Coordinating Speed and Amplitude in G-Protein Signaling

Abstract

G-protein-mediated signaling is intrinsically kinetic. Signal output at steady state is a balance of the rates of GTP binding, which causes activation, and of GTP hydrolysis, which terminates activation. This GTPase catalytic cycle is regulated by receptors, which accelerate GTP binding, and GTPase-activating proteins (GAPs), which accelerate hydrolysis. Receptors and GAPs similarly control the rates of signal initiation and termination. To allow independent control of signal amplitude and of the rates of turning the signal on and off, the activities of receptors and GAPs must be coordinated. Here, the principles of such coordination and the mechanisms by which it is achieved are discussed.

Introduction

Heterotrimeric G proteins function as GTP-gated conformational switches in all eukaryotic cells. G proteins are themselves molecular timers, and they convey exquisitely timed signals in diverse regulatory circuits. Timing — the kinetics of the on and off reactions of G proteins and their signaling partners — is central to understanding how G proteins convey information, how the information is amplified and how the information is sorted before being conveyed to cellular effector proteins.

G proteins convey information by transiting a cycle of GTP binding, hydrolysis of bound GTP, and GDP dissociation (Figure 1). Heterotrimeric G proteins are activated by GTP binding to their α subunits. Although binding is reversible, GTP dissociation is slow, and deactivation takes place when bound GTP is hydrolyzed by the G protein. GDP release is also slow and additionally delays reactivation. Each of these events is tightly regulated (Figure 1). GDP release and GTP binding are accelerated by guanine nucleotide exchange factors (GEFs). The GEFs for heterotrimeric G proteins are G-protein-coupled receptors (GPCRs), a large family of cell-surface proteins that respond to a huge variety of extracellular agonists, such as hormones, neurotransmitters, odorants, and light. Intracellular GEFs for trimeric G proteins have also been identified recently and are implicated in specific regulatory functions [1,2]. GDP binding is stabilized by proteins known as GDP dissociation inhibitors (GDIs). The principal GDI for heterotrimeric G proteins is the stable Gβγ subunit complex, but Gβγ subunits are additionally able to convey signals to multiple effectors [3]. At rest, the G-protein subunits form a Gα-GDP–Gβγ trimer. Upon activation by exchange of GTP for GDP, the structure of the complex changes such that Gα-GTP and Gβγ can regulate effectors, and Gβγ may dissociate completely.

Figure 1. G protein switches follow a monocycle composed of GTP binding, GTP hydrolysis and GDP release.

The GTP-bound state (G*-GTP) is defined as ‘active’, but G-GDP may also have its distinct regulatory and protein-binding activities. The fractional activation of G protein is described simply by the k_on and k_off rate constants, as described in the text, even though they may summarize several intermediary reactions. Each step in the GTPase cycle is unusually slow for a typical enzyme and is subject to multiple kinetic controls. Guanine nucleotide exchange factors (GEFs) accelerate GDP dissociation and, in the case of heterotrimeric G proteins, GTP binding. GEFs for heterotrimeric G proteins are cell-surface transmembrane receptors; GEFs for monomeric G proteins are heterogeneous. GTPase-activating proteins (GAPs) accelerate hydrolysis of bound GTP, which is rapidly followed by dissociation of orthophosphate. Some GAPs are also G-protein-regulated effectors. GDP dissociation inhibitors (GDIs) stabilize GDP binding and thus inhibit activation. The Gβγ subunits are GDIs, among their many other functions. ‘Effector’ refers to any protein that is regulated by a G protein, whether by the GTP-binding Gα subunit or by Gβγ. The nucleotide-free G protein, an obvious intermediate in the exchange reaction, is not shown because it is calculated to have a very short half-life at cytosolic GTP concentrations (~2 ms for Gα_q).

Hydrolysis of bound GTP is also slow. The half-life of GTP bound to the fastest trimeric G protein is about 9 seconds, and hydrolysis by the slowest can take minutes. Hydrolysis of bound GTP, and consequent deactivation, are accelerated by GTPase-activating proteins (GAPs), which can increase the hydrolysis rate by more than 10³-fold. GAPs for heterotrimeric G proteins include regulator of G-protein signaling (RGS) proteins and phospholipase Cβ (PLCβ) isoforms [4]. GAPs act allosterically by stabilizing the catalytic active site on Gα in a conformation that is favorable for hydrolysis [5]. Paradoxically, PLCβ isoforms and the p115 family of Rho GEFs (which contain a RGS domain) are both G-protein-regulated effectors as well as GAPs for the G proteins that activate them (G_q and G₁₃, respectively).

Heterotrimeric G proteins repetitively transit the GTPase cycle during stimulation by receptor. Signal output is proportional to the number of G proteins activated and the relative period of time that they spend in the GTP-bound active state. In this paradigm, the system operates at an adjustable steady state in which fractional activation for each G protein is the balance of the rates of GEF-promoted activation and GAP-promoted deactivation. Fractional activation is defined as k_on/(k_on + k_off). Signal amplitude thus quantitatively reflects the relative activities of the relevant GEFs and GAPs. However, it is not necessarily related to how fast a G protein transits the GTPase cycle, only to the fraction of each cycle during which it is active.

Formally, GEFs and GAPs can act catalytically. A single GEF molecule catalyzes GDP–GTP exchange on multiple G-protein molecules, and a GAP molecule can similarly de-activate multiple G proteins. Some GPCRs can catalyze the activation of multiple G-protein molecules during the lifetime of a single G protein’s GTP-bound active state. Thus, at steady state, one receptor can maintain the activation of many G-protein molecules and thus amplify the signal many fold on a molecular basis.

Speed versus Amplitude — What’s Wrong with This Picture?

There is an implicit problem with the simple monocycle shown in Figure 1. As written, it implies a strict linkage between the steady-state signal amplitude elicited by a stimulus and the rates of signal onset and termination when the stimulus is added or removed. Specifically, fractional G-protein activation at steady state is simply k_on/(k_on + k_off), and the rate to achieve a new steady state has the rate constant k_app = k_on + k_off. These equations make intuitive sense for the activation process: an increase in k_on caused by receptor-promoted GTP binding increases fractional signal output with an appropriately fast response time. However, when the stimulus is removed and fractional activity drops, the return to the resting steady state is predicted to be slow because now both k_on and k_off are low. The simplest statement of this problem is represented by the effects of turning a light switch on and off. When the light comes on, we perceive it in a few milliseconds because rhodopsin rapidly activates transducin (G_t), the visual G protein. However, the intrinsic rate of hydrolysis of G_t-bound GTP is slow. If this rate were the only determinant of the deactivation rate, it would take us several seconds to perceive darkness when the light is switched off. The need for a GAP to increase k_off and thus give physiologically fast deactivation when the stimulus is removed was first pointed out by Breitwieser and Szabo 20 years ago [6], and now appears obvious. However, the increase in k_off caused by a GAP might also be expected to inhibit signaling substantially even while stimulus is present. There lies the problem.

GAPs do physiologically inhibit G-protein signaling in many cells, and diminished signaling is the most common response to overexpression of exogenous GAP proteins. A prototypical GAP, Sst2p in the yeast Saccharomyces cerevisiae, is a simple feedback inhibitor whose induction terminates the G-protein-mediated mating response [7]. GAPs can also accelerate signal termination without causing substantial inhibition, however (Figure 2). Whether a GAP inhibits signaling or accelerates signal termination, or both, can be regulated simply by the amount of GAP that is present [8]. GAPs can even preferentially modulate signaling through one receptor while having relatively little effect on signaling through a different receptor that uses the same G-protein transducer [9–11]. GAPs apparently do operate in most mammalian cells because the termination rates for multiple signals are faster than predicted by the hydrolysis rate of the isolated G protein that mediates the response. How is this versatility possible given the formulation of the monocycle shown in Figure 1? If GAPs are active and k_off is thus high, how can fractional activation, k_on/(k_on + k_off), also remain high? Why don’t all GAPs inhibit all the time?

Figure 2. GAPs can accelerate response kinetics without substantially inhibiting steady-state signaling.

(A) Currents generated by mouse photoreceptor cells decay slowly after a light flash if the GAP is absent. Rod outer segments, which contain the rhodopsin–transducin G-protein module, were taken from wild-type mice (black trace) or from mice either lacking RGS9-1 (red trace), the principle photoreceptor GAP, or heterozygous for RGS9-1 (green trace): traces show responses to a single photon. While the amplitude of the downstream current spike is essentially unchanged in the GAP^−/− cells, the current decay after the flash is extremely long-lived. (Reproduced with permission from Chen et al. [54].) (B) Co-expression of a GAP is required to reconstitute the native kinetics of a Gβγ-gated potassium (K⁺) channel that is normally expressed in cardiac myocytes. K⁺ channel subunits Kir3.1/3.2 and m2 muscarinic cholinergic receptors were co-expressed in CHO fibroblasts with or without the relatively non-selective GAP RGS4. K⁺ current was monitored during exposure to acetylcholine (ACh) for the period highlighted in grey. The lower trace shows a cell that expresses RGS4. Its response to ACh resembles that of an atrial myocyte, shown in the top trace, both in amplitude and kinetics. CHO cells that do not express RGS4 (middle trace) display a markedly slow recovery after agonist is removed. Average time constants for channel deactivation (τ_deact) are shown at the right: τ_deact is the inverse of the apparent rate constant k_app for deactivation. Deactivation rates upon removal of agonist are at least 10-fold faster in the presence of RGS4. Reference bars show 0.2 nA (vertical) and 5 s (horizontal). For a simple monocycle of the sort shown in Figure 1, k_app = k_on + k_off. Because k_on is small when agonist is absent, the difference in τ_deact reflects the difference in k_off. (Reproduced with permission from Doupnik et al. [23].) These phenomena have been observed for numerous G-protein-gated K⁺ and Ca²⁺ channels [4], but are crucially dependent on the stoichiometric relationships among receptor, G protein and GAP in the membrane [8].

Other problems with the simple monocycle emerged from the kinetic characterization of GAPs, particularly PLCβ [4,12]. First and most notable, when the rate of the overall steady-state GTPase reaction was measured in the presence of agonist-bound receptor and GAP, it was found to be about 10-fold faster than the rate of receptor-promoted binding of GTP that was previously measured in the absence of a GAP. Because GTP binding to G protein is the first step in the overall hydrolysis reaction, these results were clearly contradictory. Another problem in these early studies was that, if the rates of receptor-promoted GTP binding and GAP-promoted GTP hydrolysis were used to estimate G-protein activation at steady state as described above, the GAP activity should have completely suppressed the signal. In fact, attempts to detect accumulation of the active G_q-GTP complex directly were unsuccessful. Regardless, substantial activation of PLCβ could still be observed. These problems were explained to some extent by more careful studies of GTP binding that detected a fast component of receptor-stimulated GTP binding in the presence of PLCβ1 or RGS proteins [13]. It remained unclear how GAPs exert this effect, however, because isolated GAPs had been shown not to have an effect on nucleotide binding rates [4,14]. How was it possible to generate a signal when hydrolysis was so fast? Does a GAP really make a receptor more efficient?

Speed versus Amplitude — Balancing on the Monocycle

In principle, there are two answers to the question of how a GAP can promote a fast termination of signaling without also inhibiting signaling while receptor is activated. Either GAPs must somehow potentiate G-protein activation by the receptor or the receptor must inhibit GAP activity (Figure 3). Both have experimental support, each is probably correct in one system or another, and both mechanisms may act simultaneously. I discuss possible biochemical mechanisms for each below.

Figure 3. Balancing on the monocycle.

To allow regulation of turn-on and turn-off rates while independently regulating steady-state signal output, it is necessary to coordinate activities of receptors and GAPs. In principle, there are two solutions. Either the receptor must inhibit the activity of the GAP (red) or the GAP must potentiate the activity of the receptor (green). In the first mechanism, the receptor must inhibit the GAP in a way that does not depend on its activation of the G protein and that itself turns off rapidly when agonist is removed. In contrast, potentiation of the receptor by GAP may be unregulated and may be propagated by the G-protein heterotrimer or Gα.

These two general explanations are not as vague as they may seem, and both can be demonstrated to work as proposed by simple kinetic algebra or computer simulation. If a GAP potentiates the receptor’s regulatory activity by what-ever mechanism, it may increase the receptor-promoted k_on enough that fractional activation, k_on/(k_on + k_off), remains high in the presence of agonist even though k_off is large enough to give prompt signal termination. If receptor, while it is active, can inhibit the GAP, then signaling will be turned on when receptor is active because k_off will be low, but k_off will increase upon receptor deactivation and terminate the signal.

GAP inhibition by receptor cannot use the Gα subunit as a mediator because the GAP must reactivate promptly when agonist dissociates so that it can drive the G protein turn-off reaction. Other proteins might be involved, but any acceptable mechanism must terminate quickly so that the GAP can initiate GTP hydrolysis as soon as receptor is deactivated. In contrast, potentiation of receptor regulation by the GAP can operate via the Gα subunit and need not be fast at all. Further, potentiation of a receptor’s activity as a G-protein activator is functionally identical to sensitization of the G protein to regulation by receptor. Indeed, if receptor, G protein and GAP are bound simultaneously, potentiating receptor and sensitizing G protein may be indistinguishable.

How Do GAPs Potentiate Receptors?

Three general mechanisms, with varying degrees of experimental support, appear to combine to allow GAPs to potentiate receptor activity and thus balance the monocycle of Figure 3: scaffolding mechanisms that maintain receptor–G protein binding and/or block GAP action; ‘kinetic scaffolding’, i.e. enhanced association based on the lifetimes of specific GTPase cycle intermediates; and allosteric potentiation of receptor function that is mediated by the Gα subunit itself.

Receptor–GAP association, either direct or mediated by scaffolding proteins, has now been described at varying levels of detail for diverse receptors and GAPs in many cells (see [4,11,15,16] for reviews). These interactions no doubt contribute significantly to the selectivity of GAPs for specific receptor pathways, but most studies focused on how receptor–GAP binding promotes inhibition of signaling. To some extent, this focus reflects the relative experimental ease of measuring inhibition of downsteam signaling in cells compared with the difficulty of monitoring the kinetics of isolated G-protein functions, but selective recruitment of GAPs as inhibitors is clearly widespread.

Two studies of receptor–GAP interaction do suggest that GAP may stimulate the receptor; however. Wang et al. [17] found that neurabin, a scaffolding protein that binds both RGS proteins and GPCRs, potentiated Ca²⁺ signaling by the α_1B-adrenergic receptor both in cell culture and parotid gland ducts. Supporting data were consistent with the idea that neurabin acts by blocking the inhibitory effects of an RGS protein while both are bound to the receptor, although a detailed mechanism was not available. The result is interesting also because spinophilin, a close paralog of neurabin, inhibits G-protein signaling by recruiting RGS proteins to form an inhibited complex [18]. A second but more ambiguous case is RGS4-induced inhibition of G_q-mediated signaling in pancreatic acinar cells [9,19]. Here, RGS4 inhibits m3 muscarinic cholinergic receptor signaling at only about 1% of the concentration needed to inhibit signaling by cholecys-tokinin, with the bombesin receptor displaying intermediate sensitivity. RGS2 did not display such selectivity among the three receptors. All three receptors and RGS4 acted through a single pool of G_q/G₁₁, suggesting that some specific interaction led to the resistance of cholecystokinin signaling to inhibition by this GAP, although the mechanism is unknown. Signal termination kinetics were not monitored in these studies.

An alternative mechanism for how GAPs potentiate stimulation of G protein by the receptor was proposed specifically to explain the problem of combining robust signaling with a fast turn-off rate [12,20] (Figure 4). The mechanism, called kinetic scaffolding, describes a pathway of reactions through the GTPase cycle in which a GAP promotes the continuous association of receptor and G protein (Figure 4, inner cycle). Put simply, GAP-stimulated GTP hydrolysis is fast enough that receptor does not have time to dissociate from the G protein-GTP complex, such that it is still bound and available to drive a new round of GDP–GTP exchange. Kinetic scaffolding thus obviates the slow, diffusion-limited association between the receptor and the GDP-bound G protein (Figure 4, outer cycle) and shifts the rate-limiting (i.e. slowest) step in G-protein activation from receptor–G protein binding to receptor-driven GDP dissociation, which is far faster (Figure 4). Note that kinetic scaffolding does not imply physical scaffolding or any thermodynamic enhancement of affinity, but merely a change in reaction path that is allowed because GTP hydrolysis is accelerated by the GAP. Conversely, however, if receptor and G protein remain bound as an active complex, their relative interactions might be expected to be even more efficient than if both were merely tethered close to each other by a separate protein.

Figure 4. Kinetic scaffolding limits receptor dissociation [4].

The two concentric loops describe stereotyped paths through the GTPase cycle: the fast inner cycle that predominates in the presence of a GAP and the slower outer cycle that is traversed in the absence of GAP. The GTPase reactions that describe kinetic scaffolding are shown in the inner cycle, with rates of key reactions shown as average lifetimes (τ = 1/k). The key branch-point species is the activated complex of receptor–G protein-GTP–GAP (R–G*-GTP–GAP). Because GTP hydrolysis is more than 100-fold faster than dissociation of receptor, the receptor remains bound after hydrolysis and can rapidly catalyze GDP/GTP exchange. This cycle can maintain about 25% of G protein in the active state as long as receptor is activated. The rate constants shown are for m1 muscarinic cholinergic receptor, G_q and PLCβ1 at 30°C [22], and the GTP association rate assumes 200 µM GTP (cyto-solic concentration). Detailed analysis shows that the GAP does not remain tightly bound throughout the GTPase cycle as shown in the inner cycle, but is in rapid binding equilibrium with R–G-GDP [22]. The slower collisional coupling path, shown in the gray outer cycle, proceeds in the absence of a GAP. Because the GTP-bound species has a long lifetime, about 10 s, receptor dissociates during every catalytic cycle and the rate-limiting step becomes the diffusion-limited rebinding of receptor and G protein [55].

Several enzymologic studies support the importance of the kinetic scaffolding pathway [12,13,21], but the association lifetime for receptor–G protein binding during GTPase cycle turnover has not yet been measured directly. Recently, Turcotte et al. [22] used an experimentally determined set of rate constants for the complete GTPase cycle to simulate the reaction pathways for the m1 muscarinic cholinergic receptor and G_q with and without PLC, which is both a GAP for G_q and a G_q effector. They found that PLC increases the fraction of G protein that stably associates with receptor during steady-state GTPase turnover—the basic concept of kinetic scaffolding. These simulations also describe the dependence of kinetic scaffolding on the concentration of GAP and on its maximal activity, and thus map the period over which direct measurements of binding should be made.

In addition to scaffolding mechanisms, analysis of steady-state GTPase activity has suggested that a GAP can also directly increase the rate of receptor-promoted GDP–GTP exchange. This provides yet a third mechanism for a GAP to maintain signal output by accelerating exchange to match fast hydrolysis. Turcotte et al. [22] found that the rate of dissociation of GDP from a complex of the m1 muscarinic cholinergic receptor, G_q and PLCβ1 is about 17-fold faster than its dissociation from the receptor–G_q complex alone. This enhancement of the intrinsic GDP–GTP exchange rate combined with kinetic scaffolding to allow the receptor to maintain about 20% of the G_q in the active state, despite a more than 1,000-fold increase in the rate of hydrolysis of G_q-bound GTP. The physical mechanism whereby a GAP contributes to receptor-driven GDP dissociation is not clear. For example, GAP may either make the receptor a better exchange catalyst or make the G protein more responsive to receptor. Differentiating between these two possibilities will be difficult because a complex of all three proteins is required, and there is no information on its structure. There was no effect of GAP on the nucleotide exchange rate for G_q in the absence of receptor, as is true for many GAP–G protein combinations [4]. Significantly, the GAP in this case is also the effector whose signaling activity is stimulated by G_q-GTP, and it will be interesting to see whether non-effector RGS proteins also display a similar effect.

GAPs and the Activation Rate

It may be surprising that GAPs potentiate receptors at steady state, but they also seem to increase the actual rate of onset of G-protein signaling, at least in some cases. Doupnik et al. [23] first noted that RGS4 accelerates the opening of G-protein-gated potassium channels upon addition of agonist (Figure 2B). This channel is a Gβγ-regulated effector, and Chuang et al. [24] made a strong argument that receptor and GAP somehow cooperate to prime the G protein, or at least its Gβγ subunits, for rapid activation when GTP binds (see also [8]). This positive effect of GAPs is superficially consistent with the potentiative effects discussed above, but precisely how they accelerate the turn-on reaction remains a mystery. Regardless, other investigators have observed the same phenomenon for this and other G-protein-gated channels [4]. Simulations of the rate of G_q activation by receptor also predict that GAPs slightly enhance the turn-on rate when agonist is added [22], although this prediction has not yet been verified experimentally. The simple monocycle of Figure 1 also predicts that a GAP will increase the observed activation rate, because k_app = k_on + k_off, and a GAP will increase k_off. These phenomena may relate to steady-state functions of GAPs and support the general concept of GAPs as accelerators.

How Do Receptors Inhibit GAPs?

There is not yet an established mechanism for G-protein-independent inhibition of GAP activity by agonist-activated receptor—the second general strategy for matching the rates of activation and deactivation in the presence of a GAP (Figure 3). GAPs are inhibited in their GAP activities by many diverse mechanisms, however, that include: phosphorylation of RGS proteins in the RGS and ancillary domains [25–28] (see [29–32] for stimulation or relocalization); palmitoylation of a conserved cysteine residue in RGS domains [33] (see [34] for stimulation); phosphorylation or palmitoylation of the amino-terminal helix of the G-protein substrate [35–37]; and either blockade or sequestration by binding to any of the Gβγ subunits [38,39], or to scaffolding proteins [11,16,17,40–42] or to acidic lipids [43,44]. Most of these events have been described both in cells and for purified proteins, and cellular regulation has been described for a few. However, it remains difficult to assign to these events relative quantitative importance in specific instances. It is also unclear how most might be adapted for coordinating GAP activity with receptor-catalyzed GDP–GTP exchange, and those that involve subcellular relocation of proteins or covalent modifications are probably too slow. While selective functional interactions between receptors and GAPs have been described, none seems to involve inhibition of GAP activity that would meet the criteria of rapid response to activation and deactivation of receptor combined with independence of Gα.

One plausible mechanism whereby receptor may inhibit GAP activity in a multiprotein complex is based on the ability of Gβγ subunits to inhibit the GAP activities of both RGS proteins and PLCβ isoforms [37–39,45]. Gβγ can inhibit GAPs nearly completely and with reasonably high potency. Inhibition probably depends on the binding of Gβγ to the GTP-bound Gα subunit to block access of the GAP [39], but other data suggest that binding of Gβγ and GAP are not mutually exclusive. When receptor is active, it is likely that receptor, Gα and Gβγ are at least loosely bound to each other [37].

How might receptor promote GAP inhibition by Gβγ? Gβγ is essentially required for high-efficiency regulation of Gα subunits by receptors [3]. Gβγ subunits interact selectively with receptors [46–50], either binding directly to receptors or somehow maneuvering Gα subunits to alter their receptor selectivity. The mechanism of receptor–Gβγ selectivity is not known. While it is likely, on the basis of their selective interactions, that receptor and Gβγ make contact, there is little experimental evidence for direct binding of isolated Gβγ to receptors in the absence of Gα [48,51,52]. If receptor makes Gβγ a better GAP inhibitor, it will probably be through a combination of direct but low-affinity binding that can be reversed rapidly when receptor deactivates and the loosening of interaction between Gα and Gβγ when Gα is active. A conceptual model may be the complex of Gα_q, Gβγ and the GPCR kinase GRK2 (also a GAP for G_q), in which Gα_q and Gβγ remain nearby but not touching, and appear poised to rebind [53]. Because GAPs and Gβγ subunits bind both to the switch regions on the Gα Ras-like domain and to the amino-terminal helix, it is possible that they compete functionally only for the switch regions while binding simultaneously to the amino-terminal region.

How Does the Balance Work in Cells?

The mechanisms described above for allowing GAPs to potentiate receptor function are both biochemically feasible and kinetically adequate. Less is known about inhibition of GAPs by receptors, but this too is plausible. All of these events may co-exist in varying proportions to balance the output of the G-protein monocycle while displaying appropriate kinetics. The immediate challenge is to determine which of these mechanisms apply more generally and what aspects of their regulatory properties make them more or less applicable to specific signaling paths. In many cases, we have the kinetic tools needed to approach these questions.

At a more general level, we want to define the dependence of G-protein signaling dynamics on the complex and inter-related reactions that combine to determine output in response to stimuli. How are the discrete behaviors of a G-protein module sensitive to the concentrations of proteins and to the rate constants of the reactions in which they participate? Understanding this system will depend on a clear description of how these parameters contribute to signal output in time and space.

Next, which mechanisms are important in cells? How are they regulated? Answering these questions will require in situ measurement of the individual steps of the GTPase cycle and time-resolved monitoring of the conformations and associations of multiple proteins. It will also require completing the experimental cycle of developing mechanistic models in vitro and testing the outcomes of model-based manipulations in cells. These tools are increasingly available, and should allow us to relate cellular signaling dynamics to its kinetic foundations.

Conclusions

G-protein signaling provides a wonderful example of how a conserved module composed of relatively few proteins can operate over a wide range of timescales and relative extents of amplification, and do so in a controlled and regulated way. This sophisticated behavior depends on the intrinsic kinetic behaviors of the individual constituent proteins — receptor, Gα, Gβγ, GAP and effector—and multiple regulatory interactions among them. We are now beginning to understand these interactions well enough to evaluate their contributions to native cellular signaling networks.