Main

The treatment of diseases such as cancer and inflammation is extremely challenging because the pathology involves dysregulation of endogenous and often essential cellular processes. Effective therapies typically capitalize on differences between diseased and healthy tissues that can be targeted with drugs. Many cancer drugs, for example, target the cell cycle (for example, DNA replication and cytoskeletal formation). More recently, signalling pathways have become the focus of new anticancer drugs (for example, kinases and phosphatases)1,2. Inflammation therapies focus on suppressing the immune system (for example, antagonists of tumour-necrosis factor-α (TNFα), steroids)3. Though combination therapies show improved selectivity, these therapies suffer from side effects that result from imperfect selectivity for disease tissue. The availability of novel molecular targets that distinguish diseased from healthy cells could vastly amplify therapeutic opportunities. An ongoing challenge is therefore to identify other disease-associated changes in cellular physiology that might be harnessed for more-selective treatment and diagnosis.

Glycobiologists have known for decades that the structures of glycans, which decorate all eukaryotic cell surfaces, change with the onset of cancer4 and inflammation5,6,7. The glycosylation pattern of a cell is therefore a code for cellular physiology. An understanding of this code at both molecular and functional levels is starting to emerge. Although glycobiologists are still at the early stages of deciphering this code, their results have already inspired novel methods to detect glycans characteristic of disease conditions, to prevent disease glycan formation and to destroy cells that display them.

Here we provide an overview of the link between glycan structures and disease progression, with a particular focus on cancer and chronic inflammation. We also highlight opportunities for therapeutic and diagnostic approaches that are based on the underlying glycobiology. This burgeoning field should provide considerable fuel for new academic and industrial enterprises.

Glycobiology

Although it has long been appreciated that glycan expression changes with cellular condition, progress toward delineating the molecular basis of glycan function has been rather slow relative to comparable studies of proteins and nucleic acids. This slow progress is partly due to the fact that the biosynthesis of glycans, unlike other biopolymers, is not template-driven. In mammals, glycans are constructed from nine monosaccharide building blocks that can be connected to one another through glycosidic linkages in myriad combinations. Glycans are formed on their protein or lipid scaffolds by GLYCOSYLTRANSFERASES and GLYCOSIDASES as they traffic through the secretory pathway. The combined action of these enzymes in the secretory pathway leads to a diverse array of glycan structures. The major classes of glycan structures are shown in Box 1. N-linked glycans are attached to an asparagine residue of a protein carrier, O-linked glycans are attached to serine or threonine residues, and glycosaminoglycans (GAGs) are attached to serine residues of proteoglycan molecules. By contrast, glycolipids have a lipid carrier, such as sphingolipid. Glycan structures can vary from highly branched and complex glycans (for example, N-linked and O-linked glycans, and glycolipids) to linear glycans (for example, proteoglycans). Due to the large number of possible structures, the information content of glycans is enormous. Cracking the 'glycan code' as it relates to disease states has been a major focus of glycobiologists during the past several decades8.

Description of tumour-associated glycans

Altered glycosylation patterns are a hallmark of the tumour phenotype. This phenomenon was first described by Meezan et al. in 1969 with the demonstration that healthy fibroblasts have smaller membrane glycoproteins than their transformed counterparts4. This finding was later corroborated with histological evidence that LECTINS show differential binding to healthy compared with malignant tissue5,9. More recently, cancer-associated cell-surface glycans have been directly characterized using specific monoclonal antibodies and mass spectrometry10, further illuminating the molecular changes that occur upon malignant transformation11,12.

We now know that changes in glycosylation include both the under- and overexpression of naturally-occurring glycans, as well as NEOEXPRESSION of glycans normally restricted to embryonic tissues. These structures most often arise from changes in the expression levels of glycosyltransferases in the Golgi compartment of cancerous cells. Changes in glycosyltransferase levels can lead to modifications in the core structure of N-linked and O-linked glycans. One of the most common changes is an increase in the size and branching of N-linked glycans. This increased branching is often attributed to increased activity of N-acetylglucosaminyltransferase V (GlcNAc-TV, also known as or MGAT5; the enzyme that leads to β1,6GlcNAc branching)13. The increased branching creates additional sites for terminal sialic acid residues, which, in conjunction with a corresponding upregulation of sialyltransferases, ultimately leads to an increase in global sialylation14.

In addition to changes in the core structures of glycans, altered terminal structures are also associated with malignancy. Glycosyltransferases (for example, sialyltransferases and fucosyltransferases) involved in linking terminating residues on glycans tend to be overexpressed in tumour tissue. The increase in activity of these glycosyltransferases in turn leads to the overexpression of certain terminal glycans. Examples of terminal glycan epitopes commonly found on transformed cells include sialyl Lewis x (sLex), sialyl Tn (sTn), Globo H, Lewis y (Ley) and polysialic acid (PSA) (Fig. 1)15,16,17,18. Many of these epitopes are observed in malignant tissues throughout the body, including the brain, breast, colon and prostate (Table 1)19.

Figure 1: Cancer-associated glycans.
figure 1

Altered glycosylation patterns are a hallmark of the tumour phenotype, consisting of both the under- and overexpression (for example, sialyl Lewis x (sLex)) of naturally-occurring glycans as well as the presentation of glycans normally restricted to expression during embryonic development (for example, polysialic acid (PSA)). The oligosaccharide epitopes depicted here are a small subset of glycans that are found on malignant cells. Glycans presented on a cell's surface act as a code to convey information about the physiological state the cell. For example, high levels of the capping monosaccharide sialic acid suggest a poor prognosis (high metastatic ability) in many types of cancers.

Full size image
Table 1 Common expression patterns of cancer glycans on malignant tissues*
Full size table

Another common feature of tumours is the overproduction of certain glycoproteins and glycolipids. For example, epithelial tumours often overproduce mucin glycoproteins, which are characterized by dense clusters of O-linked glycans. Mucins are used as diagnostic markers of cancer and can also function as scaffolds for most of the above-listed cancer-associated epitopes20. Additionally, cancer tissues can display an increase in GANGLIOSIDE expression. For example, complex gangliosides (for example, GD2, GD3 and fucosyl GM1) (Fig. 1) are found at elevated levels in small-cell lung carcinomas, neuroblastomas and melanomas16,21. Although gross changes in glycosylation of tumour tissues are apparent, no single change seems to distinctly differentiate normal and malignant cells. Instead, each type of malignant tissue is characterized by a distinct set of changes in glycan expression (Table 1)22,23.

Carbohydrate-based anticancer vaccines

Because cancer cell glycans differ from those found on their healthy counterparts, it might be possible to recruit the immune system to target cancer cells on the basis of their altered glycosylation. Although atypical glycans can render cancer cells mildly antigenic (that is, capable of eliciting specific antibodies), they are rarely immunogenic (that is, the antibodies are not capable of recruiting immune effector functions to kill the cells)24. Many tumour-associated glycans have an embryonic origin or are expressed at low levels in normal tissue and at elevated levels on tumours. Consequently, the glycans can be perceived as 'self' by the human immune system, in which case B cells expressing high-affinity antibodies for these structures would have been eliminated during development25,26. For this reason, many attempts at generating anticancer vaccines have focused on breaking immune self-tolerance for tumour-associated glycans27,28.

Danishefsky, Livingston and co-workers have prepared an array of carbohydrate-based anticancer vaccines by linking multiple copies of synthetic tumour-associated glycans to the immunogenic carrier protein keyhole limpet haemocyanin (KLH)27. This foreign protein provides peptide antigens that are required for T-cell help and a full-blown immune response. When administered to mice and/or humans, many of these glycoconjugate vaccines elicited antibody- and cell-mediated immune responses against the glycan27,29,30,31,32,33,34 (Fig. 2).

Figure 2: Carbohydrate-based anticancer vaccines.
figure 2

Upon stimulation with carbohydrate-based anticancer vaccines, the immune system can be stimulated to recognize cells presenting cancer glycans. a | For example, a vaccine containing the polysialic acid (PSA) epitope conjugated to the immunogenic carrier protein keyhole limpet haemocyanin (KLH) (PSA–KLH) raised a mild immune response in animals33. b | Recent reports have found that the chemically modified (unnatural) PrPSA–KLH conjugate raised a stronger immune response in animals than the natural PSA–KLH conjugate33. c | Finally, a passive immunization strategy in which mice were treated with N-propanoylmannosamine (ManNProp), a biosynthetic precursor of N-propionylated PSA, and a monoclonal antibody raised against synthetic N-propionylated PSA (anti-PrPSA), showed success in preventing metastasis of PSA-positive cells in mice39.

Full size image

Several glycan-based vaccines are presently undergoing clinical evaluation and have shown some promise27,29,30,31,32,33,34,35 (Table 2). In a Phase III clinical trial for metastatic breast cancer, the sTn–KLH conjugate Theratope (Biomira) failed to meet the endpoints of time-to-disease progression and overall survival. However, patients treated with Theratope in conjunction with hormone therapy had improved survival rates with a time-to-disease progression of 8.3 months compared with 5.8 months for those on hormone therapy alone36. This modest clinical efficacy could be due to the fact that Theratope elicits a B-cell-mediated immune response but does not seem to trigger a T-cell-mediated immune response36.

Table 2 Carbohydrate-based anticancer vaccines under development
Full size table

Successful tumour immunotherapy might require the induction of cytotoxic T lymphocytes (CTLs) in addition to antibodies. There are strategies currently being pursued to recruit T-cell help, including mimicking the cancer cell surface by displaying vaccine glycans in a multivalent context, chemically modifying cancer glycans to make them more 'foreign', and actively recruiting T-cell help by exploiting the naturally abundant human 'anti-Gal' antibody. Each of these glycan-based vaccine strategies is discussed below.

Each type of malignant tissue is characterized by a distinct set of changes in glycan expression (Table 1). A vaccine that targets several cancer-associated glycans should, in principle, lead to a stronger and more specific immune response than one that targets a single cancer glycan29,37. In an effort to mirror the heterogeneous nature of cancer-cell-surface glycans, Danishefsky and co-workers have prepared multiantigenic vaccines that contain several different glycan structures29,38. For example, a tri-antigenic vaccine containing Globo H, Ley and Tn has been shown in animal models to elicit an immune response against each oligosaccharide antigen29. In this preclinical study, antibodies that recruit T-cell help (immunoglobulin G (IgG) isotype) were generated against each glycan antigen. This promising result suggests that a multiantigenic approach might be successful at recruiting both humoral and T-cell mediated immune responses against tumours in human patients.

Chemical modification of monosaccharide structures can augment the immunogenicity of glycan-based vaccines28,33. Chemically modified sialic acid residues with unnatural N-acyl side chains, such as N-propanoyl33,39,40,41, N-butanoyl40,41,42, N-phenylacetyl40,41 and N-levulinoyl43, have been incorporated into KLH conjugates40,41 and are more immunogenic than the corresponding natural sialic acid vaccines33,40,41 (Fig. 2). N-propanoyl, N-butanoyl and N-phenylacetyl sialic acid derivatives conjugated to KLH produced a robust immune response in mice. These derivatives elicited IgG antibody production, which recruits T-cell help. By contrast, the natural N-acetyl-sialic-acid–KLH conjugate was essentially non-immunogenic40. Similarly, Livingston and co-workers conducted a preliminary clinical trial in which small-cell lung carcinoma patients were vaccinated with polysialic acid (PSA)–KLH or N-propionylated PSA–KLH (PrPSA–KLH) conjugates. The modified PrPSA–KLH conjugate elicited specific antibodies that crossreacted with PSA, whereas the natural PSA conjugate did not (Fig. 2)33.

Chemically modified sialic acid conjugates have also shown success in PASSIVE IMMUNIZATION in animal models. Jennings and co-workers demonstrated that administration of a monoclonal antibody raised against a synthetic PrPSA–KLH conjugate prevented metastasis of a leukaemic cell line when the mice were treated with N-propanoylmannosamine, a biosynthetic precursor of N-propionylated PSA (Fig. 2)39. More recently, the same group applied this passive immunization strategy to melanomas expressing the sialylated glycolipid GD342.

T-cell-mediated immune reactions against cancer glycans can be elicited using the natural human 'anti-Gal' antibody as an endogenous adjuvant44. Although cancer glycans are not normally detected by the human immune system, they can be processed and presented to T-cells if the cancer cells express another epitope on them that is detected by the immune system. Galili and co-workers have developed a vaccine strategy that takes advantage of the 'anti-Gal' antibody, which constitutes ∼1% of circulating IgG in humans. The 'anti-Gal' antibody reacts specifically with the α-Gal epitope (Galα1,3Galβ1,4GlcNAcR), which is entirely absent from human tissues. This antibody is thought to originate from widespread exposure to bacteria that express similar glycans. The α-Gal epitope can be enzymatically synthesized on human tumour cells by the use of recombinant α-1,3-galactosyltransferase (α1,3GT)45 or by infection of the tumour cells with adenovirus containing the gene encoding α1,3GT46. These α-Gal epitope-expressing tumour cells or their membranes can then be used as a vaccine to elicit an immune response against tumour-associated antigens47.

Finally, there are several examples of specific glycoproteins that undergo changes in glycosylation upon malignant transformation. These include mucin 1 (MUC1)48, MUC16 (also known as CA125)49, prostate-specific antigen50,51 and carcinoembryonic antigen (CEA)52. MUC1, a mucin-type glycoprotein containing numerous O-linked glycans, is found on healthy breast tissue and on breast cancer tissue, but it is aberrantly glycosylated in the cancerous tissue20,48. Truncation of the O-glycans in breast cancer leads to the appearance of novel carbohydrate epitopes, such as Thomas Friedrich (TF) and sialyl Tn antigens17,48,53. These changes in glycosylation reflect both the upregulation of a sialyltransferase (ST3Gal1), which transfers sialic acid onto the nascent O-glycan, and the underexpression of the glycosyltransferase core II GlcNAc-TI, which transfers another sugar (GlcNAc) to these sites53,54.Vaccines based on underglycosylated MUC1 (uMUC1) are currently undergoing clinical evaluation with encouraging results55,56,57 (Table 2).

Tumour glycans: prospects for new drugs

Carbohydrate-based anticancer vaccines can be effective irrespective of the functional significance of cancer-associated glycans. By contrast, drug development strategies based on preventing the formation of cancer glycans or interrupting their downstream interactions can only be effective if these glycans make a functional contribution to the disease.

Despite mounting clinical evidence suggesting a correlation between changes in glycosylation and poor prognosis58,59,60,61, dissecting the role of a specific glycan in the biology of a given tumour has proven to be extremely challenging14. Complications include the MICROHETEROGENEITY of glycosylation and the fact that many factors other than glycosylation contribute to malignant transformation14. In many cases, the causal link between carbohydrates and malignant function is not clearly understood. Hypotheses range from protection of tumour cells from immune surveillance by their glycan coat to promotion of tumour-cell metastasis by glycan-mediated adhesion to distant sites14,20,62. However, in some cases the causal link is clear — certain glycan structures (for example, β1,6GlcNAc-branching and sLex) contribute to oncogenic signalling63, transformation64 and metastasis13,62.

Several approaches have been undertaken to assess the functional roles of tumour-associated glycans. In each case, cells with different glycosylation profiles have been selected or created, and their physiological behaviour has been compared. Cellular glycans have been structurally perturbed by treatment of cells with glycosylation inhibitors or glycosidases, or by altering cellular glycosyltransferase expression. Summarized below are examples of experiments that provide evidence that these tumour-associated changes in glycosylation are functionally relevant.

Levels of sialic acid affect metastasis. Cells can be selected for alterations in cell-surface glycan expression using lectins65. Wheat germ agglutinin (WGA), a cytotoxic plant lectin that binds to sialic acid or β-GlcNAc residues, was used to select WGA-resistant B16 melanoma cells. These WGA-resistant cells showed reduced cell-surface sialylation and metastatic activity relative to the parental B16 cell line66. This correlation suggested a role for sialic acid in promoting metastasis. This interpretation was further supported in an experiment by Tokuyama and co-workers. They addressed the functional role of sialic acid residues in tumour metastasis by stably transfecting BL6 melanoma cells with a sialidase, an enzyme that removes sialic acid residues from glycans67. These transfected cells showed a significant reduction in lung colonization in a mouse model relative to the parental cell line.

Increased β1,6GlcNAc branching of N -linked glycans transforms cells. The role of β1,6GlcNAc-branching of N-linked glycans in cancer has been extensively studied. Demetriou et al. addressed the functional role of the branching β1,6GlcNAc residue in malignant transformation by transfecting an immortalized lung epithelial cell line (Mv1Lu) with GlcNAc-TV, the glycosyltransferase that installs this residue on N-linked glycans. GlcNAc-TV-transfected cells formed solid tumours in mice, whereas untransfected cells did not. In addition, the transfected cells showed a loss of contact inhibition of growth and increased motility in culture64. These data provide strong evidence that upregulation of this enzyme can induce malignant change. Consistent with this hypothesis, disruption of GlcNAc-TV in mice reduced malignancy68. Dennis and co-workers showed that GlcNAc-TV-deficient mice have suppressed mammary tumour growth and metastasis relative to littermates expressing the glycosyltransferase68. Finally, Pierce and co-workers reported that GlcNAc-TV-expression levels regulate cadherin-associated homotypic cell–cell adhesion and downstream signalling pathways63. Collectively, these data suggest a functional, rather than simply correlative, role for β1,6-GlcNAc branching in tumour formation.

sLex–selectin interactions can promote metastasis. Tumour metastasis is facilitated by adhesion between tumour cells and either platelets in the bloodstream or remote endothelial cells69. A well-known example of a receptor–ligand binding interaction that initiates this adhesion involves the SELECTINS. These selectins bind to sLex expressed on tumour cells and are thought to contribute to tumour-cell migration to distant tissues70.

Esko and co-workers used metabolic inhibitors, termed disaccharide decoys, to study the role of sLex in metastasis71,72,73,74. The decoys act as competitive substrates of glycosyltransferases and ultimately lead to the production of truncated glycans on the cell surface71,72,73,74. A disaccharide precursor of sLex, GlcNAcβ1,3Galβ-O-napthalenemethanol, was shown to reduce levels of cell-surface sLex (Fig. 3) on LS180 colon adenocarcinoma cells. Relative to untreated cells, these sLex-deficient cells had decreased interactions with selectins, increased susceptibility to leukocyte-mediated lysis and a reduction of lung metastasis in a murine tumour model71. Additionally, no significant difference in metastasis was observed when decoy-treated and untreated cells were injected into a P-selectin-deficient mouse, supporting the hypothesis that the sLex–P-selectin interaction participates in tumour-cell dissemination75,76,77. Experiments with metabolic decoys raise the possibility of using such compounds as antimetastatic drugs74.

Figure 3: Disaccharide decoys act as metabolic inhibitors of glycosylation.
figure 3

a | The disaccharide N-acetylglucosamine β1,3 galactose (GlcNAcβ1,3Gal) is synthesized on O-linked glycoproteins and is further elaborated by glycosyltransferases in the Golgi to form the sialyl Lewis x (sLex) epitope. b | In the presence of high concentrations of the disaccharide decoy, GlcNAcβ1,3Galβ-O-napthalenemethanol, glycosyltransferases in the Golgi are diverted away from the normal glycoprotein substrate and instead form sLex on the decoy. Ultimately, cells treated with the disaccharide decoy have reduced levels of sLex and present truncated glycans on their cell surfaces.

Full size image

In complementary work, Shirota et al. investigated the effects of an inhibitor of sLex–selectin binding on metastasis78. They used GSC-150, which is an analogue of sLex that competitively binds to the selectins. Co-administration of mice with GSC-150 and the colon carcinoma cell line KM12-HX reduced metastasis to the liver compared with mice treated with the cells alone. This result suggested that the sLex–selectin interaction is a contributing factor to metastasis and that selectin inhibitors might find use in cancer treatment.

Interestingly, the clinically approved anticoagulant drug heparin inhibits tumour metastasis in experimental animal models79 and has shown beneficial effects in human clinical trials of colon cancer80. Varki and co-workers found that heparin disrupted P-selectin-mediated interactions of platelets with carcinoma cell-surface mucins75. These cell-surface mucins are sialylated and fucosylated and bear the P-selectin-binding epitopes sLex and sLea. These findings are particularly exciting because heparin is already an approved drug.

Given the functional link between aberrant glycosylation and malignancy, therapeutics that block the formation of cancer-associated glycans or their downstream interactions should have an effect on tumour progression. A number of glycosylation inhibitors and receptor inhibitors for the treatment of cancer are currently under development in pharmaceutical companies (for example, GlycoDesign (GD39), Oxford Glycoscience (OGT719), GlycoGenesys (GCS-100)).

Diagnostics based on cancer glycans

Glycan changes that indicate malignancy can be used for diagnosis. Indeed, the commonly used 'CEA' test for colon cancer monitors serum levels of an antibody specific for a cancer-associated glycan (sLea)52. Glycan analyses of other serum markers, such as CA12549 and prostate-specific antigen50,51, reveal distinct changes in glycosylation in ovarian and prostate tumour tissue, respectively. These studies suggest that specific changes in glycosylation could be useful as diagnostic tools. For many cancers, however, there are no serum markers available. Instead, the tissue must be analysed directly. Existing diagnostic methods used to monitor tumour-specific glycosylation require surgical biopsy followed by histological analysis with lectins or monoclonal antibodies. An interesting future direction in the field is to target aberrant glycosylation with probes for image contrast17.

To date, the field of molecular imaging has focused on protein-based markers that are targeted with antibody conjugates, receptor ligands or enzyme substrates/inhibitors81,82,83. So far there is only one report of indirectly imaging a change in glycosylation84,85. Underglycosylated MUC1 antigen (uMUC1) is highly overexpressed in a number of human epithelial cell adenocarcinomas and is an early marker of tumorigenesis48,53. In its underglycosylated state, this protein can be recognized by a specific peptide (termed EPPT1) derived from a monoclonal antibody17,84. Epenetos and co-workers developed an EPPT1-99mTc probe to monitor breast cancer in vivo84. More recently, Dai and co-workers synthesized an EPPT1 conjugate capable of magnetic resonance and optical imaging85. Experiments with this multimodal probe in uMUC1-positive murine tumour models showed specific accumulation of the probe in the tumour tissue, with very low background signal in uMUC-1-negative tumours. However, uMUC1-specific imaging targets the underlying underglycosylated protein, not the glycans directly.

The direct imaging of changes in glycosylation might be achieved by tapping into the underlying metabolic processes of the cell. Monosaccharide substrates can be labelled with a non-invasive reporter, such as a radionuclide (18F or 125I), to follow the metabolic fate of these substrates by radionuclide imaging. Current non-invasive positron-emission tomography imaging technologies for the early diagnosis of cancer rely on the elevated accumulation of the simple sugar 2-fluoro-2-deoxy-glucose (FDG) in tumour tissue relative to healthy tissue83. FDG is an unnatural (18F)-containing glucose analogue that concentrates in malignant tumours owing to their enhanced glycolysis rates. Similarly, the metabolic fates of 14C-radiolabelled monosaccharides have been monitored in animal tumour models, albeit using invasive methods (for example, surgical biopsy)86.

We have recently reported the ability to tag cell-surface glycans with probes in vivo87. In a process termed metabolic oligosaccharide engineering, unnatural sugars bearing BIOORTHOGONAL FUNCTIONAL GROUPS can be metabolically introduced into cellular glycans86,88,89. We have adapted this technique for the cell-surface display of glycans bearing azido groups90. The azide is not endogenous to biological systems and is inert to biological components, yet will react with another bioorthogonal functional group, the phosphine, in a selective manner (Box 2). The covalent reaction of azides and phosphines to form a stable adduct is a reaction termed the Staudinger ligation90 (Box 2). This reaction can chemically tag azide-functionalized sugars in living animals87.

In these experiments, azide-containing sialic acid residues (SiaNAz) can be metabolically incorporated into membrane glycoconjugates in mice that are treated with the azidosugar precursor N-α-azidoacetylmannosamine (ManNAz)87. Because cancer cells are known to overexpress sialic acids and are more metabolically active than normal cells, it is possible that these cells would incorporate disproportionately high amounts of SiaNAz into their cell-surface glycans. Subsequent reaction between azide-modified cancer cells and phosphine-conjugated imaging probes would then allow discrimination of cancerous tissue from the surrounding healthy tissue using non-invasive diagnostic techniques such as magnetic resonance imaging (MRI) (Fig. 4). As a proof-of-principle, Lemieux et al. found that cells expressing high levels of sialic acid could be labelled in cell culture with unnatural sialic acids and preferentially tagged with an MRI-contrast reagent91. Furthermore, changes in glycan expression that accompany the onset or progression of other diseases, such as chronic inflammation, might be identified or tracked non-invasively by chemically tagging glycans in vivo. This chemical tagging could be useful for finding new targets for drug development.

Figure 4: Imaging chemically modified cellular glycans in vivo.
figure 4

In a process termed metabolic oligosaccharide engineering, unnatural sugars can be biosynthetically introduced into cellular glycans86,88,89. For example, azide-containing sialic-acid residues (SiaNAz) can be metabolically incorporated into membrane glycoconjugates in mice that are treated with the azidosugar precursor N-α-azidoacetylmannosamine (ManNAz)87. If sialic-acid expression is upregulated in cells with an altered physiological state, such as cancer cells, then these cells should incorporate large amounts of SiaNAz. Subsequent reaction between azide-modified cells and phosphine-conjugated imaging probes (Staudinger ligation; see Box 2) should allow non-invasive detection of altered tissue. Alternatively, reaction of azide-modified glycans with phosphine-conjugated affinity purification tags should allow enrichment and subsequent identification of glycans on cells with an altered physiological state.

Full size image

More broadly, the azide might serve as an in vivo reporter of glycan expression. In principle, any sugar could metabolically be labelled with an azide if the biosynthetic enzymes are tolerant of azido substrates. In addition to the sialic acid biosynthetic pathway, the N-acetylglucosamine (GlcNAc) and N-acetylgalactosamine (GalNAc) salvage pathways are tolerant of azido analogues92,93. N-α-azidoacetylglucosamine and N-α-azidoacetylgalactosamine are incorporated into O-GlcNAc-modified proteins and mucin-type O-linked glycoproteins, respectively. We and others are applying this phenomenon to proteomic analysis of glycoprotein subtypes92,93.

Inflammation-associated glycans

In contrast to cancer-associated glycans, where the effects of changes in cellular glycosylation on tumour progression are not fully understood, the functions of glycans found at sites of chronic inflammation are well-defined. Inflammatory diseases, such as arthritis, psoriasis, asthma and diabetes, are characterized by leukocyte homing into the affected tissues. A crucial first step in the normal entry of circulating lymphocytes into peripheral lymph nodes and leukocyte emigration into inflamed tissues is their adhesion to activated endothelial cells lining blood vessel walls94. In addition to their role in cancer metastasis (discussed above), the selectins (E-, P- and L-selectin) mediate the transient initial interaction that results in rolling of the leukocytes along the endothelial surface95,96,97.

The selectins bind sialylated and fucosylated epitopes such as sLex, often in sulphated form, which comprise a terminal component of glycans on most leukocytes, endothelial cells in the lymph node and the endothelium of inflamed tissues98,99. These epitopes have been studied extensively with monoclonal antibodies, such as MECA-79, which binds to a collection of endothelial cell glycoproteins known as peripheral node addressin or PNAd100. MECA-79 binds glycan epitopes in a sulphate-dependent manner and shares similar specificity with L-selectin. MECA-79 and L-selectin ligands (for example, 6-sulpho-sLex) have been shown to be absent from non-inflamed endothelial tissues but are prominently expressed on the endothelium of chronically inflamed tissues (Table 3)95,101,102. These glycan epitopes act as endothelial zip codes for the homing of lymphocytes, which constitutively express L-selectin, to inflamed tissues95.

Table 3 Induction of sLex and 6-sulpho-sLex in inflammatory diseases
Full size table

Inhibition of leukocyte emigration into inflamed tissue is an attractive target for anti-inflammatory therapy103. Like most inhibitors of carbohydrate-binding proteins, however, selectin blockers face the challenge of inhibiting multivalent, low-affinity interactions in vivo. Therapeutics that block selectins (for example, selectin inhibitor PSGL-1104) or the biosynthesis of their ligands (for example, inhibitors of the glycosyltransferases105,106,107 and sulphotransferases108 that synthesize sLex or 6-sulpho-sLex) are presently under development.

In addition, there is an opportunity for the development of non-invasive diagnostics that might identify sites of chronic inflammation prior to the presentation of disease symptoms. For example, the best diagnostic indicator of type 1 diabetes is the presence of autoantibodies against pancreatic islet cells. By the time these autoantibodies are detectable, a significant amount of autoimmune destruction has already occurred. Weissleder and co-workers have recently reported an improved diagnostic for inflammatory diabetes that targets an earlier event in disease progression — leaky vasculature109. Still-earlier detection of inflammatory diabetes might involve imaging the expression of L-selectin ligands, the induction of which presumably precedes inflammatory tissue damage. Sibson et al. reported a strategy for the early detection of inflammation in the brain by targeting the E-selectin–sLex interaction110. An MRI-contrast reagent coupled to an sLex mimetic that binds E-selectin revealed the presence of inflammation in the brain that was otherwise invisible by MRI110. Strategies for the early detection of other inflammatory diseases could similarly be improved by imaging the early changes in glycosylation.

Other disease-associated glycans

Although the focus of this article is on changes in glycosylation that occur in diseased tissue within humans, there are also opportunities for targeting the unique glycosylation patterns of pathogenic microbes, such as those that cause malaria or meningitis. Recently, glycan-based vaccine candidates have been developed that activate the host's immune system against a malarial toxin111 and Haemophilus influenza type b112. Efforts are currently underway to generate glycan-based vaccines that target the viral envelope protein of HIV (gp120)113,114 and other pathogenic microbes, including Salmonella typhi, Candida albicans, Shigella and Mycobacterium tuberculosis115. The unique glycosylation patterns of pathogenic microbes could potentially be targeted for diagnostic purposes as well.