Mechanistic Diversity of Cytokine Receptor Signaling Across Cell Membranes

See allHide authors and affiliations

Science's STKE  04 May 2004:
Vol. 2004, Issue 231, pp. re7
DOI: 10.1126/stke.2312004re7


Circulating cytokines bind to specific receptors on the cell outer surface to evoke responses inside the cell. Binding of cytokines alters the association between receptor molecules that often cross the membrane only once in a single alpha-helical segment. As a consequence, association of protein domains on the inside of the membrane are also altered. Increasing evidence suggests that an initial "off-state" of associated receptors is perturbed, and brought to an activated state that leads to intracellular signaling and eventually effects a change in DNA transcription. The initial detection event that transduces the change in receptor association is sensitive to both proximity and orientation of the receptors, and probably also to the time that the activated state or receptor association is maintained. Ultimately, a cascade of phosphorylation events is triggered. The initial kinases are sometimes part of the intracellular domains of the receptors. The kinases can also be separate proteins that may be pre-associated with intracellular domains of the receptors, or can be recruited after the intracellular association of the activated receptors. We focus here on each of the cases for which structures of the activated cytokine-receptor complexes are known, in a search for underlying mechanisms. The variations in modes of association, stoichiometries of receptors and cytokines, and orientations before and after activation of these receptors are almost as great as the number of complexes themselves. The principles uncovered nevertheless illustrate the basis for high specificity and fidelity in cytokine-mediated signaling.


The membranes of all cells contain receptors that harmonize external and internal environments. Reversible associations of proteins with consequent changes in structure are dominant themes in signal transduction pathways both inside and outside the cell. This review briefly introduces some of the general principles of receptor-mediated signaling and then focuses on one of the most highly characterized classes of receptors, the cytokine-receptor superfamily. The structures of the cytokine receptors show in atomic detail how they are activated and how they utilize a diversity of protein folds and oligomeric structures to transmit signals.

Fundamentals of Receptor Structure

Most receptors are activated by binding a ligand on one side of the membrane; they then initiate a response on the other side. Most ligands do not pass through the membrane, but associate with extracellular binding sites within a single receptor molecule, or through bridging interactions between multiple receptors. These ligands or effectors may be as small as a photon, in the case of photoreceptors, or as large as globular proteins, such as the cytokines.

Changes in the relative positions of transmembrane segments and their associated protein domains bring about signal transduction across the membrane. This can occur within a single multi-spanning transmembrane protein, in a mechanism that we describe as "vertical signaling," or by changing lateral associations of receptors, which we term "horizontal" signaling (Fig. 1). These two signaling paradigms have different etiologies, structures, and functions (Table 1).

Fig. 1.

The characteristics of horizontal and vertical receptor signaling mechanisms. Vertical receptor signaling (right) takes place within a single preassembled protein or oligomer of protein subunits that constitute a transmembrane receptor. Signaling requires that a conformational change be relayed from outside to inside the cell by means of the trans-bilayer elements. Horizontal receptor signaling (left) depends on diffusion of receptor molecules within the membrane plane. Increasing evidence favors the idea of receptor associations existing in an "off-state" before activation (far left). The ligand is in yellow. The bilayer is indicated as spheres that represent phosphatidyl head groups. Ligand binding to one receptor causes dissociation of inactive complexes and induces further association that involves at least one other receptor. The oligomeric state and orientation is imposed on the intracellular components of the receptors (below the membrane) where associations of kinases lead to their activation.

Vertical Versus Horizontal Signaling

Vertical signaling receptors are found in the membranes of all organisms from bacteria through higher eukaryotes. Here, the initial ligand binding event produces changes in the receptor conformation vertically through the membrane. These receptors often respond to small nonprotein ligands that cause a transient and reversible change in membrane potential or in cellular metabolism. In the case of ligand or voltage-gated ion-conducting receptors, the response-mediated gating controls are also found within the transmembrane region. The preassembled and generally unimolecular nature of these receptors allow them to respond rapidly, as is the case for neuronal signaling.

The vertical receptors include the seven-transmembrane receptor superfamily, light-sensing receptors, ion channels, and nutrient-sensing receptors, among others. These receptors undergo an intramolecular structural change in response to ligand binding (Fig. 1) (1, 2). The structurally best characterized of seven-transmembrane membrane proteins are rhodopsin (3, 4) and bacteriorhodopsin (which is structurally homologous to rhodopsin). Changes in response to activation by light have been detected in the cytoplasmic regions of rhodopsin (5-7). Electron crystallography (8-10) and x-ray crystallography (11) describe structural changes that open a short channel on the cytoplasmic surface of bacteriorhodopsin in response to activation by light; this process may be prophetic of other changes in the heterotrimeric guanine nucleotide--binding protein (G protein)-coupled receptors (GPCRs). Vertical receptors also include the chemotactic bacterial receptors for aspartate that are pre-assembled as dimeric complexes before activation. Ligand binding induces a change that is detected across the membrane (12).

Horizontal receptors are found principally in multicellular organisms. They are generally activated by initial binding of a mono- or multimeric protein ligand; this binding induces oligomerization, or an ordered association of receptors on the cell surface. These receptors generally control longer-term or irreversible changes in cells, such as transcription, translation, replication, or apoptosis.

The three-dimensional structures have been determined for almost a dozen ligands in complex with the extracellular portions of their horizontal class receptors. The vast majority of these involve cytokine-receptor complexes, which can be divided into four major groups (I-IV) on the basis of the preponderance of α-helical, β sheet, mixed α/β, or mosaic substructures within the cytokine component of the complex (13). Receptor oligomerization is a common theme in horizontal signaling, yet there is great diversity in the binding sites, folds, and stoichiometries of these signaling complexes (Fig. 2).

Fig. 2.

The oligomeric forms of cytokine signaling complexes show a rich variety of multimeric horizontal assemblies that are detected and translated into pathways such as kinase cascades inside the cell.

The distinction between horizontal and vertical receptors has been blurred by recent discoveries that a number of single-pass receptors exist in inactive but preassociated states. In some cases, these are poised to bind cytokine, as for members of the tumor necrosis factor (TNF) receptor (TNFR) class (14). In other cases, such as the erythropoietin (EPO) receptor (EPOR) (15, 16), the receptors are associated in an "off-state" that requires them to dissociate before a productive complex can be made. Conversely, there are examples of receptors in the vertical class of seven-transmembrane receptors (1), such as the monocyte chemoattractant protein MCP-1 (17), that appear to form active multimers during the signaling process.

Mechanisms of Horizontal Signaling Complexes

As horizontal receptors are drawn together on one side of the membrane, domains on the other side are oriented to initiate signaling events. In many cases, the intracellular domains of receptors are preassociated with protein kinases, either noncovalently, as is the case with EPORs (18), or by gene fusion, as in the case of epidermal growth factor (EGF)-class of receptors [reviewed in (19)]. In other cases, the intracellular domains associate with kinases after they bind hormone, as is the case for human growth hormone (hGH) receptors (hGHRs) (20). Once the receptors are appropriately oriented, the kinases act intermolecularly to transphosphorylate each other. These in turn act as docking sites for binding and activating other signaling factors.

Cytokine-receptor complexes include those in which two identical receptor molecules are dimerized by binding to two different sites on a single cytokine to produce a 2:1 complex, as seen for hGH (21) and EPO (22). Some cytokines form complexes in which two cytokine molecules bind two identical receptors such as in the gp130:interleukin-6 (IL-6) complex, or the granulocyte colony-stimulating factor and its receptor (G-CSF:G-CSFR) complex (23) and the fibroblast growth factor and its receptor (FGF:FGFR) complexes (24). Still other cytokines bind two different receptors simultaneously, as is seen with IL-4 and interferon-gamma (IFN-γ). Still higher order complexes are seen for the TNFR class.

Group 1, Four-α-Helix Cytokines

hGH and EPO receptors

HGH was the first cytokine-receptor complex for which structure and mechanism were elucidated. HGH is a four-α-helix bundle protein with up-up-down-down topology. The cytokine is characteristic of the long-chain subfamily of the group I cytokines. The extracellular portion of the receptor contains two domains that each have a seven-stranded fibronectin III-like topology [reviewed in (25)]. The domains are termed D1 and D2 (typically, domains are numbered from N to C terminus in direction, and the N terminus is generally farthest from the membrane). hGH binds the first receptor through a higher-affinity site (site 1 on hGH), and binds a second identical receptor through a lower-affinity site (site 2 on hGH) (21) (Fig. 3). The cytokine-binding sites are formed at the outside of the elbow bend formed between D1 and D2 domains, and involve six loops (L1-L6) between β strands from each of the domains, L1 to L3 from D1, and L4 to L6 from D2. The second receptor is rotated by 157° about the normal to the membrane plane relative to the first, as it binds to the opposite side of the cytokine.

Fig. 3.

In this and the next five figures of group 1 "four α-helical" cytokine:receptor complexes, the receptor chains are shown in ribbon rendition in blue colors. Different shades of blue indicate different receptor chains in the complex. The cytokines are cloaked in a molecular surface, shown in greens, above ribbon diagrams in purple colors. Slight variations of color distinguish different cytokine chains. The N- and C-terminal ends of receptor chains are indicated. In general, the C-terminal ends are oriented toward the transmembrane region, although intermediate domains are sometimes interposed between the N terminus and the membrane. Specifically, as illustrated in this figure and Fig. 4, hGH and EPO are long-chain four α-helical bundle cytokines. Both cytokines interact with two identical copies of their receptors using two completely different interaction sites on the cytokine, termed Site 1 and Site 2. Interaction sites on each of the two receptors use many of the same residues.

Mutational and biophysical experiments show that the hGHRs associate sequentially with hGH, binding hGH first at site 1 and then at site 2 on the hormone (26). This second step in the reaction, which occurs in the plane of the membrane, is subsequently stabilized by receptor-receptor contacts, as well as by the hormone-receptor interactions at site 2 [reviewed in (27)].

Alanine-scanning mutagenesis on both sides of the interface with site 1 show that binding affinity is dominated by a small group of mostly hydrophobic side-chain contacts near the center of the contact interface. In the case of the hGHR, these side chains are centered around Trp104 and Trp169 (28, 29). Such "hot spots" seem to be present at many protein-protein interfaces, and they can be used to bind multiple protein partners. Although in many cases hot spots are hydrophobic, there are some that include polar and charged residues as well. It is remarkable that the receptor uses virtually the same set of contact residues to interact with two very different contact surfaces on hGH (site 1 and site 2). Moreover, the receptor-receptor contacts are not homotypic, but involve two different sets of residues from each receptor stem. The receptor achieves this by slight domain movements and larger shifts in the surface side chains involved in binding.

Not only does the hGHR adapt to two different surfaces on hGH, but hGH can interact with multiple receptors. Complexes of hGH with the prolactin (PRL) or placental lactogen receptor conform to the same paradigm (30). Virtually the same set of contact residues in site 1 of hGH mediate interaction with either to the hGHR or PRL receptor. Interestingly, hGH requires Zn2+ to bind the PRLR, but not to bind the hGHR. Although the contact residues are the same for binding either hGHR or PRLR, the residues that constitute the hot spot for binding on the receptor differ (28). This functional and structural plasticity allows cytokines to cross-talk to multiple receptors.

EPO-EPOR complexes

The EPOR complexes provide new insights into the structural requirements for turning signaling on and off. EPO is a four-α-helix group I cytokine with the same characteristic up-up-down-down topology seen in hGH. EPORs are composed of a short N-terminal α-helix, two seven-stranded fibronectin III-like domains (D1 and D2) that are structurally homologous to those in hGHR, a single 22-amino acid transmembrane domain, and a 237-amino acid cytoplasmic domain (31, 32). EPO dimerizes two receptors on the cell surface in a sequential manner; site 1 of EPO binds a first receptor with high affinity (Kd ~ 1 nM), then EPO binds a second receptor through site 2 with lower affinity (Kd ~ 1 μM ) (33, 34).

The structure of the EPO:EPOR2 complex shows that EPO is bound at the outside of the elbow formed by the D1 and D2 domains (22, 35), as in the hGH:hGHR2 complex. However, the second receptor chain in the EPO:EPOR2 complex is rotated by 110° perpendicular to the membrane relative to the first, which is 40° less than that seen in the hGH:hGHR2 complex. The EPO:EPOR2 complex also differs from the hGH:hGHR2 complex in that there are no receptor-receptor contacts. These crystal structures show there is a difference in the way the receptors for group I cytokines can be oriented for signaling to occur (Fig. 4A).

Fig. 4.

Activating and inhibitory complexes of EPO receptors. In these structures, the EPOR molecules were truncated at the C-terminal positions (C) to remove the single-spanning transmembrane segment and the intracellular portions of the receptor. Thus the positions labeled C indicate the approximate separation of the single membrane crossing helices. (A) The EPO:EPOR2 complex shows a 40° difference in orientation (about the vertical axis) of the receptor that binds to Site 2 relative to that seen in the hGH:hGHR2 complex. The Site 1 interactions (right side) produce a similar orientation of the cytokine:receptor interface in both the EPO:EPOR and hGH:hGHR complexes. Thus the interaction at site 2 produces a very different angle of the Site 2-binding receptor around the perpendicular to the membrane surface. The EPO:EPOR2 complex also differs from the hGH:hGHR2 complex in that there are no receptor-receptor contacts between EPORs. (B) The structure of EPOR2 brought about by binding so-called "EPO-mimetic" peptides. Discovered by selection from bacteriophage-displayed peptides, these peptides themselves dimerize to bring receptors close together, although in a completely different manner than that induced by actual EPO binding (as in A). The relative orientation of one receptor chain to the other in this complex is 180° around the vertical axis perpendicular to the membrane. Although the peptide shown (EMP1) activates the EPO response, other similarly discovered peptides can be activating or inactivating for EPOR. (C) The structure of the EPOR2 dimeric complex may represent a means of ensuring that EPOR transmembrane domains are held apart so that EPORs do not come close enough together to become activated by chance. Shown is a view looking down onto the plane of the membrane plane. Thus, the greater distance between the positions of the C-termini (labeled C, top left, bottom right) may be a factor in preventing the association of the transmembrane and intracellular domains inside the cell.

As for the hGHR, virtually the same residues of the EPOR chain are buried on binding at both site 1 and site 2 of EPO, even though the epitopes on EPO to which they bind are entirely different from one another. Binding to the two different sites by the same residues of EPOR is achieved by a small twist of the D1 and D2 domains relative to one another. The side chains involved in binding EPO all derive from the loops L1 to L6. Phe93, the structural counterpart of Trp104 in hGH, has a similarly large effect on affinity when mutated (36). The hot spot on EPOR is remarkable, because it also provides the binding site for synthetic peptides (see below) and the interacting surface for unliganded receptor-receptor dimers that may form an off-state of receptors (15).

Evidence suggests that bringing the two extracellular portions of the EPO receptors close together can activate the intracellular pathway. For example, point mutations that introduce cysteine residues into any of three positions in the EPOR (residue numbers 129, 131, and 132) form disulfide-linked intermolecular dimers between the cysteines introduced into each receptor and the same cysteine on the other receptor and cause constitutive signaling (37-40). These cysteines are arrayed around the D2 domains near the membrane spanning region. Disulfide bonds here would place the receptors about 8 Å closer together than in the EPO complex with EPOR, and rotated by at least 60° from their normal EPO-activated position in the membrane plane. As a second example, a construct of two EPO molecules linked from C-terminal of the first to N-terminal of the second, and both mutated to abrogate the weaker site binding at site 2, leads to activation of the EPOR response, and with almost the same efficiency as for wild-type EPO (41). Third, some but not all antibodies to EPOR can also lead to activation (42). Fourth, several "EPO-mimetic" peptides (EMPs) have been discovered by selection to bind, dimerize, and activate the EPOR (43, 44) (Fig. 4B).

These data, along with the structure of the EPO:EPOR2 complex, suggest that close proximity of receptors is necessary for activation. However, the efficiency of activation is highly dependent on the way in which receptors are dimerized, and on the length of time that they are held together. For example, the median effective dose (ED50) for EPO activation is 10 pM, just 5.6% of its median inhibitory concentration (IC50), implying that only about 60 receptors are dimerized for a 50% response in cells. The ED50 for EMP is typically greater than the IC50, where 600 receptors would be dimerized for a 50% response (22, 35). Some dimerizing peptides, for example, clearly antagonize the response (45, 46), showing that dimerization is not sufficient for activation of the receptor. This data and others showing that covalently dimerized peptides are 1000-fold more active (46) suggest that activation depends on correct orientation and residence time.

Residence time has also been shown to be important for signaling in the hGH receptor. For example, a systematic set of mutations in hGH that decease the residency time at site 1 below a half-life (t1/2) of about 1 min proportionately diminish signaling as detected by Janus kinase 2 (JAK-2) phosphorylation and cell proliferation (47). Residency times longer than t1/2 ~ 1 min do not further enhance signaling, presumably because the signal is maximal. Interestingly, the residency time for the wild-type hormone is about 30 min, which is more than ample for maximal signaling.

The intracellular domain of the EPOR contains eight tyrosine residues that can be phosphorylated after EPO binds. These phosphorylation sites allow the intracellular domains to interact with cytosolic signaling factors that include signal transducer and activator of transcription 5 (STAT-5), the adaptor protein Grb-2, and phosphatidylinositol (PI) kinase (48). Mutagenesis shows that no particular phosphotyrosine is essential for signaling (49). Indeed, a single tyrosine is sufficient to support differentiation of the erythroid lineage (50). This suggests that any of the phosphotyrosine sites can serve as a docking site for STAT-5. Once bound to the phosphorylated EPO receptor, STAT-5 is phosphorylated by JAK-2 protein kinase, which leads to STAT-5 activation through dimerization and transport to the nucleus (51). Thus, regulated oligomerization reactions appear to be a common theme both at the level of the membrane receptor and at the level of downstream effectors.

"Off-state" of receptors

The cell must have a means of distinguishing receptors associated in response to cytokine binding, as opposed to random collisions. Recent x-ray studies suggest that the extracellular domains of the EPOR might dimerize in an inactive antiparallel fashion in the absence of activating ligand (15). In this state, the predicted distance between the membrane insertion points would be about 70 Å, which would prevent the dimerization of intracellular domains. Such a receptor off-state would provide an active means of reducing nonspecific "noise" by preventing ligand-less signaling through random receptor association.

To test this model, a fusion protein was constructed that contained complementary portions of the enzyme dihydrofolate reductase (DHFR) that were inactive when separate but became active when brought together. These portions were fused to the transmembrane segment on the cytoplasmic side of EPOR (16). Addition of EPO was required for DHFR activity. However, when long glycine-rich linkers (lengths of 5, 10, or 30 amino acids) were placed between the EPOR membrane-spanning and DHFR fragments, the receptors could be activated without EPO. This suggests that, without EPO, the intracellular domains of the EPORs are held apart, possibly in the manner seen in the structure of the [EPOR]2 dimeric complex (15) (Fig. 4C).

Dimerization or oligomerization of a "resting state" of EPOR has also been established by immunofluorescence (52). Mutagenesis shows that the correct orientation of L253, I257, and W258, located in the juxtamembrane region of the transmembrane α-helix in EPOR, is required for activation of the receptors. The transmembrane- (52) and ectodomain-associated dimers may represent redundant mechanisms for ensuring that free receptors remain in their off state but available in a cluster. When the EPO ligand binds, it changes the orientation to an active state.

Recently, a peptide was discovered that activates the EPO response but binds to a noncompeting, therefore different, region of the EPOR than does EPO (53). The site is close to the membrane surface on the extracellular portion of EPOR, and appears to bind to EPOR in a 1:1 ratio. Amazingly, the peptide has the same sequence as the sequence in EPOR that it binds to. One plausible function for this peptide is that it may antagonize an EPOR-EPOR (twofold symmetric) inactive dimer. However, its binding site on EPOR differs from that seen in EPOR-EPOR dimers by crystallography, so any interference in that association may be indirect.

Other receptors can exist in preassociated states. Fluorescence techniques have suggested preassociated oligomeric forms for IL-1 receptors (IL-1Rs) (54), IL-2Rs (55), erbB2 receptors (56, 57), and EGF receptors (EGFRs). Crystallographic studies show that TNFR can form parallel and antiparallel dimers (58-60). The antiparallel dimer would block TNF binding, as is seen in the EPOR2 complex.

Peptide mimetics and receptor hot spots

Several investigators have discovered small peptides from phage display libraries that were optimized for binding receptors other than the EPOR; interestingly, they do so at the same hot spots used by the natural ligand. One peptide selected for binding to the extracellular domain of the thrombopoietin receptor (TPOR) was found to have agonist activity (61), and when covalently dimerized, it produced an agonist as potent as TPO itself (62). More potent agonists have been developed for EPOR (63) and G-CSFR by similar routes. In another case, a 19-residue peptide was selected for binding an FGFR. When the peptide was dimerized by fusion to the c-Jun leucine zipper, it activated the FGFR in presence of a synergistic binder, heparin, with a potency similar to that of FGF (64). Thus, small mimetics can be found for the cytokine receptors.

Peptides that compete with the normal ligand have also been discovered by phage display for vascular endothelial receptor (VEGFR) (65, 66), FGFR (67), gp120 (68), and the Fc region of IgGs (69). The fact that all these phage-derived peptides target the natural ligand-binding sites on these receptors suggests that the surface chemistry of these receptor hot spots is somehow primed for binding. Generally, these sites are hydrophobic, well exposed on the surface, and positioned at hinge regions in the structure, allowing adaptability, or conformational promiscuity such that they are the sites most suited to binding ligands (69).

Heterodimeric complexes in IL-4 signaling

IL-4 illustrates how a helical bundle can heterodimerize two receptors. Many cytokines bind two entirely different receptors in their signaling complex. IL-4 is a group I cytokine belonging to the short-chain subfamily. It binds first to the IL-4αR, then to a multipurpose receptor termed the common gamma chain (γC), which is common to several other cytokine complexes, such as those containing IL-2, IL-7, and IL-9 (70). The use of different receptor chains opens the way to broader intersections of different signaling pathways.

There are important differences in the way that IL-4 binds its IL-4αR, compared to the way that hGH and EPO bind their high-affinity receptors. The high-affinity site on IL-4 that binds IL-4αR (Kd = 150 pM) (71) corresponds structurally to the weaker site 2 interface of the hGH or EPO complexes (Fig. 5). The association rate of IL-4 for the IL-4αR shows an unusually high on-rate (that is the rate at which IL-4 associates with IL-4R) that can be ascribed to a positively charged surface that is attracted by the negative surface on IL-4αR (72, 73). Mutational analysis of the IL-4αR shows that it has two independent binding regions. Each region has a single charged side chain, important for binding IL-4, that forms hydrogen-bonded pairs with IL-4; these regions are positioned between Asp72 and Arg88, and between Tyr183 and Glu9, respectively (74). The binding site on the receptor has been further mapped by alanine-scanning mutagenesis to identify and evaluate the contributions of four hydrophobic side chains (Ile100, Leu102, Tyr103, and Leu208) that interact with similarly hydrophobic side chains on the cytokine (75). Interestingly, IL-13, a close homolog of IL-4, binds first to the IL-13αR and then recruits the IL-4αR for signaling. Presumably, IL-13 uses sites similar to those used by IL-4, but in opposite order of binding.

Fig. 5.

IL-4 is also a group 1 four α-helical cytokine, although in the short-chain subclass. The IL-4:IL-4αR structure mimics the Site 2 interface of the hGH or EPO complexes, and is one of the tightest binding cytokine-receptor interfaces, assisted by oppositely charged interfaces.

The 2:2 G-CSF:G-CSFR and IL-6:gp130 complexes

Although G-CSF is structurally similar to EPO and hGH, it activates its receptor through a higher stoichiometry (2:2), and by binding on the "back" side of the receptor. G-CSF is a four α-helical bundle cytokine that first binds one receptor, and then forms a dimeric complex with another 1:1 G-CSF:receptor complex (23). In the 2:2 complex generated by receptor ectodomains that lack their N-terminal Ig domains, the receptors make contact, but there is no interaction between the two G-CSF molecules (Fig. 6). Each G-CSF forms a receptor contact of about 820 Å2 at the site 2 interface (helices A and C) similar to that for IL-4. The interface also involves polar amino acid side chains. Incorporated into the interface are water molecules that are proposed to fill cavities and reduce the penalty of desolvating a charged protein interface. A second smaller interface of about 384 Å2 forms between the normally flexible N-terminal regions of G-CSF in one of the 1:1 complexes, and the D2 domain of the second receptor, in a novel second interface.

Fig. 6.

G-CSF is also a short-chain, group 1 four α-helical cytokine. A dimer of G-CSF binds to two receptors using two identical interactions that are each the mimic of Site 2 interfaces of the hGH or EPO complexes. (A) A side view that shows the lack of contact between G-CSF molecules. (B) A top view (looking down on the membrane) that shows the lack of contact between G-CSF receptors or G-CSF molecules.

The region in G-CSF that corresponds to site 1 in hGH is not involved in initial receptor binding, and it remains exposed in the 2:2 complex. However, alanine-scanning mutagenesis throughout this region suggests that site 1 does play a role in the biological response (76). Higher order complexes up to 4:4 stoichiometry also may form (77), suggesting that higher order complexes may play a role in signaling. It is possible that the vacant site 1 region on G-CSF seen in the 2:2 complex is used in these higher order complexes. The structure of the extracellular portion of the G-CSFR contains additional domains not included in the x-ray structure, including an N-terminal Ig domain that stabilizes the signaling complex (78) and three additional fibronectin III domains that lie between D2 and the transmembrane segment. Thus the exact nature of the active G-CSF:receptor complex still needs to be elucidated.

The structure of the complex formed between IL-6 and the full-length ectodomain of gp130 with the N-terminal Ig-like domain has been determined (79). Two gp130:IL-6 complexes associate so that the N-terminal Ig domains of gp130 also bind to a third site on the opposite IL-6 of the complex to make a twofold symmetric "ring" of gp130:IL-6:gp130:IL-6:. Signaling may depend on bridging of two receptors by the ligand, but differently than for EPO or hGH.

The 2:2:2 IFN-γ receptor complex

The IFN-γ receptor complex is structurally analogous to two hGH-receptor complexes glued together by domain swapping between two ligands. IFN-γ is composed of six helices and forms a V-shaped, interlocked dimer in solution. The fifth and sixth helices of one monomer pack along with helices 1 to 4 of the twofold related monomer to form a "domain-swapped" dimer (80). The cytokine dimer binds with very high affinity (Kd ~ 10-10 M) to two IFN-γRα to form a 2:2 complex (81). This 2:2 complex then binds to two accessory β-receptor chains from the group II cytokine receptor family that are specific to the type of responding cell, activating a 2:2:2 signaling complex (82) (Fig. 7).

Fig. 7.

IFN-γRα is a short-chain group 1 four α-helical cytokine that is intimately dimerized in a domain-swapped form. The dimer recruits a 2:2:2 signaling complex.

IFN-γRα contains two fibronectin III-like domains, D1 and D2, that are related in sequence to those of hGHR or EPOR. In complex with IFN-γ, it utilizes the loops between β strands of D1 and D2, as in the hGH:receptor complex, and buries a surface area of 960 Å2 in the interface, of which one-third is provided by three aromatic side chains on the receptor (Tyr52, Trp210, and Trp85). One notable aspect of the interface is the conservation of charged residues that pack up against tryptophan and tyrosine side chains, where they account for a significant portion of the binding energy. Helices A, B, and F, and the A-B loop form the IFN-γ surface in an interface region that is very different from that of the EPO or hGH complexes. In making the complex, the A-B loop in the cytokine undergoes a conformational change, as is seen for G-CSF (83).

The twofold symmetric IFN-γ dimer evokes a twofold symmetric dimerization of the receptors such that the D1 domains of the receptor are rotated ~40° from corresponding positions in the hGH:hGHR2 complex. The are two identical sites for binding two IFN-γRα molecules in the 2:2 complex. However, this leaves the D2 domains of the IFN-γRα receptor at the membrane entry point, distant from each other.

The structure of the complex of IFN-γ with IFN-γRβ is needed to reveal the atomic details of the active signaling complex. Mutagenesis of IFN-γ suggests that it uses determinants close to the C terminus of the IFN-γ molecule to bind to the IFN-γRβ. A 3:1 IFN-γRα:IFN-γ complex seen crystallographically suggests a binding site for IFN-γRβ. In this complex, the crystallography shows a third IFN-γRα that binds to the spot that is expected to bind IFNRβ, and places the C-terminal regions of the two chains close together on the cytoplasmic side of the membrane (84). This notion is supported by an experiment in which IFN-γ was made into a continuous single-chain dimer by gene fusion, and then mutated to remove just one of the binding sites for IFN-γRα. This 2:2:2 complex activates both JAK-1 and JAK-2 protein kinases. This construct shows that the 1(IFN-γ dimer):1:1 complex can still evoke a cellular response achieved through single copies of the receptor-bound JAK-1 and JAK-2. This transient IFN-γ (dimer):IFN-γRα:IFN-γRβ complex, by analogy with the IL-4:IL-13αR:IL-4αR complex, indicates that the cross-phosphorylation steps are very efficiently carried out, presumably by just one side of the complex. Ligand binding releases receptors from any restraint such as that brought about by inactive oligomeric states of the receptors (85).

The 2:2:2 IL-10:IL-10R1:IL-10R2 complex

Structurally, the IL-10:receptor complex is closely analogous to the IFN-γ complex. IL-10 is a natural immune suppressant, and so is hijacked by viral mimetics of IL-10 to anaesthetize the immune system. Interestingly, the ways that the viral mimetics activate IL-10 so as to enhance viral infectivity are completely different in each case, and produce different spectrums of downstream signaling (86).

IL-10 is a six-helix cytokine that forms a twofold symmetric, intertwined dimer in which four helices (A-D) from one monomer conjugate with two (E and F) from the other to form a 12-helix twofold symmetric bundle. IL-10 and IFN-γ contain several conserved residues in their dimeric cores, which suggests a possible "fingerprint" for detection of other members of this dimerized family (87, 88). The IL-10 dimer first forms a twofold symmetric dimer with high affinity (Kd ~ 8 nM) for IL-10R1 (Fig. 8), and then recruits the second, low-affinity site on IL-10R2 that activates the signaling pathway mediated by JAK1 or TYK2 protein kinases and STATs.

Fig. 8.

IL-10 is a long-chain group 1 four α-helical cytokine that first recruits IL-10R1 and then recruits the second, low-affinity site on IL-10R2 that leads to activation.

There are viral homologs of IL-10, for example in cytomegalovirus (CMV) and Epstein-Barr virus (EBV). Their structurally distinct scheme is paralleled by biological differences that highlight receptor orientation in the activating complex (86). EBV, though it presents a homolog of IL-10, is unable to stimulate thymocyte production or mast cell proliferation, for example, whereas human IL-10 (hIL-10) can. Unlike hIL-10, in CMV IL-10 the dimer is crosslinked by a cysteine bridge. Complexes of CMV IL-10 with IL-10R1s show that the CMV IL-10 and human IL-10 have a 40° difference (viewed parallel to the membrane plane) in their intramolecular angle as a dimer (86). The structures of 2:2:2 activating complexes have not yet been determined. But the markedly different activation spectra emphasize the evolutionary tendency to raise the most appropriate of pleiotropic responses, and they show that aspects of the orientation of IL10-R1/R2 receptors after cytokine dimer binding are key to effector signaling. Differential activation of various signaling pathways is presumably due to different complexes formed on the cell surface; the key is that each complex is distinct enough in geometry and residence time that it induces varied downstream signaling.

Group 2, β Sheet-Rich Cytokines

TNF:receptor complexes and the formation of higher order oligomers

TNF and its family members step up the degree of oligomerization to trimeric structures, where the ligand serves as the bridging scaffold, with no receptor-receptor contacts. TNF belongs to the group 2 (β-sheet rich) cytokines and is implicated in many inflammatory processes and apoptosis (89). TNFα is a 17-kD "β-jellyroll" that forms a wedge-shaped homotrimer (90, 91). Signaling is mediated by either of two TNF receptors, TNFR1 (55 kD) and TNFR2 (75 kD). The structure of the extracellular portion of the TNFR1 has been solved by crystallography and shows that it is composed of four cysteine-rich domains. The extracellular domains of the TNFR1 crystallize as a dimer [reviewed in (92)].

Upon binding the TNF trimer, the receptor trimerizes in an active configuration (92). TNFR-associated factors (TRAFs) are a family of cytoplasmic proteins that assist in transducing signals for the TNFR superfamily inside the cell (93). These proteins are also trimeric structures in isolation (94). Thus, the TNF trimer itself dictates the assembly of the TNF:TNFR trimer (59) (Fig. 9A) that is detected by a "trimer detection" system in the TRAFs.

Fig. 9.

Group 2 "β-sheet rich" cytokines. (A) TNF is in the β-jellyroll subclass. A TNF trimer recruits three TNFRs to activate the response. (B) Several forms of the dimeric receptor complex show both a parallel (left) and an antiparallel association (right). The parallel form of the receptor dimer may promote clustering of TNF:TNFR complexes on the cell surface. (C) The antiparallel mode may be the means of maintaining receptors in their off-state.

Several forms of the dimeric receptor complex have been determined. At pH 7.5, the elongated receptor chains dimerize in both a parallel manner and an antiparallel association. In the antiparallel dimer, the TNF-binding interfaces are buried, suggesting that this may be the means of maintaining receptors in their off-state (58), which is reminiscent of the case with EPORs (Fig. 9B). In the parallel off-state form of the receptor dimer, the TNF binding site is exposed and open to bind the cytokine, suggesting that this mode may promote clustering of TNF:TNFR complexes on the cell surface.

The activity of TNF receptors is regulated in several ways; this is essential because overexpression of the intracellular domains alone leads to activation, mediated by an intrinsic aggregation mechanism (95). First, the extracellular concentration of TNF is reduced by the presence of the cleaved extracellular domain of TNFR in plasma. Second, the TNFRs themselves aggregate in at least two ways, as described above. Third, after TNFRs are activated on the cell surface, the complexes are endocytosed and the signal is truncated. The structure of TNFR dimers at pH 3.5 shows that they form dimers that share a large area of contact (2900Å2), and compete directly for the TNF binding surface, suggesting that the low-pH endosome can terminate signaling by this "competitive" dimerization mechanism (96). Fourth, silencer of death domains (SODD) associate with the cytoplasmic "death domains" of TNFR1s. Binding of TNF releases SODD from TNFR1, permitting recruitment of proteins such as TNFR-associated death domain (TRADD) and TRAF2 to the active TNFR1 signaling complex. SODD association may exemplify a general mechanism for preventing spontaneous signaling by death domain-containing receptors (95).

The receptors for FGF, nerve growth factor (NGF), VEGF, insulin, and EGF include intrinsic tyrosine kinase domains on the cytoplasmic side of their receptor chains.

FGF receptor and the role of extracellular matrix

FGF illustrates how a polysaccharide can facilitate binding between ligand and receptor. Crystal structures of the FGF:receptor complexes show that heparan sulfate chains in the extracellular matrix enhance the binding between FGF and its receptor. This synergy between carbohydrate and protein has important roles in development and organ homeostasis. The FGF growth factor family consists of 23 cytokine isotypes. There are four receptor isotypes, each with a different specificity for the FGFs. Five FGFs have both a hydrophobic receptor-binding epitope and a somewhat variable heparin-binding surface that localizes the FGF to the extracellular matrix (97). The receptors are composed of two or three ~110 amino acid IgG-like domains (D1-D3), each containing seven β strands. The D2 domain binds heparin with high affinity (98).

FGF2 forms a twofold symmetric 2:2 complex with the receptor FGFR1 (24, 99), as is also found for a FGF1:FGFR2 complex (100). Each monomeric FGF cytokine binds the D2 and D3 domains of one FGFR, which induces a change in the FGFR that then leads to (FGF:FGFR):(FGF:FGFR) dimerization mediated primarily by the D2 domains within the FGFRs. The dimer interaction area between receptors is surprisingly small, about 300 Å2, and the FGF molecules have no contact with each other in the complex. In contrast, the FGF:FGFR interaction surface is large (2700 Å2). The cytokine's contacts with the D2 domain of the receptor are primarily hydrophobic, but also have important polar interactions. For example, Arg250 in the D2-D3 linker is oriented by neighboring residues to make two conserved hydrogen bonds to the cytokine. The preponderance of central hydrophobic interactions, along with key polar interactions to encode specificity, are reminiscent of the EPO and hGH complexes with their receptors.

Dimerization of the receptor also depends on the heparin of extracellular matrix. An optimal concentration exists for heparin to assist in cross-linking receptors (92). Higher than saturating concentrations of heparin compete for all the sites and antagonize crosslinkage. Thus, heparin can either promote or prevent the proliferation and migration of cells involved in vascular development.

The structure of the initial 2:2 complex between FGF and FGFR shows a basic canyon that extends from FGF2, across the D2 dimer, and across the D2-D3 linker of the FGFR, that could accommodate polyanionic heparin and stabilize the dimer (24). A crystal structure of the ternary complex of FGF:FGFR:FGFR:FGF with a decasaccharide bound shows how an activating heparin analog actually binds to the 2:2 complex (101). Surprisingly, two decasaccharides bind predominantly to the FGF molecules (perhaps reflecting the FGF association with the matrix before activation), and increase the affinity of each FGF for its receptor in 1:1:1 FGF:FGFR:heparin complexes (Fig. 10A). Each decasaccharide adopts a helical structure like that seen by fiber diffraction of heparin alone. The helix is generated by repeating D-glucosamine (GlcN)-L-iduronic acid (IdoA) disaccharide units, each one linked to the next by a α 1-4-glycosidic bond. Each disaccharide unit is sulfated at three positions, and the sulfate and carboxylate groups form the negatively charged edges of the heparin helix that appear every 17 to 19 Å on each side of the helix. Heparin polysaccharides have a nonreducing end (O4) and a reducing end (O1). The decasaccharides bind with their nonreducing ends toward the center of canyon, preserving the overall twofold symmetry of the complex. These are each bound by 30 hydrogen bonds. Of these, 25 are found within one FGF:FGFR pair (16 from EGF, nine from EGFR), and five more help bind the neighboring FGFR, augmenting the cross-linking and hence stability of the quaternary 2:2:2 complex. These include salt bridges from positive charges of five lysines on each D2 of the FGFRs to sulfates of heparin.

Fig. 10.

Other examples of group 2 "β-sheet rich" class cytokines. (A) FGF is in the β-trefoil subclass. The 1:1:1 FGF-FGFR-heparin complex illustrates how intracellular matrix contributes to the activation of the FGFR complexes. Heparin is rendered in stick model colored red. This is a side view; (B) is a top view onto the membrane.

The structure of the 2:2:2 FGF:FGFR:heparin complex shows how heparin draws together the protein components and increases affinity of FGF for its receptor (101). It explains how smaller oligosaccharides, including hexasaccharides, can bind and dimerize FGFRs in spite of not being able to reach from one FGF:FGFR complex to the other, and do so even while lacking the five-hydrogen bond saccharide-FGFR contact. It also can explain why excess heparin can antagonize the receptors through competition for the secondary five-hydrogen bonded site on the "other" receptor. If this model is correct, the optimal ratio of oligosaccharide to receptor would be 1:1. Another structure of the complex formed with a bound decasaccharide confirms a role for the heparin molecules in drawing the complex together (102). The conformational flexibility of IdoA, a crucial saccharide that becomes sulfated and is part of the ligand, must play a role in specific recognition of various FGFs or FGFRs that are variable in sequence in this portion of the heparin site. The model suggests that heparin serves to bind in the interface between FGF and FGFR in a 1:1:1 complex to bring about a conformational change that leads to productive dimerization of FGFR, and hence to intracellular autophosphorylation as the first step in the signaling cascade (103) (Fig. 10A).

Another model proposes a preexisting inactive complex of two FGFRs. When FGF binds to this complex. it alters the conformation of the preformed, inactive dimer to initiate signaling (97, 104, 105). Heparin, through its positional determination in the intracellular matrix, ensures fibroblast growth once such cells are in close contact with matching neighboring cells.

VEGF: a simpler symmetric paradigm

VEGF represents a simpler symmetrical complex where binding occurs at subunit interfaces. VEGF is a member of the "cystine knot" subclass of group 2 cytokines that all act as homodimers. Three subfamilies are recognized that are distinguished by their different ways of forming homodimers. VEGF stimulates growth of endothelial cells (106). It is usually attached to proteoglycans in the intracellular matrix by heparan sulfate, and it binds to two different tyrosine kinase-containing receptors, Flt-1 and KDR, to effect signaling. These receptors are in the PDGF receptor family, and are composed of five or seven Ig-like domains. In the VEGF-Flt-1 complex, only Ig-like domains 2 and 3 provide the binding interface from the receptor side. Based on sequence analysis, the platelet-derived growth factor receptor (PDGFR) family of receptors may all use domains 2 and 3 in a similar way, with domain 4 adding stability by receptor-receptor contacts. The interface from the hormone is rich in leucine and isoleucine residues, which constitute nine of 47 interface residues (107) (Fig. 11).

Fig. 11.

VEGF is in the cystine knot subclass of group 2 cytokines. The 1:1 VEGF-Flt-1 complex is shown. Another receptor chain, KDR, is required for activation.

The NGF-TrkA complex suggests convergent evolution in group 2 cytokines

Like VEGF, NGF is in the dimeric "cystine knot" subgroup of the group 2 cytokine family. It is bound by two receptors of different types, TrkA and p75. Like the VEGFR chains, TrkA is composed of multiple Ig-like domains (five in total), although only the one closest to the membrane is required for binding NGF. Binding occurs with little conformation change (66).

p75 is a member of the TNFR class, and it promotes apoptosis when NGF binds in the absence of TrkA. Despite the many similarities to the VEGF:receptor complexes that reflect underlying homology, the NGF ligands dimerize in quite different ways, and the binding domains of the NGF versus TNF receptors are at almost opposite poles of the receptor molecules (Fig. 12), suggesting convergent evolution of ligand induced dimerization, but mediated by a different domain in the receptor, and in a different way. In the NGFR, the complex brings membrane-proximal domains of the receptor chains within 20 Å to effect signaling, whereas in the VEGFR, several domains of undetermined function intervene between ligand-binding domains and the membrane.

Fig. 12.

NGF-TrkA uses just the membrane-proximal Ig-like domain and must bind a second receptor p75 to activate.

BMP-2, TGF-β, and their serine-threonine kinase receptors

The cystine knot hormones form hetero-oligomeric receptor complexes with receptors that have unique folds. This third subfamily of the group 2 β-sheet dimeric cystine knot cytokines also includes bone morphogenic growth factors (BMPs) and tissue growth factors (TGFs) that control proliferation and differentiation of bone and connective tissue vis-á-vis their position. These are dimeric proteins where each monomer is based on a nine-stranded β sheet and one α helix. These cytokines act as homodimers, and recruit two pairs of heterodimeric receptors to form a 2:2:2 signaling complex. The BMP-2 in 2:2 complex with only one of its receptor chains, BRIA (108) (Fig. 13), and the transforming growth factor TGFβ3 in complex with TGFβR-IIec (109), illustrate these quaternary interactions.

Fig. 13.

BMP-2 in 2:2 complex with one of its receptor chains, BRIA, labeled "Type I". (A) A top view; (B) a side view. The signaling complex is a 2:2:2 association with two different receptors.

Unlike the cytokine receptors, these receptors have binding domains resembling neurotoxins that bind with high affinity to the acetylcholine receptor (110-113) and are like domains found in the low-density lipoprotein (LDL) receptor (114). Each domain is composed of six extended strands that form a three-finger-like motif stabilized by six cystine bridges in a knot at one end of the molecule. The cystine bridges hold the extensively β-sheet structures together to present a large binding surface in the form of a groove that is about 15 Å by 5 Å by 5 Å in dimension. Although the BMP-2 ligand is similar to VEGF and NGF, the neurotoxin-like binding domains in the receptor are completely different from the Ig-like folds in the VEGF and NGF receptor family. Moreover, the internal domains are serine-threonine kinases rather than tyrosine kinases.

BMPs are multifunctional signaling molecules. Dimeric BMP-2, for example, signals through two different pathways. A small proportion of BMP receptors (BMPRs) rest in preformed inactive hetero-complexes containing one BMP type II receptor (BRII) and one of the two BMP type I receptors, either BRIA or BRIB (115). The signals induced by binding of BMP-2 to these preformed receptor complexes activate the SMAD pathway. By a separate mechanism, BMP-2 dimers can bind first to BRI, and induce recruitment of BRII receptors that then activate a different, SMAD-independent pathway, resulting in the induction of alkaline phosphatase activity through p38 mitogen-activated protein kinase (MAPK). Thus, initial binding of BMP-2 to receptor heterocomplexes can activate different signaling pathways than does binding first to one of the monomeric BRI proteins, which have greater affinity for the ligand (116).

In the 2:2 BMP-2:BRIA portion of the whole complex, the dimeric ligand binds receptors such that the receptors are not in contact with each other, although each receptor chain contacts both monomers of the dimeric ligand. A principal element in the BMP-2:BRIA interface is a highly conserved phenylalanine ring (Phe85 in type 1 receptors) that emanates from the BRIA receptor site and is bound in a remarkably hydrophobic cavity between the monomers of the dimeric ligand. This cavity is invariant or highly conserved across the entire TGF-β family. This suggests that this class of ligands should be a tractable target for drug design that takes advantage of the highly structured pocket for affinity and builds selectivity from polar contacts in the neighboring regions.

Bone growth is highly regulated, and antagonists of BMPs are essential to control development and dissolving of bone, for example, between the digits in the developing hands or feet of mammals. The first three-dimensional structures of an antagonist, Noggin, when bound to BMP-7, shows that these negative regulators bind to the BMP cytokine and occupy both of the receptor-binding sites on BMPs. The inhibitory Noggin molecule has a topography that is surprisingly similar to that of BMPs themselves, suggesting that both ligand and antagonist may have evolved from a common ancestral gene, but later acquired opposite effects on regulation of bone growth in development.

Group 3, Small α/β Cytokines

EGF receptor complexes

The epidermal growth factor family of receptors includes EGFR, Erb2, Erb3, and Erb4. These receptors are activated by EGF, TGFα, or some 10 other ligands according to their location. These receptors are important in tissue remodeling, and mutations in the receptors can underlie several cancers (117). Dimerization of EGFRs occurs upon binding two EGF molecules in a 2:2 complex. Two structures of these complexes (118, 119) show a back-to-back association of the EGFR:EGF 1:1 complexes in which conformational change in the receptors leads to their association (Fig. 14). The receptors are inherently three or four domain structures, in which domains I and III (the thumb and fingers) clasp the EGF molecule, and the structural change relayed through domain II of each receptor molecule, at the backs of the "hands," leads to their dimerization. This process depends on a dimerization loop of 17 amino acids at the back of each of the domain II.

Fig. 14.

Group 3 "small α/β" cytokines. Dimerization of EGFRs occurs on binding two EGF molecules to form a 2:2 complex in a back-to-back association of the EGFR-EGF 1:1 complexes in which conformational change in the receptors leads to their association.

The structure of the extracellular domain of ErbB3 (120) suggests that deletion of domain IV of ErbB3 releases a kind of auto-inhibition imposed by interactions between domain IV and the dimerizing domains II, as if this intramolecular interaction competes with the dimerizing intermolecular domain II-domain II interactions.

Deletion of domain IV is accompanied by a reduction in binding of EGF, supporting a reciprocal allosteric feedback in which lack of "dimerization-mimetic" binding of domain IV against domain II lowers the affinity for EGF between domains I and III. Intracellularly, intermolecular phosphorylation of particular tyrosines in the activation loops of the protein tyrosine kinase (PTK) domains opens them up to bind adenosine triphosphate (ATP), whereupon they become active as tyrosine kinases.

The insulin receptor complex

The insulin receptor and its related insulin-like growth factor (IGF-I and IGF-II) receptors are unusual in presenting covalently preassembled α2β2 oligomers that are inactive. Signaling is mediated by an insulin-induced conformational change within the pre-formed complex of two α subunits that provide for binding of insulin, and two β subunits that span the membrane and include intracellular tyrosine kinase domains. The two α chains are disulfide cross-linked together, and each α chain is in turn disulfide-linked to a β chain. The insulin receptor is an example of a horizontal receptor that has evolved vertical receptor characteristics; it is preassociated and, for the most part, regulates reversible changes in cell metabolism. The three-dimensional (3D) profile structure of the complete covalent heterodimeric insulin receptor complex in the membrane, bound to a single insulin, was visualized to limited resolution by electron cryomicroscopy (121, 122). The 3D profile was then fitted with available known high-resolution domain substructures to deduce the structure of the ligand-bound, activated receptor complex. The structure of a fragment of the homologous IGF-IR that includes the two homologous L1 and L2 domains and the intervening cysteine-rich domain, comprising about 450 amino acids of the ectodomain of the α chain (123, 124), and the imaging analysis of Fab fragments bound to specific regions of the receptor (125), provide a sound basis for assessing and improving this model.

The insulin family is based on cysteine-rich, class III, α/β ligands. α and β chains of the receptor complex bind insulin, and the β chains each carry a single membrane-crossing sequence and an intracellular kinase domain. These cytoplasmic domains have intrinsic activity that can transfer the γ-phosphate of ATP to tyrosine residues in the neighboring domain, once brought together by the insulin-induced conformation change (121). The binding of insulin by the entire α2β2 complex is negatively cooperative, with one high-affinity site, and a second of low affinity (126, 127). Binding to a single αβ monomer occurs only with low affinity and without cooperativity, implying that the four-chain complex is required to present the high-affinity site.

Peptides selected for binding to the insulin receptor, optimized by phage display, bound to any of three different sites on the receptor (128). All competed with insulin, with Kd values in the high nanomolar to low micromolar range. Some peptides that bound to site 1 alone activated the tyrosine kinase activity of the receptor, suggesting that site 1 corresponds to a region involved in insulin-induced activation of the insulin receptor. Peptides that bound to site 2 or site 3 acted as antagonists in phosphorylation assays and assays of glucose incorporation into lipids in adiopocytes.

The structure of both the inactive intracellular insulin receptor tyrosine kinase domains (129) and the activated phosphorylated forms, resolved to 1.9 Å (130, 131), define the changes that activate the kinase. In the active state structure, the kinase domains are dimerized in a well-ordered structure and bound to a peptide substrate and an ATP analog. This dimer shows how the activation loop (the A-loop) of the kinase undergoes a major conformational change upon autophosphorylation. Three tyrosines within the loop are phosphorylated: Tyr1158, Tyr1162 and Tyr1163. The resulting structural change then pulls the tyrosine residues away from the active site and so allows unrestricted access of the kinase active site to peptide substrates and ATP. Phosphorylated Tyr1163 (pTyr1163) plays the central role in determining the conformation of the tri-phosphorylated A-loop. pTyr1158 is then fully exposed to solvent, suggesting a possible role in interacting with other proteins downstream in the signaling process. The structure of the IGF-IR kinase domain activated by phosphorylation shows that the degree of phosphorylation controls the degree of activation in successive stages (132, 133).

In the insulin receptor kinase, the peptide substrate forms an antiparallel β strand with the C-terminal portion of the activating A-loop. The two methionine side chains in the YMXM substrate sequence are bound in two hydrophobic pockets on the C-terminal lobe of the kinase. The phosphotyrosine sites either enhance catalytic efficiency of the kinase domain or provide a target for recruitment of other signaling proteins downstream in the signaling pathway. The structure thus reveals the molecular basis for autophosphorylation of the insulin kinase and more generally suggests the basis for target specificity in the tyrosine kinases as well as the mechanism of phosphotransfer from ATP.

Receptor oligomerization in the family

Some of the earliest biochemical evidence for the importance of receptor clustering came from studies of the EGFR family (134). The receptors for erbB2 (HER2, Neu) are believed to form both homodimers and heterodimers with the EGF and heregulin receptors (erbB1 or EGFR; and erbB3-4). A tremendous amount of biochemical and cell biological evidence indicates that oligomerization is critical for signaling; however, because of their size and complexity, it has been difficult to study these receptors structurally and thereby discover their mode of oligomerization.

The structure of the entire HER3 EGFR (ERB3) (128) shows four domains structurally homologous to those found in the type I insulin-like growth factor receptor (IGFR). Contacts between domains II and IV constrain the orientations of ligand-binding domains and provide a structural basis for understanding both multiple-affinity forms of EGFRs and conformational changes induced in the receptor by ligand binding during signaling.

"Inside-out" receptors in the "RGD-activated" integrin families as spring-loaded heterodimers

αβ heterodimeric integrins provide connections between cells and between cells and the extracellular matrix. These proteins signal bidirectionally and are coupled to intracellular signaling pathways through activation of tyrosine kinases and through connections to the cytoskeleton. The extracellular domains of the integrins bind an RGD sequence in the ligand, configured as a β turn (135). The integrins are αβ heterodimers composed of two single-transmembrane receptors in which the α chain is typically composed of four or five domains. The β domain is composed of about five unique domains, followed by about five EGF-like domains. Both chains end with a transmembrane helix and a short intracellular platform for assembly of intracellular factors. These termini are seemingly sufficient to hold the α and β chains together (136). In bidirectional signaling, they relay information from the extracellular environment of the cell to intracellular pathways (137), and in the reverse direction, intracellular connections to the submembranous cytoskeleton modulate external binding to the extracellular matrix.

Association of dimers in the membrane plane is sufficient to activate intracellular signaling by integrins (138). Other kinase-associated membrane proteins, called syndecans and tetraspans, sometimes augment the activation effect (139). The crystal structure of the αVβ3 integrin ectodomain (140) is switched back on itself and is presumed to be the inactive conformation. Upon binding the RGD sequences as would be presented by a ligand, the integrins also undergo a large, spring-loaded extension of several external domains (141) that transform them from a low-affinity state to a high-affinity state for ligand binding.

At the hairpin bend of the resting-state integrin β chain is a domain that in its structure resembles the α subunit of G proteins. This domain is tightly bound to a "seven-bladed propeller" domain on the integrin α-chain that in its structure resembles a G protein β subunit (Fig. 15). These two domains each bind fibronectin (142). In structural terms, this extracellular interaction mimics the interaction between intracellular α and β subunits of G proteins. In the integrin case, the G protein homolog is not a GTPase; instead, the site corresponding to the GTP-binding site provides the invariant binding site at which the RGD sequence of the integrin ligand interacts with the integrin β chain. Binding of the RGD sequence is mediated by a Mg2+ metal ion-binding site (a MIDAS domain). Dissociation of the Gβ-like and Gα-like domains triggers the response. In a curious way, this pairing of domains that resemble intracellular mates that transmit intracellular signals from the vertical receptors--but in the case of integrins occur on the outside of the cell--reflects the bidirectional nature of integrin signaling. The binding of GTP normally dissociates Gα proteins from the propeller domain in Gβ. One of several proposals is that binding of the RGD ligand may dissociate the integrin α and β chains, leading to the "switchblade" activation of the β chain (143).

Fig. 15.

Integrin signaling involves a large change in conformation upon activation, which activates binding to the intracellular matrix in a mechanism of bidirectional signaling. (A) The resting state complex. (B) A model--made to match electron images--that indicates the extent and type of change involved when activation is triggered to generate the binding site.

Table 1.

Comparison of "vertical" and "horizontal" signaling paradigms.


The structures of cytokines and their receptors reveal a eukaryotic system of intercellular communication that is increasingly in focus, as revealed by the growing number of signaling paradigms seen in three-dimensional structures of both the functional and the off-state receptor complexes. Central principles are that there is control over lateral associations of receptors in the plane of the membrane that can activate inside the cell cascades of intracellular enzymes that are predominantly protein kinases. In general, the functionally important surface areas of the binding epitopes on cytokines span the range of 700 to 1200 Å2.

The diversity of mechanisms relies on three principles. First, many receptors are maintained in some ordered off-state in which their transmembrane domains are held apart. Second, binding of ligands leads to activation by changing the way homo- or heterodimers of receptors are formed as partners on the outside surface of the cell. Third, these structural changes in proximity and orientation of the outside domains are detected by associated intracellular kinases whose actions are determined by their proximity and orientation.

The diverse interactions that can result from different combinations of ligands and co-receptors allows integration of multiple incoming signals at the level of receptor activation. Uncovering the keys to this diversity in protein-protein interactions remains one of the most exciting challenges in modern biology, and it promises a rich field for intervention in human disease by targeting of receptor-protein interactions.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
View Abstract

Navigate This Article