Nội dung

T H E T H E O D O R B Ü C H E R L E C T U R E Molecular basis of cytokine signalling – theme and variations Delivered on 8 July 2009 at the 34th FEBS Congress in Prague Walter Sebald1, Joachim Nickel1, Jin-Li Zhang2 and Thomas D. Mueller3 1 Department of Physiological Chemistry II, Theodor-Boveri Institute for Life Sciences (Biocenter), University of Wuerzburg, Germany 2 Institute for Developmental Biology, University of Cologne, Germany 3 Department of Molecular Plant Physiology and Biophysics, Julius-von-Sachs Institute, University of Wuerzburg, Germany Keywords bone morphogenetic proteins (BMP); drug development; interleukins; molecular recognition; receptor oligomers Correspondence W. Sebald, Department of Physiological Chemistry II, Theodor-Boveri Institute for Life Sciences (Biocenter), University of Wuerzburg, Am Hubland, 97074 Wuerzburg, Germany Fax: +49 931 8884113 Tel: +49 931 3188322 E-mail: sebald@biozentrum.uni-wuerzburg.de (Received 27 July 2009, revised 28 September 2009, accepted 4 November 2009) doi:10.1111/j.1742-4658.2009.07480.x Cytokine receptors are crucial for the maintenance, regulation and growth of cells in multicellular organisms. As a common theme in cytokine signalling, single-span receptor chains are assembled in the cell membrane by a ligand enabling cross-activation of the aligned cytoplasmic receptor domains. Nature has created many variations of how this general principle is realized in a cell. Here we focus on cytokines of the four-helix bundle (interleukins) and cystine knot (transforming growth factor-b ⁄ bone morphogenetic proteins) families. Upon activation, receptor chains can form duos, trios, quartets and even larger assemblies. The structure of the extracellular ligand-binding domain of a number of these receptor complexes has now been elucidated, providing the molecular basis for understanding the functional relevance of mechanistic diversity in a cellular context. Biochemical and structural data have revealed ligand recognition mechanisms. Contact sites are usually large and rather flat. A limited number of contact residues provide most of the binding free energy (hot spots). Leaks in hydrophobic seals appear to provide a mechanism for adjusting the affinity of a hot spot interaction (scalability). Bone morphogenetic protein ligands are often promiscuous and interact not only with receptors, but also with a multitude of modulator proteins, which inhibit or enhance bone morphogenetic protein signalling. Cytokine receptor systems offer promising targets for drug development. Information on the structure and the activation mechanism provides leads for developing biologicals, such as engineered cytokines, cytokine mutants acting as receptor antagonists and receptor extracellular ligandbinding domain–Fc fusion proteins. Possible indications exist in the areas of haematology, immunology, inflammation, cancer and tissue regeneration. Introduction In 1968 I attended the FEBS meeting in Prague as Doktorand. I have lasting memories of the opening ceremony in the opera house, which was initiated by the fanfare of Janacek’s Sinfonietta. This is one of the reasons why when I think of Prague I also think of music. In the late 1960s, Theodor Bücher gave a well- Abbreviations BMP, bone morphogenetic proteins; CV-2, crossveinless-2; GDF, growth and differentiation factor; IL, interleukin; SMAD, homologs to the protein from Caenorhabditis elegans SMA and Drosophila mothers against decplentaplegic; STAT, signal transducers and activators of transcription; TGF, transforming growth factor; VWC, Von Willebrand factor type C; cc, common c chain. 106 FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS W. Sebald et al. received traditional Christmas lecture for the medical students about the storage and realization of genetic information. As a sounding illustration of this topic, one of the Brandenburg Concertos was played and in parallel the single pages of the partitur, the musical score, were projected; I had to change the slides in harmony with the music. So this is one of the reasons why when I think of Theodor Bücher I also think of music. Another reason, of course, is his close association with the Munich Bach-Chor, which has been catalysed by Ingrid Bücher, who I would like to thank for attending the Theodor Bücher Lecture at the 34th FEBS Congress. The present Theodor Bücher Lecture on the molecular basis of cytokine receptor signalling – theme and variations – has three movements, like a sonata. First, we will look at the basic mechanism and the many variations realized in diverse receptor systems. Second, we will discuss molecular recognition in these receptors; this means the structural basis for affinity and specificity. And, third, we will see how the accumulated data on structure and mechanism aid in the development of drugs. Basic mechanism – receptor oligomerization For a long time it was a mystery how single-span membrane proteins, like cytokine receptors, can signal into a cell. These receptors have an extracellular binding domain, which is connected to a cytosolic domain by only a short peptide segment probably folded in the membrane as a single a-helix. It is difficult to conceive how such a segment can transduce a signal from the outside to the inside of a cell. How, therefore, can an extracellular signal initiated by ligand binding be propagated across the membrane? It is clear now that single-span receptor chains cannot signal alone. They function as oligomers. Binding of the ligand leads to an oligomeric state of the extracellular domains, which is transmitted to the cytosolic domains inside the cell. This general theme ‘signalling by oligomerization’ has been the ‘Leitmotiv’ of receptor research for many years. It was called ‘horizontal signalling’ in a 2004 review by Stroud & Wells [1] to set it apart from the ‘vertical signalling’ of multi-span membrane receptors, such as G-protein coupled receptors, which employ a transmembrane conformational change (Fig. 1). In the most simplistic model, the receptor chains diffuse freely in the membrane and are bound together – oligomerized – in the presence of the ligand. Recently, evidence has accumulated that some single-span receptor chains can form complexes by Molecular basis of cytokine signalling A B Fig. 1. Horizontal versus vertical receptor signalling [1]. Signalling across membranes requires either a conformational change in a receptor or a change in the oligomerization state of the receptor. (A) Single-span transmembrane receptors are examples of so-called horizontal signalling. Upon ligand binding to one receptor subunit a binary complex intermediate is formed, in the subsequent step a second (or further) receptor subunit is recruited into the complex, leading to the activation of the cytoplasmic receptor parts, e.g. by transphosphorylation of inherent or receptor-associated kinases. (B) Vertical receptor signalling is initiated in a single receptor (or preformed oligomer) by transducing a ligand-induced conformational change from the extracellular to the intracellular side. themselves, so-called preformed complexes, which are inactive without a ligand [2–4]. Here, ligand binding probably initiates a conformational change, which is transmitted across the membrane. On the other hand, some G-protein coupled receptors have been found to oligomerize during signalling [5]. Thus, it seems that there exist a variety of intermediate receptor states between pure ‘horizontal’ and pure ‘vertical’ signalling. This simple and elegant horizontal signalling mechanism integrating the membrane as the organizing principle was very successful during the evolution of multicellular organisms. It is therefore not surprising that the signalling receptor oligomers vary considerably, differing in stoichiometry and topology. An oligomerization mechanism was postulated for the first time by Schlessinger [6] for the epidermal growth factor receptor. Here, the formation of homodimeric receptors is triggered by the binding of two ligands. However, Cunningham et al. [7] showed that a homodimeric growth hormone receptor is formed by binding to a single ligand. Another renowned example for a 1 : 2 stoichiometry is the receptor for erythropoietin [8]. In the growth hormone receptor, the two receptor chains differ; they are bound to different ligand epitopes in a high- and a low-affinity mode. It is therefore not unexpected that heterodimeric oligomers exist, where two different receptor chains are bound by one ligand, as in the interleukin-4 (IL-4) receptor [9,10]. This division of labour between different chains opens FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS 107 Molecular basis of cytokine signalling W. Sebald et al. up a whole range of new possibilities for cellular signalling. Even more complex oligomers are assembled by dimeric ligands, such as the bone morphogenetic proteins (BMPs) and other members of the transforming growth factor-b (TGF-b) superfamily. Here, twice heterodimeric receptors are assembled by the dimeric ligand [11,12]. This can lead to avidity effects, where ligand affinity is increased by binding simultaneously to two receptor chains. The formation of heterodimeric ligands and ⁄ or multiple receptor chains might allow specific signalling modes, for instance during development. Receptor structures Here we will discuss a few receptor structures, and focus on the extracellular domains only, in particular on the binding domains for the ligand. Although the first structures were elucidated in the early 1990s, the more complex ones have only recently been described. The homodimeric complex of the growth hormone receptor represents the prototype and the reference structure for many other systems [13]. The growth hormone ligand consists of a helix bundle. Site 1 constitutes a high-affinity epitope and site 2 a low-affinity epitope. Both bind the same receptor species. It is unclear why this polarization into high- and low-affin- ity sites originated. However, as a consequence, the oligomerization is often considered an ordered sequential process (Fig. 2). Step 1 is the binding of the solute ligand at the high-affinity site and in step 2 the second chain is recruited in the membrane to form the signalling oligomer. The cytosolic parts of the homodimer carry tyrosine kinases, which transphosphorylate and thus activate the twin chain. This creates docking sites for signal transducers and activators of transcription (STAT) proteins, which initiate and propagate the signal within the cell. For the intracellular part, the homodimer is symmetrical. Each chain can function as a trigger, which transactivates, or as a driver, which initiates intracellular signalling. This symmetry is broken in the heterodimeric receptors, as shown in Fig. 2 for the IL-4 receptor [14–16]. One chain, called the common c chain (cc), is the trigger, which can only transactivate. The other chain, IL-4Ra, is the driver, which can only initiate the intracellular signal. The division of labour is indicated by the cytosolic domains. The trigger, cc, contains only a binding site for the tyrosine kinase Janus kinase 3 (JAK3), which transactivates. The driver, IL-4Ra, contains a large cytosolic domain with binding motifs for Janus kinase 1 (JAK1), the intracellular signalling protein STAT6, the insulin-receptor-substrate 2, and others. Again there exist a high-affinity chain, IL-4Ra, A B 108 Fig. 2. A two-step sequential binding mechanism allows for a simple design of antagonists [9]. Signal transduction of single transmembrane receptors, e.g. cytokine receptors, often follows a sequential binding mechanism. (A) In the first step, the ligand binds to its high-affinity receptor subunit forming an intermediate binary complex. In the second step, the low-affinity receptor subunit is recruited into a ternary complex (higher oligomeric states are also possible), leading to intracellular receptor activation (indicated by the star). (B) A mutated variant, which is not capable of binding to the second receptor subunit but with unaltered binding to its first receptor subunit, will still form the binary complex, but cannot proceed to the second step and thus is unable to activate the receptor [58,59,61,67]. This antagonist is most efficient in blocking receptor activation if binding affinity to the second receptor subunit does not contribute significantly to the overall ligand–receptor binding affinity. FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS W. Sebald et al. and a low-affinity chain, cc. Therefore, the assembly of the signalling receptor heterodimer proceeds in two steps: First, solute IL-4 binds to IL-4Ra. The solute IL-4 is concentrated 100- to 1000-fold at the membrane surface. This concentration effect and also probably the two-dimensional diffusion in the membrane, facilitate the following recruitment of cc. The assembly of the ternary IL-4 receptor complex can be simulated at a biosensor surface [17]. The solute IL-4 at 1–10 nm concentrations associates rapidly with the immobilized IL-4Ra chain. Buffer alone results in a very slow dissociation with a half-life of  5 min. When the immobilized IL-4Ra has been first saturated with the IL-4 ligand, more and more of the ternary complex can be formed after the addition of increasing concentrations of cc. Dissociation of cc is fast and its affinity to IL-4 corresponds to a dissociation constant (KD) of 3 lm. This is more than 10 000-fold lower than the affinity for IL-4Ra. The IL-4Ra chain is shared by three receptor–ligand complexes: two IL-4 receptors containing either cc or IL-13Ra1 as a second chain, and one IL-13 receptor containing IL-13Ra1 [18]. As a consequence, genetic or pharmacological inactivation of the shared IL-4Ra will abolish not only IL-4, but also IL-13 signalling. This will be discussed further below. The cc family is larger, with cc being shared by at least five receptors, including the IL-2 receptor [14]. The receptor for IL-2 exists in two forms. A mediumaffinity heterodimeric receptor exists in natural killer cells. Its architecture corresponds to the IL-4 receptor. The driver is IL-2Rb, and cc again functions as the trigger. A second high-affinity IL-2 receptor exists in activated T-lymphocytes. It also contains the coreceptor IL-2Ra, also called Tac [19]. This coreceptor enhances affinity specifically for IL-2. In other cells, a different coreceptor, IL-15Ra, co-operates with the same heterodimer to provide enhanced affinity for IL-15. The structure of the tetrameric high-affinity IL-2 receptor shows that the coreceptor IL-2Ra interacts only with the IL-2 ligand. It has no contacts with the other two chains. This is a telling example of the importance of concentrating the ligand at the surface of the membrane. A soluble IL-2Ra without membrane anchor functions as an inhibitor of IL-2 signalling. Finally, as a further variation of horizontal signalling we will discuss the hexameric BMP receptors (Fig. 3). These complexes are not true hexamers, as the BMP ligand is a disulfide-bonded homodimer [11,12]. The dimeric ligand assembles a heterodimeric receptor at each end. The extracellular domains are small and linked to the membrane-spanning segment by a short peptide segment. This places the binding domains close Molecular basis of cytokine signalling A B Fig. 3. The ternary complex of BMP-2 ⁄ BMPR-IA ⁄ Act-RIIB forms a heterohexameric complex. (A) A side view of the ternary complex of BMP-2 (UniProtKB P12643; the BMP-2 dimer is indicated in blue and yellow) bound to the extracellular domains of its type I receptor BMPR-IA (UniProtKB P36894; green) and its type II receptor ActRIIB (UniProtKB Q13705; red). The membrane surface is indicated by yellow spheres. The membrane-proximal C-termini of the receptor ectodomains were missing in the crystal structure of the ternary complex (PDB entry 2H64 [11]) and were therefore not modelled. (B) A top view of (A) showing the two-fold symmetry of the ligand– receptor complex imposed by the symmetrical ligand homodimer. to the membrane. The binding domains of the receptor chains have no contact with each other. They are bound together solely by the BMP ligand. The BMP receptors are set apart from the cytokine receptors described above by employing a serine ⁄ threonine kinase (and not tyrosine kinases) in their cytoplasmic domains and homologs to the protein from Caenorhabditis elegans SMA and Drosophila mothers against decplentaplegic (SMAD) proteins (and not STAT proteins) as intracellular signalling proteins. However, BMP receptors obey the general rule that one chain (type II) is the transactivating trigger and the other chain (type I) is the driver activating the SMAD proteins by phosphorylation [20]. Several proteins have been identified that qualify as FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS 109 Molecular basis of cytokine signalling W. Sebald et al. coreceptors. For instance, repulsive guidance molecule proteins determine affinity and specificity for certain members of the BMP family [21], or b-glycan functions as a coreceptor for TGF-b2, which belongs to the same family as the BMPs [22,23]. However, no structures comprising such coreceptors have been determined and therefore we do not know in molecular detail how they function. The binding of two trigger and two driver chains to a dimeric ligand has profound consequences for BMP signalling. Multiple interactions of the ligand with membrane receptor chains provide new opportunities for a cell to determine and tune receptor affinity and, therefore, specificity. Combinatorial assemblies of heterodimeric BMPs and mixed receptor chains are possible [24]. A B Molecular recognition The structures of the complexes provide a wealth of information on the mechanism of cytokine receptor signalling. As Theodor Bücher put it: ‘Function is structure in action’. Of particular importance is the structural definition of the interfaces between a cytokine and a receptor. In principle, these contact sites, called structural epitopes, carry all the determinants for the molecular recognition among these proteins, i.e. for the affinity and the specificity of their interaction. However, it is still a big challenge to understand or even to predict how these structural epitopes create binding free energy during association. One problem is that these epitopes are large and flat [25]. They have sizes between 800 and 1500 Å2 and comprise 20–25 residues. This is similar to the interfaces of antibody– antigen complexes. Often there exist no obvious knobs or holes that could suggest geometric complementarity and therefore binding. It was an influential new concept that contact residues are not of equal importance for binding. Clackson & Wells [26] performed a mutational analysis of growth hormone and receptor and could demonstrate that a few contact residues contribute the major part of the binding free energy. They coined the term ‘hot spots’, which is now regularly used in the field. The functional binding epitope defined by alanine mutations is smaller than the structural epitope defined by the residues buried in the contact. In the functional epitopes of the growth hormone and the receptor exists one hot spot created by two tryptophan residues (104 and 169) interacting with complementary hydrophobic residues of the hormone. The difference between a structural and a functional epitope has now been established in numerous cytokine–receptor contacts [27]. However, epitopes can be mosaic in comprising several independent hot spots. 110 Fig. 4. The hot spot of binding determinants in the IL-4 ⁄ IL-4Ra complex are formed by a so-called ‘avocado cluster’ [16]. Two polar bonds (a hydrogen bond or a salt bridge) comprise the main binding determinants of the IL-4 ⁄ IL-4Ra ligand–receptor interaction, contributing more than 80% of the overall binding free energy. (A) The side chain guanidinium group of Arg88 of IL-4 (UniProtKB P05112) forms a bidentate salt bridge with the carboxylate group of Asp72 of IL-4Ra (UniProtKB P24394). This salt bridge is shielded from solvent access due to the surrounding hydrophobic residues from the receptor (Leu39, Phe41, Leu43 and Val69) as well as the ligand (Y56 and K84). (B) The side chain of Glu9 of IL-4 forms several hydrogen bonds to the main and side chain groups of IL-4Ra (Tyr13 OH, Ser70 main chain amide, Tyr183 OH). Similar to the salt bridge formed by Arg88 of IL-4, the hydrogen bonds emanating from Glu9 are effectively shielded by the hydrophobic environment provided by Ile5 (IL-4), Tyr13, Val69, Tyr127 and Tyr183 of IL-4Ra. The shielding from solvent embeds the polar bonds into a vacuum-like environment, thereby dramatically increasing the contribution of these noncovalent bonds to the overall binding energy. Because the embedding of a polar bond into a surrounding hydrophobic environment is reminiscent of the placement of seeds in a fruit, this setup was called the avocado cluster [16]. Also, there exist strong polar bonds. As an example, the IL-4 receptor system will be discussed (Fig. 4), in particular the interface between IL-4 and the highaffinity IL-4Ra chain [16,28,29]. Two main binding determinants are identified in IL-4: the acidic residue Glu9 and the basic residue Arg88. Mutation of either residues to alanine leads to  1000-fold loss in receptor affinity. The crystal structure of the complex shows that the Arg88 forms a perfect salt bond with receptor Asp72 and that the Glu9 forms a hydrogen bond FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS W. Sebald et al. network with three tyrosines of the receptor. These bonds represent the hot spots in the receptor epitope. A more thorough analysis by a double mutant cycle indicated that the two hot spots bind independently of each other and that each of them is surrounded by a shell of hydrophobic side chains, which co-operate with the polar core in binding. This motif has been called an ‘avocado cluster’ in order to suggest that the polar bond of the hot spot has to be shielded from the bulk solvent by a hydrophobic shell. It has also been called the ‘O-ring model’ by Bogan & Thorn [30] or ‘core ⁄ rim patches’ by Conte et al. [25]. The IL-4 contact with the IL-4Ra chain contains an additional third element, which is positively charged at IL-4 and negatively charged at the receptor [31]. Molecular dynamics calculations suggest that the very highly charged interfaces of IL-4 and IL-4Ra – not the avocado nature of the site – lead to electrostatic steering during the association of the two proteins and, thus, to an 10-fold increase in the association rate constant. This unusually fast association can be measured by Biacore interaction analysis, as described above, and contributes to the high affinity of the IL-4 receptor corresponding to a very low dissociation constant KD of  100 pm. Sharing receptor chains is common among cytokines [14,32]. cc functions with IL-2, IL-4 and several other ILs, as discussed above. Other receptor families employ the common b chain or the common gp130. Promiscuity and sharing receptor chains also exist in the BMP ⁄ growth and differentiation factor (GDF) ⁄ activin ⁄ TGF-b superfamily [33]. Of particular interest are the type II activin receptor chains IIA and IIB. They bind with high affinity to activins and certain GDFs and with low affinity to BMPs. The structural epitopes at the interfaces are largely hydrophobic with a single serine at the core [11]. According to the structure, this serine establishes a hydrogen bond with the receptor Leu61 main chain. However, mutational analyses indicate that this bond does not contribute to the binding affinity of BMP-2. It does not represent a hot spot, not even a minor determinant. Surprisingly, this hydrogen bond is conserved in the receptor complexes with activin A and BMP-7. In the complex with BMP-2 and BMP-7 it does not contribute to binding affinity. However, in the activin complex it is a hot spot of binding energy, and it is responsible for the high-affinity interaction with this ligand. What makes this bond binding? When the residues surrounding Ser88 are compared in BMP-2 and activin A, a few differences are found. Fortunately, swapping two activin residues, an aspartic acid and a lysine, yielded a BMP-2 with activin-like affinity. We know the structure of the complex Molecular basis of cytokine signalling between the aspartic acid ⁄ lysine mutant of BMP-2 and ActR-IIB. The structure does not indicate any new bonds in trans between the ligand and the receptor. The swapped side chains form an ion pair in cis, which fixes the hydrophobic parts of the lysine in such a way that it seals the Ser88 from the bulk solvent. Evidence is accumulating that the sealing effect in an avocado cluster is used by some receptors to scale affinity according to the signalling requirements [18]. Inherited diseases demonstrate that small changes in receptor affinity can be crucial for in vivo function (Fig. 5). Human BMP-2 and human GDF-5 bind with high affinity to the BMP receptor IB. BMP-2 has an even slightly higher affinity for the IA subtype, whereas GDF-5 affinity for IA is nearly 20 times lower. Nickel et al. [34] identified the determinant for this specificity as Arg57 occurring in a loop region of GDF-5. A mutation of this large basic residue to an alanine in GDF-5 causes a 20-fold gain in IA affinity. A substitution of Arg57 by a leucine residue produces an intermediate effect. In Berlin, Seemann et al. [35] studied a family with inherited symphalangism. They identified the very same Arg57Leu substitution in the GDF-5 of the afflicted individuals. These observations suggest that the gain of affinity in the GDF-5 mutant leads to an inappropriate high signalling by the IA subtype. The outcome is a hyperproliferation of chondrocytes and, as a consequence, a loss of certain joints. The recently established structure of GDF-5 in complex with the IB receptor [36] reveals the molecular A B Fig. 5. (A) Familial symphalangism caused by a gain-of-function mutation in GDF-5 (UniProtKB P43026) [35]. Joints are replaced by bone in finger V and defective in finger IV (see arrows). The R438L mutation is located in the wrist epitope of GDF-5 (R57L in the mature protein). The mutant GDF-5 has a several-fold increased affinity for the BMPR-IA receptor. (B) A similar phenotype is produced by loss-of-function mutations in the NOG gene coding for the BMP and GDF-5 inhibitor Noggin (UniProtKB Q13253). (Reproduced with kind permission of The Journal of Clinical Investigation via the Copyright Clearance Center.) FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS 111 Molecular basis of cytokine signalling W. Sebald et al. basis of receptor specificity and discrimination. A rigid disulfide-stabilized loop has different orientations in the subtypes. In the IA receptor, the loop occludes the binding site and allows the binding of only a small alanine side chain. In the BMP receptor IB, the loop is oriented away and gives room for the bulky arginine of GDF-5. In summary, small structural variations leading to small and selective changes in affinity can be of high functional importance and result, in the case of GDF-5, in profound chondrodysplasias of skeletal elements in vivo. BMPs not only interact with receptors. A large variety of proteins occur in the extracellular compartment that bind BMPs and regulate their activity [37,38]. These proteins provide fascinating paradigms for molecular recognition, as they often interact with the same epitope. Well-known representatives are Noggin, follistatin and the members of the differential screening-selected gene aberative in neuroblastoma (DAN) family. Numerous proteins belong to the Chordin family, which typically contain one or multiple Von Willebrand factor type C domains (VWC domains) [39]. Members are Chordin itself, the Chordin-like proteins 1 and 2, crossveinless-2 (CV-2), connective tissue growth factor and others. These proteins are essential during gastrulation for dorsal–ventral patterning and neural induction [40]. They occur in the Spemann organizer (Chordin) and in the ventral centre (CV-2, twisted gastrulation). Later in development they regulate organ formation; in the adult they regulate the regeneration of organs and tissues. The VWC domain is a versatile protein module that occurs in many forms. Some of them can bind BMPs or other proteins; some seem to exert a purely structural role. Of particular interest is VWC1 of CV-2. Zhang et al. [41,42] demonstrated that, with zebrafish CV-2, out of the five modules present, only VWC1 binds BMP-2. The affinity is high, comparable with the BMP receptor IA. Two CV-2 proteins can bind one BMP molecule. The complex of VWC1 and BMP-2 has been isolated. The crystal structure revealed how VWC1 inhibits BMP signalling [43] (Fig. 6). The small module of only 66 residues is tripartite. A short N-terminal segment of eight residues occupies the binding epitope for the IA receptor; a subdomain SD1 of 34 residues binds to the epitope for the type II receptor; the C-terminal subdomain SD2 points away from the complex and has no contacts with BMP-2. Most of the binding energy is provided by the SD1 part. This hydrophobic interaction alone has a micromolar KD. The N-terminal segment extends across the small ridge, like a paper clip, onto the other side of BMP-2 and provides a 1000-fold increase in affinity. The SD1 and the clip 112 together compete efficiently for receptor binding and therefore prevent BMP-2 signalling. The BMP inhibitor Noggin uses a similar trick for the generation of high-affinity binding [44]. This beautiful structure has been elucidated by Groppe et al. [44]. It shows that Noggin also uses an N-terminal extension to block the binding epitope of BMP-7 for the type I BMP receptors. Thus, a clip-like extension to generate an additional binding epitope might represent a more general mechanism to increase affinity. Drug design and development When working in the Bücher Institute, I experienced not only the atmosphere of competitive and ambitious basic research, but there was also always a readiness to improve or to invent something. A major stimulus, of course, was the invention and the design of the Eppendorff system. The Eppendorff caps, pipettes, centrifuges, incubators and photometers have established a worldwide standard for equipment in academic, industrial and clinical laboratories. A keen sense for industrial applications is also a hallmark of cytokine research. Cytokine signalling is vital for the growth, maintenance and repair of cells and tissues in our body. Dysregulation of cytokine function can result in serious and widespread diseases. Not surprisingly, therefore, cytokines and cytokine receptors are promising targets for drug design and development. Basic research has generated a remarkable spin-off of new drugs. Several of them are already very successful on the pharmaceutical market. Most of these therapeutics are, however, biologicals; this means they are recombinant proteins. The development of synthetic drugs is made difficult by the architecture of the binding epitopes and the activation mechanism, in particular of heteromeric receptors, as discussed above. Recombinant erythropoietin [45] and granulocyte colony-stimulating factor (Neupogen) [46] are now well-established therapeutics. New players in tissue engineering and regenerative medicine are the BMPs [47], which induce the formation of new bone at critical size defects that otherwise would not heal. Recombinant BMP-2 is a powerful protein that allowed the repair of a 5 cm defect in the mandible of a Göttingen minipig [48] (Fig. 7). A functional, mechanically stable and vascularized new bone formed in situ within 8–12 weeks. Spinal fusion, bone augmentations and the treatment of nonhealing fractures represent major clinical applications of BMPs. In the USA alone, more than 100 000 patients with unstable or collapsed vertebral bodies were treated last year. Mechanical load during healing is essential. After ectopic application of FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS W. Sebald et al. Molecular basis of cytokine signalling A B C Fig. 6. Clip-like structures gain binding strength by co-operative interactions. (A) A schematic representation of the binding mechanism of the BMP modulator proteins ⁄ domains Noggin and CV-2 (UniProtKB Q5D734) VWC1 to BMPs. An N-terminal extension (clip) binds into the epitope for the type I receptor of the ligand, whereas the main core structure binds into the epitope for the type II receptor of the BMP ligand. Therefore, the binding of the receptors of both subtypes is blocked and BMP activity is effectively suppressed. Because of the strong co-operativity of both interfaces (clip and core structure) the contribution of the individual binding interfaces can be small. (B) The binding of two N-terminal VWC domains of CV-2 (grey, left in surface representation) to the dimeric BMP-2 (blue and yellow) resembles the stacking of a paperclip (VWC1 of CV-2) to a sheet of paper (BMP-2) (PDB entry 3BK3 [43]). (C) The binding of Noggin to BMP-7 (PDB entry 1M4U [44]) follows a similar mechanism as in (B). An N-terminal clip folds into the type I receptor-binding site of BMP-7, whereas the core structure blocks the type II receptor binding. The much higher binding affinity of Noggin for BMP ligands can possibly be explained by the homodimeric nature resulting in four binding interfaces for a single Noggin molecule. BMP, for instance in a muscle pouch, the induced bone is resorbed at later stages when transplanted in a functional site under mechanical stress. Thus, the cultivation of artificial bone with a certain desired shape is science fiction at the present state of the art. Soluble receptor ectodomains are specific inhibitors of their genuine cytokine ligands. Fusion proteins consisting of the constant Fc part of an immunoglobulin and two receptor domains are even more potent, as the cytokine can be bound at two sites. They function as efficient ligand traps. The Fc-fusion protein with the ectodomain of the activin receptor IIA is a powerful inhibitor of its high-affinity ligands, in particular activin A. ActRIIA–Fc induces an increase in bone mass in ovariectomized mice [49]. A clinical study has recently shown that the human fusion protein provides an effective treatment of osteoporotic bone loss in postmenopausal women [50]. Most importantly, the inhibition of ActR-IIA ligands stimulates bone formation by osteoblasts and therefore increases bone mass. Treatment with, for instance, biphosphonates inhibits bone degradation by osteoclasts and thus at best preserves the status quo. Following the same approach, an Fc-fusion protein with the ectodomain of the activin receptor IIB was developed. The IIB receptor subtype has two ligands: GDF-8 and the very similar GDF-11. These GDFs are bound with even higher affinities than the activins. The signalling of GDF-8 and -11 is inhibited by the fusion protein ActR-RIIB ⁄ Fc at the very low IC50 of 100 pm [51]. GDF-8 has also been called myostatin. This protein became well known because disruption of the myostatin gene in mice [52], cattle [53] and man [54] leads to a dramatic increase in muscle mass, the so-called double-muscling phenotype. The ActR-IIB fusion protein when injected into mice produces an FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS 113 Molecular basis of cytokine signalling W. Sebald et al. A B C Fig. 7. Direct reconstitution of the mandible bone of a minipig [48]. (A) X-ray control taken immediately postoperative. (B) A critical size 5 cm defect in the mandible was treated with carrier material plus recombinant BMP-2. Full regeneration of the mandible with a mechanically stable bone is visible in the X-ray taken after 8 weeks. The control defect treated with carrier alone formed a pseudarthose and the defect was filled with connective tissue. (C) Explanted mandible bone shown in (B) (12 weeks postoperative) demonstrates complete reconstitution of the bone. (Reproduced with kind permission of Springer Science+Business Media.) even more pronounced muscle phenotype, possibly because it neutralizes both GDF-8 and GDF-11 [53]. The fusion protein also increases muscle mass in an mdx mouse, an animal model of muscular dystrophy. 114 Thus, it represents a promising drug candidate for the treatment of diseases associated with muscle loss or wasting. Another type of inhibitor has been generated by mutating cytokine ligands. An IL-4 mutein, Aerovant, is now in clinical phase IIB trials as a drug candidate for the treatment of allergic asthma [55]; a growth hormone mutein, Pegvisomant, is already in clinical use for the treatment of acromegaly [56]. Allergies and asthma represent a nuisance in the case of seasonal rhinitis or conjunctivitis and a life-threatening condition in anaphylactic shock and asthma. IL-4 and IL-13 are the hormones that make us allergic. During the sensitization phase, IL-4 triggers the formation of type 2 T helper lymphocytes. Type 2 T helper cells then secrete cytokines that initiate the formation of IgE in B cells, which finally leads to the symptoms of a delayed hypersensitivity reaction. In the effector phase, IL-4 co-operates with IL-13. A rational drug design is straightforward on the basis of the activation mechanism (see Fig. 2) and of the functional epitopes [57] (Fig. 8). As discussed above, there exist two IL-4 receptors and one IL-13 receptor, all of which use the IL-4 receptor a chain as the essential driver. An inhibition of the a chain will therefore inhibit IL-4 as well as IL-13 signalling. Two mutations of IL-4 are necessary to disrupt the interaction with the low-affinity chains cc and IL-13Ra1 [58]. The double mutein binds with nearly unchanged affinity to the cellular IL-4 receptor, as the low-affinity chains contribute only marginally to the affinity. The double mutein, Aerovant, is therefore a potent antagonist of IL-4 and IL-13. Animal studies have shown that the IL-4 mutein effectively inhibits an anaphylactic shock in mice when applied during the sensitization phase [59]. Recently, clinical trials have shown that Aerovant can also ameliorate allergic asthma in human patients [55]. Following the same rationale, an antagonist of growth hormone has been designed and developed [60]. Increased growth hormone production by, for instance, a pituitary adenoma, leads to a phenotype called acromegaly, which is typically associated with large body size and, among other symptoms, a prominent supraorbital ridge and a large nose and jaw. In the homodimeric growth hormone receptor, the second chain is bound with low affinity to the ligand, as described above. This interaction can be abolished by introducing a mutation in the functional epitope, substituting a small glycine with a large arginine. This mutein has efficiently inhibited growth hormone action in an animal model. However, large amounts had to be applied, as the affinity of the mutein for the cellular FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS W. Sebald et al. A Molecular basis of cytokine signalling B C D E F Fig. 8. An electrostatic mismatch is the basis of the antagonistic property of the IL-4 variant Y124D [67]. (A) The first step of IL-4 receptor activation is the binding of IL-4 (green) to its high-affinity receptor IL-4Ra (cyan). (B) The binary complex then recruits the lowaffinity receptor subunit cc (orange surface representation) into a heterotrimeric complex (C) (PDB entry 3BPL [15]). In the case of the IL-4 antagonist variant Y124D the formation of the ternary complex is blocked (D). Circles mark the interaction of the tyrosine residue of IL-4 with residues of cc. (E) Closer inspection of this area reveals that the side chain of Tyr124 of IL-4 is embedded in a hydrophobic cleft formed by the residues His159, Cys160, Leu208 and Cys209 of cc, with both cysteine residues forming a disulfide bond. (F) A model of this interaction with IL-4Y124D instead of wild-type IL-4 shows that the negatively charged carboxylate group of Asp124 would be placed in the centre of the hydrophobic interface, thereby causing electrostatic repulsion, which explains the loss of binding of IL-4Y124D to cc [67]. receptor was severely reduced compared with normal growth hormone. Therefore, six additional mutations were introduced, which increased affinity of the mutein to wild-type levels. In addition, the mutein was pegylated (i.e. covalently modified with polyethyleneglycol), in order to prolong the half-life of the protein in the body. This engineered and modified growth hormone antagonist (pegvisomant) is in clinical use for the treatment of acromegaly. Cytokine signalling still provides a fertile ground for the development of biologicals – protein drugs. However, it is still a big challenge to find chemical compounds that bind to functional epitopes of cytokines or their receptors. It appears that small peptides can function as agonists in homodimeric receptors, such as in the receptor for erythropoietin [61]. Chemicals have been found that inhibit IL-2, but, surprisingly, they bind outside the functional epitope. The compound Ro26-4550 distorts the conformation of IL-2 and therefore destroys the receptor-binding epitope [62]. An elegant method called ‘fragment tethering’ has been invented by Erlanson et al. [63] to screen for ligands with very low affinities. The future will show whether such ligands may be used as lead structures for further drug development. Other approaches involve large synthetic chemicals, such as dendromers or foldamers [64,65], which can expose large surfaces similar to the binding epitopes of cytokine receptors. So, the quest continues to reach high-hanging fruit [66]. Acknowledgement W. Sebald wishes to thank the organizers of the 34th FEBS Congress. It was a great privilege to present the Theodor Bücher Lecture. FEBS Journal 277 (2010) 106–118 ª 2009 The Authors Journal compilation ª 2009 FEBS 115