephritis, the expression of sialoadhesin correlates with proteinuria68,69, and in individuals infected with HIV and that carry high viral loads, the sialoadhesin gene is induced up to tenfold in circulating monocytes70. It was recently shown that under normal conditions, sialoadhesin-deficient mice show only subtle alterations in the haematopoietic and immune systems, but in a model of peptide-induced experimental autoimmune uveoretinitis, the deficient mice showed reduced retinal inflammation, and T cells isolated from draining lymph nodes showed lowered proliferative responses in vitro71. Furthermore, in two genetically determined models of peripheral and central nervous system demyelination, disease in sialoadhesin-deficient mice was ameliorated and numbers of infiltrating CD8+ T cells and macrophages at the sites of inflammation were reduced72,73. These new findings are consistent with a potentially important role of sialoadhesin in modulating T-cell function and activation during immune responses. An additional possibility is that sialoadhesin functions as a phagocytic receptor to clear sialylated pathogens.

Although the cellular and molecular bases for these effects require further investigation, sialoadhesin can mediate both sialic-acid-dependent and sialic-acid-independent interactions with cells of the immune system. Sialic-acid-dependent sialoadhesin interactions might be mediated by mucin-like molecules presenting high densities of sialylated O-linked glycans. For example, the sialomucins CD43 and mucin-1 were previously identified as putative T-cell and breast-cancer cell counter-receptors for sialoadhesin, respectively74,75. Sialic-acid-independent sialoadhesin interactions could involve the mannose receptor and macrophage galactose-type N-acetylgalactosamine-specific lectin 1 (MGL1). Both of these membrane lectins are expressed on dendritic cells (DCs) and have been shown to bind sialoadhesin extracted from lymphoid tissues76,77. Although DCs themselves do not normally express sialoadhesin, it can be induced on human monocyte-derived DCs following exposure to rhinoviruses in vitro78. Interestingly, these DCs were poor stimulators of T cells in mixed lymphocyte reactions, a feature that was partly attributed to the expression of sialoadhesin78. It is possible that when sialoadhesin is expressed by macrophages it is immunostimulatory, whereas on DCs it is immunosuppressive.

Role of CD22 and its ligands

CD22 is a well-documented regulator of B-cell signalling, homeostasis and survival7,8,9,79. This Siglec is best known for helping to set a threshold for antigen-induced activation of B cells80, an activity that involves as many as six tyrosine-based motifs in the cytoplasmic domain of CD22, including three ITIMs. B-cell receptor (BCR) ligation leads to increased phosphorylation of the ITIMs of CD22 by the SRC-family kinase LYN, which results in the recruitment of SHP1 (SRC homology 2 (SH2)-domain-containing protein tyrosine phosphatase 1) and the downregulation of BCR signalling8,79. However, this oversimplifies the complexity of CD22 signalling, as it can also recruit positive effectors of cell activation, including GRB2 (growth-factor-receptor-bound protein 2), SHC (SH2-domain-containing transforming protein C), PI3K (phosphoinositide 3-kinase) and PLCγ2 (phospholipase Cγ2)79. This results in activation of alternative signalling pathways that contribute to the regulation of B-cell activation. The impact of CD22 on these pathways probably depends on the manner of B-cell activation. Ligation of the BCR with either antigen or IgM-specific antibody, or simultaneous ligation of the BCR and CD40 (with IgM-specific and CD40-specific antibodies) result in the differential phosphorylation of CD22 tyrosine-based motifs both quantitatively and qualitatively81,82. Furthermore, CD22 does not seem to affect B-cell signalling when activated by ligation of cell-surface IgG83. A detailed molecular understanding of the role of CD22 will continue to evolve as B-cell signalling pathways become better defined8,79.

Recent work on CD22 has provided important insights into how sialic-acid recognition can modulate its signalling functions. B cells of CD22-deficient mice exhibit hyperimmune responses in vitro and in vivo8,79, consistent with the loss of negative regulation by ITIMs of CD22. Several CD22 functions, including BCR-dependent proliferation and B-cell turnover rates, depend on the ligand-binding function of CD22, as shown using mice that carry knock-in mutations of CD22 that ablate its ability to bind sialic acid84. In contrast to CD22-deficient mice, ST6GAL1-deficient mice (which lack CD22 ligands)35,55,85 exhibit hypo-immune responses. B cells from mice that are deficient in both CD22 and ST6GAL1 behave similarly to those from CD22-deficient mice, which indicates that the immunodeficiency of ST6GAL1-deficient mice depends on the presence of CD22 (Refs 35, 55, 85). Following BCR ligation in vitro, the immunodeficiency caused by the absence of cis ligands in ST6GAL1-deficient mice is manifest by reduced B-cell proliferation and calcium flux, and increased CD22 phosphorylation and recruitment of SHP1 (Refs 35, 55). Grewal et al. recently showed that ablating St6gal1 in the LYN-deficient mouse model of lupus abrogated autoimmune responses55, which indicates the therapeutic implications of inducing a CD22-ligand deficiency.

Insights into how the absence of CD22 cis ligands reduces B-cell signalling have been obtained by assessing the microdomain localization of CD22 in the membrane, relative to that of the BCR35,55 (Fig. 5). In resting B cells, there is minimal colocalization of CD22 with the BCR. Most of the CD22 molecules (80%) are associated with clathrin domains, which is consistent with the presence of cytoplasmic tyrosine-based motifs for the clathrin adaptor protein AP50 (Ref. 54). By contrast, the BCR is minimally associated with clathrin domains. Following B-cell activation, the BCR moves into activation rafts, which subsequently fuse with clathrin domains before they are endocytosed86,87. Because CD22 is excluded from activation rafts, it is probable that CD22 exerts its regulation of BCR signalling in the fused raft/clathrin domains34,55,86,87. In ligand-deficient mice, there is no change in the localization of CD22. However, the amount of cell-surface IgM that is colocalized with CD22 increases twofold (from 20–25% to 40–50%), which is associated with a dramatic shift in the distribution of IgM to the raft/clathrin domains. This shift in BCR localization also results in increased endocytosis and reduced half-life of IgM35,55. The redistribution and reduced half-life of IgM do not occur in mice deficient in both CD22 and ST6GAL1, indicating they are mediated by CD22, and are not due to another effect of the ST6GAL1 deficiency. So, for ligand-deficient mice, the increased colocalization of the BCR and CD22 in fused raft/clathrin domains of resting B cells probably accounts for the reduced BCR signalling mediated by CD22 (Fig. 5). These results also indicate that cis ligands of CD22 reduce BCR localization in raft/clathrin domains of resting B cells of wild-type mice by an as-yet-undefined mechanism.

Figure 5: Cis ligands of CD22 influence microdomain localization of the B-cell receptor.
figure 5

In resting B cells, CD22 is mainly localized in clathrin-coated pits, whereas the B-cell receptor (BCR) shows minimal association with these microdomains. In the absence of CD22-ligand binding (such as in ST6GAL1-deficient mice), the BCR shows increased colocalization with CD22, which is coincident with the reduction of BCR signalling. This mainly occurs in fused raft/clathrin domains that typically form following BCR ligation before endocytosis of the BCR.

In addition to the effect on BCR signalling, deficiency in CD22-ligand binding leads to reduced levels of cell-surface CD22 and IgM, increased apoptosis and B-cell turnover, and a reduction in the number of marginal-zone B cells84. These properties are also observed in CD22-deficient mice, which suggests that they are regulated by CD22–ligand interactions35,55,84,85,88. So, it is probable that other important roles of both cis and trans ligands in CD22 function are likely to be revealed using appropriate mouse models.

Functions mediated by CD33-related Siglecs

The CD33-related Siglecs are mainly expressed by mature cells of the innate immune system, such as neutrophils, eosinophils, monocytes, macrophages, NK cells, DCs and mast cells (Fig. 1). CD33 itself is well known as a marker of myeloid progenitor cells, indicating a potential role for CD33 in the regulation of cellular proliferation and/or differentiation. Other CD33-related Siglecs seem to be expressed at later stages of haematopoiesis89,90,91. Numerous studies point to important roles of CD33-related Siglecs in modulating leukocyte behaviour, including inhibition of cellular proliferation92,93, induction of apoptosis94,95, inhibition of cellular activation96,97,98,99,100, induction of pro-inflammatory cytokine secretion101 and, in the case of Siglec-H on plasmacytoid DCs (pDCs), suppression of interferon-α (IFNα) production13. In general, these functions have been defined using selected Siglecs (Fig. 6) and the extent to which they can be extrapolated to the other CD33-related Siglecs is unknown. The signalling pathways are poorly understood but in most cases are assumed to involve the ITIM and ITIM-like motifs and recruitment of tyrosine phosphatases (Fig. 6). It should be emphasized that many of these studies have been carried out using Siglec-transfected cells and/or antibody crosslinking of cell-surface proteins, and the activities need to be confirmed in more physiological systems.

Figure 6: Signalling mediated by CD22 and the CD33-related Siglecs.
figure 6

On phosphorylation of cytoplasmic tyrosine–based signalling motifs by SRC-family tyrosine kinases, sialic-acid-binding immunoglobulin-like lectins (Siglecs) recruit and activate SRC homology 2 (SH2)-domain-containing proteins, notably the tyrosine phosphatases SHP1 (SH2-domain–containing protein tyrosine phosphatase 1) and SHP2 or the SOCS3 (suppressor of cytokine signalling 3) protein. This initiates a range of functional activities that are indicated for the Siglecs listed. In the case of Siglecs without cytosolic signalling motifs, charge-dependent transmembrane region interactions with DAP12 can provide ITAM (immunoreceptor tyrosine-based activation motif)-based signalling functions that are typically initiated by the recruitment and activation of spleen tyrosine kinase (SYK). ↑, increased; ↓, decreased; IFNα, interferon-α.

An important function of CD33-related Siglecs is their ability to regulate cell growth and survival, either by inhibition of proliferation or induction of apoptosis. Inhibition of proliferation was first shown for CD33 and Siglec-7 using both normal haematopoietic cells and myeloid leukaemic cells92,93 and it was recently confirmed using transfected Ba/F3 cells102,103. With the Ba/F3 cells, inhibition of cytokine-dependent proliferation depended on crosslinking Siglecs with primary and secondary antibodies and required an intact ITIM102,103. Induction of apoptotic and non-apoptotic cell death has been shown for human eosinophils and neutrophils through antibody-induced ligation of Siglec-8 and Siglec-9, respectively94,95. Interestingly, cell-death induction by these Siglecs is enhanced in the presence of cytokines that normally promote survival, which suggests a complex interplay between cytokine receptor and Siglec signalling pathways104. Regulation of ITAM-dependent cellular activation by CD33-related Siglecs has been shown in various cell types, including transfected T cells, mast cell lines and myeloid cell lines. Interestingly, both primary human T cells and Jurkat cells, which normally lack significant levels of CD33-related Siglecs, showed inhibition of T-cell-receptor-dependent activation following overexpression of Siglec-5, Siglec-7 or Siglec-9 (Refs 99, 105). In comparison, chimpanzee T cells express several CD33-related Siglecs and this may account for their lower proliferative responses compared with human T cells105.

The most direct evidence for a role of CD33-related Siglecs in the modulation of leukocyte functions has been provided by studies of mice lacking Siglec-F107, which is a functionally convergent paralogue of human Siglec-8 and is expressed by eosinophils23,106 and by activated T cells. In a model of induced lung allergy, these mice show increased bone-marrow, blood and tissue eosinophilia, which is consistent with an inhibitory role of CD33-related Siglecs in controlling leukocyte expansion during inflammatory responses107. Taken together, these data indicate the possible existence of a negative-feedback loop that controls allergic responses of eosinophils and helper T cells, through Siglec-F and upregulated expression of Siglec-F ligands in the inflamed tissue.

CD33-related Siglecs can also function as endocytic receptors that could be important in the clearance of sialylated antigens and/or in promoting or inhibiting antigen presentation90,91,108,109,110. Their endocytic capacity can also be exploited for therapy. For example, gemtuzumab ozogamicin (Mylotarg; UCB and Wyeth-Ayerst Laboratories (Wyeth)) is a humanized CD33-specific monoclonal antibody coupled to the potent antibiotic calicheamicin-γ1, which is currently being used for the treatment of relapsed acute myeloid leukaemia (AML) following chemotherapy. Recent screens of CD33-related Siglec expression by AML cells have shown that several are expressed at variable levels and may therefore provide additional targets for the treatment of haematological diseases90,91.

Signalling by CD33-related Siglecs. Most CD33-related Siglecs have two conserved cytoplasmic tyrosine-based motifs, comprising a membrane-proximal ITIM and a membrane-distal ITIM-like motif (consensus sequence (Glu/Asp)-Tyr-X-Glu-(Val/Ile)-(Arg/Lys); where X denotes any amino acid) (Fig. 1). The distal motif was originally described111 as being similar to a signalling motif in the SLAM (signalling lymphocytic activation molecule) family112 of receptors that recruit SAP (SLAM-associated protein) and/or its homologue EAT2 (Ewing's sarcoma-associated transcript 2). However, more recent studies on the consensus sequence for SAP and EAT2 binding (Thr-Ile-Tyr-X-X-(Val/Ile))112 make it seem unlikely that CD33-related Siglecs interact with these adaptors. Mutagenesis experiments with CD33-related Siglecs have shown that the ITIM dominates over the ITIM-like motif, both for the recruitment of SHP1 and SHP2 and for inhibitory signalling functions96,97,98,100,113,114. However, the ITIM-like motif was required for optimal recruitment of SHP1, but not of SHP2, and could therefore be important in fine-tuning downstream signalling from CD33-related Siglecs.

The ITIMs of CD33-related Siglecs are important for other functions, including the suppression of Siglec-dependent adhesion to sialylated ligands and endocytosis91,98,100,110,113. The degree of tyrosine phosphorylation is likely to be crucial in determining which function predominates. Robust binding to SHP1 and SHP2 requires tyrosine phosphorylation of both the ITIM and ITIM-like motif. However, following mutation of both tyrosines to alanine, Siglec-5 was capable of weak SHP1 binding and activation and was still a potent inhibitor of cellular activation100. By contrast, mutation of the ITIM alone reversed the suppression of Siglec-dependent adhesion and this correlated with the loss of SHP2 binding98,100.

Orr et al. have recently shown that suppressor of cytokine signalling 3 (SOCS3) can bind efficiently to the phosphorylated ITIMs of CD33 and Siglec-7 and compete with SHP1 and SHP2 for ITIM-dependent inhibition of cytokine-induced proliferation102,103. Recruited SOCS3 was shown to function as an E3 ligase, leading to proteasome-dependent degradation of both Siglec and SOCS3. As SOCS3 is a key negative regulator of cytokine signalling in myeloid cells115, CD33–SOCS3 interactions could affect both SOCS3 and Siglec-dependent inhibitory functions. Taken together, these findings indicate a finely balanced mechanism of leukocyte activation and growth regulation involving CD33-related Siglecs, SHP1, SHP2 and SOCS3 proteins (Fig. 6).

Siglec-H is coupled to DAP12. Siglec-H is a CD33-related Siglec that is expressed by mouse and rat pDCs13,109,116. This receptor can mediate endocytosis and cross-presentation of antigens109 and can also function as a negative regulator of IFNα production by pDCs13. This observation is surprising because Siglec-H lacks ITIM-like motifs and depends on the presence of the ITAM-containing adaptor DAP12 for cell-surface expression13. However, there is growing evidence that under certain circumstances ITAM-associated receptors can mediate inhibitory signalling through poorly characterized mechanisms15. So far, Siglec-H has not been shown to bind sialic acids109, and the rat orthologue has a mutation in the arginine residue that is required for sialic-acid binding10. Strictly speaking, therefore, Siglec-H may not fulfil the criteria to be defined as a bona fide Siglec117. One interesting possibility is that this pDC-restricted receptor has evolved from sialic-acid binding to function as a pattern-recognition molecule that binds viral or other pathogen ligands and delivers them to endosomal compartments for subsequent triggering of Toll-like receptor (TLR)-dependent cytokine responses and antigen presentation118. So far, there are no reports addressing the possibility that the more conserved C2-set immunoglobulin domains of CD33-related Siglecs might mediate protein–protein interactions that are important in immune functions.

Siglec-5 is paired with Siglec-14, a DAP12-coupled receptor. Several inhibitory receptors of the immune system are paired with activating counterparts that share highly related extracellular domains but differ in the transmembrane and cytoplasmic regions. Paired activating receptors often have a positively charged residue within the transmembrane region that is required for binding to the ITAM-containing adaptors, such as DAP12 and the Fc receptor γ-chain, or to the DAP10 adaptor that contains the sequence Tyr-X-X-Met (where X is any amino acid), which recruits PI3K119. As shown for the NK-cell receptor Ly49H, which binds the MHC-like protein m157 of mouse cytomegalovirus120,121, some paired activating receptors may have evolved as a counter-strategy to bind host-derived pathogen ligands. Although Siglecs were not previously thought to include paired receptors, recent data indicate that the inhibitory receptor Siglec-5 is paired with a putative activating receptor, Siglec-14, and both seem to be expressed by myeloid cells in a coordinated manner14. These proteins share more than 99% sequence identity in their first two immunoglobulin domains and then diverge in the rest of their coding sequences. Siglec-14 has an arginine residue in its transmembrane region that is required for its association with DAP12 (Ref. 14). Siglec-5 and Siglec-14 would therefore be expected to deliver opposing signals through ITIM- and ITAM-dependent pathways, respectively. A different arginine residue (of the first immunoglobulin domain) that is required for sialic-acid recognition by both Siglec-5 and Siglec-14 is present in humans (but is mutated in almost all great-ape alleles), indicating that these two proteins may work in a cooperative manner, balancing activating and inhibitory signalling through sialic-acid recognition14. Indeed, repeated gene-conversion events occurred between the 5′ regions of the Siglec5 and Siglec14 genes in each primate lineage, assuring maintenance of a paired receptor status for sialic-acid binding.

Concluding remarks

Siglecs are emerging as important regulators of the immune system. What links this family of proteins at a molecular level is their ability to bind sialic acids in a range of glycoconjugates, both in cis and in trans. Challenges for the future are to understand in precise terms which ligands and counter-receptors are important for mediating the biological functions of Siglecs, to elucidate the role of sialic-acid recognition in Siglec biology and to dissect the signalling pathways that are triggered. Siglecs are also recognized as endocytic receptors and this function is likely to be regulated by the same tyrosine motifs that regulate signalling. The interplay between the regulation of receptor signalling and endocytosis may prove to be a fruitful area of investigation for the future. A related challenge will be to understand the impact of pathogen sialylation on Siglec-mediated host immune responses, an issue that may give insights into the evolutionary pathways that have led to the diversification of this family. With growing data linking inhibitory receptor polymorphisms and autoimmune disease, studies of Siglec polymorphisms among human populations are clearly warranted. Finally, the development of Siglec-specific agonists and antagonists may provide new approaches to the treatment of certain autoimmune and inflammatory conditions.