The TGF-β (transforming growth factor-β) cytokine superfamily in mammals contains some 30 members. These dimeric proteins are characterized by a strongly conserved cystine knot-based structure. They regulate the proliferation, differentiation and migration of many cell types, and therefore have important roles in morphogenesis, organogenesis, tissue maintenance and wound healing. Thus far, around one-quarter of these cytokines have been shown to bind to heparin and heparan sulphate. Well-established examples are the TGF-β isoforms 1 and 2, and the BMPs (bone morphogenetic proteins) -2 and -4. In studies in my laboratory, we have shown that GDNF (glial-cell-line-derived neurotrophic factor) and its close relatives neurturin and artemin bind to heparin and heparan sulphate with high affinity. We have reported previously that binding of GDNF is highly dependent on the presence of 2-O-sulphate groups. More recently, we and others have been investigating the heparin/heparan sulphate-binding properties of BMP-7, which is a representative of a distinct BMP subgroup from that of BMPs -2 and -4. Interestingly, several of the various specific BMP antagonist proteins also bind to heparin and heparan sulphate. Much remains to be learnt about the nature and role of glycosaminoglycan interactions in the TGF-β superfamily, but current work suggests that these cytokines do not share a single highly conserved heparin/heparan sulphate-binding site.
The TGF-β (transforming growth factor β) cytokine superfamily
The TGF-β-related cytokines constitute a major cytokine superfamily, the first members of which appeared with the evolution of metazoans. With the emergence of higher vertebrates, there was a dramatic radiating divergent evolution of superfamily members, leading to some 30 proteins in the mammals. Despite this divergence, there has been a strong conservation of structure. The primary sequences contain seven cysteine residues with a characteristic spacing pattern. Six of these cysteines form intrachain disulphide bridges so as to give rise to an eight-residue-membered cystine ring knot . The remaining cysteine forms an interchain covalent bridge which links the two monomers together in the dimeric cytokine molecule. The cystine knot structure provides the basis for a conserved polypeptide fold. The high-resolution structures of nine of the superfamily members show a tertiary structure comprising two β-strand loops stacked one above the other and extending away from one side of a short α-helix . The exceptions within these structures are activin A, which lacks an α-helix , and TGF-β3, which has an extra-short α-helix . The TGF-β cytokines are synthesized as the C-termini of large pre-proteins from which they are released by proteolytic cleavage.
The members of this cytokine superfamily have diverse biological activities, regulating cell proliferation (often negatively), inducing cell differentiation, promoting cell migration and enhancing the deposition of extracellular matrix. For instance, the subfamily of bone morphogenetic factors have roles in embryogenesis and organogenesis which go well beyond an eponymous role in skeletal tissues, and are therefore considered to be developmental morphogens. Such roles clearly require localized paracrine activity within the embryonic tissues. A key question here is how such small soluble glycoproteins, being dimers of approx. 20 kDa subunits, can achieve such locally restricted activity. One answer is that around one-quarter of the superfamily members are now known to bind to heparin and HS (heparan sulphate). Such interaction with cell-surface and extracellular matrix GAG (glycosaminoglycan) would retain the cytokines close to their sites of secretion within the tissues.
An early insight into the heparin-binding properties of TGF-β superfamily members emerged from the studies of Lyon et al. . These workers showed that human TGF-β1 and TGF-β2, but not TGF-β3, bind to heparin and highly sulphated HS. By examining the partial conservation of basically charged amino acids between these three isoforms, and considering the locations of these residues in the high-resolution structures of the TGF-βs, they proposed that the heparin-binding site might constitute the arginine and lysine residues at positions 25, 26, 31 and 37, and His34, which are located on the tip of the first β-strand loop, together with Arg/Lys94, which is adjacent on the tip of the other β-strand loop. This site would be present twice in the dimer, but the two sites might be engaged by a single heparin/HS chain . Interestingly, the TGF-β cytokines bind to their receptors near the tips of their β-strand loops, placing the receptor-binding sites close to the proposed heparin-binding site. However, Lyon et al.  found that heparin and highly sulphated liver HS potentiated the activity of TGF-β1, while having no effect on the activities of TGF-β2 and TGF-β3 . These findings indicate that there is no competition between GAG binding and receptor binding for TGF-β2, and, indeed, with TGF-β1, there appears instead to be some form of co-operativity.
BMPs (bone morphogenetic proteins)
The 14 BMPs are an important subfamily of TGF-β type cytokines. It is well established that both BMP-2  and BMP-4  bind to heparin and HS, in both cases via clustered basic residues in the short N-terminal sequences lying in advance of the first of the conserved cysteine residues (see Figure 1). In the case of Xenopus BMP-4, the removal of three consecutive basic resides in this region not only abolished heparin binding, but also resulted in long-range BMP-4 signalling within the embryo, in contrast with the more localized signalling observed with wild-type BMP-4 . The distribution of basic residues within the N-terminal sequences of BMP-4s is highly conserved between Xenopus and mammals (Figure 1). BMPs -2 and -4 show similarity to the Drosophila cytokine Dpp (decapentaplegic), which is a morphogen that controls anteroposterior patterning in the wing. Recombinant Dpp has also been shown to bind to heparin , and a number of studies have shown an involvement in Dpp activity of the two HS proteoglycans, Dally and Dly, both homologues of mammalian glypicans (see  and references therein). Most recently, these two proteoglycans have been shown to be essential for transporting Dpp across fields of cells in order to establish the morphogenetic gradient which establishes the anterior–posterior boundary in the developing wing discs . It would seem likely that similar HS-dependent mechanisms will operate to establish morphogenetic gradients of BMPs -2 and -4 within vertebrate and mammalian tissues.
The N-terminal sequences of selected BMP type cytokines of the TGF-β superfamily
A further BMP homologue in Drosophila is 60A. This appears to be an evolutionary prototype of the mammalian BMPs -5, -6, -7 and -8. Compared with BMPs -2 and -4, these have considerably longer N-terminal sequences ahead of the first conserved cysteine. The distribution of basic residues within these N-terminal sequences is quite different, with an absence of clustered basic residues close to the first cysteine (Figure 1). However, work in several laboratories, including my own, has shown that BMP-7 also binds to heparin and HS ([10,11], and S. Chau and C.C. Rider, unpublished work). The location of the heparin-binding site within BMP-7 remains to be established, as do the consequences of such GAG interactions. Given the close sequence homology of the mammalian 60A-type BMPs, one can predict that all of these will bind similarly to heparin and HS. Therefore, taken overall, GAG binding is likely to be a widespread property across the various members of the vertebrate BMP subfamily of TGF-β cytokines.
The activity of the BMPs is tightly regulated, in part through a number of secreted antagonist proteins which bind BMPs and other TGF-β superfamily cytokines with high affinity. Such antagonist binding blocks the subsequent engagement between the cytokine and its cell-surface receptors. Many of these antagonists, including noggin and chordin, are distantly related to the TGF-β superfamily. Noggin in particular has a similar cystine knot and protein fold [1,12], Other antagonists, such as follistatin and follistatin-like 3 are quite distinct . It is emerging that a number of antagonists also bind to GAGs. Follistatin binds heparin and cell-surface HS via a cluster of seven basic residues located in the most N-terminal of three consecutive characteristic domains [13,14]. Noggin is likewise able to bind to cell-surface HS, the binding site being a basic-residue-rich sequence located in the central stem region of the dimer, where the two monomers are in close contact with each other [12,15]. This heparin-binding region is therefore well removed from the BMP-binding site on the exposed tips of the monomers. Noggin bound to cell-surface HS is therefore still able to bind BMP-4 . Chordin binds to heparin via multiple binding sites that are distributed throughout its four cysteine-rich domains . Chordin binds to cell-surface HS proteoglycans, in particular the syndecans, but not to the HS proteoglycans of basement membranes. In an in vitro cell assay of BMP-4 activity, the antagonistic activity of chordin depended on the presence of sulphated HS . In part, this was because chordin was no longer retained on the cell surface when expressed by the transfected cells in the presence of chlorate, a sulphation inhibitor . Other BMP antagonists have to date received little biochemical study; however, it is notable that recombinant sclerostin, a further cystine-knot-type antagonist, binds to heparin-affinity columns .
GDNF (glial-cell-line-derived neurotrophic factor) and the GFLs (GDNF family ligands)
A further TGF-β cytokine subfamily for which there is clear evidence are the GFLs. These are four neurotrophic factors, GDNF, neurturin, artemin and persephin. Each has a specific receptor polypeptide, GFRα-1 to GFRα-4 respectively, but all share the cell-surface tyrosine kinase c-Ret, as a common signalling component of the cytokine–receptor complex . We have shown previously that GDNF binds strongly to heparin derivatives and highly sulphated HS . Using a series of heparins subjected to selective chemical modifications, we showed that this binding is highly dependent on the presence of 2-O-sulphate groups . This particular finding is of interest because homozygous deletion of the single gene encoding HS 2-O-sulphotransferase results in several developmental defects, but most notably an absence of kidneys . As argued elsewhere , since homozygous gene deletions of GDNF, GFRα-1 and cRet all produce similar phenotypes of kidney agenesis, this strongly suggests that the development consequences of the loss of HS 2-O-sulphotransferase activity arise to a large extent from disruption of GDNF–HS interactions. However, the lack of 2-O-sulphate groups within HS may also disrupt the signalling of other developmentally important cytokines, such as the fibroblast growth factors . Davies and co-workers have shown that GDNF signalling is dependent on cell-surface HS , and that such signalling is inhibited in the presence of exogenous soluble heparin . More recently, we have shown that neurturin and artemin also bind heparin, in fact with higher affinities than GDNF ; the current lack of a suitable antibody prevents the similar study of persephin. Since the distribution of basic residues within the various GFLs shows little conservation, consideration of the primary sequences of the GFLs sheds little light on the possible location of the heparin-binding sites within this family. Indeed, it is quite plausible that, even within the closely related members of the GFL superfamily, the heparin-binding sites are divergent.
Although heparin-binding studies have yet to be reported for most of the TGF-β superfamily cytokines, it is already emerging that many of them bind to heparin and related GAGs. However, since TGF-β3 has already been shown not to bind in this way , such interactions are clearly not universal in this superfamily. Nonetheless, binding to HS would appear to be an important mechanism underlying the biological outcomes of many of these cytokines. Arguably, in any biological system, Dpp during Drosophila wing development is the clearest example that we have at present for the role of HS proteoglycans in establishing and maintaining a morphogenetic gradient through restricting the diffusion of an otherwise soluble growth factor. In mammalian systems, a strong case is emerging that 2-O-sulphate-rich HS acts in a similar way by localizing growth factors such as GDNF which are critical in the organogenesis of the mammalian kidney, although the experimental evidence base here is currently less complete.
Given that the hallmark of this large TGF-β cytokine superfamily is a conserved pattern of cysteines giving rise to a characteristic protein fold, it is likely that these cytokines will eventually provide much insight into evolution of heparin-binding sites. To date, few studies have delineated heparin-binding sites in these cytokines, but what has already emerged suggests divergent binding sites rather than a common conserved site. This being the case, we should expect that the functional roles and importance of heparin/HS binding will be similarly divergent between the different family members.
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.).