Chapter 32I-type Lectins

Varki A, Crocker PR.

Publication Details

I-type lectins are glycan-binding proteins that belong to the immunoglobulin superfamily (IgSF), excluding antibodies and T-cell receptors. Bioinformatics analyses of mammalian genomes predict more than 500 proteins of the IgSF, other than antibodies and T-cell receptors. Thus, there is considerable potential for assignment to the I-type lectin family. In fact, the Siglec family of sialic acid–binding lectins is the only well-characterized group of I-type lectins, both structurally and functionally. These proteins are thus the major focus of this chapter. Details of their discovery, characterization, binding properties, and biology are provided, along with discussions of their functional implications in mammalian biology.


Members of the IgSF must contain at least one immunoglobulin (Ig)-like fold, but often contain other structural features such as fibronectin type III repeats. The Ig fold was first discovered in antibodies and is made up of antiparallel β-strands organized into a β-sandwich containing 100–120 amino acids and usually stabilized by an intersheet disulfide bond. Three types or “sets” of Ig domains have been defined on the basis of similarities in sequence and structure to the domains of antibodies: the V-set variable-like domain, the C1- and C2-set constant-like domains, and the I-set domain that combines features of both V- and C-set domains.

Prior to the 1990s, it was thought that antibodies were the only IgSF members capable of recognizing glycans. The first direct evidence for non-antibody IgSF glycan-binding proteins was from independent studies on sialoadhesin (Sn), a sialic acid (Sia)-dependent binding receptor on certain mouse macrophage subtypes, and on CD22, a molecule already previously cloned as a B-cell marker. A variety of techniques were used to show that Sn functions as a lectin, including loss of binding following sialidase treatment of ligands, inhibition assays with sialylated compounds, and Sia-dependent binding of the purified receptor to glycoproteins and to red blood cells derivatized to carry Sias in α2-3 linkages. In the case of recombinant CD22, loss of cell adhesive interactions caused by sialidase treatment led to the discovery that it was a Sia-binding lectin, with a high degree of specificity for α2-6-linked Sias. The cloning of Sn then showed that it was an IgSF member that had homology with CD22 and with two other previously cloned proteins, CD33 and myelin-associated glycoprotein (MAG). Demonstration of Sia recognition by CD33 and MAG resulted in the definition of a new family of Sia-binding molecules, which were initially termed the “sialoadhesins.” Meanwhile, evidence for glycan binding by additional IgSF members had emerged, and a suggestion was made to classify all of these molecules as “I-type” lectins. However, it became clear that the Sia-binding molecules were a distinct group sharing both sequence homology and Ig domain organization and that they were not all involved in adhesion. The term Siglec (sialic acid–binding, immunoglobulin-like lectin) was therefore proposed in 1998. Subsequently, most of the CD33-related Siglecs (CD33rSiglecs) were discovered as a direct result of the large-scale genomic sequencing projects, which allowed in silico identification of novel Siglec-related genes and cDNAs.


Several IgSF members other than Siglecs have been claimed to bind glycans, but in many cases, the evidence is indirect. The best evidence is probably for L1 cell-adhesion molecule (L1-CAM) in the nervous system. However, further studies have indicated that this molecule does not bind Sias through an IgSF domain and it therefore does not strictly qualify as an I-type lectin. The neural cell-adhesion molecule (NCAM) has been claimed to recognize and bind oligomannose-type glycans on adjacent glycoproteins in the nervous system. Similar findings have recently been reported for another IgSF molecule called basigin. The cell-adhesion molecule ICAM-1 has been shown to bind hyaluronan and possibly certain mucin-type glycoproteins. Hemolin is an IgSF plasma protein from lepidopteran insects that binds lipopolysaccharide (LPS) from Gram-negative bacteria and lipoteichoic acid from Gram-positive bacteria. Hemolin appears to have two binding sites for LPS, one that interacts with the phosphate groups of lipid A and another that interacts with the O-specific glycan antigen and the outer-core glycans of LPS. There is indirect and less convincing evidence for interactions of other IgSF molecules with glycans, such as P0 with HNK-1, CD83 with Sias, PILR with Sias on CD99, and CD2 with Lewisx. Further studies are needed to ascertain whether these are indeed I-type lectins. The rest of this chapter will be devoted to the Siglecs, which are the best-characterized I-type lectins.


The Siglecs can be divided into two major subgroups based primarily on sequence similarity (Figure 32.1) and on conservation between different mammalian species. The first group comprises Sn (Siglec-1), CD22 (Siglec-2), MAG (Siglec-4), and Siglec-15 for which there are clear-cut orthologs in all mammalian species examined and which share only about 25–30% sequence identity among each other. The second group comprises the CD33rSiglecs, which share about 50–80% sequence similarity but appear to be evolving rapidly and undergoing shuffling of Ig-domain-encoding exons, making it difficult to define orthologs even between rodents and primates (see details below).

FIGURE 32.1. Domain structures of the known Siglecs in humans and mice.


Domain structures of the known Siglecs in humans and mice. There are two subgroups of Siglecs: One group contains sialoadhesin (Siglec-1), CD22 (Siglec-2), MAG (Siglec-4), and Siglec-15, and the other group contains CD33-related Siglecs. In humans, Siglec-12 (more...)


The Amino-terminal V-set Sialic Acid–binding Domain

All Siglecs are type-1 membrane proteins that contain a Sia-binding, amino-terminal V-set domain and varying numbers of C2-set Ig domains that act as spacers, projecting the Sia-binding site away from the plasma membrane. The V-set domain and the adjacent C2-set domain contain a small number of invariant amino acid residues, including an “essential” arginine on the F β-strand that is required for sialic acid binding and an unusual organization of cysteine residues. Instead of the typical intersheet disulfide bond between the B and F β-strands, Siglecs display an intrasheet disulfide bond between the B and E β-strands, permitting increased separation between the β-sheets. The resulting exposure of hydrophobic residues allows specific interactions with constituents of Sia. All Siglecs also appear to contain an additional unusual disulfide bond between the V-set domain and the adjacent C2-set domain, which would be expected to promote tight packing at the interface between the first two Ig domains. Although the significance of this for ligand recognition is unclear, it has been noted that the Sia-binding activity of some Siglecs (e.g., CD22 and MAG) appears to require the adjacent C2-set domain, probably for correct folding.

Absolute Requirement for Sialic Acids in Glycan Ligands

Siglecs differ from most other mammalian Sia-binding lectins (such as selectins) with respect to their absolute requirement for Sia. Whereas the selectins use Sias as carriers of negative charge to make ionic interactions, the Siglecs make more extensive molecular contacts, exploiting not only the negatively charged carboxylate group, but also the glycerol side chain, the N-acyl group, and the C-4-hydroxyl group. In contrast to selectins, substitution of Sia in oligosaccharides with a sulfate moiety results in loss of binding to Siglecs (although addition of sulfate esters to other parts of a the underlying glycan can enhance affinity). In addition, treatment of target cells or glycoconjugates with broad specificity sialidases is an effective way to destroy glycan recognition by Siglecs. Similar to many lectins, the affinity of Siglecs for sialylated ligands is low, with binding constants typically in the high micromolar to low millimolar range, as revealed in surface plasmon resonance, equilibrium dialysis, nuclear magnetic resonance (NMR), or thermal calorimetry measurements (see Chapter 27). Multimerization of Siglecs is likely to occur naturally on plasma membranes, leading to high avidity binding to clustered glycan ligands.

Masking and Unmasking

The cell-surface glycocalyx of most mammalian cells is richly decorated in glycoconjugates that contain Sias. The high local concentration of Sias is likely to greatly exceed the Kd value of each Siglec, resulting in “masking” of the Sia-binding site. Consequently, the Sia-dependent binding activity of most naturally expressed Siglecs is difficult to demonstrate unless the cells are first treated with sialidase to eliminate the cis-interacting sialylated glycans. A notable exception is Sn, which was discovered as a Sia-dependent cell-adhesion molecule on macrophages isolated from various tissues. The “masked” state of most Siglecs is a dynamic equilibrium with multiple ligands. Thus, an external probe or cell surface bearing high-affinity ligands or very high densities of Sia residues can effectively compete even with the binding domains of “masked” Siglecs. In addition, changes in expression of glycosyl-transferases or sialidases could influence masking and unmasking of Siglecs at the cell surface, especially during immune and inflammatory responses.

Expression in a Cell-type-restricted Manner

Siglecs show restricted patterns of expression in unique or related cell types. This is most striking for Sn, CD22, and MAG, which are expressed on macrophages, B lymphocytes, and myelin-forming cells, respectively. This theme also extends to some of the CD33rSiglecs (most notably in humans), Siglec-6 on placental trophoblasts, Siglec-7 on NK (natural killer) cells, Siglec-8 on eosinophils, and Siglec-11 on tissue macrophages, including brain microglia. In the mouse, Siglec-H and CD33 are excellent markers of plasmacytoid dendritic cells (DCs) and neutrophils, respectively, and Siglec-F is a useful marker of eosinophils. These cell-type-restricted expression patterns are thought to reflect discrete, cell-specific functions mediated by each of these Siglecs. However, certain key cells of the immune system such as monocytes and conventional DCs express multiple CD33rSiglecs in humans.

Cytoplasmic Tyrosine-based Signaling Motifs

Most Siglecs have one or more tyrosine-based signaling motifs. Exceptions are Sn, Siglec-14, Siglec-15, mCD33, and Siglec-H. The most prevalent motif is the immunoreceptor tyrosine-based inhibitory motif (ITIM) with the consensus sequence (V/I/L)XYXX(L/V), where X is any amino acid. More than 100 ITIM-containing membrane receptors have been identified in the human gemone, and many of these are established inhibitory receptors of the hematopoietic and immune systems. They function by recruiting certain SH2-domain-containing effectors, the best characterized of which are the the protein tyrosine phosphatases SHP-1 and SHP-2 or the inositol 5′ phosphatase SHIP. These counteract activating signals triggered by receptors containing immunoreceptor tyrosine-based activatory motifs (ITAMs). Some Siglecs without a prominent cytoplasmic domain instead have a positively charged residue within the transmembrane region, which can associate with the DAP-12 (DNAX activation protein-12) ITAM-containing adaptor.


The three-dimensional structures of the mouse (m) Sn and human (h) Siglec-7 V-set domains have been determined by X-ray crystallography, in the presence and absence of Sia ligands (Figure 32.2). These provide a structural template for Sia recognition by Siglecs that is likely to be shared by other family members. In both instances, the “essential” arginine residue is located in the middle of the F β-strand, making a bidentate salt bridge with the carboxylate of Sia. In the absence of bound ligand, the essential arginine of both Sn and Siglec-7 is masked by a basic residue (either arginine or lysine). On binding ligands, this basic residue moves away to allow access of the essential arginine to the Sia carboxylate group. Siglecs also contain a conserved hydrophobic amino acid (either tryptophan or tyrosine) on the G β-strand that interacts with the glycerol side chain of Sia. The C-4 hydroxyl of Sia makes a hydrogen bond either directly or indirectly via a water molecule, and the N-acetyl group interacts with a tryptophan of Sn via hydrophobic interactions and with a similarly positioned tyrosine of Siglec-7 via hydrogen bonding. Although all Siglecs probably share this common template for binding glycosidically linked Sias, their binding preferences for extended glycan chains vary greatly. The peptide loop between the C and C′ β-strands is highly variable among Siglecs and has a key role in determining their fine sugar specificity. For example, molecular grafting of the C-C′ loop between Siglecs-7 and -9 resulted in switched sugar-binding specificities. Structural studies have shown that this loop appears to be highly flexible, being able to make specific and varied interactions with long glycan chains.

FIGURE 32.2. Structural basis of Siglec binding to ligands.


Structural basis of Siglec binding to ligands. X-ray crystal structures of the V-set domains of sialoadhesin (Sn) (A) and Siglec-7 (B) are shown complexed with sialic acid. (C,D) Molecular details of the interactions of sialic acid with Sn and Siglec-7. (more...)


Sialoadhesin (Siglec-1, CD169)

Sn was identified in 1983 as a Sia-dependent sheep erythrocyte receptor (SER) expressed by mouse stromal macrophages isolated from various tissues. Studies with anti-Sn monoclonal antibodies established that expression in humans and mice is highly specific for macrophage subsets, especially those in lymphoid tissues and those recruited to inflammatory sites, as seen in inflamed joints in individuals with rheumatoid arthritis and other autoimmune inflammatory disorders. Full-length Sn is predicted to be a transmembrane protein with 17 Ig-like domains. Multiple splice variants encoding truncated and secreted forms of Sn exist, but their biological significance is unknown. Sn can be expressed at very high levels on macrophages, with up to 1 million molecules per cell, generating the potential to mediate cell–cell and cell–matrix interactions. Microscopy studies provided evidence for Sn clustering in vivo, especially in the bone marrow where Sn is localized at the contact sites of resident stromal macrophages and developing granulocytes. Several glycoprotein ligands have been identified from cell lysates using affinity chromatography methods with soluble recombinant forms of Sn. In all cases, the glycoconjugates that bind in a Sia-dependent manner are transmembrane mucins known to display multiple clustered O-linked glycans. These include MUC1 expressed by breast cancer cells and P-selectin glycoprotein ligand-1 (PSGL-1) and CD43 expressed on myeloid cells and T cells. It is unclear if these are indeed specific ligands or if they simply have the highest densities of cognate Sia residues.

The unusually large number of 17 Ig domains in Sn appears to be conserved in mammals and is thought to be important for extending the Sia-binding site away from the plasma membrane to promote intercellular interactions. Electron microscopy shows that Sn is an extended molecule of about 50 nm. Besides sialic acid binding mediated by the V-set domain, the C2-set domains of Sn display N-linked glycans that can act as ligands for other mammalian lectins such as the cysteine-rich domain of the mannose receptor and the macrophage galactose-type C-type lectin-1 (see Chapter 31). Although the biological significance of these interactions is uncertain, it is interesting that both lectins are found on dendritic cell populations that traffic into lymph nodes where Sn is abundantly expressed (Figure 32.3).

FIGURE 32.3. Biological functions mediated by sialoadhesin: Interactions of sialoadhesin on macrophages with cells and pathogens.


Biological functions mediated by sialoadhesin: Interactions of sialoadhesin on macrophages with cells and pathogens. (Right) Red staining shows a ring of sialoadhesin expressed by macrophages in the marginal zone of the spleen and green staining shows (more...)

A role for Sn in interactions with pathogens has also been suggested. On the one hand, Sn can act as a phagocytic receptor for bacterial and protozoal pathogens such as Neisseria meningitidis and Trypanosoma cruzi, which coat themselves with Sias in an attempt to evade other forms of immune recognition. On the other hand, Sn expressed on alveolar macrophages can be “hijacked” as an endocytic receptor for the porcine reproductive and respiratory syndrome virus, which derives its envelope surface Sias from the mammalian cells from which it originally buds.

Sn-deficient mice appear essentially normal under specific pathogen-free conditions, with only subtle alterations in numbers of CD8 T cells and some subsets of B cells and reduced levels of circulating IgM. Granulocyte numbers appear normal in the bone marrow, in the blood, and following acute inflammation, suggesting that Sn–granulocyte interactions are not essential for maintenance of myelopoiesis. Interestingly, in mouse models of inherited neuropathy in which Sn-deficient mice were bred onto a P0 heterozygous background or proteolipid protein transgenic mice, reduced infiltration of CD8 T cells and macrophages was observed, accompanied by attenuated demyelination. In a mouse model of autoimmune uveoretinitis, Sn-deficient mice exhibited reduced inflammation, lowered T-cell proliferative responses, and a reduced time of onset to disease. Taken together, these findings suggest that Sn is important for the fine-tuning of adaptive immune responses, but the mechanisms remain to be established.

CD22 (Siglec-2)

CD22 was initially identified in 1985 as a developmentally regulated cell-surface glycoprotein on B cells. It is expressed at approximately the time of Ig gene rearrangement and is lost when mature B cells differentiate into plasma cells. CD22 was cloned in 1990 and was shown to have seven Ig-like domains. The intracellular region of CD22 has six tyrosine-based signaling motifs, four of which function as ITIMs.

CD22 is a well-established negative regulator of B-cell activation, making an important contribution toward the threshold for signaling via the B-cell receptor (BCR) complex. Following BCR cross-linking, CD22 is rapidly tyrosine-phosphorylated on its ITIMs by the Lyn tyrosine kinase. This leads to recruitment and activation of the SHP-1 tyrosine phosphatase and subsequent inhibition of downstream signaling mediated via the BCR. Besides SHP-1, the Ca++ pump PMCA4 is also recruited and activated and plays a key role in efflux of intracellular Ca++, which results in dampening of signals that depend on elevated intracellular Ca++ concentration. Although additional activatory signaling molecules are recruited to the phosphorylated tyrosine motifs in CD22, the net phenotype of CD22-deficient mice is consistent with a primary role of CD22 in negative regulatory signaling, manifest by up-regulated MHC class II expression, enhanced B-cell turnover, reduced numbers of recirculating B cells in the bone marrow, reduced numbers of marginal zone B cells, and reduced anti-IgM-induced proliferation.

Of all the Siglecs, CD22 has the highest specificity for sialylated ligands, binding primarily to α2-6-linked Sias of the type Neu5Ac(Gc)α2-6Galβ1-4GlcNAc, which are common capping structures of many N-glycans. Additional specificity can be conferred by the nature of the Sia moiety: Neither hCD22 nor mCD22 binds 9-O-acetylated Sias; mCD22 has a strong preference for Neu5Gc over Neu5Ac, whereas hCD22 binds both of the latter forms. Binding of mCD22 to NeuGcα2-6Galβ1-4GlcNAc has a Kd value of 250 μM at 37°C and displays very rapid dissociation kinetics. Recombinant soluble CD22 can precipitate a subset of glycoproteins from cell lysates including CD45, a major sialoprotein of T and B cells carrying up to 18 N-linked glycans. Detailed kinetic studies with a range of native and enzymatically derivatized glycoconjugates have established that binding of CD22 to ligands is dependent on Sia density and linkage and appears not significantly influenced by the nature of the glycan carrier.

In common with many Siglecs, the Sia-binding site of CD22 on B cells is “masked” by cis-interactions with α2-6-sialylated ligands. However, when a B cell contacts another cell expressing high levels of α2-6-sialylated ligands, CD22 can redistribute to the points of cell contact, suggesting that trans-cellular communication can occur under physiological conditions. This could be important for altering B-cell activation thresholds and may help to ensure that signaling through the B-cell IgM receptor can only occur in lymphoid tissues where CD22 α2-6-sialylated ligands are particularly abundant on both T cells and B cells. In addition, some evidence for “unmasking” of the CD22-binding site has been observed following B-cell activation in vitro, and the specialized subset of B1 marginal zone cells that respond to carbohydrate antigens independently of T cells have constitutively unmasked forms of CD22. This could be important for lowering their activation thresholds, allowing efficient antibody production to foreign carbohydrate antigens.

Mouse mutants have shed light on how the lectin function of CD22 regulates its immunomodulatory role in B cells. ST6Gal-I-deficient mice lack ligands for CD22 and their B cells exhibit an anergic phenotype, essentially the opposite of the phenotype observed with CD22-deficient mice, which have hyperactivated B cells and enhanced signaling via the BCR. Crossing the ST6Gal-I-deficient mice with CD22-deficient mice restores B-cell signaling function, suggesting that CD22 is essential for the reduced BCR signaling seen in CD22-ligand-deficient mice. CD22 is known to associate with the BCR and this still occurs in ST6Gal-I-deficient mice, as well as in B cells transfected with a non-Sia-binding mutant of CD22. This argues strongly for a conserved non-Sia-dependent association of CD22 with BCR. These results, together with biochemical data showing that CD22 is homomultimerized via protein and Sia interactions, suggest that CD22 at the cell surface is the major counterreceptor for itself, possibly to the exclusion of other cell surface sialylated glycoproteins such as CD45. Furthermore, CD22 can associate with BCR in distinct microdomains of the B-cell plasma membrane, and its potential association with clathrin-rich microdomains and recycling at the cell surface are important topographical issues that require more study in the future. Thus, the anergic phenotype of ST6Gal-I-deficient mice could be due to excessive interactions of BCR with CD22, leading to mislocalization and aberrant signaling. In this model, the role of α2-6-sialylation could be to promote CD22 homomultimerization and sequestration of CD22 away from the BCR (Figure 32.4). Complementary data have arisen from the use of mouse mutants in which the sialic acid–binding capacity of CD22 has been selectively inactivated, but conflicting results were obtained using potent Sia-based inhibitors of CD22. Overall, the somewhat inconsistent data on functions of CD22 Sia recognition may reflect the reality of the biological role of this interaction in “tuning” the B-cell response appropriately to a given circumstance.

FIGURE 32.4. Proposed biological functions mediated by CD22: CD22 glycan-dependent homotypic interactions in equilibrium with CD22–BCR interactions.


Proposed biological functions mediated by CD22: CD22 glycan-dependent homotypic interactions in equilibrium with CD22–BCR interactions. The actual situation seems to vary between different cell types and analysis conditions. (BCR) B-cell receptor; (more...)

Myelin-associated Glycoprotein (Siglec-4)

MAG was identified in 1972 as a minor constituent of central nervous system (CNS) and peripheral nervous system (PNS) myelin. Cloned in 1987, it has five Ig-like domains and is highly conserved in mammals. MAG has a clear-cut ortholog in fish and may represent a primordial Siglec that gave rise to the immune Siglecs via gene duplication and “exon shuffling.” Alternative transcripts of MAG (L-MAG and S-MAG) contain long (105 amino acids) or short (74 amino acids) cytoplasmic tails, respectively. The two isoforms are developmentally and anatomically regulated: Equivalent amounts are found in the adult CNS, whereas S-MAG predominates in the adult PNS. MAG is expressed by myelin-forming cells, oligodendrocytes in the CNS, and Schwann cells in the PNS. In mature myelinated axons, it is found on the innermost (periaxonal) myelin wrap but not in the multilayers of compacted myelin. MAG has features that indicate a role in adhesion and signaling in axon–glia and/or glia–glia interactions. L-MAG can recruit the nonreceptor tyrosine kinase Fyn, the calcium-binding protein S100β, and the phospholipase Cγ, whereas S-MAG has been reported to bind to tubulin and microtubules, supporting a role for IT as a cell-adhesion molecule linking the axonal surface and the myelinating glial cell cytoskeleton.

MAG-deficient mice develop normal myelin, but defects in myelin and axons increase as animals age, indicating a role for MAG in the maintenance of myelin and myelinated axons, rather than in the process of myelination. MAG-null mice display late-onset progressive PNS and CNS axonal atrophy and increased “Wallerian” degeneration. Myelinated axons of MAG-null mice fail to show characteristic myelin-induced increases in neurofilament phosphorylation and axon diameter, indicating that MAG signaling is required for appropriate myelin–axon communication.

MAG also makes an important contribution to the inhibitory activity of myelin on axon outgrowth and repair, a major factor in poor recovery from nervous system injury. MAG extracted from myelin, as well as expressed forms of soluble and cell-surface MAG, inhibits neurite outgrowth from a wide variety of neuronal cell types in vitro.

Genetic and biochemical evidence indicates that gangliosides are important physiological ligands for MAG, mediating both myelin-axon stability and inhibition of axon outgrowth. Recombinant forms of MAG bind selectively to the abundant axonal gangliosides GD1a and GT1b, and they bind with higher affinity to the minor ganglioside GQ1bα (see Chapter 10). The phenotype of MAG-deficient mice is similar to that of mice lacking an N-acetylgalac-tosaminyltransferase (GalNAcT) required for synthesis of gangliosides. In addition, the disialyl T antigen (NeuAcα2-3Galβ1-3[NeuAcα2-6]GalNAc-R), a structure found on O-glycans and gangliosides, effectively inhibited the binding of soluble MAG to a target ligand and reversed MAG-dependent inhibition of axon outgrowth. Binding of soluble MAG to certain neurons is Sia-dependent and blocked with antibodies to gangliosides. Neurons from GalNAcT-null mice are less responsive to MAG, whereas MAG still inhibits neurite outgrowth from mice lacking the“b-series”gangliosides (lacking GT1b but expressing GD1a; see Chapter 10) due to a mutation of a sialyltransferase, GD3 synthetase. These findings suggest that gangliosides GD1a or GT1b both act as functional docking sites for MAG on neuronal cells.

MAG also binds specifically and with high affinity to a family of GPI-anchored proteins, the Nogo receptor (NgR) family. Gangliosides and NgRs may act independently or interactively as MAG receptors, linking MAG binding to axonal signaling in different neuronal cell types (Figure 32.5). In one model, a lipid-raft-associated signaling complex on the surface of neurons mediates MAG-dependent inhibition of axon outgrowth. Binding of MAG to NgRs (and perhaps gangliosides) is accompanied by movement of the p75 neurotrophin receptor into glycolipid-enriched membrane microdomains. The p75 is then cleaved by α-and γ-secretase, leading to activation of the small GTP-binding protein Rho A in a protein kinase C–dependent manner. GTP-bound Rho A then activates a Rho-A-dependent kinase leading to changes in actin filaments and microtubules, resulting in inhibition of axon outgrowth. It remains unresolved whether MAG binding to NgRs is Sia-dependent. One possibility is that MAG has two binding sites, one in the amino-terminal V-set domain, which mediates Sia-dependent binding, and another in domain 4 and/or 5, which binds a protein determinant on NgRs. Thus, MAG appears to mediate its effects on neurite outgrowth via protein–glycan and protein–protein interactions with gangliosides and NgR, converging on Rho-A activation and the control of the axon cytoskeleton.

FIGURE 32.5. Proposed biological functions mediated by myelin-associated glycoprotein (MAG): Interactions between MAG and molecules of the axonal membrane lead to inhibition of neurite outgrowth.


Proposed biological functions mediated by myelin-associated glycoprotein (MAG): Interactions between MAG and molecules of the axonal membrane lead to inhibition of neurite outgrowth. For a full explanation, see text. (NgR) Nogo receptor. Symbol Key: Image symbol_key_small.jpg


Genes encoding most of this Siglec subfamily are clustered together on human chromosome 19q13.3-13.4 or the syntenic region of mouse chromosome 7. They include CD33 (Siglec-3) and Siglecs-5, -6, -7, -8, -9, -10, -11, -12, and -14 in humans and CD33, Siglecs-E, -F, -G, and -H in the mouse. Similar clusters are found in other primates and rodents. It is difficult to assign definitive orthologs between Siglecs in primates and rodents, resulting in the current use of different nomenclatures. One reason for this is that most IgSF domains are encoded by exons with phase-1 splice junctions. This permits exon shuffling without disrupting the open reading frame, resulting in generation of species-restricted hybrid genes that are difficult to distinguish from similarly organized genes in other species. A second reason is that the Sia-binding sites in the V-set domains of the Siglecs appear to have been rapidly evolving, presumably to change their binding specificity in response to the rapid evolution of the endogenous host sialome (see Chapter 14). There is also evidence for gene conversion events between adjacent genes and pseudogenes within this cluster in a species-specific manner. Of particular interest is the finding that humans show many CD33rSiglec differences compared with their closest evolutionary cousins (the chimpanzees), more than the differences between mice and rats (which shared a common ancestor much earlier). For example, Siglec-13 is specifically deleted in humans, but present in chimpanzees and baboons.


The CD33rSiglecs share sequence similarity and certain structural features, for example, the presence of a linker region encoded by a separate exon between domains 2 and 3. Most contain a membrane-proximal ITIM and a membrane-distal ITIM-like motif. Although the latter is similar to the ITSM (switch motif) found in other signaling receptors of the immune system, there is no evidence currently that it is essential for phosphatase recruitment or signaling functions. The ITIM recruits and activates both SHP-1 and SHP-2 and is thought to be important for negative inhibitory signaling and modulating CD33rSiglec-dependent adhesion. Most CD33rSiglecs are restricted in expression to the immune system. A shared property may therefore be to regulate leukocyte functions during inflammatory and immune responses, including cell proliferation, differentiation, activation, and survival, perhaps via recognition of sialic acids as “self.” In addition, CD33rSiglecs are actively endocytic and could be important in the clearance and/or antigen presentation functions of myeloid cells, especially when involving sialylated pathogens. Brief notes on the properties and putative functions of each of the CD33rSiglecs in humans and mice are provided below.

CD33 (Siglec-3)

CD33 was identified with monoclonal antibodies in 1983 as a marker of early human myeloid progenitors also found on mature monocytes and some macrophages and cloned in 1988. The later cloning of Sn and the recognition of Sia-binding by CD22 prompted experiments to investigate CD33′s Siglec activity. CD33 has some preference for α2-6- over α2-3-sialylated glycans. It was the first of the CD33rSiglecs to be characterized as an inhibitory receptor. Cross-linking of CD33 with the activating FcγRI was shown to reduce Ca++ signaling and CD33 was able to recruit and activate the inhibitory tyrosine phosphatases SHP-1 and SHP-2 (Figure 32.6). Mutation of the proximal ITIM prevented recruitment of these phosphatases and also resulted in increased Sia-dependent binding of RBC. CD33 is also rapidly phosphorylated on serine residues as a downstream consequence of protein kinase C activation, but the biological significance of this is not clear. Antibodies to CD33 have been reported to inhibit the proliferation of both normal and leukemic cell populations and the differentiation of DCs from bone marrow precursors, suggesting a role of CD33 in regulating hematopoiesis. The endocytic property of CD33 is currently being exploited in the treatment of acute myeloid leukemia using Gemtuzumab, a humanized anti-CD33 monoclonal antibody coupled to the toxic antibiotic calicheamicin. The endocytic pathway is as yet poorly characterized for any member of the CD33-related family, but it requires intact tyrosine motifs and appears to be via a clathrin-independent mechanism, in contrast to endocytosis mediated by CD22.

FIGURE 32.6. Proposed biological functions mediated by CD33-related Siglecs: A generic CD33-related Siglec is represented, showing the location of the immunoreceptor tyrosine-based inhibitory motif (ITIM) and the potential for inhibitory signaling.


Proposed biological functions mediated by CD33-related Siglecs: A generic CD33-related Siglec is represented, showing the location of the immunoreceptor tyrosine-based inhibitory motif (ITIM) and the potential for inhibitory signaling. Symbol Key: Image symbol_key_small.jpg

Prior to the discovery of the CD33rSiglec family, a murine ortholog of CD33 was isolated with two Ig-like domains (61% amino acid identity), yet with a cytoplasmic domain showing considerably less homology. Two alternatively spliced forms of mCD33 that differ in the cytoplasmic region have been described, but neither contains the typical ITIM found in most other CD33rSiglecs. This may explain the lack of a robust phenotype in CD33-null mice. Furthermore, mCD33 has a lysine residue in the transmembrane sequence and may therefore couple to the DAP-12 transmembrane adaptor, as shown recently for mouse Siglec-H and human Siglecs-14 and -15. In contrast to hCD33, mCD33 is expressed mainly on neutrophils rather than monocytes, which also suggests a nonconserved function of this receptor.

Siglec-5 (CD170) and Siglec-14

Siglec-5 contains four Ig domains and mediates Sia-dependent binding to α2-3Gal, α2-6Gal(NAc), and α2-8Sia linkages, indicating a potentially broad binding specificity for protein- and lipid-bound glycoconjugates. Siglec-5 is prominently expressed in the myeloid lineage, but at a later stage in differentiation than CD33. Rather than being lost from neutrophils on exit from marrow into the blood, Siglec-5 is retained and is also expressed on circulating monocytes and subsets of tissue macrophages. It seems to be the only Siglec in humans expressed on plasmacytoid DCs, a specialized cell type responsible for rapid production of type I interferons following viral infections. Furthermore, Siglec-5 can bind and internalize sialylated strains of N. meningitidis, raising the possibility that it could play a role in recognition of these pathogens in humans. Siglec-5 has been shown to function as an inhibitory receptor following co-cross-linking with the activating high-affinity IgɛR in transfected rat basophilic leukemia cells.

Siglec-14 was recently characterized as a novel human Siglec with three Ig domains and no ITIM-like motifs (Figure 32.7). The first two Ig domains are greater than 99% identical to those of Siglec-5 with only one amino acid difference. In contrast, the third Ig domain of Siglec-14 is much less similar, and its transmembrane region contains an arginine residue that can associate with the DAP-12 ITAM-containing adaptor. This curious partial homology is explained by the finding that the Siglec-5 and Siglec-14 genes appear to be undergoing concerted evolution in multiple primate species, suggesting that these receptors may function as paired activating and inhibitory receptors. It should be noted that almost all antibodies currently available against Siglec-5 cross-react with Siglec-14, raising uncertainty about the expression patterns of Siglec-5 described above.

FIGURE 32.7. Chromosomal organization of CD33-related Siglec clusters in some rodents and primates.


Chromosomal organization of CD33-related Siglec clusters in some rodents and primates. Note that the locus marked as 5* is now known to encode Siglec-14 in primates. (Reprinted, with permission, from Angata et al. 2004. Proc. Natl. Acad. Sci. 101: 13251–13256.) (more...)


Siglec-6 was cloned from a human placental cDNA library and also during a screen for proteins that bind leptin, a hormone that regulates body weight. It has three Ig-like domains and the typical arrangement of ITIM and ITIM-like motifs in its cytoplasmic tail. Siglec-6 binds dimeric leptin with a Kd of approximately 90 nM, which reflects an affinity for leptin that is tenfold weaker than the leptin receptor (Ob-R), yet tenfold stronger than the related Siglecs-3 and -5. These findings indicate that although it is unlikely to mediate leptin signaling, Siglec-6 could function as a leptin “sink” and contribute to regulation of leptin plasma concentration. Another unusual feature of Siglec-6 is its high expression in placenta, being localized to cytotrophoblastic and syncytiotrophoblastic cells, where its levels seem to increase during the progress of labor and delivery. Lower levels are expressed on B cells. Siglec-6 does not have an obvious ortholog in mice, but one is present in the chimpanzee and baboon.

Siglecs-7, -8, and -9

These three Siglecs have three Ig domains, share a high degree of sequence similarity, and appear to have evolved by gene duplication from a three-domain ancestral Siglec. Siglec-7 is the major Siglec expressed by NK cells and functions as an inhibitory receptor when expressed naturally on these cells, both in antibody-directed killing assays and in assays using target cells overexpressing GD3, which is a Siglec-7 ligand. Siglec-7 also mediates selective interactions with sialylated lipooligosaccharides of the human pathogen, Campylobacter jejuni, suggesting a role in host–pathogen interactions. Siglec-9 is prominently expressed on neutrophils, monocytes, and to a much lesser extent on subsets of NK cells and CD8 T cells. In contrast, Siglec-8 is expressed on eosinophils, with weaker expression on basophils. The original Siglec-8 cDNA clone lacked the typical ITIM and ITIM-like motifs. However, subsequent analyses showed that this represents a minor alternatively spliced form and that a transcript encoding the longer form is more abundantly expressed in eosinophils.

Inhibition of cellular activation by Siglecs-7 and -9 can be demonstrated following co-cross-linking with ITAM-coupled activating FcRs (U937 and RBL cells) or following T-cell receptor engagement (in Jurkat cells). A novel proapoptotic function for Siglec-8 expressed by eosinophils was discovered following antibody-induced cross-linking. This depended on generation of reactive oxygen species and caspase activation and was paradoxically enhanced in the presence of cytokine “survival” factors such as GM-CSF and interleukin-5 (IL-5). Similar observations were made with Siglec-9 expressed on neutrophils. In addition to inducing apoptosis, ligation of Siglec-9 in the presence of GM-CSF (granulocyte-macrophage–colony-stimulating factor) also resulted in a caspase-independent nonapoptotic form of cell death. Autoantibodies to Siglecs-8 and -9 have been shown to be present in human serum, and their induction of granulocyte cell death may be linked to the anti-inflammatory properties of intravenous immunoglobulin, a pooled human serum preparation used to treat certain human autoimmune disorders. The role of the Siglec tyrosine-based motifs and SHP recruitment in triggering cell death are currently unknown.

Analyses of the glycan-binding specificities of these proteins revealed striking differences among Siglecs-7, -8, and -9. Siglec-8 was found to bind best to a single unique ligand, 6′-sulfo-SLex, whose natural expression pattern is only partly known. In contrast, Siglec-9 preferred a related structure, 6-sulfo-SLex, and Siglec-7 bound well to both forms.

The closest mouse relative to Siglecs-7, -8, and -9 is Siglec-E, which falls into the same evolutionary clade and shares about 70% sequence similarity with all three proteins. Siglec-E appears to exhibit a combination of features found in Siglec-7 and Siglec-9, being expressed on similar cell populations and exhibiting similar sugar-binding activity. Although there is no ortholog of Siglec-8 in mice, the four-Ig domain mouse Siglec-F is expressed in a similar way to Siglec-8 on eosinophils and has a similar glycan-binding preference. It therefore appears to have acquired the same functions through convergent evolution. Siglec-F-null mice show exaggerated eosinophilic responses in a lung allergy model, suggesting that its normal role is to dampen such responses. Interestingly, Siglec-F ligands in the airways and lung parenchyma were also up-regulated during allergic inflammation.

Siglecs-10 and -11

Siglec-10 has five Ig-like domains and displays an additional tyrosine-based motif in its cytoplasmic tail. It is expressed at relatively low levels on several cells of the immune system, including monocytes, eosinophils, and B cells. It is the only CD33-related human Siglec that has a clear-cut ortholog in mice, designated Siglec-G. This, combined with phylogeny analyses, suggests that Siglec-10 represents an ancestral CD33rSiglec that may have given rise to the other members of this subgroup via gene duplication, exon loss, and exon shuffling. Mice deficient in Siglec-G show a tenfold increase in numbers of a specialized subset of B lymphocytes, the B1a cells, that make natural antibodies and fast T cell–independent antibodies to some bacteria. In addition, the B1a cells show exaggerated Ca-fluxing following BCR cross-linking. Siglec-11 also has five Ig domains that are 90% identical to Siglec-10, but with much lower similarity in the transmembrane and cytoplasmic regions. Siglec-11 in humans appears to be a chimeric molecule that underwent gene conversion with an adjacent Siglec pseudogene. Despite the high sequence similarity in the extracellular region, the sugar-binding properties are distinct, with Siglec-10 binding to both α2-3- and α2-6-linked Sias and Siglec-11 binding only weakly to α2-8-linked Sias. Siglec-11 also exhibits different expression patterns, being absent from circulating leukocytes, but expressed widely on populations of tissue macrophages, including resident microglia in the brain, where high levels of α2-8-linked Sias are present on gangliosides. Interestingly, this microglial expression appears unique to humans.


Siglec-15 has two Ig-like domains: a short cytoplasmic tail and a transmembrane domain containing a lysine residue that allows association with the activating adaptor proteins DAP-12 and DAP-10. Siglec-15 preferentially recognizes the Neu5Acα2-6GalNAcα-(Sialyl-Tn) structure and is expressed on macrophages and/or dendritic cells of human spleen and lymph nodes. While Siglec-15 has the potential to be an activating receptor, it does not have an inhibitory counterpart like Siglec-14. Siglec-15 has been conserved throughout vertebrate evolution, and it presumably plays a conserved, regulatory role in the immune system. As with Sialyl-Tn in a tumor marker, one suggested possibility is that it functions in tumor surveillance.


As mentioned earlier, an “essential” arginine residue in all of the known Siglecs is important for binding Sia-containing ligands. Surprisingly, this arginine is frequently mutated in nature, resulting in loss of binding ability. Examples of this include Siglec-12 in humans, Siglecs-5 and -14 in the chimpanzee, gorilla, and orangutan, Siglec-6 in the baboon, and Siglec-H in the rat. The common arginine codon (CGN, where N is any nucleotide) tends to be highly mutable because of the CpG sequence. However, the frequency with which such events occurs is surprising, suggesting that it might be a natural mechanism to eliminate Sia binding of a given Siglec when such activity becomes inappropriate under changing evolutionary pressures without requiring a complete loss of the Siglec.


The ancestral condition of some hominid Siglecs (e.g., Siglec-7 and Siglec-9) appears to have been preferential binding to Neu5Gc, a Sia that was specifically lost in human evolution about 2–3 million years ago. The loss of Neu5Gc could thus have resulted in extensive Siglec unmasking, possibly leading to a state of heightened innate immune reactivity. Some human Siglecs have undergone an adjustment to allow increased Neu5Ac binding, and the question arises as to whether the adjustment is yet complete. Possibly as a consequence of this event, several Siglecs seem to have undergone human-specific changes in comparison to our great ape evolutionary cousins. For example, Sn seems to be expressed on most human macrophages, whereas only subsets of chimpanzee macrophages are positive for Sn. This may be related to the fact that Sn in humans has a strong binding preference for Neu5Ac over Neu5Gc, similar to that seen in other species. Human Siglec-5 and Siglec-14 appear to have undergone a restoration of the “essential” arginine residue needed for Sia recognition, which is mutated in chimpanzees, gorillas, and orangutans. As mentioned earlier, the gene encoding Siglec-11 has undergone a human-specific gene conversion, resulting in a new protein with altered binding properties and new expression in brain microglia. Siglec-12 has suffered a human-specific inactivation of the essential arginine residue in humans, with subsequent permanent pseudogenization by a frame-shift in some humans. Siglec-13 has undergone a human-specific gene deletion. Expression patterns of some Siglecs also appear to have undergone changes, with the placental expression of Siglec-6 being human-specific and a general suppression of all CD33rSiglecs on human T cells compared with the chimpanzee. The functional implications of these human-specific changes in Siglec biology for physiology and disease deserve further exploration.


  1. Powell LD, Varki A. I-type lectins. J Biol Chem. 1995;270:14243–14246. [PubMed: 7782275]
  2. Crocker PR, Feizi T. Carbohydrate recognition systems: Functional triads in cell–cell interactions. Curr Opin Struct Biol. 1996;6:679–691. [PubMed: 8913692]
  3. Kelm S, Schauer R. Sialic acids in molecular and cellular interactions. Int Rev Cytol. 1997;175:137–240. [PubMed: 9203358]
  4. Varki A. Sialic acids as ligands in recognition phenomena. FASEB J. 1997;11:248–255. [PubMed: 9068613]
  5. Crocker PR, Varki A. Siglecs, sialic acids and innate immunity. Trends Immunol. 2001;22:337–342. [PubMed: 11377294]
  6. Angata T, Brinkman-Van der Linden E. I-type lectins. Biochim. Biophys. Acta. 2002;1572:294–316. [PubMed: 12223277]
  7. Crocker PR. Siglecs: Sialic-acid-binding immunoglobulin-like lectins in cell–cell interactions and signaling. Curr Opin Struct Biol. 2002;12:609–615. [PubMed: 12464312]
  8. Crocker PR. Siglecs in innate immunity. Curr Opin Pharmacol. 2005;5:431–437. [PubMed: 15955740]
  9. Nitschke L. The role of CD22 and other inhibitory co-receptors in B cell activation. Curr Opin Immunol. 2005;17:290–297. [PubMed: 15886119]
  10. Varki A, Angata T. Siglecs—The major subfamily of I-type lectins. Glycobiology. 2006;16:1R–27R. [PubMed: 16014749]
  11. Crocker PR, Paulson JC, Varki A. Siglecs and their roles in the immune system. Nat Rev Immunol. 2007;7:255–266. [PubMed: 17380156]