Microfilament-Membrane Interaction

456 TIBS - November 1985

microvilli (see below), etc. Trans- Microfilament-membrane membrane interactions which involve cytoskeletal elements are commonly interaction found in adherens and desmosomal junctions, in surface patches and caps Benjamin Geiger induced by multivalent ligands, etc. (see Ref. 1 and discussion below). Surveys of a large variety of biological One of the major topics of research in modern cell biology involves the structure and systems suggested that the induction of organization of the cytoskeleton and, in particular, the interaction of its various surface specialization due to the interac- components with the plasma membrane. Studies carried out in many laboratories over tion between the cytoskeleton and the the last several years pointed to the complex and heterogeneous nature of membrane- inner surfaces of the plasma membranes cytoskeleton interactions in different systems. This is manifested by a remarkable cell- can play multiple roles of great physio- type'specificity as well as distinct differences between the mode of anchorage of the logical significance in processes such as various cytoskeletal components. Moreover the assembly and maintenance of each of cell motility, attachment, division and the cytoskeletal systems within individual cells appears to be a dynamic process which recognition. is probably spatially and temporally regulated and modulated. Microfilaments are associated with the Here, examples are presented to illus- and topology of the latter. Many studies, plasma membrane via specific linker trate some of the common features of both biophysical and structural, have proteins the interactions which occur at the mem- suggested that membrane constituents Microfilaments constitute one of the brane-microfilament interphase, and may sometimes display non-random dis- major cytoskeletal systems in living cells4 their physiological significance is distribution, forming specialized domains cussed (for a recent review see Ref. 1). with distinct topology, structure and One of the most important con- functions. Many such domains have sequences of membrane-microfilament been identified, for example, the synap- interaction is the local perturbation of tic clusters of acetyicholine receptors, the membrane and the outcoming alter- receptors to low-density lipoprotein, ations in its biochemical and biophysical LDL (or to other ligands) which are pre- properties. The generation of molecular clustered in coated pits, surface microdomains in biological membranes glycoproteins in budding enveloped by such mechanisms has central roles in viruses, and specific cellular junctions. cellular physiology, as discussed below. Moreover, morphological specializations and irregularities of the free Microdomains in membranes: the surface could often be seen by transmis- involvement of the cytoskeleton sion or scanning electronmicroscopy 3. The widely accepted fluid mosaic In principle, there are several possible model depicts biological membranes as mechanisms for microdomain forma- two-dimensional lipid bilayers in which tion, some of which are outlined in 'integral proteins' are embedded2. It was Fig. 1. proposed that these membrane proteins The starting state is the cell depicted are free to laterally diffuse through the on the left of Fig. 1 with different types fluid phase and thus reach a uniform of homogeneously distributed surface distribution. It was, nevertheless, molecules (the squares and circles). This pointed out in the original formulation cell might undergo various changes of the model that the lateral mobility of which could lead to the segregation of the integral membrane constituents may the 'squares' and 'circles' into specific, be restricted by interactions with specialized and mutually exclusive Fig. 1. Four different mechanisms for the genera- immobile matrices at the cytoplasmic or microdomains (see caption to Fig. 1). tion of non-homogenous distribution of integral extracellular interphases of the plasma The mechanisms shown in Figs lc and d membrane proteins. The cell on the left has its membrane. Such interactions may be are of direct interest to the topic of this 'square' and "circular' receptors homogenously dis- article; they may involve either a uni- tributed. (a) Formation of a cluster of the 'square' mediated by 'peripheral membrane pro- receptors by intramembranal interactions. This teins', namely proteins which are tightly lateral association of the membrane aggregation may be triggered by interaction with bound to either surface of the mem- receptors with intracellular matrices, another membrane protein, modification of the brane without being actually inserted namely the cytoskeleton (c) or a trans- "squares', changes in medium conditions (pH, into its bilayer. Furthermore, it was con- membrane interaction with insoluble ionic environment), etc. (b) Exofacial interactions ceivable that if the peripheral elements elements at the two faces of the plasma- with extracellular matrices leading to immmobiliza- lemma (d). Examples of the former tion and aggregation of a particular class of recep- form mechanically stable networks, tors. (c) Endofacial interactions, namely their attachments to the membrane may might be the association of band-3 microdomain formation induced by an interaction have long-range effects on the dynamics protein with the cytoskeletal shell in of the 'square" integral proteins with cytoskeletal erythrocytes which is mediated by filaments. (d) Transmembrane interactions involv- B. Geiger is at the Department of Chemical Immu- ing an integral membrane element attached to both nology, The Weizmann Institute of Science, ankyrin, the attachment of microfila- extracellular and intracellular matrices. Rehovot 76100, Israel. ments to the membrane in epithelial t~) 1985, Elsevier Science Publishers B.V., Amsterdam 0376 - 5(~67/85/$02.(D TIBS - November 1985 457

and are present predominantly in the As pointed out, a more widely the linkage between F-actin and the cortical cytoplasm, subjacent to the plas- accepted mechanism for the anchorage membrane in intestinal microvilli. The malemma. Many biochemical and ultra- of actin may involve peripheral associa- advantage of this system which attracted structural studies indicated that tion with the cytoplasmic faces of the the attention of Mooseker and Tilney microfilaments are not only enriched membrane, mediated by specific integral over a decade ago 6, lies in its relatively near the plasma membrane but are often receptor(s) and peripheral linking- regular structure and the presence of actually attached to its endofacial sur- protein(s). only a few cytoskeletal components7. faces (see Ref. 1). The cell types and An excellent example which may help Electronmicroscope analyses of intesti- subcellular assemblies in which mem- to illustrate this mode of anchorage is nal epithelium indicated that a bundle of brane-microfilament interactions occur are however quite diverse in their origins and molecular properties. They include sites such as microvilli of polar- ized epithelia, lamellipodia and ruffling membranes, stereocilia of the sensory epithelium in the cochlea, contractile ring in dividing cells, end-on and lateral connections between myofibrils and the sarcolemma, dense plaques of smooth muscle, and intercellular junctions of the adherens type in a large variety of cells. In these locations, membrane-microfilament interaction probably affects such cardinal processes as cell morphogenesis, sensation, mobility, division, force generation, intercellular adhesion and cell communication. The involvement of membrane-microfilament interaction in such an important battery of cellular activities has motivated a general search for ubiquitous linkers of actin to the membrane. However, as one might have antici- pated on the basis of the apparent structural variability, direct biochemical and ¢. immunochemical data revealed con- b siderable heterogeneity in the molecular mechanisms of membrane anchorage in the different systems. One of the common features, however, was the presence of specific linker proteins capable: of bridging between F-actin and the membrane5. The necessity of accessory proteins for membrane anchorage has raised some controversy since the pos-. sibility has been considered over the', years that actin filaments may be', directly inserted into the lipid domain of the plasmalemma. This suggestion wa,; supported by occasional reports on the: presence of 'cell surface actin' or on the: effects of anti-actin antibodies on the physiology of living cells. This view, however, is usually regarded as unlikely or, at least, physiologically irrelevant; spontaneous penetration of a hydro- philic protein such as actin into the core of the membrane seems to be extremely unfavorable thermodynamically and the incidental immunocytochemical detec- tion of actin on the outer cell surface Fig. 2. The structure and molecular organization of intestinal microvilli. (a) and (b): Transmission could be attributed, in most cases, to an electronmicrographs of cross-sections (a) and longitudinal sections (b) through chicken intestinal micro- artefactual binding of actin released villi. (c) Scheme depicting the locations of the different cytoskeletal proteins of microvilli: F, fimbrin; V, from dead cells. villin; 110 kDa protein; C, calmodulin. The nature of the electron-dense material of the tip is not known. 458 TIBS - November 1985

microfilaments runs from the tip down villi belong to the third category of molecular processes involved in this type to the terminal web along the core of membrane microdomains formation of transmembrane linkage, very little is each microvillus (Figs 2a and b). This depicted in Fig. 1. Actin, in this case, is presently known. The difficulties bundle is associated with the membrane attached to the cytoplasmic faces of the encountered are partially attributable to in two distinct modes; at the tip, micro- plasmalemma without any indication for the limited spatial resolution of the filaments are apparently attached corresponding local specialization at the methods used, to the possibility that dif- end-on to the membrane through an exofacial cell surfaces. In these systems ferent molecular interactions occur at electron-dense plaque, while along the one may assume that the exact topology the various phases of the process and to microvilli periodic cross-bridges extend and timing of membrane-microfilament the many ligands and actin-associated from the core bundle to the lateral rhem- contact formation (of which we know proteins which might be involved. None- brane. Biochemical analyses of isolated nearly nothing) are intrinsically deter- theless, it has been shown that several microvilli revealed that beside actin mined within the cells. In the next sec- actin-associated proteins such as myosin there are only four major cytoskeletal tion I discuss other systems in which and ct-actinin accumulate under surface (detergent insoluble) proteins: villin (95 anchorage of actin filaments to the patches and caps, although the exact kDa), fimbrin (68 kDa), a 110 kDa pro- membrane is locally induced by an nature of their involvement is still not tein and calmodulin. Selective extraction extracellular signal. clear. It is noteworthy that in our experi- combined with immuno-electronmicro- ence ct-actinin was the most prominent scope studies indicated that the lateral Transmembrane induction of component of the subpatches and sub- bridges connecting actin to the mem- membrane-microffiament association caps. The recent results which demon- brane consist of a complex between the Transmembrane induction of micro- strate the capacity of a-actinin to 110 kDa protein and calmodulin. Villin filament binding to the plasmalemma directly interact with natural and syn- and fimbrin, on the other hand, were appears to be a common way of trans- thetic membranes (see below) may thus localized within the core filament bundle ducting signals into cells. Two appar- be of great significance for understand- where they are probably involved in the ently interrelated systems are briefly ing the capping process and justify direct bundling of actin filaments or in their discussed here: the transmembrane experimental examination. Among the structural modulation (Fig. 2c). The interaction of ligand-clustered surface many yet unanswered questions are: Do exact mode of interaction of the 110 receptors with actin; and the formation receptor-ligand complexes become kDa protein with the membrane is not of adherens junctions of various forms. attached to actin via additional linking entirely clear yet; one view is that it The former process is initiated by the proteins (integral, peripheral or both)? associates with a specific integral mem- specific binding of di-, oligo- or poly- Do the various ligands perturb a normal brane protein with an apparent molecu- valent ligands (an antibody, lectin, flow of surface proteins? Are the dif- lar mass of 140 kDa (Ref. 8). An toxin, etc.) to an integral surface recep- ferent actin-binding proteins detected in alternative possibility proposed recently tor. Subsequent to the binding, the indi- subpatches and subcaps (cx-actinin, myo- is that the 110 kDa protein itself is an vidual ligand-receptor complexes sin) involved in the anchorage to the amphipathic molecule which is embed- become clustered and are eventually membrane, or in the energy-dependent ded in the lipid domain of the mem- endocytosed. In lymphoid cells the force-generating system involved in the brane and associated with the core actin 'patching' step is often followed by mobilization of the surface clusters dur- filaments through a cytoplasmic exten- aggregation of the complexes into caps ing capping and endocytosis? tion (Ref. 9 and see below). at one pole of the ceiiHL Early studies on Another system in which trans- The molecular identity of the mem- the capping phenomenon (nearly 15 membrane linkage to actin occurs is a brane linking proteins in intestinal years ago), indicated that cytoskeletai family of cell contacts known collectively microvilli seems to be quite unique to structures and, in particular, microfila- as adherens junctions. These sites may that cellular system but the principle also ments were somehow involved in the be of two major classes: contacts with seems to hold for other cell types. One process. Initially it was shown that drugs non-cellular surfaces such as basement fine example is that of mammalian which disrupt microfilaments such as membranes; or tissue culture substrates erythrocytes in which the cytoskeletal cytochalasin b either alone or in con- and direct intercellular attachments submembrane network is composed pre- junction with colchicine inhibit capping. between neighbouring cells. Several dominantly of spectrin filaments which Later, after immunocytochemical examples for adherens junctions as visu- are cross-linked by actin oligomers and methods were developed for visualizing alized by electron microscopy are shown several associated proteins. This dense cytoskeletal proteins, the relationships in Fig. 3. The distinctive property of all web of filaments is attached to the cyto- between newly-formed caps and patches adherens junctions is their association plasmic extension of integral membrane and the microfilament system could be with the microfilament system through a proteins (mainly band-3 protein) via the more directly appreciated. Such studies cytoplasmic, membrane-bound plaque. bifunctional membrane-linking protein, confirmed (both visually and by quan- Attempts made by us and others to ankyrin. Other systems of membrane titative methods such as fluorescence characterize the molecular nature of this microfilament contact, such as junc- recovery after photobleaching) that anchoring plaque indicated that it con- tional specializations, exhibit a similar membrane proteins which were initially tains a specific protein, named vin- general mechanism (see below) free to diffuse laterally became immo- culinH-13 and that c~-actinin is usually although the components involved are bilized after ligand-induced clustering enriched in its vicinity~4. Extensive again quite different and their molecular and were apparently associated with immunocytochemical studies indicated topology is less clear (see below). submembrane microfilaments. Unfor- that while vinculin was ubiquitous in Interaction of microfilaments with tunately, despite the efforts invested in adherens junctions, it was not detected membranes such as in intestinal micro- the detailed characterization of the at or near any other site of actin-mem- TIBS - November 1985 459

these sites and provided better under- standing of the common features of different adherens junctions and of their molecular heterogeneity. Intercellular adherens junctions, for example, have recently been shown to contain a 135 kDa surface glycoprotein which is apparently absent from contacts with non-cellular surfaces ~5. In contrast, the recently-discovered protein, talin ~6, is apparently present in the membrane- bound plaques of cell substrate contacts or in associations with connective tissues only and absent from intercellular junctions (for details see Ref. 17). Although the information so far available is still incomplete, several general conclusions may be drawn from what we already know: (1) the linkage of actin to the membrane in adherens junctions involves peripheral linking-proteins which are the components of the junctional plaque, including vinculin and talin; (2) these proteins are unique to contact sites and differ from the linker proteins in non-junctional areas; and (3) on the basis of kinetic studies as well as dissection of the junction to its molecular domains, we proposed that the assembly of the various junctional elements is a vectorial process, triggered by local contact with extracellular surfaces. Contact-induced changes in the plasma membrane could lead to binding of vinculin to the membrane and to the subsequent induction of actin bundle assembly in the same areas.

Fig. 3. Transmission electronmicrographs of different adherens junctions: (a) a cross-section through cultured chickens lens cells showing focal contact (FC) with the substrate with numerous submembrane microfilaments, as well as intercellular adherens junction (double arrowhead); (b) intercellular zonula adhaerens (ZA) between neighbouring intestinal epithelial cells; (c) intercellular fascia adhaerens (FA) between adjacent cardiac myocytes; (d) membrane-bound dense plaques (DP) of chicken gizzard smooth muscle. Notice the intercellular material between the two smooth-muscle cells. brane association in which no contact they contain three major molecular with exogenous substrates or cells was domains: (1) integral 'contact receptors' present, such as microvilli, contractile in the membrane; (2) peripheral, mem- ring, ligand-induced surface patches, brane-bound plaque at the endofacial etc. Unfortunately, not much is known surfaces of the contact area; and (3) a about the biochemical properties of vin- bundle of actin-containing microfila- culin and even previous reports suggest- ments (Fig. 4). Only limited information ing that vinculin can bind directly to is presently available on the molecular actin raise controversy. In addition, vin- constituents of the various domains. It is culins can bind in the test-tube to not known, for example, whether there another protein named talin, as well as are ubiquitous exogenous components to the junctional plaque, in an actin- which induce adherens junction forma- independent manner. tion and whether all adherens junctions Closer examination of adherens junc-. share the same 'contact receptors'. tions has shed some light on the molecu.- Neither is it clear how the different if R\ lar substructure and dynamics of these domains interact with each other at the Fig. 4. The major molecular domains o] adherens sites. Experimental manipulation of molecular level. Recent studies, how- junctions including a microfilament bundle (M), adherens junctions suggested that beside: ever, have shed light on additional com- membrane-bound plaque (P) and integral mem- adhesive extraceUular surface molecules ponents involved in the assembly of brane receptors (R). 460 TIBS - November 1985

may, under appropriate conditions, culin which is readily extractable in low- a become embedded in the lipid domain ionic strength buffers, the solubilization of the membrane, without the require- of the closely related metavinculin ment for additional integral membrane required the presence of both detergent constituents. This view is still not widely and highly concentrated salt solutions J accepted, nevertheless an increasing suggesting that the latter might be number of examples suggest that such a embedded in the lipid bilayer21. This mechanism may indeed exist and that suggestion is still based on indirect some actin-binding proteins might observation and direct biochemical evi- 4 become integrated in the membrane. dence to support the view is yet to be These include recent studies on supplied. et-actinin, metavinculin and the 110 kDa The last example for an integral actin b protein of microvilli. linker may involve the 110 kDa protein A few years ago, attempts were made of intestinal microvilli. As discussed to identify integral membrane proteins above, one view was that this protein in activated platelets using the hydro- mediates the binding of the core bundle phobic photoactive probe, iodonaphthyl of microfilaments to an integral mem- azide (INA). It was shown that beside brane component. Recent studies, how- known membrane constituents, INA ever, indicated that the 110 kDa, actin- / became specifically bound also to a 100 binding protein has several biophysical kDa polypeptide identified as platelet properties typical of integral membrane a-actinin~8. This suggested that during proteins including partitioning into the platelet activation, concomitantly with detergent phase of Triton X-114 and C dramatic shape changes and polymeriza- incorporation into artificial lipid ves- tion of actin, c~-actinin becomes avail- icles 9. able to the hydrophobic probes. One of the interpretations offered was that Concluding remarks ct-actinin might have become inserted The detailed molecular structure of into the lipid domain of the membrane. most of the systems studied so far, their This possibility was later corroborated j diversity and the mechanisms of their by important experiments in which pure assembly, are still poorly characterized. a-actinin was shown to interact in vitro Nevertheless, the more we know about with lipid bilayers containing dia- membranes and membrane dynamics, cylglycerol and palmitic acidlg.zL More- cell morphogenesis and mobility, etc., over these lipids could promote the Fig. 5. Two alternative mechanisms for the incor- the more we appreciate the major interaction between oc-actinin and actin. poration of an actin binding protein into the lipid physiological importance of these asso- bilayer of the membrane. The starting state in (a) Conceptually, these results imply, ciations. Of particular interest and shows a microfilament bundle (m]) associated with (though indirectly) that a cytoplasmic, importance is the recent suggestion that a linker protein (lp), both residing in the cyto- water-soluble protein such as a-actinin some cytoskeletal proteins whose dis- plasm. Interaction of the linker protein with the may, under specific circumstances, be membrane can be triggered by a modification in the tribution was formerly believed to be translocated from the cytoplasm and protein itself which renders it compatible with inser- restricted to the cytoplasm may, in fact, inserted into the membrane. It is still tion into the bilayer (b). Alternatively local changes translocate and partition into the mem- unclear what might trigger such trans- in the membrane may create membrane domains brane proper. Hopefully, future with a high affinity for the linker protein (c). location. Are there post-translational research directed at the various ques- changes in the a-actinin molecule itself tions discussed here, and the isolation of which lead to the exposure of hydro- new 'linker proteins', will shed light on phobic domains on the surface of the the mechanisms of membrane-micro- molecule, or are there changes within Direct insertion of cytoskeletal proteins filament interactions and their precise the bilayer leading to the binding of into the lipid domain of the plasma involvement in cellular process. membrane a-actinin? (see Fig. 5.) At least in plate- The view that actin commonly inter- lets the latter possibility is attractive in acts with membranes through other view of the documented turnover of References cytoplasmic 'linking proteins' which bind phospholipids which occurs upon activa- 1 Geiger, B. (1983) Biochim. Biophys. Acta 737, tion. 305 to integral membrane receptors implies 2 Singer, S. J. and Nicolson, G. L. (1972) that actin (a cytoplasmic component), Another relevant system involves the Science 175, 720 the 'linker' protein (a cytoplasmic/ smooth-muscle protein metavinculin 3 Fawcett, D. W. (1981) in The Cell, pp 1-35, peripheral membrane component) and which is a vinculin-like protein by anti- W. B. Saunders the receptor (an integral membrane genic and peptide map criteria with a 4 The Cytoskeleton (1981) Cold Spring Harbor component) interact without being molecular mass of about 160 kDa (larger Syrup. 46 5 Pollard, T. D. and Craig, S. W. (1982) Trends than vinculin by 30 kDa). Biochemical grossly translocated from their regular Biochem. Sci. 7, 55-58; 88-92 compartments within the cell. studies indicated that metavinculin is 6 Mooseker, M. S. and Tilney, L. G. (1975) An alternative possibility is that the quite resistant to extraction with con- J. Cell Biol. 67, 293 'linking protein(s)' or even actin itself ventional aqueous buffers. Unlike vin- 7 Bretcher, A. (1981) Cold Spring Harbor TIBS - November 1985 461

Syrup. Quant. Biol. 46, 871 12 Geiger, B., Tokuyasu, K., Dutton, A. H. and 17 Geiger, B.~ Volk, T. and Volberg, T. J. Cell. 8 Coudrier. E., Reggio, H. and Louvard, D Singer, S. J. (1980) Proc. Natl Acad. Sci. USA Biol. (in press) (1981) Cold Spring Harbor Symp. Quant. 77, 4127 18 Rotman, A., Heldman, J. and Linder, S. Biol. 46, 881 13 Geiger, B. (1981) Cold Spring Harbor Syrup. (1982) Biochemistry 21, 1713 9 Glenny, J. R. and Glenny, P. (1984) Cell 37, Quant. Biol. 46, 671 19 Meyer, R. K., Schindler, H. and Burger, 743 14 Lazarides, E. and Burridge, K. (1975) Cell 6, M. M. (1982) Proc. Natl Acad. Sci. USA 79, 10 De Petris, S. (1977) in Dynamic Aspects of 289 4280 Cell Surface Organization (Poste, G. and 15 Volk, T. and Geiger, B. (1984) EMBO J. 3, 20 Burn, P., Rotman, A., Meyer, R. K. and Nicolson, G. L. eds), pp. 64, Elsevier/North 2249 Burger, M. M. (1985) Nature 314, 469 Holland 16 Burridge, K. and Connell, L. (1983) J. Cell 21 Siliciano,J. D. and Craig, S. W. (1982) Nature 11 Geiger, B. (1979) Cell 18, 193 Biol. 97, 359 and (1983) Cell Motility 3, 405 300, 533

called 'Disaggregation Theory of Hor- Programmable messengers: a mone Action '5. This theory is now extended and modified in the light of new theory of hormone action information acquired recently. M. Rodbell The disaggregation theory Briefly, this theory suggests that various classes of receptors are complexed Many hormone receptors are linked to GTP-regulatory proteins in membranes. When with a family of oligomeric GTP-regula- these proteins are activated by hormones and GTP, the ct-subunits are released from tory proteins. When the receptors are the membrane as soluble proteins. It is proposed that these et-subunits are modified by occupied by agonists and the G units by kinases, proteases and other protein-modifying enzymes to give new forms with GTP, the oligomers dissociate into differing functions. This proviaes a way of explaining the multiple actions of a monomers. In the process, the receptors hormone on its target cell, and the released et-subunits of G TP-regulatory proteins can are transformed from a high affinity be called 'programmable messengers'. state when they can bind physiological concentrations of hormones, into a low Two ideas have dominated the field of whether the pleiotypical responses affinity state in which they are no longer signal transduction over the past 2:5 induced by a hormone are due solely to active. At the same time, the G units are years. One is that hormone/neuro- any of these molecules. If not, what type transformed to a 'monomeric' structure transmitter receptors interact with vari- of molecule might be more closely that reacts specifically with an effector ous effector enzymes in the plasma linked to receptors that could serve as unit (E) such as adenylate cyclase. The membrane to generate signals in the primary messengers of hormone action? theory is thermodynamically soundr; it form of small molecules. The classical As for the concept of receptor mobility, explains the apparent paradox of recep- example is the receptor-controlled there is evidence that membrane receptors undergoing transitions from high to adenylate cyclase system in eukaryotic tors can be induced by agonists to move low affinity states during concerted cells. The other is that receptors exist about in the plane of the membrane. activation of G by hormone and GTP; it either in membranes or in the cytosol as However, there is no compelling evi- explains the findings of target analysis 'mobile' elements which, when com- dence that mobility is necessary or that the ground-state structure of recep- bined with the activating hormone, causal for signal transduction to take tors coupled to G exhibits a much higher induce the receptor to collide with or place. Indeed, there is a report that molecular weight than the activated move to the site(s) of the effector sys- increasing the fluid environment to adenylate cyclase. This theory predicts tems. Examples of theories that have enhance receptor mobility in mem- that the putative monomeric form of G evolved from the mobile-receptor branes is detrimental to hormone is the primary messenger of hormone theory are the 'collision-coupling'~ and action 3. For the estrogen receptor, action, whereas the product of the effec- 'two-step '2 theories proposed for the recent studies indicate that most of the tor unit(s) is a secondary signal. coupling of [3-adrenergic receptors to receptors are bound to the nuclear the adenylate cyclase system. Another matrix prior to their occupation by hor- G units are oligomeric proteins example is the estrogen receptor; it has mone; receptor release into the cytosolic In recent years, G units have been been thought that the receptor first compartment is an artifact of the purified and structurally analysed 7. It is reacts with the steroid in a cytosolic methods used for isolating the nucleus 4. now clear that G units coupled to rho- compartment, and that the activated This article proposes an alternative dopsin (termed transducin) and those receptor then enters the nucleus where it view of the function of membrane recep- coupled to receptors (R) that stimulate regulates gene expression. tors and develops a logical framework or inhibit adenylate cyclase (termed G~ There is ample evidence that cyclic for a theory that the primary messengers and G~, respectively), and a newly dis- AMP and other small molecules (cyclic of hormones acting on membrane recep- covered G unit of unknown action GMP and inositol trisphosphates are tors are proteins that bind and degrade (termed Go) are composed of three dis- recent examples) mediate some of the GTP. These are the so-called GTP-reg- tinct protein subunits, only one of effects of hormones. The question :is ulatory proteins (G) that are linked to which, the o~-unit, binds GTP. The type numerous receptor types in eukaryotic of ~t-subunit coupled depends on the M. Rodbell is at the National Institute of Environ- mental Health Sciences, Research Triangle Park, cells. The fundamental aspects were type of G unit (and associated R) to NC 27709, USA. presented five years ago in a theory which it is attached. The other two sub-