Fibrillar collagens provide the most fundamental platform in the vertebrate organism for the attachment of cells and matrix molecules. We have identified specific sites in collagens to which cells can attach, either directly or through protein intermediaries. Using Toolkits of triple-helical peptides, each peptide comprising 27 residues of collagen primary sequence and overlapping with its neighbours by nine amino acids, we have mapped the binding of receptors and other proteins on to collagens II or III. Integrin α2β1 binds to several GXX′GER motifs within the collagens, the affinities of which differ sufficiently to control cell adhesion and migration independently of the cellular regulation of the integrin. The platelet receptor, Gp (glycoprotein) VI binds well to GPO (where O is hydroxyproline)-containing model peptides, but to very few Toolkit peptides, suggesting that sequence in addition to GPO triplets is important in defining GpVI binding. The Toolkits have been applied to the plasma protein vWF (von Willebrand factor), which binds to only a single sequence, identified by truncation and amino acid substitution within Toolkit peptides, as GXRGQOGVMGFO in collagens II and III. Intriguingly, the receptor tyrosine kinase, DDR2 (discoidin domain receptor 2) recognizes three sites in collagen II, including its vWF-binding site, although the amino acids that support the interaction differ slightly within this motif. Furthermore, the secreted protein BM-40 (basement membrane protein 40) also binds well to this same region. Thus the availability of extracellular collagen-binding proteins may be important in regulating and facilitating direct collagen–receptor interaction.
- discoidin domain receptor
- glycoprotein VI
- peptide Toolkit
- von Willebrand factor
The 28 human collagens form a protein family that is characterized by the triple-helical domains (COL domains) that result from the presence of repetitive (GXX′) sequences within each α-chain. The degree of repetition may vary from four or five in short COL domains to 300 or more. Long triple-helical COL domains with short N- and C-terminal telopeptides dominate the structure of the fibrillar collagens and facilitate the assembly of the triple-helical monomers into the staggered three-dimensional quasi-linear structure of the collagen fibre that forms the fundamental mechanical support of the vertebrate organism. Protein interactions with fibrillar collagens, central to the organization of connective tissue, form the main subject of this article.
Approx. 50 collagen genes encode individual collagen chain types: some collagens, for example the fibrillar collagens II and III, are homotrimeric, whereas others, for example collagens I, V and XI, are heterotrimeric and usually contain two specific α1 chains and one copy of the α2 chain that together define each collagen type. Non-fibrillar collagens may have much longer and more elaborate telopeptides that preclude fibrillar assembly and support less regular network-like structures. Chief among these is collagen IV in which two or three of six α-chains combine to form the main tissue-specific structural component of the basement membrane.
Collagens function most obviously as the principal component of the extracellular matrix in the connective tissues. Collagens provide tensile strength to skin, bone, cartilage, tendon and blood vessel wall, for example, as well as having a more subtle role as the major constituent of the vitreous humour of the eye. These largely physical roles dominated the understanding of collagen that arose from intensive study in the 1960s and 1970s. In the subsequent two or three decades, however, it has become clear that collagens can regulate cell function through the engagement and activation of specific cellular receptors, the collagen-binding integrins, collagen-binding immune receptors and the DDRs (discoidin domain receptors) (reviewed in [1–5]). Structural studies have provided insight into some of these [6,7]. Collagens can also interact with the cell surface by binding an intermediate species that itself is recognized by a specific cellular receptor. Thus both direct and indirect receptor interactions can be supported by collagen. Several examples will be discussed below.
Our work has focused on the interaction of collagen with its receptors. Initially, reductive studies used partial digests, CNBr peptides of collagen, that retained the capacity to assemble in triple-helical form . These proved useful to map the binding of cells and other molecules to specific regions of the fibrillar and other collagens, and to establish the need for the triple-helical structure that remains essential for their recognition by the true collagen receptors. CNBr, however, cleaves collagen α-chains at methionine residues, producing a break in any recognition motif that contains methionine. In the last decade or so, the availability of automated peptide synthesis has allowed the investigation of the cell–collagen interaction in a more systematic way, through the preparation of large sets of overlapping triple-helical peptides (Toolkits) that contain the primary sequence of the homotrimeric fibrillar collagens II and III. Using these, we have mapped in detail the recognition of collagen by the collagen-binding integrin, α2β1  and DDR2, and the indirect interaction of the human platelet with collagen mediated by plasma vWF (von Willebrand Factor) and platelet Gp (glycoprotein) Ib . Studies of the immune receptors, platelet GpVI and LAIR-1 (leucocyte-associated Ig receptor-1), suggest that their recognition motifs in collagen are more complex (G.E. Jarvis, N. Raynal, J. Langford, D.J. Onley, A. Andrews, P.A. Smethurst and R.W. Farndale, unpublished work, and R.J. Lebbink, T. de Ruiter, N. Raynal, R.W. Farndale and L. Meyaard, unpublished work).
Interaction of collagen with other proteins
The platelet–collagen interaction
Collagens of the blood vessel wall are concealed by an intact contiguous endothelial cell layer, but may be revealed during disease or injury. After cytokine insult  or mild injury, endothelial cells may retract, exposing collagen IV, the principal component of the basement membrane, which may support adhesion and activation of circulating platelets. Deeper injury will expose the fibrillar collagens I and III, which are more fully understood in terms of their platelet-binding properties and subsequent thrombus deposition. These collagens are considered the prime stimulus for haemostasis and its pathological expression, arterial thrombosis, which is precipitated by the rupture of an atherosclerotic plaque. Integrins α1β1 and α2β1 may have other roles in the vessel wall, including supporting the vascular smooth muscle–collagen interaction which contributes to atherogenesis during migration, proliferation and intimal thickening , and lymphocyte targeting in inflammation (reviewed in ).
Three different axes support the adhesion of platelets: (i) the indirect interaction of collagen-bound vWF which supports GpIb-mediated platelet rolling across the collagen surface; (ii) a tighter interaction with integrin α2β1 (the only collagen-binding integrin expressed on platelets); and (iii) the stimulatory interaction with GpVI that leads to a phosphorylation cascade mediated by the associated ITAM (immunoreceptor tyrosine-based activation motif)-containing FcRγ (Fc receptor γ-chain) (reviewed in [14,15]). Together, these co-operative interactions lead to the activation of platelets primarily through GpVI via phospholipase Cγ and the secretion of thromboxane A2, ADP and ATP, amplifying the activation process through prostanoid and purinergic receptors, including P2Y12 and P2X1 [16–18].
Ultimately, the fibrinogen receptor (integrin αIIbβ3) is activated, allowing fibrinogen to bind and cross-link adjacent activated platelets, and vWF to form supplementary inter-platelet links by binding both αIIbβ3 and GpIb. These two, plasma protein-mediated, platelet–platelet interactions are fundamental to the processes of platelet aggregation and thrombus deposition. The detail of these collagen-independent events is beyond the scope of the present article, but has been reviewed extensively elsewhere [18a–18c].
The recognition of collagen by α2β1
The early study of CNBr peptides revealed that multiple sites within the fibrillar collagens support integrin α2β1 binding . Platelets, especially, provided an accessible cell surface in which α2β1 was found to be competent to bind collagen even in the resting state. In this respect, α2β1 differs from the fibrinogen receptor, which is maintained in an inactive conformation in the resting platelet, consistent with the need to avoid inappropriate and dangerous aggregate formation in the absence of haemostatic stimulus. One CNBr peptide in particular, α1(I)CB3, exhibited good Mg2+-dependent adhesion of platelets, but did not support platelet activation , now taken to indicate the presence of integrin-recognition sites, but the absence of those for GpVI. A set of overlapping peptides of α1(I)CB3 was synthesized, allowing the location of the sequence GFOGER as its integrin-recognition site [20,21]. As a short triple-helical peptide , this sequence supported the first integrin-ligand co-crystal . Similar GXX′GEX″ sites were reported, all of lower affinity. A combination of rotary shadowing and prospective peptide synthesis located the mid-affinity GLOGER sequence and low-affinity GASGER sequence within the collagen I α1 chain . Subsequent sequence-searching revealed conserved GXX′GER sites within the collagens, proved once more to be integrin-binding by synthesizing the corresponding triple-helical peptides . One of these in particular, GMOGER, occurs quite widely in the collagens, but is cleaved by CNBr, and the site is thus destroyed during CNBr peptide production.
The completion of the collagen Toolkits allowed comprehensive and exact location of integrin-binding sites, and revealed two unusual motifs: one proposed previously  to be platelet-reactive, GROGER, in which the conserved hydrophobic residue (X) is replaced by arginine, and GLOGEN, from which a positively charged arginine residue at X″ that forms an electrostatic link with the integrin I-domain surface is absent . GROGER and GLOGEN are too close within the native collagen sequence to be resolved by rotary shadowing, although prospective peptide synthesis had confirmed the former as an integrin-binding sequence . These GXX′GEX″ motifs are able to support binding of all the collagen-binding integrins, although binding of α10β1 has not yet been reported [21,27].
GFOGER, found in collagens I, II, IV and very few others, is unique among these motifs in being able to induce integrin activation. The affinity of all other GXX′GEX″ motifs for α2β1 is markedly increased by inside-out signalling, for example by activating platelets with G-protein-coupled or other activatory stimuli [24,28]. This calls into question their role in the native collagen: are these lower-affinity motifs active in cell biology? Collagen III, for example, lacks GFOGER, but, by virtue of being able to activate platelets (and thus α2β1) through GpVI, it is able to support firm integrin-dependent adhesion through the lower-affinity motifs.
In the setting of the blood vessel wall, the presence of GFOGER only in collagens I and IV implies that these species are essential components of the haemostatic system: the high affinity of GFOGER for α2β1 allows arrest of platelets from the circulation that have been slowed by collagen-bound vWF, with subsequent activation of GpVI bringing into play the lower-affinity GXX′GER motifs (indicated in Figure 1) in collagen III and other collagens, and securing firm α2β1-dependent adhesion.
The recognition of collagen by GpVI and other immune receptors
Most of our work on immune collagen receptors has addressed the platelet activatory receptor, GpVI, which operates in concert with α2β1 to secure firm adhesion and activation of the platelet. GpVI is expressed together with the dimeric FcRγ, to which it is bound by electrostatic forces . The receptor possesses tandem Ig domains and a long glycosylated stem, with the collagen-binding site centred on the N-terminal domain 1, and embracing the hinge region between the two Ig domains. The collagen-binding site has been modelled from docking on to its crystal structure [7,30] and defined functionally by site-directed mutagenesis and antibody mapping .
We have produced CRPs (collagen-related peptides) typically comprising ten GPO triplets that will bind GpVI, and, when chemically cross-linked to introduce polymeric structure, these CRPs are potent activatory ligands . This underscores clustering of GpVI on the platelet surface by its multivalent ligand as an essential mechanism of signal propagation and subsequent platelet activation , which occurs through Src family kinase-mediated phosphorylation of the ITAMs of FcRγ (reviewed in ).
The recognition of collagen is more complex in GpVI than in α2β1: although our model polymers are potent ligands, for which their GPO triplets are essential , studies with the Toolkits suggest that the presence of single GPO triplets, present in many peptides, or tandem GPO triplets, present in five peptides, is not sufficient to secure GpVI binding, and nearby primary sequence may be involved (G.E. Jarvis, N. Raynal, J. Langford, D.J. Onley, A. Andrews, P.A. Smethurst and R.W. Farndale, unpublished work). The highest expression of GpVI-binding activity is found in Peptide III-30, which contains a single essential GPO triplet close to a pair of tandem GPO triplets, of which either one is essential. It is not yet clear whether GpVI exists as a dimer on the platelet surface, as has been proposed [7,36,37], or whether dimerization might occur as part of the activation process. It is attractive to propose that GpVI dimers dock on to this extended site on Peptide III-30, since it is too long to interact with just a single copy of GpVI . The elucidation of this interaction requires further structural study. Binding data are summarized in Figure 1.
The leucocyte collagen receptor LAIR-1  possesses just a single extracellular Ig domain, and, like GpVI, can signal through associated immune receptors, in this instance ITIM (immunoreceptor tyrosine-based inhibitory motif)-containing species that inhibit the activation of immune cells [4,39]. Although LAIR-1, like GpVI, can recognize the GPO triplets found in CRPs, it seems for both receptors that more complex GPO-containing sequences within the collagens provide their natural ligands.
The interaction of collagen with vWF
The indirect interaction between the platelet GpIb complex and collagens of the blood vessel wall is mediated by vWF . vWF exists as a polymer in the plasma, and both the degree of polymerization and its activity increase with shear rate, so that its crucial role in securing platelet adhesion to the damaged vessel wall is most readily observed at high arterial shear. vWF binds GpIbα on the platelet surface through its A1 domain, and collagen through its A3 domain. The A3 domain has been crystallized , and its collagen-binding surface identified by site-directed mutagenesis and epitope mapping [41,42], in conjunction with co-crystallization with a blocking antibody . Collagen III was found to contain just a single high-affinity site for vWF, in Peptide III-23 (see Figure 2). Truncation and substitution of III-23 identified four triplets and four specific amino acids (underlined and bold) that were essential for binding vWF: GPRGQOGVMGFO . This motif is conserved in collagen II, but not in an intact form in collagen I in either the α1 chains (which lack the crucial hydroxyproline residue) or the α2 chain (which lacks the crucial arginine residue). Thus we have proposed that both chains of collagen I contribute to a competent vWF-binding site. This composite structure can be docked on to the collagen-binding site of vWF A3. Such motifs do not occur in any other collagen, underscoring the crucial role of the fibrillar collagens I and III in the blood vessel wall.
Discoidin domain receptor 2
DDR1 and DDR2 are receptor tyrosine kinases that are activated by collagen [44,45]. DDRs control development of mammary tissue (DDR1)  and long bones (DDR2) , and may be involved in human fibrotic diseases, atherosclerosis and several types of cancer . The use of Toolkit II revealed three major sites of interaction of recombinant DDR2 [47a]: one close to the collagenase cleavage site, one in the overlap region between Peptides II-22 and II-23, and another closer to the N-terminus, also in the D2 period of collagen II (Figure 2B). DDR2 also bound Peptide III-23, containing the highly conserved vWF-binding site in collagen III. The use of truncated and substituted peptides revealed that DDR2 bound to the same site as vWF, GPRGQOGVMGFO. The last six amino acids were able to support binding, with the methionine and phenylalanine residues proving critical for the interaction, providing a hydrophobic region that may complement the proposed collagen-binding site of DDR2 [47a,48]. A modest reduction in binding was observed after substitution of arginine and the C-terminal O residue, suggesting that the full DDR2-binding site may be up to four triplets in length. Remarkably, the mapping of the secreted protein BM-40 [basement membrane protein 40, also known as osteonectin or SPARC (secreted protein acidic and rich in cysteine)] on to collagens located a binding site in Peptide III-23 that overlaps with the vWF/DDR2-binding sequence (C. Giudici, N. Raynal, H. Wiedemann, W. Cabral, J.C. Marini, R. Timpl, H.P. Bächinger, R.W. Farndale, T. Sasaki and R. Tenni, unpublished work).
Functional modulation of cells by interaction-specific collagen peptides
Synthetic agonists at platelet GpVI
GpVI-specific peptides can be used to activate platelets . The requirement for higher-order structure has historically been fulfilled by chemical cross-linking of triple helices using cysteine residues, incorporated into the ends of the peptide for this purpose. Such cross-linked CRPs, designated CRP-XL, are potent agonists for platelet GpVI and have found wide application in the investigation of the properties of the receptor. An alternative strategy, which avoids the difficulty of controlling the cross-linking process, is to extend the length of the GpVI-reactive triple helix so that multiple copies of the receptor can assemble upon it. This possibility arises because the collagen-binding trench of GpVI appears to be two or three GPO triplets in length  and thus long peptides that might accommodate six or more copies of the receptor, sufficient to signal, can be synthesized without too much difficulty. The potency of CRPs increases with their length (Figure 3A), providing insight into the size of the cluster of GpVI molecules needed to initiate a useful signal in the platelet. Moreover, the inclusion of an integrin-recognition motif within such peptides enhances their potency, presumably by increasing their avidity for the platelet surface. These defined peptides may find application in diagnosis of platelet dysfunction.
There is a need to study deposition of thrombi from whole blood on to collagen. Glass or plastic surfaces are coated with collagen fibres, usually of bovine or equine origin, and blood is passed across the surface at defined shear rate, so that platelet function can be investigated microscopically under conditions that more closely represent the high shear occurring in stenotic arteries . Such systems allowed the shear-dependent activity of vWF to be explored. The availability of reproducible collagen preparations is not always secure, and the use of collagens derived from sources other than vessel wall has been questioned, since they may contain irrelevant or misleading associated materials.
The receptor- and vWF-specific peptides described here allow platelet-reactive surfaces to be constructed in which the composition of the surface can be controlled, allowing independent investigation of GpVI, integrin α2β1 and vWF/GpIb (N. Pugh, N. Raynal and R.W. Farndale, unpublished work). An advantage of this system is that, although vWF coatings will serve this purpose , they are less easily defined and there is no guarantee that the fully active conformation is achieved without interaction with a collagenous substrate. We have found that the creation of a 1:1:1 mixture of peptides specific for each axis supports good thrombus deposition, and that, for thrombi to be stable under high shear, the vWF-binding motif is essential (Figure 3B).
Cell migration and spreading
The study of integrin-mediated cell attachment and migration can be considered from the perspective of both the cell and the extracellular matrix. At the cellular level, the expression of integrin can be modulated or its activity blocked by antagonists, or, alternatively, its activation state can be increased in several ways: by using activatory antibodies, by inducing inside-out signalling or by applying activatory cations such as Mn2+ or Co2+, for example. From the matrix perspective, the only recourse available to date has been to modulate the surface density of the ligand to which the integrin may attach. The availability of this series of collagen-derived integrin ligands offers the opportunity to manipulate both the affinity and density of ligand independently.
We have recently observed that the α2β1-expressing fibrosarcoma cell line HT1080 can bind and spread on GXX′GER-containing peptides, and this activity varies with the affinity of the peptide for the receptor. The capacity to migrate upon surfaces coated with these peptides at fixed density, however, varies inversely with their affinity, so that GFOGER supports little migration at all, and the weak-binding GAOGER allows rapid cell migration (Figure 4). Successive reduction in density of GFOGER on the substrate has a biphasic effect, with migration increasing initially, peaking at approx. 10% surface coverage, then declining as density decreases further. Very low density indeed (∼1% of saturation) can still support rapid cell migration even when cell adhesion is negligible (C. Smerling, N. Raynal and R.W. Farndale, unpublished work). This suggests that the cell integrates its motility over a significant area of membrane, so that a large number of contacts of low affinity can support the same net traction and locomotion as a smaller number of high-affinity interactions.
External regulation of the cell can modify cell behaviour on such peptides, so that inhibition of the activity of the integrin permits migration on the high-affinity GFOGER, whereas activation of the integrin supports adhesion on the low-affinity GAOGER. Thus the native collagen fibre, containing quite widely separated, mainly mid- to low-affinity motifs, can support both cell adhesion and migration: the activation state of the cell determines which behaviour predominates. This offers insight into the regulation of cell homing and targeting, by locating activatory stimuli at the correct density and position to permit migration and to arrest the migrating cell once it has arrived.
The application of Peptide III-23 and the shorter vWF-binding motif in monomeric but triple-helical form was able to induce autophosphorylation of DDR2, expressed in HEK-293 (human embryonic kidney) cells [47a]. The activity of monomeric univalent peptides represents a departure from the accepted model for receptor tyrosine kinase activation in which application of ligand induces receptor dimerization and consequent transactivation. However, the shortest triple-helical peptide (GVMGFO) with the capacity to bind DDR2 did not induce receptor autophosphorylation (Figure 5). This may imply two nearby binding sites within a single copy of the receptor that support overall conformational change and signalling. Alternatively, since the DDR2-binding site in collagen is likely to be too short to accommodate two copies of the receptor side-by-side, the latter may instead interact pincer-like with the same peptide and transactivate each other as a consequence. Such a model requires the two DDRs to be antiparallel, and thus to interact in opposite senses with the triple helix. This model differs from that applied to most receptor tyrosine kinases, where ligand dimerization can support parallel ligand–receptor interactions. Structural studies may help to resolve this issue. It seems very likely, however, that fibrous collagens are not the natural agonists for DDR2, but rather may express their activity as nascent collagen monomers that are secreted by the cell, or may do so after partial degradation of fibres which might allow DDR2 activation to contribute to either the repair or the resorption of the extracellular matrix.
Common themes in protein-binding motifs in the collagens
GPO triplets are often presented as the consensus collagen triplet, and, in view of the high thermal stability of GPO polymers, may have been crucial to the evolution and extension of the protocollagens, being necessary for their assembly as triple helices. With this in mind, it is not surprising that some collagen receptors make use of GPO triplets as specific ligands, including GpVI and LAIR-1, the only examples of immune receptors studied to date. This may represent an ancient receptor lineage.
The GXX′GER motifs form a group of regularly spaced ligands that are recognized by all the collagen-binding integrins. The co-ordination of a metal ion within the I-domain MIDAS (metal-ion-dependent adhesion site) by ligand glutamate is necessary, but not sufficient, to support their interaction: many glutamate residues within COL domains do not bind integrin, and several GER triplets are either inactive or expressed at low affinity only after integrin activation. Hence, the role of the X residue defines the affinity of the motif, with the bulky hydrophobic phenylalanine ring offering a highly complementary ligand for the I-domain surface. The affinity of α2β1 for GFOGER, unique, in our hands, among these GXX′GER motifs, is not increased by cell activation.
The same mode of binding does not apply universally to A domains, although the integrin I-domains may be derived from an archetypal vWF A domain: the cation-replete vWF A3 domain binds collagen in a cation-independent manner, and at a site near to but not overlapping with its MIDAS . vWF and DDR2 both interact with the short hydrophobic motif GVMGFO, as, it seems, does BM-40. These various collagen-binding proteins have few structural features in common, and therefore had different evolutionary development.
Coincidence of sites
The discovery that the vWF site in collagen is identical with a major DDR2-binding site was entirely unexpected given the dissimilar structures of their collagen-binding domains. The addition of BM-40 to this list is remarkable. DDR2 binds one or two other sites within collagen II (and the same may apply to collagen III), including some other GFO-containing motifs. GFOGER, however, does not support DDR2 binding, and thus far remains specific for collagen-binding integrins. The specificity of BM-40 appears similar to that of DDR2, but remains to be established. The determinants of DDR2 and vWF binding within the collagen III sequence GPRGQOGVMGFO differ sufficiently, in their requirement for arginine, the first hydroxyproline, and valine (vWF) or methionine (DDR2), to allow specific synthetic ligands to be constructed that can avoid possible cross-talk between pathways in an experimental setting.
The developmental and physiological consequences of coincident binding sites in collagen are of great interest. It may be that DDR2 expression and/or function is normally sufficiently remote from the vasculature that the collagenous substrate for DDR2-expressing cells is never exposed to vWF, so that DDR2 remains free to ligate exposed collagen. Alternatively, the other DDR2 sites in collagen III (and most likely in collagen I) may be able to compensate for vWF blockade of the Peptide 23 region. DDR2 is expressed in vascular smooth muscle, and may be up-regulated in atherosclerotic plaque, so competition between DDR2 and vWF may be important in vessel wall disease, since exposed collagens would be rapidly saturated with circulating vWF and unavailable to ligate DDR2. DDRs are also overexpressed in breast cancer metastases , and metastasis is reported to be inhibited by endogenous BM-40 expression . In the same vein, a reciprocal relationship between vWF levels and metastasis has been described [53,54]. It is interesting to speculate that competition for collagen may have a role in the regulation of metastasis.
Binding sites for approx. 50 different collagen-binding species on collagen I have been collated . This low-resolution map, based largely on data from rotary shadowing and CNBr peptides, and from the effects of collagen mutation on disease, is powerful enough to identify putative cell-binding and matrix-binding domains of collagen, repeated across the four intact D-periods of the collagen sequence. Since few of these sites have been located with precision upon the collagen molecule, there is considerable scope for the application of Toolkits and derivatives to define sites and to discover other possible coincidences such as that described here between DDR2, BM-40 and vWF. Our data are summarized in Figure 6.
Structure and Function in Cell Adhesion: Biochemical Society Annual Symposium No. 75 held at The Palace Hotel, Manchester, U.K., 3–5 December 2007. Organized and Edited by David Garrod (Manchester, U.K.).
Abbreviations: BM-40, basement membrane protein 40; CRP, collagen-related peptide; DDR, discoidin domain receptor; FcRγ, Fc receptor γ-chain; Gp, glycoprotein; HEK-293, human embryonic kidney; ITAM, immunoreceptor tyrosine-based activation motif; LAIR-1, leucocyte-associated Ig receptor-1; MIDAS, metal-ion-dependent adhesion site; vWF, von Willebrand factor
- © The Authors Journal compilation © 2008 Biochemical Society