Chemokines exert their biological activity through high-affinity interactions with cell-surface receptors, thereby activating specific signalling pathways, and a second low-affinity interaction with proteoglycans. Proteoglycans consist of a protein core, to which GAG (glycosaminoglycan) chains are attached. The GAGs are long, linear, sulphated and highly charged heterogeneous polysaccharides that are expressed throughout the body in different forms depending on the developmental or pathological state of the organ/organism. Mechanistically, the GAG interaction is thought to facilitate the retention of chemokines on cell surfaces, thereby forming a high local concentration required for cell activation. Recently, we demonstrated that certain chemokines require interactions with GAGs for their in vivo function. Additionally we have shown that chemokines oligomerize on immobilized GAGs, and this ability to form higher order oligomers has also been shown to be essential for the activity of certain chemokines in vivo. We believe that interference with the chemokine–GAG interaction provides a novel anti-inflammatory strategy, exemplified by a variant of RANTES (regulated upon activation, normal T-cell expressed and secreted) that has abrogated GAG binding and oligomerization properties.
The chemokine family
Chemotactic cytokines, or chemokines, are a large family of small proteins, which are distinguished from other cytokines in that they are the only members of the cytokine family that act on the receptor superfamily of G-protein-coupled, seven-transmembrane-spanning (7TM) receptors [1,2]. Their principal role is that of mediating directional cellular migration, principally of leucocytes, whereby they play a major role in the development and maintenance of a functional immune system. They also play a pivotal role in diseases such as autoimmune and allergic inflammatory disorders, cancer and organ transplant where excessive cellular recruitment plays a deleterious role. Their function is not limited to recruitment, since they also play a role in cellular activation, differentiation, degranulation and in processes such as organ development, angiogenesis, lymphogenesis and wound repair. Although chemokines have a relatively low level of sequence identity, their three-dimensional structure shows a remarkably similar topology in that they all possess the same monomeric fold . This fold, consisting of three β-strands, a C-terminal helix and a flexible N-terminal region, is conferred on these proteins by a four-cysteine motif, which forms two characteristic disulphide bridges in most of the chemokines. The flexible N-terminal region is believed to be important in receptor activation since modification of this region has been shown by many laboratories to affect activity [4–6].
It was initially thought that the second important interaction that chemokines form with GAGs (glycosaminoglycans) was spatially separated from the receptor-binding domains [7,8]. However, as more GAG-binding sites are mapped on chemokines, several have been shown to partially overlap with receptor-binding sites [9–13]. However, while the interaction of the chemokines with GAGs on the leucocytes is not essential for in vitro bioactivity, this interaction is thought to play a role in sequestration of chemokine and subsequent presentation to the receptor expressed on the leucocyte cell surface .
Characterization of protein–GAG interactions
GAGs represent the largest diversity among biological macromolecules. They consist of repeating disaccharide units with variations in basic composition of the saccharide in acetylation and N- and O-sulphation patterns, and with respect to their linkage, and moreover the chain lengths may range from 1 to 25000 disaccharide units. The different combinations could give rise, if one considers a hexasaccharide, to over 12 billion sequences, representing 100 times the variation exhibited by hexapeptides and an astounding two million times more than a 6-mer of DNA . They are found either as soluble molecules in plasma, or surface-bound in the form of proteoglycans . They are known to interact with hundreds of proteins including proteases, cytokines, adhesion molecules and growth factors, and of particular interest to the present review, to chemokines. However, their molecular characterization and a full identification of their interacting partners are still in infancy, but the era of glycomics has begun [17–19].
There are several classes of GAGs, the most ubiquitous being HS (heparan sulphate), a polysaccharide that is expressed on virtually every cell in the body and comprises 50–90% of total endothelial proteoglycans. Other classes of GAGs include heparin, which is produced almost exclusively by mast cells and has been exploited therapeutically as an anticoagulant, chondroitin sulphate and dermatan sulphate, which, like HS, are found on cell surfaces and the extracellular matrix, keratan sulphate, a major component of the cornea and cartilage and hyaluronic acid, often referred to as a biological glue that is involved in joint lubrication and holding together gel-like connective tissue by virtue of its high molecular viscosity, and ability to order water.
A common feature of GAGs is their overall negative charge due to the density of sulphate and carboxylate groups on the GAG chains, which suggests an electrostatic interaction with basic proteins, such as the chemokines. However, while the driving force for binding is principally electrostatic, this interaction is not merely based on overall charge interactions, as best exemplified by the fact that the acidic chemokines CCL3 and CCL4 bind GAGs . The XBBXBX and XBBBXXBX motifs, where B is a basic amino acid, are common heparin-binding motifs for several proteins . Indeed, the common heparin-binding motif for several chemokines has been described as a classical BBXB motif for example on the 40s loop for RANTES (regulated upon activation, normal T-cell expressed and secreted) , MIP (macrophage inflammatory protein)-1α  and MIP-1β  and in the 20s loop for SDF-1α (stromal-derived factor-1α) .
There are several methods used to measure the GAG–protein interactions and, to date, we believe that none of these methods are devoid of caveats. However, the most commonly used screen, which demonstrates that the interaction has a degree of selectivity, involves determining the amount of salt required to elute protein from a heparin–Sepharose column. This assay should be coupled with a comparison of the amount of salt required to elute from a cation exchange column, and when compared with elution from the heparin–Sepharose column, provides a measure of the specificity of the interaction [7,13]. Using this technique, the acidic chemokines MIP-1α and MIP-1β bind to the heparin column but not to the cation exchanger – despite their pI values of 4.5. Another simple and inexpensive assay involves binding of soluble, tritiated heparin to the protein of interest [11,23]. In contrast with the electrostatic bias of the chromatographic assays, this assay should reflect the overall binding capacity of a given protein. However, in our hands although we find that this assay is of qualitative value, it is difficult to quantify, and is only inexpensive for laboratories that produce large quantities of recombinant chemokines since each assay uses approx. 50 μg.
An assay we have used extensively involves binding of chemokine to immobilized heparin, using commercially available heparin–Sepharose beads in a classic equilibrium competition assay, but a major drawback of this strategy is the requirement for the use of radiolabelled proteins. This method has been used to determine the affinities of certain chemokines for different GAG families where the ranges in affinity were shown to vary over a range of three logs for RANTES, but only one log for MIP-1α . It has also been used to demonstrate that oligomerization occurs on immobilized heparin . This assay showed an increase in radioactivity bound to the beads when unlabelled chemokine was used as the competitor, which could only be explained by oligomerization occurring on the beads, a hypothesis that was validated by the use of an obligate monomeric form of IL-8/CXCL8, N-Me-Leu25-IL-8 .
An alternative format for binding to immobilized heparin is a scintillation proximity assay where heparin–BSA is bound to 96-well Flashplates . This assay allowed the identification of virally encoded chemokine-binding proteins that interfere with chemokine–GAG binding.
Other more biophysically based approaches include IFT (isothermal fluorescence titration) where one monitors the change in fluorescence as a function of added GAG but is only useful for proteins that have a fluorophore  and ITC (isothermal titration calorimetry) that quantifies the heat released upon binding . Both IFT and ITC seem to be more robust than either the affinity chromatography or 3H-heparin assay for defining GAG-binding epitopes. The caveat of ITC, however, is that it requires high concentrations of interacting species and in the case of protein–GAG complexes, precipitation is often observed. Plasmon resonance is also a powerful technique for measuring binding kinetics and thermodynamics without the need for large amounts of proteins or radioactivity [9,28,29]. However, this technique requires immobilization of one of the binding partners that may introduce artefacts. Other biophysical methods that can lead to insight into the contacts between GAGs and proteins are NMR and X-ray crystallography, but again high concentrations are required (>50 μM) again often resulting in precipitation. To date, there is only one report of the structure of a complex between a chemokine and GAGs, which was limited to heparin disaccharides due to the problem of precipitation with larger fragments .
The role of GAG interactions in chemokine function
It has long been hypothesized that chemokines must bind to GAGs to create a high local concentration of chemokine since without such a mechanism, chemokine accumulation would be disrupted by diffusion, especially in the presence of flow in blood vessels and draining lymph nodes (Figure 1). However, formal demonstration of the biological relevance of the GAG interaction was only recently reported. This was accomplished by engineering mutants with impaired GAG-binding capacity . Specifically, by characterizing alanine mutants of RANTES/CCL5, MIP-1β/CCR4 and MCP-1/CCR2, with several of the methods described above, the key residues involved in the interaction with heparin were defined. Importantly, the mutants showed robust chemotaxis in vitro where flow forces are not operative, and receptor activation is achieved through the construction of increasing artificial concentrations of chemokine. In this assay, GAG binding is not directly involved, although it may contribute to some extent to sequestration of chemokine and subsequent presentation to the receptor . However, when tested in an in vivo peritoneal recruitment assay, the mutants were unable to induce cell migration, even at a dose 10000-fold higher than the dose at which the wild-type variants achieved statistically significant cell recruitment, confirming that RANTES/CCR5, MIP-1β/CCR4 and MCP-1/CCR2 require interactions with GAGs to elicit cell migration in vivo .
Chemokines require immobilization on cell-surface GAGs in order to initiate the transmigration process from the circulation
GAG-binding specificity may also be implicated in differences in the ability of chemokines to oligomerize on GAGs, and the types of oligomeric structures they form. Indeed, chemokines do not need to oligomerize to cause cell migration in vitro as demonstrated by monomeric variants of RANTES/CCL5 , MIP-1β/CCL4  and MCP-1/CCL2 ; these mutants are indistinguishable in their chemotactic properties from the wild-type proteins. Accordingly, for quite some time it was unclear if chemokine oligomerization had any functional role, or was merely an artefact of the high concentrations of protein used for structural studies. However, when these mutants were examined in the in vivo peritoneal recruitment assay, they did not induce cellular recruitment, suggesting that in addition to GAG binding, oligomerization is required (Figure 2) . This assay provided several fascinating, and as yet not fully understood, observations. Using mutants for RANTES/CCL5 that had been previously described as having obligate quaternary structures, such as the dimeric mutant E66A-RANTES and the tetrameric E26A-RANTES mutants , it was shown that RANTES has a minimal tetrameric quaternary structure for cell attractant activity in vivo, whereas other chemokines that have been shown to be monomeric in vitro by structural studies such as eotaxin/CCL11, MCP-3/CCL7 and I-309/CCL1 were able to recruit cells . However, we cannot rule out the possibility that these chemokines may dimerize or oligomerize under physiological conditions in vivo in the presence of GAGs. With respect to HIV infectivity, the propensity of RANTES to oligomerize plays a role in the enhancement of HIV infection which is observed at high concentrations in vitro and is independent of receptor usage [34,35].
RANTES requires a minimal oligomeric structure in order to provide a directional signal to circulating leucocytes
Interfering with chemokine–GAG interactions as a novel therapeutic approach
It has been known for some time that heparin not only possesses anticoagulant activities, but also has anti-inflammatory properties and other disease-ameliorating affects. The anti-inflammatory effect of heparin and heparin derivatives has been demonstrated not only in animal models, but also in humans . Improved disease symptoms in several animal models of inflammation have been observed following treatment with heparin devoid of anticoagulant activity. For example, effective abrogation of clinical score in rat adoptive transfer EAE (experimental autoimmune encephalomyelitis) was observed following prophylactic dosing with heparin administered subcutaneously on a daily basis . Diminished airway responses in neonatally immunized rabbits and naturally sensitized sheep, measured as lung function in response to bronchoconstricting agents, also supported a role for heparin in improving symptoms in inflammatory disease [38,39]. In the rabbit model, improved lung function was accompanied by diminished neutrophil and eosinophil infiltration, similar to that seen in untreated animals. The results from the sheep study indicated that heparin may have a disease-modifying effect on asthmatic responses through modulation of mast cell mediator release. These anti-inflammatory effects of heparin have also been extended to patients, ameliorating symptoms in IBD (inflammatory bowel disease) sufferers  and asthmatics  in controlled clinical trials.
Although the precise mechanism(s) of the action of heparin in these studies was not established, it has been suggested that inhibition of the interaction between pro-inflammatory cytokines such as IFN-γ (interferon-γ) and membrane-associated GAGs may provide a mechanism for inducing clinically useful immunosuppression . Whereas immobilized heparin is essential for the biological activity of chemokines, soluble heparin has been shown to inhibit the biological effects of chemokines as demonstrated in vitro [23,43] and in vivo [30,44]. It is therefore likely that the anti-inflammatory effects of heparin are mediated, at least in part, by interference with the chemokine system. With this in mind, different strategies that interfere with chemokines and GAGs for therapeutic benefit have been explored.
GAG binding-deficient mutants
Abrogation of the heparin-binding site of RANTES in the variant A44ANA47-RANTES results in a protein not only unable to recruit cells in vivo, but one that has inherent anti-inflammatory properties . It is able to inhibit RANTES and thioglycollate-induced recruitment into the peritoneal cavity, and this inhibitory activity has been shown to translate into anti-inflammatory activity in animal models of disease. In the ovalbumin-induced murine model of lung inflammation, A44ANA47-RANTES reduced the number of infiltrating cells into the bronchoalveolar fluid, and in a murine model of multiple sclerosis, MOG (myelin oligodendrocyte glycoprotein)-induced EAE, it significantly reduced the clinical symptoms from hind leg paralysis to hind leg weakness. It was not clear initially how abrogation of heparin binding resulted in this inhibitory property, until the relationship between oligomerization was demonstrated. The crystal structure of the RANTES–disaccharide complex shows an electrostatic interaction between Arg47 and one of the residues previously identified as playing a role in RANTES oligomerization, Glu66 . Effectively, the A44ANA47-RANTES variant behaved as a dimer, and moreover was able to create non-functional heterodimers with wild-type RANTES by dissociating the preformed oligomer, thus acting in a dominant-negative manner .
These results represent proof of concept that interference with the chemokine–GAG interaction is a viable therapeutic strategy. Exploring this further, we attempted to find small molecules that bind to RANTES and which would prevent the binding of GAGs. Such small molecules have been found, and are more potent than a heparin disaccharide, as they were actually capable of inhibiting RANTES function in vivo, whereas the disaccharide was not (A. Proudfoot, J. Shaw and M. Schwarz, unpublished work). Dose-related inhibition has been observed for longer heparin fragments such as tetra-, hexa- and octa-saccharides. These results suggest that it may be possible to identify small oligosaccharide fragments that have anti-inflammatory properties, but not the anticoagulant properties of the heparin pentasaccharides used for inhibition of the clotting cascade, thereby opening up a potential new source of anti-inflammatory compounds. As chemokines are not unique among the cytokine superfamily members in their ability to bind to GAGs, this strategy may be extended to inhibit the function of other proteins, e.g., IFN-γ effects have been shown to be ameliorated in vivo using a similar strategy .
Cytokine–Proteoglycan Interactions: Biology and Structure: Biochemical Society Focused Meeting held at Royal Holloway University of London, Egham Hill, U.K., 9–10 January 2006. Organized and edited by B. Mulloy (NIBSC, U.K.) and C. Rider (Royal Holloway University of London, U.K.).