GSLs perform many functions on the surface of cells. One main characteristic of GSLs that allows them to perform such versatile roles is their ability to form clusters. These clusters, termed ‘glycosynapses’, are able to interact with functional molecules on the cell surface . Significantly, they have contrasting properties with lipid rafts (Table 1). Glycosynapses rely on the ability of GSLs to bind via cis-carbohydrate to carbohydrate interactions, forming clusters in membrane and interacting with functional proteins.
Alterations in glycosylation in cancer
Glycolipids play an important role in many pathways of cellular physiology, and changes in the sugar head of the glycolipid can confer advantages to cancer cells. It has long been known that alterations of GSL, O- and N-linked oligosaccharides provide a survival advantage to tumours [23,24].
Incomplete synthesis. Studies have shown that some cancers lack enzymes involved in the complete synthesis of some carbohydrate determinants due to silencing of the genes responsible. This is due to DNA methylation of transferases involved in glycan production and histone deacetylation . Incomplete synthesis of the normally expressed glycans sialyl- 6-sulpho Lewis x and disialyl Lewis a leads to the increased expression of sialyl Lewis x and sialyl Lewis a [26,27]. Over-expression of sialyl Lewis a on either glycolipids or glycoproteins increases tumour cell adhesion and motility, resulting in increased metastasis [28,29].
Neo-synthesis. As well as the incomplete synthesis of glycans on tumour cells, aberrant glycans can be expressed through neo-synthesis . This is mediated by hypoxic regions of solid tumours, resulting in the enhanced transcription of glycosyltransferases [31,32]). The increase in these factors in response to hypoxia leads to the increase in glucose uptake (glucose-transporter type-I) leading to increased cell growth as well as an increase in glycan synthesis, leading to the increased expression of glycans including E-selectin ligands sialyl Lewis x and sialyl Lewis a .
Abnormal fucosylation. Fucosylation is a common modification of glycoproteins. Up-regulation of enzymes involved in the addition of fucose residues to glycans can lead to an increase in fucosylation in cancer . For example, α-fetoprotein (AFP) is increased in chronic hepatitis and liver diseases, but fucosylated AFP is over-expressed only in hepatocarcinomas (HCC). For this reason, it was approved as a tumour marker for HCC by the Food and Drug Administration (FDA) in 2005 .
Over-expression. Gangliosides are sialylated glycolipids present on normal tissues, but have been shown to be over-expressed in malignant melanoma and tumours from neuroectodermal origin [35–37]. The glycolipid isoglobotetraosylceramide (IsoGb4) has been shown to be a marker of metastasis and over-expressed on tumours, with limited expression on normal tissue . Sialyl Lewis a has been shown to be over-expressed in a range of cancers, including colorectal , breast  and ovarian [41,42] cancer. It is used as a serum marker in a range of cancers, including colorectal cancer to measure a patient's response to therapy . GD2 is a disalylganglioside antigen expressed on the surface of tumours of neuroectodermal origin, including neuroblastoma and melanoma . In addition, GD2 is also expressed on glioma and non-small cell lung cancer . It is expressed abundantly on 100% of neuroblastoma tumours, regardless of stage . Normal tissue expression is limited to neurones, skin melanocytes and peripheral fibres. Furthermore, GD2 has been shown to be involved in cell growth and differentiation and apoptosis . The ganglioside GD3 has been described as a melanoma marker due to its over-expression  in this cancer and restricted normal expression. GM2 is over-expressed on a range of cancers, including malignant melanoma and neuroblastomas . Enhanced expression of the Gb3Cer/CD77, a globotriaosylceramide, has been observed in most pancreatic and colon adenocarcinomas. As it is the receptor for Shiga toxin, the B-subunit of this toxin has been used a novel therapeutic strategy . Studies have shown that some gangliosides can also be shed into the tumour microenvironment . For example, GM2, GM3 and GD1a gangliosides have been shown to be shed by medulloblastoma cell lines . Secreted gangliosides can be internalized by activated T cells triggering their apoptosis  and providing a mechanism for cancers to evade the immune system.
Abnormal sialylation. Sialic acid is a common terminal addition to glycans in both normal and cancer tissues. As discussed previously, an increase in sialylated Lewis antigens can be observed on a range of cancers. The two most common forms of sialic acid expressed on mammalian cells are N-acetlyneuraminic acid (NeuAc) and N-glycolylneuraminic acid (NeuGc). The enzyme cytidine monophosphate (CMP)-N-acetylneuraminic acid converts CMP-NeuAc to NeuGc through hydrolysis. However, this enzyme is not expressed by humans due to a genetic mutation, resulting in the inability to synthesis NeuGc. However, it has been shown to be expressed by some cancers [7,51–53]. It has been shown that some cancers have a greater ability to incorporate the NeuGc from the diet into glycans through a scavenger pathway [54,55]. A possible advantage and explanation of tumour cells expressing NeuGc may be due to immunosuppression, with the ability of NeuGc to down-regulate CD4 on T cells in the tumour environment .
Glycolipids as targets for immunotherapy
The alteration of glycan structures on glycolipids suggests that they may be good targets for vaccines or mAbs.
Glycolipids as cancer vaccines. Many issues complicate the use of glycan vaccines. First, isolation of the glycolipid from cancer cells is very difficult due to the heterogeneity of cell surface glycosylation. More recently, synthetic organic chemistry provides a solution to this problem . Secondly, as glycolipids are often T cell-independent antigens (TI) they often elicit short-lived, low-affinity immunoglobulin (IgM) responses that lack memory. In addition, as most glycolipids are ‘self-antigens’ expressed at low levels on normal cells, they tolerize the immune system and their immunogenicity is low. Glycolipids are also shed into the blood by growing tumours, further reinforcing their immunotolerance. Studies have shown that TI antigens cross-link the B cell receptor (BCR) multi-valently, giving B cells a primary signal which results in proliferation [58,59]. However, Ig affinity maturation and production requires further signals, including ligation of CD40 and production of cytokines. This can be achieved by linking a synthetic glycan to a T cell carrier such as keyhole limpet haemocyanin (KLH), bovine serum albumin (BSA) or tetanus toxoid (TT) .
Vaccination of GD2–KLH gave little antibody response, whereas an alternative molecule GD2–lactone–KLH gave an antibody response in 83% of patients. Interestingly, patients immunized with the highest dose of GD2–lactone–KLH gave a high antibody response with the ability to induce CDC in GD2-positive cell lines . Patients immunized with synthetic fucosyl GM1–KLH in small cell lung cancer plus QS-21 showed IgM responses, but only one patient mounted an IgG response at low titre . A GD3–KLH conjugate failed to raise an antibody response in humans . However, melanoma patients injected with GD3–lactone–KLH plus QS-21 stimulated IgM responses. The most promising KLH conjugate was GM2. Early vaccines utilized GM2 alone plus bacillus Calmette–Guérin (BCG) or CX (cyclophosphamide), which gave a predominantly IgM response in Phase I trials, with some reports of an IgG response. Despite the lack of an affinity-maturated, class-switched antibody response, a highly significant increase in disease-free interval and overall survival in 17% patients was observed in patients treated with GM2/BCG . In order to increase the immunogenicity of GM2, it was conjugated to the T cell carrier, KLH. This resulted in increased titres of both IgM and IgG responses in patients . As well as increasing the immunogenicity of GM2 by adding KLH, other adjuvants were tested, including Dextox [monophosphoryl lipid A (MPL) and purified mycobacterial cell-wall skeleton-based adjuvant] and QS-21 (a saponin-based adjuvant) which out-performed other adjuvants . The GM2–KLH vaccine showed promise in early clinical trials . In the Phase III study, although there was a clear correlation between antibody response and survival, there was no clinical benefit from vaccination when correlated with placebo or no treatment groups .
In an attempt to raise an IgG response to this ganglioside, an anti-idiotypic monoclonal antibody that mimics GD3, BEC2  was used as an immunogen. Despite early encouraging results in lung cancer patients, a Phase III trial failed to show any survival advantage for vaccinated patients .
NeuGc has been studied as a possible target for therapy [71–73] due to its absence from normal human tissue. There are two NeuGcGM3 ganglioside-based vaccines currently in Phase II clinical trials. The first racotumumab (known formerly as 1E10) is an anti-idiotype murine monoclonal antibody, which is a mirror image of an antibody that specifically recognizes NeuGcGM3. Thus, anti-idiotype antibodies can act as antigens inducing a response against the original antigen. The alternative vaccine is NeuGcGM3/VSSP, which results from conjugation of the ganglioside into very small-sized proteoliposomes (VSSP) derived from Neisseria meningitides. Both vaccines had acceptable safety outcomes and were able to induce specific humoral and cellular immune responses in patients. The response to vaccination was stronger in patients with lower tumour burden, better performance status and a good response to previous treatment [74–77]. The current Phase III trials will confirm any survival benefit.
Immunization with GM3 plus Salmonella minnesota mutant R595 vaccine did not induce an antibody response in mice . However, incorporation of GM3 into very small-sized proteoliposomes produced by using anionic detergents to incorporate gangliosides into the outer membrane protein complex (OMPC) of N. meningitidis produced an IgG response in chickens, mice and monkeys . Unfortunately, when this vaccine went into Phase 1 study in 26 patients with metastatic melanoma, it showed IgM responses in only some patients. There were, however, signs of tumour regression in two patients .
A sialyl Lewis a–KLH conjugate  and a Lewis y–KLH conjugate have also induced anti-carbohydrate antibodies. The latter was of interest, as several patients made responses to the Lewis y hapten that only recognized glycolipid and not glycoproteins . Human antibodies recognizing tumour cells that could induce complement-mediated lysis were induced with all these vaccines; however, they were of low titre and predominantly of the IgM subtype, which have reduced ability to penetrate solid tumours compared to IgG.
To enhance further the immunogenicity of carbohydrate vaccines, new approaches are currently being tried. Attachment of a protein carrier to a glycan is often problematic, as the chemistry is difficult to control and gives wide batch variations . Another major drawback is that the carrier proteins are highly immunogenic, leading to suppression of the anti-glycan response. A more potent vaccine targeting glycolipids is therefore required. Glycolipids can be processed by B cells and presented on CD1d , a major histocompatibility complex (MHC)-class 1-like molecule, to natural killer (NK) and NK T cells [84,85]. These antigens need to be multimerized and are often presented within liposomes [38,76,79]. Evidence from a study immunizing mice with bacterial glycolipids incorporated within liposomes and mixed with an anti-CD40 mAb displayed an increased antigen-specific antibody response to the pathogen with an increase in class-switching to IgG, showing that the anti-CD40 mAb could substitute for T cell help . More recently, fully synthetic carbohydrate vaccines incorporating a glycan, the Toll-like receptor (TLR)-2 activator Pam2CysSK4, and a T cell epitope incorporated within liposomes, stimulated high IgG antibody titres .
However, whether these new approaches can overcome tolerance in humans and stimulate high titre, potent IgG antibody responses remains to be tested. It seems more logical to develop human monoclonal IgG antibodies, which can be administered repeatedly in high amounts.
Monoclonal antibody targeting of tumour-associated glycolipids. In contrast to the thousands of mAbs recognizing protein antigens, only a limited number of anti-tumour glycolipid antigens have been identified. Of great interest, however, is that many of these mAbs induce direct killing by oncosis as well as mediating potent antibody-mediated cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC).
Lewis antigens. A range of Lewis y antibodies have been identified, but a consistent problem with Lewis antibodies has been a degree of cross-reactivity with Lewis x and H type 2 structures, causing red blood cell agglutination and gastrointestinal toxicity [87–89]. More recent studies have shown that even this cross-reactivity of anti-glycan mAbs has been underestimated . We have raised a new mAb, FG27, against Lewis y expressing glycolipids. In contrast to anti-Lewis y mAbs raised against cells, they are very specific and do not cross-react with other Lewis antigens such as Lewis-X (BR96 mab), Lewis b (SC101), B blood group (BR96), H blood group (BR55) or bi-antennary Lewis y antigens. FG27 failed to stain liver, lung, colon, jejunum, breast, kidney and the ileum, which contrasts with the other Lewis y cross-reactive mAbs. Indeed, its only cross-reactivity with normal tissues is against duodenum and stomach. Moreover, our monoclonal antibodies recognize Lewis y exposed at the cell surface and mediate oncosis, ADCC and CDC, which does not occur in the mAbs raised against Lewis y conjugated directly to the T cell carrier KLH .
CA19·9 is a mAb recognizing sialyl Lewis a [also known as carbohydrate antigen 19·9 (CA19·9)]. SC104 is a novel mAb inducing direct killing as well as ADCC/CDC. It recognizes sialyltetrasoyl ceramide  and is shortly to enter Phase I clinical trials. A human anti-sialyl Lewis a mAb was produced using peripheral blood mononuclear cells (PBMCs) isolated from a breast cancer patient undergoing sialyl Lewis a–keyhole limpet haemocyanin (KLH) treatment . This mAb has shown specific binding to sialyl Lewis a alone, and promisingly induces ADCC and CDC of antigen-positive cell lines as well as anti-tumour activity in a xenograft model.
GD2. A number of anti-GD2 mAbs have been produced, including 14.G2a, ch14·18 [93,94], ch.60C3 , 3F8  and KM8138 , to target neuroblastoma, melanoma and non-small cell lung cancers. The benefit of targeting GD2 on the surface of cells is that, unlike other gangliosides studied, GD2 is not shed by the cells into the microenvironment . 14·18 is an IgG3 murine mAb targeted to GD2 and 14.G2a is a class-switch variant developed to enhance the ADCC effect of the mAb. 14.G2a has also been shown to induce CDC in vitro. When administered with IL-2, 14.G2a showed minimal effectiveness and suffered similar human anti-mouse antibody (HAMA) responses to 3F8 . To overcome the HAMA responses seen with the murine mAbs, chimeric mAbs ch14·18 and c.60C3 were produced. ch14·18 was shown to be more effective than its murine counterpart, with overall survival greater than with maintenance therapy . The humanized mAb Hu18K322A was created from ch14·18 in order to increase the half-life of the mAb in vivo. As well as being humanized, Hu18K22A was engineered with an amino acid change in the Fc region at position 322 and produced in the YB2/0 cell line rather than Chinese hamster ovary (CHO) lines, which lacks fucosylation in the Fc region, with the purpose of increasing the efficacy of CDC and ADCC in vivo. Currently, the use of anti-GD2 mAbs in clinical trials in conjunction with chemotherapy is the mainstay of neuroblastoma therapy .
GD3. The ganglioside GD3 has been described as a valid target for mAb therapy due to anti-GD3 mAb-mediated melanoma cell lysis being observed with the mAb MB3·6 . R24 is a mouse mAb that recognizes GD3 and has undergone numerous clinical trials . It has been used to treat melanoma patients, and in one trial showed a complete response in one patient who lasted 2 years and a partial response in one patient who lasted 2 months . It also displayed a human anti-mouse antibody response in patients, but the low level of efficacy in patients meant that the antibody was not humanized .
GM2. DMF10·167·4 is a hamster mAb raised against a murine T cell lymphoma cell line and has been shown to induce apoptosis of that cell line in vitro. Subsequently it was found to bind to GM2 and could bind to both melanoma and small cell lung cancer cell lines, but showed minimal binding to normal tissues . Its ability to induce apoptosis in a range of cell lines in vitro also transferred to in vivo studies, where it was shown to inhibit the formation of tumours in murine models .
NeuGcGM3. An anti-NeuGcGM3 mAb, 14F7, was produced in mouse and shown to bind to NeuGc GM3 and not NeuAc GM3 specific to cancer tissues [52,106]. It has been shown to kill NeuGc–GM3-positive cell lines by CDC, ADCC and directly, although high concentrations of mAb were required .
Galβ1-3GlcNAcβ1-3Gal. RAV12 is a chimeric mAb that has been shown to bind a minimal epitope of Galβ1-3GlcNAcβ1-3Gal, which has been observed on 90% of intra-abdominal tumours, but also on mucosal and glandular/ductal epithelium [4,108]. RAV12 has been shown to directly kill the colorectal cancer cell, Colo 205, in vitro by oncosis . A study has shown that RAAG12 is present on insulin-like growth factor-I receptor (IGF-IR) and binding of RAV12 leads to increased phosphorylation of IGF-IR in RAAG12-positive cell lines, leading to accelerated desensitization of the Akt/PKB pathway . A recent Phase I study in 33 recurrent adenocarcinoma patients showed some anti-tumour activity of RAV12, although toxicity of the mAb precluded the delivery of maximal doses .
F77. Interestingly, F77, a mAb targeted at an as-yet unidentified glycolipid target, is also able to induce direct cell death by oncosis. The mAb has been shown to bind to a large proportion of both primary and metastatic prostate cancer specimens by immunohistochemistry . F77 is proposed to induce oncosis by recruiting antigen to lipid rafts and through the production of large membrane pores .