ReviewG Proteins

Dissociation of Heterotrimeric G Proteins in Cells

See allHide authors and affiliations

Science Signaling  24 Jun 2008:
Vol. 1, Issue 25, pp. re5
DOI: 10.1126/scisignal.125re5


Heterotrimeric G proteins dissociate into their component Gα and Gβγ subunits when these proteins are activated in solution. Until recently, it has not been known if subunit dissociation also occurs in cells. The development of optical methods to study G protein activation in live cells has made it possible to demonstrate heterotrimer dissociation at the plasma membrane. However, subunit dissociation is far from complete, and many active [guanosine triphosphate (GTP)–bound] heterotrimers are intact in a steady state. This unexpectedly reluctant dissociation calls for inclusion of a GTP-bound heterotrimeric state in models of the G protein cycle and places renewed emphasis on the relation between subunit dissociation and effector activation.


One of the most ubiquitous transmembrane signaling mechanisms involves activation of heterotrimeric guanosine triphosphate (GTP)–binding regulatory proteins (G proteins) by G protein–coupled receptors (GPCRs). These receptors detect light, odorants, hormones, and neurotransmitters, and active G proteins regulate diverse effector molecules such as enzymes and ion channels. G protein heterotrimers consist of a Gα subunit, which binds and hydrolyzes GTP, and a Gβγ dimer (see Table 1). The idea that Gα and Gβγ physically dissociate from one another during signaling arose when hydrodynamic studies showed that activation decreased the apparent size of G proteins in solution (1), as well as that active Gα could be resolved from Gβγ by gel filtration (2). The findings that an isolated Gα subunit (now known as Gαs) could stimulate adenylyl cyclase (AC) and that isolated Gβγ inhibited this stimulation suggested that subunit dissociation was important for effector activation and, thus, was not just an experimental artifact. Abundant evidence consistent with this model has accumulated over the years, and the idea that subunit dissociation is an integral part of G protein signaling is now widely accepted. Accordingly, many schematic diagrams of the G protein cycle combine GTP binding and subunit dissociation into a single step (Fig. 1A).

Fig. 1.

Two versions of the G protein cycle. (A) A common simplified model in which GTP binding and heterotrimer dissociation are shown as a single irreversible step. GPCR activity accelerates guanine nucleotide exchange [guanosine diphosphate (GDP) for GTP], whereas RGSs accelerate GTP hydrolysis. (B) A more complete model in which GTP binding and heterotrimer dissociation are shown as separate reversible steps. RET occurs between labeled Gα and Gβγ subunits in a GαGDPβγ heterotrimer. A conformational change associated with nucleotide exchange (producing GαGTPβγ) could either increase or decrease the RET efficiency between Gα and Gβγ (ΔRET), whereas physical dissociation would likely prevent RET altogether (↓ RET). Therefore, the magnitude and direction of the net RET change associated with activation will depend on the fraction of labeled heterotrimers and subunits in each state. In a saturation BRET experiment, the amount of BRET acceptor (either Gα or Gβγ) needed for maximal BRET (an index known as the BRET50) is predicted to increase with heterotrimer activation, which indicates a decrease in affinity between Gα and Gβγ (37). Pi, inorganic phosphate; k, rate constant.

Fig. 2.

Gα-Gβγ interfaces. Surface renderings of Gαi1 (left) and Gβ1γ2 (center and right) rotated to show contact residues at the αN helix (yellow) and switch region (blue) interfaces; blue residues on Gαi1 contact blue residues on Gβ1γ2, and yellow residues on Gαi1 contact yellow residues on Gβ1γ2. Surface residues that contact only GRK2 (red), GRK2 and the αN helix of Gαi1 (orange), and GRK2 and the switch regions of Gαi1 (magenta) are shown at the right. GRK2 and Gαi1 bind to overlapping residues in both interfaces; thus, it is unlikely that Gβ1γ2 interacts with GRK2 and Gα simultaneously. [See interactive figures (]

Table 1.

G protein classification.

However, this acceptance was gained without a direct demonstration of G protein dissociation in cells. The authors of the original reports of subunit dissociation were careful to point out that it was not safe to assume that G proteins would also dissociate in the plasma membrane (2, 3). At least two factors were cause for concern. First, solubilizing membrane-associated G proteins required the use of detergent, which could have artificially promoted subunit dissociation. Second, simple dilution of the proteins in vitro compared with the native state might have favored dissociation. In addition, activation in solution was most often achieved by adding either fluoride or a nonhydrolyzable GTP analog (GTPγS) instead of GTP (the endogenous ligand), and the magnesium concentration was often much higher than what would be found in cells. Thus, despite its status as a pivotal event in G protein signaling, there were legitimate reasons to wonder whether subunit dissociation actually occurred in vivo (4, 5).

Do G Proteins Dissociate in Cells?

An experimental test of subunit dissociation was made possible by the application of resonance energy transfer (RET) techniques to study protein-protein interactions in live cells. Both fluorescence (or Förster) (FRET) and bioluminescence (BRET) resonance energy transfer occur when two fluorophores are located within ~10 nm of each other and are arranged in a permissive orientation. Therefore, energy transfer would be expected between labeled Gα subunits and Gβγ dimers, and dissociation of these subunits would be expected to decrease this signal. The first such experiments were carried out in the social amoeba Dictyostelium discoideum with cyan fluorescent protein (CFP)– and yellow fluorescent protein (YFP)–tagged Gα and Gβ subunits (6). Activation of chemoattractant receptors in these cells produced a large decrease in FRET between the subunits, consistent with dissociation of many of the labeled heterotrimers. However, a seminal paper reported the unexpected finding that FRET between labeled Gαi (the α subunit of G protein the inhibits AC) and Gβ or Gγ subunits increased after receptor activation (7). This report also showed that the direction of the FRET change was reversed if the position of the fluorescent reporter was changed from one end of the Gγ subunit to the other. Subsequent reports have shown similar activation-induced increases or decreases in BRET or FRET depending on the Gα isoform studied and the precise location of the reporter moieties (812). In some cases, these results were interpreted as evidence that activation produced a rearrangement of G protein subunits but not physical dissociation.

However, a limitation of most RET studies is that they are carried out on populations of proteins and so report only the summed behavior of the population. Therefore, a net increase in FRET or BRET does not rule out the possibility that some proteins dissociate, while the majority remain associated. In addition, as noted above, a decrease in RET could be caused by a conformational rearrangement rather than physical dissociation. Complementary approaches, however, have provided persuasive evidence that some active G protein heterotrimers do, in fact, dissociate in live cells.

Much of the evidence supporting G protein dissociation in cells has been provided by studies of G protein translocation between cellular compartments. In rod photoreceptor cells, activation of the GPCR rhodopsin by intense light stimulates movement of the G protein transducin (Gαtβtγt) from the outer segment disks to the inner segment (13). Both Gαt and Gβt translocate between segments; however, they do so at different rates (14). Intact transducin heterotrimers associate with outer segment disks because Gαt is acylated and Gγt is prenylated. Either of these lipid modifications alone is insufficient to permanently anchor a protein to a membrane, but the combined effects of both keep intact transducin heterotrimers attached to outer segment disks. Thus, translocation occurs when active subunits dissociate first from each other and then from the disk membrane (13). In keeping with this mechanism, translocation is markedly impaired when the attachment of Gγt to disk membranes is strengthened by the substitution of a 20-carbon geranylgeranyl group for the normal 15-carbon farnesyl group (15). The resolution of these experiments is not sufficient to detect subunit dissociation during normal photoreception, but these results make it difficult to avoid the conclusion that active transducin subunits dissociate from each other in intact rods after intense illumination.

Similar studies have now been carried out with other G protein isoforms in other cell types. In Chinese hamster ovary (CHO) cells, for instance, Gβγ dimers containing Gγ11, another farnesylated subunit, dissociate from the plasma membrane and translocate to an intracellular compartment after GPCR stimulation (16). The kinetics of this translocation are consistent with subunit dissociation during normal signaling. Not surprisingly, the geranylgeranylated Gγ5 subunit dissociates less readily, and the relevant Gα subunits stays attached to the plasma membrane, presumably because they are strongly anchored there by the dual acyl modifications characteristic of most Gα subunits. Similarly, experiments using a technique that detects lateral Gβγ movement between regions of the plasma membrane have demonstrated physical dissociation of G proteins in human embryonic kidney (HEK) cells (17). In these experiments, Gα subunits were modified with a transmembrane domain and immobilized with an extracellular cross-linking agent. This manipulation decreased the lateral diffusion of labeled Gβγ dimers in the plasma membrane, which showed that the modified subunits formed stable heterotrimers. Activation of a GPCR partially relieved the constraint on Gβγ mobility, which indicates that some of the heterotrimers had physically dissociated.

The combined results of these live-cell studies can be accommodated by a more complete model of the G protein cycle, one similar to that proposed in some of the original biochemical studies (3, 18). This model separates GTP binding and subunit dissociation into discrete and reversible steps (Fig. 1B). GTP binding is associated with a conformational change (activation) that weakens the interaction between Gα and Gβγ, and may lead to an increase or decrease in RET (depending on the position of the fluorophores). Individual GαGTPβγ heterotrimers may then dissociate or stay intact, and individual subunits might participate in multiple dissociation and reassociation events before GTP hydrolysis terminates activity. The combination of robust activation and reluctant dissociation would lead to an accumulation of GαGTPβγ and could produce a net increase in population RET, even though some heterotrimers would be dissociated in a steady state. This model predicts that binding of free Gα or Gβγ subunits to an effector molecule would shift the dissociation-reassociation equilibrium toward dissociation. In a RET experiment, this would tend to enhance a RET decrease or convert a RET increase into a decrease.

The existence of somewhat stable GαGTPβγ heterotrimers is by no means a new idea. Indeed, one of the early studies of purified Gi reported two discrete changes in hydrodynamic properties, the first reflecting a conformational change associated with activation by ligand, and the second (which occurred only at a physiological temperature) reflecting actual subunit dissociation (3). More recent biochemical studies have also detected a clear delay between heterotrimer activation and subunit dissociation (19).

Not surprisingly, there is evidence that some GαGTPβγ combinations are more stable than others and, therefore, are less likely to dissociate. For example, unlike transducin in rod cells, cone transducin fails to translocate after intense illumination (20). This failure is most likely due to a difference in the Gβ and Gγ isoforms expressed in these two cell types. In addition, we have found that under some conditions Gs heterotrimers are less efficient donors of free Gβγ than GoA heterotrimers (21). Therefore, it is likely that the propensity to dissociate will vary for different subunit combinations.

Is Subunit Dissociation Necessary for Activation of Effectors?

If activation of GPCRs can lead to accumulation of GαGTPβγ heterotrimers, then it is important to know whether or not these heterotrimers can interact with effector molecules and participate in signaling. In other words, even if some G protein heterotrimers dissociate on activation, is this necessary for signaling? Both Gα and Gβγ can interact with effector molecules, and biochemical reconstitution experiments have repeatedly shown that isolated Gα or Gβγ subunits are capable of signaling to effectors (18). Because many effectors are activated by overexpression of either Gα or Gβγ alone, it seems likely that free subunits are also able to signal in cells. But what of intact or "semi-intact" heterotrimers, and what might a semi-intact heterotrimer look like?

Crystal structures of Gαβγ heterotrimers show that two noncontiguous surfaces of Gα are covered by contact with Gβγ (22, 23). Approximately two-thirds of the total Gα contact area (the switch region interface) is located at the so-called "switch regions," the part of the subunit that changes conformation as a result of nucleotide binding and hydrolysis (Fig. 2). The other third of the contact area (the αN helix interface) is located on the amino-terminal αN helix. Cocrystallization of Gα subunits and effector molecules has revealed a common mode of interaction between switch 2 (in the switch region interface) and hydrophobic residues in the effector (24). Similarly, regulator of G protein signaling (RGS) proteins accelerate the intrinsic guanosine triphosphatase (GTPase) activity of Gα subunits by binding to the switch region interface (25), and binding of RGS proteins and Gβγ dimers to Gα subunits is competitive (26). Therefore, the switch region interface must be disrupted for Gα to interact with effectors or RGS proteins. However, none of the available structures have identified a role for the Gα–αN helix interface in effector binding. These observations suggest that it might be possible for a Gα subunit to bind simultaneously a Gβγ dimer (through the αN helix interface) and an effector (through the switch regions). This semi-intact heterotrimer model would require considerable flexibility of the αN-β1 linker and has been invoked to explain instances of signaling by heterotrimers (see below) (27, 28).

Would maintained association at the αN helix–Gβγ interface allow activation of Gβγ effectors? The surface of Gβγ has been extensively mapped by alanine-scanning mutagenesis and cocrystallization with Gα subunits and Gβγ effectors (2931). Again, the Gβγ surfaces that contact Gα largely overlap those that contact (or are important for interaction with) effectors. The most precise information regarding overlap comes from cocrystals, such as that of Gβ1γ2 and G protein–coupled receptor kinase 2 (GRK2). GRK2 makes contact with several residues on the Gβ1γ2 side of the switch region interface, but also with a few residues in the αN helix interface (32) (Fig. 2). This dual overlap makes it highly unlikely that Gβγ dimers could maintain contact with Gα subunits at either interface while bound to a GRK. Scanning mutagenesis suggests that Gβγ residues in both interfaces are also important for activation of inwardly rectifying potassium (GIRK) channels, adenylyl cyclase, and other Gβγ effectors (29). This suggests that activation of these Gβγ effectors would require disruption of both Gα-Gβγ interfaces, and thus complete dissociation of heterotrimers.

Despite this prediction, there is, as yet, no direct evidence that heterotrimer dissociation is absolutely required for transduction of a particular signal. On the other hand, there is evidence that semi-intact heterotrimers can signal (27, 28). Perhaps the most compelling evidence for signaling by heterotrimers comes from studies of mating in the budding yeast Saccharomyces cerevisiae. Mating signals are carried by Gβγ dimers, which activate both mitogen-activated protein kinases (MAPKs) and specific proteins that induce the growth of polarized mating projections. Strains of yeast that lack both the Gα subunit (Gpa1) and the Gβ subunit (Ste4) are defective in mating, but can be rescued by expression of a Gβ-Gα fusion protein (33). In addition to preventing complete dissociation, structural considerations suggest that this fusion would help to stabilize the αN helix interface. Moreover, mutations in the switch region and αN helix interfaces of Gβ subunits have distinct effects on mating (34). Specifically, control of polarity by pheromone requires an intact αN helix interface in Gβ but not an intact switch region interface; mutations in the switch region interface that completely prevent heterotrimer formation in in vitro assays do not prevent pheromone-induced polarization. These results suggest that polarity signals can be transmitted by heterotrimers associated at the αN helix only and that association at the αN helix is actually necessary for polarity signaling. Therefore, it may not only be possible for heterotrimers to signal while maintaining contact with each other, in some cases, it may be essential for them to do so.

Although the evidence that Gβγ dimers can interact simultaneously with effectors and Gα subunits in yeast is convincing, there is reason to suspect that the situation will be different for other Gβγ dimers and effectors. As described above, there is considerable overlap of Gα-binding and effector binding surfaces for effectors such as GRKs and ion channels. In contrast, the Gβγ effectors responsible for mating and polarity in yeast apparently bind to the amino-terminal coiled-coil of the yeast Gβγ, a much more accessible location in an intact or semi-intact heterotrimer (34, 35). Therefore, it would not be surprising if some Gα or Gβγ effectors required heterotrimer dissociation, whereas others did not.

The Functional Consequences of G Protein Dissociation

It is now clear that at least some active G protein heterotrimers do dissociate in cells. We also have some idea about the importance of dissociation for signaling by some pathways, and it is likely that more information related to this question will be forthcoming. In contrast, the biological significance of signaling by dissociated, as opposed to intact, heterotrimers is largely the subject of speculation. The idea that dissociation might occur at the plasma membrane was initially of interest because of the possibility that free Gβγ dimers would act to inhibit Gα subunits and so would provide a mechanism of heterologous inhibition (crosstalk) between different Gα isoforms (18). Indeed, there is evidence that some forms of crosstalk between G proteins are mediated by Gβγ exchange (36). Perhaps the most obvious benefit of subunit dissociation is the removal of steric hindrance between effector molecules and still-associated Gα or Gβγ subunits.

On the other hand, one can easily imagine reasons why signaling by intact heterotrimers would be advantageous. For example, maintained association would allow Gα subunits to confer specific signaling properties (e.g., receptor fidelity) onto their associated Gβγ dimers, which could explain the observation that Gβγ signaling sometimes occurs in a Gα-specific manner. Because it is reasonable to expect that reassociation from a semidissociated state would be faster than reassociation from a completely dissociated state, the maintenance of intact G protein heterotrimers might also speed the termination of signals. It should be pointed out, however, that qualitatively similar advantages would apply to a situation in which subunits dissociated and reassociated rapidly during signaling. In other words, there is little difference between maintained association and very transient dissociation, apart from the improved access to binding sites offered by the latter.

Of course, it is entirely possible that cells use both intact and dissociated heterotrimers to carry signals and that the likelihood of subunit dissociation is optimized to suit a particular signaling need. In any case, it is clear that the cautious interpretation of biochemical experiments showing G protein dissociation was justified; the phenomenon is much easier to observe with proteins in solution than with proteins in cells. As such, G protein dissociation represents a good example of a case in which protein-protein interactions are influenced by cellular context.


  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.
View Abstract

Stay Connected to Science Signaling

Navigate This Article