Human sialoglycan ligands for immune inhibitory Siglecs

Abstract

Most human Siglecs (sialic acid binding immunoglobulin-like lectins) are expressed on the surfaces of overlapping subsets of immune cells, and most carry immunoreceptor tyrosine-based inhibitory domains on their intracellular motifs. When immune inhibitory Siglecs bind to complementary sialoglycans in their local milieu, engagement results in down-regulation of the immune response. Siglecs have come under scrutiny as potential targets of drugs to modify the course of inflammation (and other immune system responses) and as immune checkpoints in cancer. Human Siglecs bind to endogenous human sialoglycans. The identities of these endogenous human sialoglycan immune regulators are beginning to emerge, along with some general principles that may inform future investigations in this area. Among these principles is the finding that a cell type or tissue may express a sialoglycan ligand for a particular Siglec on a single or a very few of its sialoglycoproteins. The selected protein carrier for a particular Siglec may be unique in a certain tissue, but vary tissue-to-tissue. The binding affinity of endogenous Siglec ligands may surpass that of its binding to synthetic sialoglycan determinants by several orders of magnitude. Since most human Siglecs have evolved rapidly, this review focuses on human Siglec ligands for human immune inhibitory Siglecs. As the identities of these immune regulatory sialoglycan ligands are defined, additional opportunities to target Siglecs therapeutically may emerge.

1. Immune inhibitory Siglecs and their ligands

Of the 14 human Siglecs, 13 are expressed on overlapping sets of immune cells, and 9 of those have immunoreceptor tyrosine-based inhibitory motif(s) in their intracellular domains (Duan and Paulson, 2020). Immune inhibitory human Siglecs include Siglec-2, -3, -5, -6, -7, -8, -9, -10 and -11. In the complex regulatory environment of the immune system, immune inhibitory Siglecs are thought to play important roles in keeping immune responses in check to avoid tissue damage and to resolve ongoing inflammatory events. As sialoglycan binding proteins, Siglecs are triggered by binding to complementary Siglec ligands, sialoglycoproteins and/or sialoglycolipids in the tissue environment (Gonzalez-Gil and Schnaar, 2021). Although different inhibitory Siglecs have different biochemical mechanisms, a generalized model proposed by Pröpster (Propster et al., 2016) is a basis for many investigations in this area (Fig. 1). Since clustering of Siglecs induces immune inhibition (van Houtum et al., 2021; Youngblood et al., 2020), it has been proposed that, at the tissue level, cells express multivalent sialoglycans that engage specific subsets of inhibitory Siglecs to modulate immune responses. Evidence for this model includes studies in which Siglecs are crosslinked with anti-Siglec antibodies to inhibit immune responses, an approach currently under clinical evaluation (O’Sullivan et al., 2020). Immune inhibitory responses are also induced by clustering immune cell surface Siglecs with polyvalent sialic acid mimetics either independently or by co-clustering with activating immune cell surface receptors (Carroll et al., 2017; Duan et al., 2021; Islam et al., 2022; Macauley et al., 2013).

Fig. 1.

Model of inhibitory Siglec function in modulating immune responses. Ongoing inflammation results in up-regulated expression of specific endogenous sialoglycan ligands (represented here as a secreted, high-molecular weight glycoprotein) in the inflamed tissue. Ligand binding to the N-terminal binding domain(s) of the Siglec induces clustering and initiates intracellular signaling cascades, which, through phosphorylation of cytoplasmic ITIM/ITIM-like motifs and recruitment of downstream effector proteins, ultimately lead to immune inhibition. Modified from Pröpster et al., 2016.

This review focuses on endogenous ligands of immune inhibitory Siglecs. Insights into the structures and functions of endogenous Siglec ligands have the potential to inform development of therapeutics to inhibit pathologies of overactive immune responses (e.g. eosinophils and mast cells in allergic inflammation) or disinhibit quiescent immune cells (e.g. microglia in neurodegenerative proteinopathies). A closely related therapeutic focus relates to Siglec ligands as immune checkpoint inhibitors in cancer (van Houtum et al., 2021). When cancer cells express cell surface or secreted sialoglycans that engage immune inhibitory Siglecs, they can suppress immune clearance. Targeting immune inhibitory Siglecs may enhance cancer immunotherapy. Pathogen sialoglycans that engage Siglecs and synthetic Siglec-targeting mimetics are not reviewed here, but are covered in other recent reviews (Chang and Nizet, 2020; Gonzalez-Gil and Schnaar, 2021; Movsisyan and Macauley, 2020).

A technical challenge in Siglec ligand research relates to the rapid evolution of the Siglec family in vertebrates. Among the inhibitory Siglecs, only human Siglec-2 has a clear ortholog in rodents (Angata et al., 2001). Although human Siglec-3 (CD33) shares ~50% polypeptide sequence identity with mouse Siglec-3 in its N-terminal sialic acid binding Ig-like domain, it fails to bind to the same sialoglycans to which human CD33 robustly binds (Brinkman-Van der Linden et al., 2003; Gonzalez-Gil et al., 2022). Furthermore, the transmembrane and intracellular signaling domains of mouse CD33 share only 26% polypeptide sequence identity with their human counterpart and have different immune regulatory functions (Bhattacherjee et al., 2019). Since most of the other human inhibitory Siglecs (Siglec-5 through Siglec-11) do not have clear mouse orthologs (Angata et al., 2004), several mouse Siglecs are designated by letters (Siglec-E through Siglec-H). Although some of these have been designated as functional paralogs of human Siglecs, their cellular distributions, glycan binding specificities, and functions are not congruent with their human counterparts (McCord and Macauley, 2022). Because of the evolutionary divergence of human Siglecs, the study of endogenous human ligands requires that they be identified (or confirmed) in human cells and tissues.

2. Siglec-2 (CD22)

Siglec-2 (CD22) is expressed on B cells in humans and other mammals (Bednar et al., 2017; Nitschke and Tsubata, 2004). It is an immune inhibitory receptor with multiple intracellular ITIM motifs and high selectivity for α2–6 linked sialic acids. On the outer leaflet of the B cell membrane, CD22 migrates laterally, spontaneously associates with the B cell receptor (BCR) in a sialic acid independent manner, and down-regulates B cell activation (Walker and Smith, 2008). Sialic acid dependent binding of CD22 functions in both activation and suppression of B cell function depending on the location of the sialoglycan ligands. When CD22 and its sialoglycan ligand are on the same cell membrane, cis-ligation sequesters CD22 away from the BCR and, counterintuitively, enhances B cell activation. In this case, the immune inhibitory Siglec is effectively neutralized by its sialoglycan binding (Muller et al., 2013).

The search for endogenous CD22 ligands on human B cells using in situ photoaffinity labeling revealed primarily CD22, which is itself a sialoglycoprotein (Enterina et al., 2019; Han et al., 2005). High resolution microscopy revealed sialic acid dependent clusters of CD22 on the B cell surface (Gasparrini et al., 2016). Mutation of CD22 at its sialic acid binding site (R130E) resulted in smaller clusters, higher lateral mobility, and more effective inhibition of the BCR. These data imply that CD22 on the B cell surface is in equilibrium between relatively stable homotypic clusters via its sialic acid binding domain and sialic acid independent binding and inhibition of the BCR. When sialic acid dependent homotypic clustering by CD22 is diminished, it is released to diffuse laterally and inhibit the BCR.

Investigation of another B cell surface ligand of CD22, CD45, further supports the role of sialic acid dependent CD22 homotypic clustering in its function (Gasparrini et al., 2016). In the absence of CD45, a relatively abundant sialoglycan on B cells, CD22 entered into yet larger (presumably homotypic) clusters with lower diffusion and less BCR inhibition. Treatment of these cells with sialidase or mild periodate (to block sialic acid dependent binding) released CD22 into smaller clusters, supporting the concept of sialic acid binding as a mechanism to sequester CD22 into homotypic clusters on the B cell surface. Whether CD22 binding to CD45 is fully sialic acid dependent remains in question (Meyer et al., 2018; Zhang and Varki, 2004).

Whereas sialic acid dependent cis clustering of CD22 diminishes inhibition of the BCR, trans ligands – particularly sialoglycans on antigen-presenting cells – have the potential to recruit CD22 to the immunological synapse and inhibit the BCR (Enterina et al., 2019; Ramya et al., 2010). The search for endogenous trans ligands of CD22 using cell surface installed photoaffinity sialic acids revealed robust interaction of CD22 primarily with IgM (Ramya et al., 2010). CD22 and IgM were then shown to colocalize at sites of cell-cell contact, implying a trans interaction. Co-localization was dependent on CD22 and dissipated when the sialyltransferase responsible for α2–6 sialylation (St6gal1) was deleted. A particularly important observation of this and related studies is that many glycoproteins bearing α2–6 linked sialic acids bind to CD22 in vitro, whereas very few preferentially bind in cellular contexts. This is a recurring theme in Siglec ligand identification, indicating that factors other than the sialic acid per se (clustering, orientation, polypeptide contributions) drive Siglec specificity.

Evidence that biological context is essential in evaluating Siglecs and their ligands was revealed in the study of CD22 in mouse and human brains (Pluvinage et al., 2019; Pluvinage et al., 2021). Whereas CD22 is expressed by mouse microglia and is upregulated with age, in human brains it is exclusively expressed by oligodendroglia, implicating a different function and/or mechanism for this evolutionarily conserved Siglec depending on the tissue and the species.

3. Siglec-3

Human Siglec-3 (CD33) is expressed on myeloid progenitor cells, monocytes, macrophages, mast cells and microglia, where it down-regulates immune cell function (Bhattacherjee et al., 2021; Bhattacherjee et al., 2019; Duan et al., 2019; Paul et al., 2000). Independent of its immunomodulatory activities, CD33 is overexpressed on acute myeloid leukemia cells, where it is exploited clinically as the target of an anti-CD33 antibody-drug conjugate (Gbadamosi et al., 2018). Evaluation of CD33 as a sialic acid binding immune modulator is confounded by conflating it with mouse CD33, which has different functional domains (Bhattacherjee et al., 2019; Cao and Crocker, 2011), different glycan binding (Gonzalez-Gil et al., 2022; Jung et al., 2021; McMillan and Crocker, 2008), and only an indirect phylogenetic relationship to human CD33 (Angata et al., 2001). For this review, we consider mouse CD33 to be sufficiently distinct from its human counterpart that its ligands and functions are not discussed.

Reports of flexible sialoglycan binding specificity of CD33 are common in the literature. The initial report of CD33 as a Siglec reported human red blood cell binding that was completely reversed by sialidase pretreatment (Freeman et al., 1995). Re-sialylation using 3-linked and 6-linked sialyltransferases restored binding. Subsequent studies using synthetic glycans likewise showed binding to 3- and 6-linked sialic acids (Blixt et al., 2003; Brinkman-Van der Linden and Varki, 2000; McMillan and Crocker, 2008; Padler-Karavani et al., 2014). This elastic view of CD33 sialoglycan specificity is challenged by recent findings. On a printed array having 113 synthetic sialoglycan structures, CD33 bound just two (Fig. 2A), both of which shared the minimum terminal glycan Neu5Acα2–3[6-SO4]Galβ1–4GlcNAc (Büll et al., 2021). Similar sialoglycans without a sulfate or with the sulfate on the GlcNAc instead of the Gal residue failed to bind. Likewise, CD33 bound to a synthetic glycolipid with the same terminal structure (not shown), and pretreatment with sialidase completely eliminated binding. Stringent binding specificity for CD33 was supported by studies using genetically modified human cells (Fig. 2B). The human embryonic kidney cell line HEK293 failed to bind CD33 until it was genetically modified to express the galactose-6-sulfotransferase CHST1. When the α2–3 sialyltransferase ST3GAL4 was selectively knocked out, binding was reversed. The data support the conclusion that robust CD33 binding requires a terminal Neu5Acα2–3(6-SO4)Gal structure. A more extensive study of carbohydrate sulfotransferases (Jung et al., 2021) likewise found that CHST1 greatly enhanced CD33 binding to a series of human cancer cell lines, including U937 (monocytic lymphoma), K562 (myeloid leukemia), Jurkat (T cell leukemia), and A549 (epithelial carcinoma).

Fig. 2.

CD33 binds specifically to Neu5Acα2–3[6SO4]Galβ1–4GlcNAc. A) CD33 binding to a sialoglycan array. Recombinant CD33-Fc chimera was tested for binding to a 113 sialoglycan array printed on a glass slide. The slide carried comparable numbers of linear and branched α2–3 and α2–6 synthetic sialoglycans. Only two sialoglycans bound CD33, both bearing an α2–3 linked sialic acid bound to a 6-sulfated galactose (insert). B) Radar chart showing CD33-Fc binding (center = no binding; periphery = maximum binding) to wild type (WT, top) HEK293 human embryonic kidney cells and the same cells genetically modified to express the human galactose-6-sulfotransfersase gene CHST1 (red font). WT cells failed to bind CD33 unless modified to express CHST1. Deletion of α2–3 sialyltransferase ST3GAL4 (blue font) in cells expressing CHST1 eliminated binding. Data modified from Büll et al, 2021. Glycan structures represented by Symbol Nomenclature for Glycans (Varki et al., 2015).

The identity of endogenous CD33 ligand from human brain is consistent with the data from glycan arrays and ectopic expression of glycan biosynthetic enzymes (Gonzalez-Gil et al., 2022). CD33 is of special interest in human brain because genome wide association studies consistently identify CD33 alleles that associate with Alzheimer’s disease (AD) susceptibility. Alleles that result in increased CD33 expression associate with higher susceptibility to AD whereas a rare allele that truncates the sialic binding domain of CD33 results in lower AD susceptibility (Griciuc and Tanzi, 2021). CD33 is expressed on microglia, phagocytic cells that remove debris in the brain. An hypothesis that fits the genomic data is that inhibition of microglical phagocytosis and debris clearance by an overabundance of CD33 results in buildup of neurotoxic proteins like amyloid beta and hyperphosphorylated tau (Chatila and Bradshaw, 2021; Griciuc et al., 2013). Based on these findings, the identity of the CD33 ligand in human brain, which may engage and inhibit microglial phagocytosis, is of potential significance in the progression of human proteinopathies.

Stringent (guanidinium hydrochloride) extraction of total proteins from human cerebral cortex was followed by electrophoretic resolution and probing for CD33 ligands by blotting and overlaying with CD33-Fc chimera (Gonzalez-Gil et al., 2022). Remarkably, only a single species migrating at 1 MDa (one million daltons) supported CD33 binding in extracts from multiple human donors (Fig. 3A). These data contradict the concept that CD33 has weak and elastic sialic acid binding specificity, since human brain extract contains many N-linked and O-linked sialoglycoproteins (Lee et al., 2020; Wilkinson et al., 2021) but only one bound to CD33. Glycohydrolases identified the glycan as a sialylated keratan sulfate proteoglycan (Fig. 3B) consistent with the terminal sialylated sulfated target for CD33 binding (Fig. 2). Affinity purification and proteomic mass spectrometry revealed that the unique protein carrier of the target glycans is the transmembrane glycoprotein receptor protein tyrosine phosphatase zeta (RPTPζ), also known as phosphacan in its extracellular released form. Double label immuno- and Siglec-blotting revealed that the human brain CD33 ligand is a unique isoform and glycoform of RPTPζ (Fig. 3C). The CD33 ligand is designated RPTPζ^S3L (Siglec-3 ligand) to distinguish it from other forms of RPTPζ. RPTPζ^S3L is upregulated (>2-fold) in extracts of AD donor brains compared to age-matched controls (Fig. 3D).

Fig. 3.

Endogenous CD33 ligand in human brain. A) Total protein extracts from human cerebral cortex from four different tissue donors (AD brains) were resolved on composite agarose–acrylamide gels, blotted to PVDF, probed with CD33-Fc and detected by enhanced chemiluminescence. Sample lanes are flanked by prestained crosslinked IgM (950 kDa major band, 1.9 MDa minor band) detected by white light. The entire length of the gel blot is presented, with the front denoted by an arrow. B) Cerebral cortex extract from a single AD donor was dialyzed against sodium phosphate buffer and incubated under matched control conditions (without enzyme) or with enzymes prior to resolution on composite agarose–acrylamide gels, blotting to PVDF, and probing with CD33-Fc. Sample lanes are flanked by prestained crosslinked IgM detected by white light. Lanes are as follows: (1) sialidase control buffer; (2) 120 mU/ml sialidase; (3) keratanase I control buffer; (4) 8.4 mU/ml keratanase I. C) Equal aliquots of human cerebral cortex total protein extract from four donors (numbered) were resolved on a composite agarose–acrylamide gel and blotted to PVDF. The blot was double-label probed with CD33-Fc (red) and anti-RPTPζ (green). D) Quantification of CD33-Fc band densities normalized to total protein for 5 AD donors and 5 age-matched control donors (*p = 0.028). Total proteins were extracted, resolved, probed with CD33-Fc as in Panel C. Band intensities were quantified and normalized to total protein quantified on a separate gel. Data from Gonzalez-Gil et al., (2022).

Mouse brains also express a single isoform and glycoform of RPTPζ that binds mouse Siglec-F and cross reacted with human CD33 and Siglec-8 (but not mouse CD33). Genetically engineered mice lacking expression of RPTPζ, galactose 6-sulfotransferase Chst1 or sialyltransferase St3gal4, no longer express the brain Siglec ligand.

These data support the conclusion that in human brain, the unique CD33 (Siglec-3) endogenous ligand is a sialylated keratan sulfate carried on a single minor glycoform and isoform of a single protein, RPTPζ. This interaction is likely to be exclusively a trans interaction, since of all cell types in the brain parenchyma microglia express the lowest levels of mRNA for RPTPζ (PTPRZ1 gene), CHST1, and ST3GAL4. These genes are much more highly expressed in neurons, astrocytes, and oligodendrocyte progenitor cells (Mathys et al., 2019). The data support the conclusion that RPTPζ^S3L is secreted by cells other than microglia into the brain parenchyma, where it encounters microglia and inhibits their phagocytosis (Fig. 1).

A cis ligand for CD33 on human cells was reported in a different context. CD33 is broadly distributed on cells of the myeloid lineage (Duan and Paulson, 2020), including monocyte-derived immature dendritic cells, where antibody-mediated CD33 clustering reduces toll-like receptor (TLR) signaling (Ishida et al., 2014). Cell surface proximity crosslinking revealed association of CD33 with the TLR co-receptor CD14. CD33-CD14 direct binding was sialic acid dependent, suggesting that specific glycans on CD14 engage CD33 in cis to regulate dendritic cell signaling. In general, different ligands may engage Siglecs in cis or trans to drive different biological outcomes.

4. Siglec-7

Siglec-7 is found most prominently on human natural killer (NK) cells, and less so on subsets of myeloid and dendritic cells (Nicoll et al., 1999). Since NK cells perform cancer immune surveillance, Siglec-7, an immune checkpoint receptor, has become a target for anti-cancer drug discovery (Daly et al., 2019). If sialoglycans on cancer cells engage Siglec-7 on NK cells, the cancer cells may avoid immune surveillance and expand. This concept is supported by the consistent finding of enhanced sialylation in many cancers (Hugonnet et al., 2021), which has led to the search for Siglec-7 ligands on human cancer cells.

Glycan specificity screening with Siglec-7 revealed enhanced binding to linear and branched disialoglycans (Anwar et al., 2022; Avril et al., 2006; Gieseke et al., 2012; Yamaji et al., 2002). Structural studies revealed the sites responsible for enhanced binding (Attrill et al., 2006a; Attrill et al., 2006b; Yamakawa et al., 2020). Sulfation also appears to support and/or enhance Siglec-7 binding (Avril et al., 2006; Campanero-Rhodes et al., 2006; Ito et al., 2001; Jung et al., 2021).

The disialo moiety Neu5Acα2–8Neu5Ac is abundantly expressed on endogenous gangliosides (sialylated glycosphingolipids) such as GD3 and GD2, which are overexpressed on certain cancers (Kasprowicz et al., 2022). Evidence that cancer cell disialogangliosides engage Siglec-7 to halt immune surveillance was obtained using human neuroblastoma cells (Theruvath et al., 2022). Anti-GD2 antibody, by itself and in combination with a second immune checkpoint inhibitor (anti-CD47), sensitized human cancer cells to robust human macrophage-mediated phagocytosis. Binding studies supported the conclusion that anti-GD2 blocked Siglec-7 binding, resulting in immune checkpoint inhibition and increased immune attack. Given that anti-GD2 antibody is clinically approved to treat neuroblastoma (Qiu and Matthay, 2022), the combined immune checkpoint inhibitor approach has entered clinical trials for human neuroblastoma and osteosarcoma (Theruvath et al., 2022).

Given the diversity of cancer cell sialoglycans, it is not surprising that different human cancer cells might express different ligands to effectively engage Siglec-7 and reduce immune surveillance. The K562 human chronic myelogenous leukemia cell line is widely used in cancer research and robustly binds to Siglec-7 (Jandus et al., 2014). Notably, two independent approaches identified CD43 (leukosialin), an ~120 kDa glycoprotein, as a major Siglec-7 ligand in K562 cell. In one study (Wisnovsky et al., 2021), the cells were tested for Siglec-7 binding after CRISPRi genetic screening. Consistent with expectations, loss of Siglec-7 binding was seen when sialoglycan biosynthesis was disrupted, but also when a single glycoprotein, CD43, was knocked out (Fig. 4). In a separate study, the same target glycoprotein, CD43, was identified by Siglec-7 affinity capture (Yoshimura et al., 2021). Together, the data support the conclusion that on K562 human cancer cells, Siglec-7 binds selectively to a subpopulation of CD43 that carries branched disialoglycans. Screening other human cancer cell lines revealed that some shared the CD43 glycoform as a ligand of Siglec-7 and others did not. A surprise from these studies is that CD43, rather than any of many other sialoglycoproteins on K562 cells, selectively binds Siglec-7. The structural components responsible for this selectivity have yet to be revealed.

Fig. 4.

Hits for Siglec-7 and Siglec-9 were plotted and ranked by hit score (−log10[positive selection score]), where a higher value indicates a stronger enrichment of sgRNAs in the low-staining population. The screen identified a number of sialic acid biosynthesis genes (purple), glycotransferase genes specific for Siglec-7 and Siglec-9 (green and orange), as well as a single-cell surface glycoprotein, CD43, specific for Siglec-7 (red). From Wisnovsky et al, 2021.

5. Siglec-8

5.1. Siglec-8 ligands of human airways

Siglec-8 is expressed by eosinophils, mast cells and basophils, human effector cells implicated in the pathogenesis of asthma, allergy and other atopic diseases (Bochner, 2009; Floyd et al., 2000; Kikly et al., 2000; Youngblood et al., 2020). It is detected late in differentiation of these cells, and in parallel with FcεRIα surface expression in mast cells (Yokoi et al., 2006). In human diseases characterized by elevated eosinophils, including chronic eosinophilic leukemia, hypereosinophilic syndrome, and chronic myeloid leukemia, all eosinophils express Siglec-8 (Hudson et al., 2011). Similarly, bone marrow mast cells in patients with mastocytosis and aplastic anemia also express Siglec-8. Siglec-8 crosslinking on human eosinophils using anti-Siglec-8 antibodies or synthetic polyvalent glycans induces apoptosis through well-characterized non-canonical pathways (Carroll et al., 2021; Carroll et al., 2017). On mast cells, Siglec-8 crosslinking inhibits release of inflammatory mediators. These effects are potentiated in the presence of activating cytokines and interleukins that promote cell survival (Nutku-Bilir et al., 2008; Youngblood et al., 2020), suggesting that Siglec-8 engagement by endogenous sialoglycans is a mechanism to resolve allergic inflammation. This led to therapeutic targeting of Siglec-8 for various types of atopy and other human diseases (Dellon et al., 2020; Youngblood et al., 2020).

Siglec-8 is the most stringent human Siglec in its glycan binding specificity, with high selectivity for a sulfated sialylated trisaccharide Neu5Acα2–3[6-SO4]Galβ1–4GlcNAc (Büll et al., 2021; Propster et al., 2016; Yu et al., 2017). The same structure is also a major glycan binding ligand for human Siglec-3 (CD33, see above), which has less stringent specificity than Siglec-8 (Büll et al., 2021; Yu et al., 2017). The molecular basis for this selective affinity was revealed by structural determination of Siglec-8 bound to a version of the above glycan (Propster et al., 2016). The precisely arranged sialic acid carboxylate and sulfate interact with complementary spaced arginines and glutamine at the Siglec-8 binding site. When endogenous human Siglec-8 sialoglycan ligands were isolated, they were found to be members of a family of sulfated proteoglycans, keratan sulfates, a portion of which are terminated with 6-sulfated and 3-sialylated galactose (Funderburgh, 2000).

On human airways, Siglec-8 ligands are expressed robustly in submucosal glands and their ducts and in cartilage (Fig. 5). They are comprised of sialylated keratan sulfate chains expressed prominently on a single type of O-linked glycoprotein that differs depending on the airway tissue compartment. In airway cartilage, minor glycoforms of the major cartilage protein aggrecan carry sialylated, highly sulfated keratan sulfate chains that bind Siglec-8 (Gonzalez-Gil et al., 2018). These minor isoforms and glycoforms are designated aggrecan^S8L. Addition of affinity-purified aggrecan^S8L to freshly isolated human eosinophils in culture increased their apoptosis in a sialic acid dependent manner.

Fig. 5.

Siglec-8 ligand is expressed on human postmortem airway submucosal glands and cartilage. A bronchial tissue section from an asthmatic donor was stained for Siglec-8 ligand by Siglec-8-Fc overlay histochemistry (red). Whereas the epithelium (arrow) is devoid of staining, submucosal glands (arrowhead) and ducts (double arrowhead) are robustly stained, along with cartilage (asterisk). Scale bar, 100 μm. From Gonazalez-Gil et al., 2021.

Aggrecan is absent from tracheal submucosal glands. In airway submucosal glands and ducts (Fig. 5), and in mucus layer proteins collected by saline flush from airways of patients undergoing sinus surgery, Siglec-8 sialoglycan ligands are carried prominently on a single 1 MDa protein, a minor isoform and glycoform of DMBT1 (Gonzalez-Gil et al., 2021). DMBT1 (Deleted in Malignant Brain Tumor 1), also known as salivary scavenger and agglutinin (SALSA or SAG) and gp340 is an abundant secreted glycoprotein in the airway. As with aggrecan, only a minor isoform of DMBT1 carries the sialylated keratan sulfate chains that bind Siglec-8 (DMBT1^S8L). Double label staining for DMBT1 and Siglec-8 ligand demonstrated co-expression in submucosal glands and ducts, and comigration via gel electrophoresis. Intact DMBT1^S8L has picomolar affinity for Siglec-8, and its binding is eliminated by pretreatment with sialidase or keratanase (Gonzalez-Gil et al., 2021).

Expression of DMBT1 and Siglec-8 ligand are increased in chronically inflamed human upper airways in patients with chronic rhinosinusitis (Jia et al., 2015; Liu et al., 2004). Quantitative analysis of saline lavage from human sinuses and nasal pathways revealed that patients with chronic rhinosinusitis with nasal polyposis, who have increased airway eosinophil infiltration, express an increased ratio of Siglec-8 ligand to DMBT1 compared to patients without polyposis. These data imply human airway induction of functional glycosylation of DMBT1 to carry increased Siglec-8 sialylated keratan sulfate ligand (Lee et al., 2021), perhaps as a mechanistic response to eosinophilic inflammation. As in airway tissue (Fig. 5), Siglec-8-Fc overlay and DMBT1 immunohistochemistry reveal robust co-staining of polyp submucosal glands (Fig. 6). Induction of DMBT1^S8L expression is selective for polyp tissue compared to neighboring nasal airway tissue (Lee et al., 2021). How this selective increase in Siglec-8 binding glycans is regulated, and the degree to which it impacts disease progression, remain questions for future studies.

Fig. 6.

Colocalization of DMBT1 and Siglec-8 ligand in nasal polyp tissue. Tissue was surgically excised, fixed, sectioned, and immunostained with anti-DMBT1 (green, left panel) and overlaid with Siglec-8-Fc (red, center panel). Submucosal glands are intensely double-labeled (yellow, right panel). Scale bar, 40 μm. From Lee et al., 2021.

5.2. Siglec-8 ligand in human brain

In human brain, Siglec-8 is expressed robustly and selectively by microglia, the specialized brain-resident phagocytic immune cells (Galatro et al., 2017; Gonzalez-Gil et al., 2022). Siglec-8 expression is upregulated in Alzheimer’s disease (AD), where it may inhibit microglial phagocytosis (Morshed et al., 2020). Human brain Siglec-8 ligand copurifies with the Siglec-3 (CD33) ligand described above and has identical molecular properties (Gonzalez-Gil et al., 2022). As in the airway, the ligand is a sialylated keratan sulfate, is ~1MDa molecular weight, and is carried by a single glycoprotein. In brain, the carrier protein is a minor isoform and glycoform of RPTPζ (phosphacan). Expression of the Siglec-3/Siglec-8 cross-reactive ligand is upregulated in AD brain compared to age-matched tissue donors. There is a single mouse brain protein with the same characteristics that binds to the mouse paralog of Siglec-8, Siglec-F. It is a 1 MDa isoform and glycoform of mouse RPTPζ that carries Siglec-F reactive sialylated keratan sulfate chains that cross react with human Siglec-8. When RPTPζ is knocked out in mice, Siglec-F binding disappears, indicating that biosynthesis of the Siglec-F-binding sialylated keratan sulfate is selective for a single protein, RPTPζ (Gonzalez-Gil et al., 2022). The role of the shared Siglec-3/Siglec-8 ligand in AD progression has yet to be determined.

6. Siglec-9

Siglec-9 is broadly expressed by human blood leukocytes including neutrophils, monocytes, myeloid progenitor cell, natural killer cells and T-cells, among others (Duan and Paulson, 2020). Crosslinking Siglec-9 with antibodies or synthetic ligands on neutrophils leads to neutrophil death and on macrophages it inhibits phagocytosis (Delaveris et al., 2021a; Delaveris et al., 2021b; von Gunten et al., 2005). Siglec-9 ligands have been studied as targets in inflammation and cancer. Siglec-9 ligand expression in the tumor microenvironment is thought to help tumor cells evade immune clearance.

Glycan arrays reveal that Siglec-9 binds to sialyl Lewis X (Neu5Acα2–3Galβ1–4(Fucα1–3)GlcNAc), 6-sulfo LacNAc (Neu5Acα2–3Galβ1–4[6SO4]GlcNAc), and gangliosides with multiple terminal Neu5Acα2,3Gal groups like GD1a and GT1b (Yu et al., 2017). Siglec-9 also binds to the non-sialylated anionic glycosaminoglycan hyaluronic acid (Secundino et al., 2016).

A search for endogenous human Siglec-9 ligands that modulate inflammation led to the discovery of the sialoglycoprotein glycophorin on erythrocytes as a Siglec-9 ligand that inhibits neutrophil activation without suppressing bacterial killing (Lizcano et al., 2017). In sickle cell disease, this modulatory effect of erythrocytes on neutrophils via Siglec-9 is reduced, neutrophils are more prone to activation, and inflammation is increased (Kiser et al., 2020). A similar inhibitory effect was observed when neutrophils were incubated with high molecular weight hyaluronic acid, a mechanism that has been exploited by bacterial pathogens to avoid the immune clearance (Secundino et al., 2016). Differences in outcomes between crosslinking Siglec-9 with a high affinity polyvalent ligand versus glycophorin or high molecular weight hyaluronic acid suggest that affinity and/or avidity of a ligand can lead to different outcomes (Lizcano et al., 2017; Secundino et al., 2016).

Mucins have also been identified as carriers of Siglec-9 ligands. In human airway tissue and the airway cell line Calu-3, Siglec-9 ligand is carried on mucin MUC5B (Jia et al., 2015). Siglec-9 ligand expression is upregulated in chronic airway inflammation, and in Calu-3 cells treated with the Toll-like receptor agonist LPS through the NF-κB pathway. In tumors and cancer cells other mucins, including MUC1 and MUC16, are post-translationally modified to carry Siglec-9 ligands (Beatson et al., 2016; Tanida et al., 2013). Sialylated glycoforms of MUC1 have been shown to alter macrophage differentiation resulting in reduced phagocytosis (Beatson et al., 2020). Heavily glycosylated mucins carrying Siglec-9 ligands inhibit tumor clearance by natural killer cells, enhance tumor cell survival through tumor associated macrophages, and support metastasis. Siglec-9 ligands have also been observed in human aorta and are upregulated by high glucose levels in human umbilical vein endothelial cells (HUV-EC-C) (Zhang et al., 2019).

The major airway Siglec-9 ligand (MUC5B^S9L) is upregulated in chronic rhinosinusitus (Jia et al., 2015). This increase is selective for the protein carrier, MUC5B. A portion of DMBT1 recovered from human airways by nasal lavage also binds Siglec-9. DMBT1^S8L and DMBT1^S9L are overlapping subsets, with a minority of DMBT1^S9L co-purifying with affinity purified DMBT1^S8L, but most expressed as a separate glycoform. Densely O--glycosylated DMBT1 can carry Siglec-8 binding glycans, separate Siglec--9 binding glycans, or both. In an example of carrier-specific biosynthetic regulation, DMBT1^S8L increases in chronic rhinosinusitis with polyposis, whereas DMBT1^S9L does not (Fig. 7). These data provide evidence that selective regulation of glycan biosynthetic genes drive immune glycoform presentation to regulate different leukocyte subpopulations depending on their Siglec repertoire.

Fig. 7.

Siglec-8 ligand on DMBT1 is increased in chronic rhinosinusitis with nasal polyposis; Siglec-9 ligand on DMBT1 is not. Presurgical nasal lavage was collected from patients without inflammatory disease (NI), chronic rhinosinusitis without polyps (CRSsNP) or chronic rhinosinusitis with nasal polyps (CRSwNP). Equivalent volumes were resolved by gel electrophoresis, blotted, and double label probed with anti-DMBT1 and Siglec-Fc chimera. Band intensities were quantified and are reported as a ratio of Siglec-Fc binding (Siglec ligand) to DMBT1, normalized to the average NI value for each Siglec. NI, n = 16–17; CRSsNP, n = 11–12; CRSwNP (n = 27–28. *, p < 0.05; **, p < 0.02 Kruskall–Wallis test.

7. Siglec-10 and Siglec-11

The search for endogenous ligands for other immune inhibitory Siglecs is ongoing. One interesting example is Siglec-10 engagement with endogenous sialoglycoprotein CD24, a potential immune checkpoint ligand involved in immune-mediated tissue injury (Chen et al., 2009), placental immune suppression (Sammar et al., 2017), and cancer immune evasion (Barkal et al., 2019). Siglec-10 is expressed on several immune cell types including monocytes, microglia, B cells, eosinophils, and tumor-associated macrophages (Barkal et al., 2019; Duan and Paulson, 2020; Gonzalez-Gil et al., 2022). It binds well to model sialoglycans with either α2–3 or α2–6 sialic acid linkages (Forgione et al., 2020). Multiple studies have identified CD24 as a major and selective ligand for Siglec-10 (Barkal et al., 2019; Chen et al., 2009; Sammar et al., 2017). Mature CD24 is a very small (32 amino acid) GPI-anchored sialoglycoprotein that is heavily glycosylated at a dozen O-linked and two N-linked glycosylation sites (Li et al., 2022). Convincing data indicate that Siglec-10 binding to CD24 is sialic acid dependent and that functional interactions of Siglec-10 and CD24 result in immune suppression. Nevertheless, one study (Barkal et al., 2019) suggested that functional interactions between Siglec-10 and CD24 are sialic acid independent based on relative insensitivity to sialidase. The reason for the relative specificity of Siglec-10 for CD24 engagement among a sea of α2–3 and α2–6 linked sialoglycans, and whether the binding is truly sialic acid independent or only sialidase resistant, remain to be resolved (Yin and Gao, 2020).

Siglec-11 is expressed on macrophages and microglia and binds selectively to α2–8 linked oligo- and polysialic acids (Angata et al., 2002). Although full length Siglec-11 has 5 Ig-like domains, human brain microglia express a 4 Ig-like domain splice variant with the inner-most Ig domain missing. The shorter variant demonstrates enhanced binding to polysialic acid in vitro, and its binding to intact human neuroblastoma cells is diminished when polysialic acid is removed (Hane et al., 2021). Data support polysialic acid on nerve cells as endogenous human Siglec-11 ligands (Wang and Neumann, 2010). Since neural cell adhesion molecule (NCAM) is the dominant carrier of polysialic acid on neurons (Thiesler et al., 2022), designation of polysialylated NCAM as an endogenous ligand for Siglec-11 is consistent with these data.

8. Principles, insights and challenges

The study of endogenous human ligands for immune inhibitory Siglecs is at an early stage. Every new discovery in this active research area provides valuable direction and insights for ongoing studies. When viewed in a broad perspective, findings to date have provided initial principles and insights, and raised challenging additional questions.

8.1. Altered Siglec ligand expression in human diseases

Glycosylation changes are a hallmark of cancer (Bellis et al., 2022). In particular, hypersialylation, the enhanced expression of sialoglycans, is associated with both metastasis and immune evasion (Hugonnet et al., 2021). Since sialoglycans are ligands for immune inhibitory Siglecs, it is reasonable to propose that hypersialylation is a selective force in cancer progression due, at least in part, to immune evasion. This area of basic and translational research is under increasing scrutiny (Laubli et al., 2021; Lim et al., 2021).

Expression of human airway ligands for Siglec-8 and Siglec-9 are increased in chronic rhinosinusitis a chronic inflammatory disease of the upper airways (Jia et al., 2015; Lee et al., 2021). One hypothesis is that biosynthesis of Siglec ligands is increased as a response to chronic inflammation in a biological pathway to damp ongoing inflammation and resolve inflammatory events. The outstanding challenge is to understand why increased Siglec ligand expression does not resolve neutrophilic and eosinophilic inflammation in this disease. Likewise, Alzheimer’s disease patients had twice as much of the shared ligand for Siglec-3 and Siglec-8 than age matched control donors (Gonzalez-Gil et al., 2022). How brain expression of Siglec ligands is controlled remains to be determined.

8.2. Siglec ligands are tissue and cell specific

An intriguing finding in the study of Siglec ligands is that binding of a particular Siglec is often to glycans carrier on a single major carrier protein, and that the carrier protein may vary tissue-to-tissue. Notable examples are the endogenous ligands of Siglec-8 (Gonzalez-Gil et al., 2021; Gonzalez-Gil et al., 2022; Gonzalez-Gil et al., 2018). Airway submucosal gland cells secrete the Siglec-8 sialoglycan ligand, a sialylated keratan sulfate, nearly exclusively on the protein DMBT1, whereas airway cartilage cells express the same general sialoglycan structure on aggrecan, and brain cells on RPTPζ. The biosynthetic flexibility of this selectivity has not been tested in humans, but has in mice (Gonzalez-Gil et al., 2022). In mouse brain RPTPζ exclusively carries a sialylated keratan sialoglycan ligand for Siglec-F, a Siglec-8 paralog. In mice lacking RPTPζ, all Siglec-F binding in the brain disappears. This is remarkable, since the enzymes required to synthesize the Siglec-F sialoglycan ligand remain intact yet are not able to synthesize Siglec-F ligand on any alternate protein carrier. Expanding the list of examples, knockout of CD43 alone on K562 human myeloid leukemia cells reduced Siglec-7 binding by half (Wisnovsky et al., 2021), and binding of Siglec-10 to CD24-null splenocytes (albeit mouse splenocytes) was completely lost in comparison to wild type mouse splenocytes (Chen et al., 2009). The latter example is particularly compelling, since Siglec-10 binds to α2–3 and α2–6 linked sialoglycans (Forgione et al., 2020), of which there are many diverse alternatives. The structural features that lead to selective expression of Siglec ligands on a single protein carrier that varies among tissues are as yet unknown.

8.3. Siglec ligands as quantitatively minor isoforms and glycoforms of abundant proteins makes nomenclature challenging

Since a human cell type or tissue may express a particular Siglec ligand on a single protein backbone, it is facile to refer to the ligand by its protein backbone name. However, this is incomplete and can be misleading. A portion of human tracheal cartilage aggrecan carries the Siglec-8 sialoglycan ligand, but aggrecan, per se, is not the ligand. A quantitatively minor isoform and glycoform of RPTPζ in human brain carries the glycan determinant that binds CD33, whereas many human brain isoforms and glycoforms of RPTPζ do not. To refer to aggrecan as the Siglec-8 ligand or RPTPζ as the CD33 ligand is misleading. It is proposed to distinguish the ligand from other protein isoforms and glycoforms using a superscript, such as aggrecan^S8L (Siglec-8 ligand) or RPTPζ^S3L (Siglec-3 ligand) (Taylor et al., 2022). This nomenclature emphasizes that an essential part of the ligand is it’s Siglec-binding glycan, which is not shared by all proteins of that name.

8.4. Experimental challenges for studying endogenous human ligands for inhibitory Siglecs

Human Siglecs are not well conserved in other mammals, including laboratory animal models (Angata, 2006; Duan and Paulson, 2020). This makes human tissues and cells essential for the study of Siglec ligands. For example, whereas Siglec-8 ligands are well characterized in human airways (Gonzalez-Gil et al., 2021; Gonzalez-Gil et al., 2018), Siglec-8 fails to bind at all to mouse airways (Yu et al., 2017). In contrast, mouse Siglec-F, a paralog of human Siglec-8, binds abundantly to the epithelium and submucosal glands of mouse airways, indicating alternative mouse Siglec ligands are present (Kiwamoto et al., 2015; Yu et al., 2017). The situation in the brain is different, in that human CD33 and Siglec-8 and mouse Siglec-F all bind to the same ligand from mouse or human brain (Gonzalez-Gil et al., 2022). From these examples, the similarity of non-human Siglec ligands to their human counterparts requires experimental determination.

An alternative to human tissues are primary human cells, established human cell lines, or human induced pluripotent stem cell (hiPS) derived cells. There are examples of recapitulation of human tissue Siglec ligand expression by related cell lines, for example human airway Siglec-9 ligand expression by the Calu-3 human airway cell line (Jia et al., 2015). However, the same cell line failed to express Siglec-8 airway ligands. Although primary human trachea submucosal gland cells in culture expressed both Siglec-8 and Siglec-9 ligands, Siglec-8 ligand expression disappeared quickly in culture (Jia et al., 2015; Yu et al., 2017). There are insufficient data to reveal whether hiPS-derived human cells will more consistently recapitulate human tissue expression of Siglec ligands. While human cancer cell lines are meaningful models of human cancer, cell-to-cell variability in Siglec ligand expression and comparison with cancer tissues in vivo are valuable additions to this literature.

8.5. Siglec ligand affinity may far exceed that of its sialoglycan binding determinants

The affinity of endogenous human Siglec ligands for their Siglecs has seldom been reported. One example serves to raise an unanswered question of how ligand affinity compares to sialoglycan binding determinant affinity. Siglec-8 ligand isolated from human airway lavage (DMBT1^S8L) binds to chimeric Siglec-8-Fc with a K_D of 60 pM (Gonzalez-Gil et al., 2021). In contrast, monovalent site affinity for the optimal binding determinant (Neu5Acα2–3[6SO4]Gal[6SO4]GlcNAc) is ~200 μM (Propster et al., 2016). The structural basis for the million-fold affinity increase of the intact ligand has not been established. It may be based on additional structural determinants on the sialoglycan chains of DMBT1^S8L, contributions from the polypeptide, and/or multivalent spacing and presentation of sialoglycan determinants on the polypeptide backbone. In terms of multivalent presentation, it is notable that several Siglec ligands contain heavily O-glycosylated domains, including CD24, CD43, DMBT1, aggrecan, and RPTPζ. Whether sialoglycan clustering or other structural components contribute to high affinity binding between Siglecs and Siglec ligands awaits further studies. The mechanisms of enhanced affinity may provide knowledge useful in designing Siglec-binding mimetics.

9. Conclusion

The study of endogenous human Siglec ligands is emerging as a complement to the study of Siglecs and Siglec mimetics to therapeutically target immune regulation (Gonzalez-Gil and Schnaar, 2021). As effective therapeutics emerge to target other classes of immune checkpoint receptors and their ligands alike (Gaikwad et al., 2022; Kaushik et al., 2022), it is anticipated that enhanced understanding of the molecular structures, biosynthesis, expression and regulation of endogenous ligands that engage Siglecs will provide additional mechanistic and potentially therapeutic insights.