Upon stimulation, certain protein kinases, phos-phatases and other players in signal transduction relo-cate to membranes, cytoskeletal structures, scaffoldingproteins or organelles1–3. Here we take receptor tyro-sine kinase (RTK) signalling as our main example(Fig. 1). Stimulation of RTKs is linked to the activationof mitogen-activated protein kinase (MAPK) cascadesthrough a cytoplasmic protein Sos (a homologue ofthe Drosophila melanogaster ‘Son of sevenless’ protein)and the small GTP-hydrolysing protein Ras, anchoredto the cell membrane4. Sos is a GDP/GTP exchange fac-tor that catalyses the conversion of inactive Ras (i.e.its GDP-bound form) to active Ras (its GTP-boundform). The adaptor protein Grb2 (growth-factor-re-ceptor-binding protein 2) mediates the binding ofSos to activated RTKs, such as the epidermal growthfactor receptor (EGFR). Grb2 binds to the activatedEGFR directly or through another adaptor protein,tyrosine-phosphorylated Shc (src homology andcollagen domain protein). The EGFR does not phos-phorylate Sos, nor does the catalytic activity of Sostowards Ras change upon Sos binding to the receptor5.When Sos is recruited to the membrane by activatedEGFR, Sos can interact with the membrane poly-phosphoinositides through an N-terminal pleckstrinhomology (PH) domain.This interaction pattern raises a number of ques-tions about the role of the plasma membrane re-location in signal transduction. Why should theGrb2–Sos complex bind to the membrane receptorif Sos catalytic activity is not activated by the recep-tor? What prevents direct interaction of cytosolicSos with the membrane-bound Ras from activatingthe latter? Why is Ras anchored to the membrane?Should anchoring itself be a regulatory event? Whatis essentially different in the activated versus thenonactivated RTK? It has been proposed that the re-cruitment of Sos to the proximity of the membrane-bound Ras is a key feature in the activation of Ras by phosphorylated EGFR6–8. But what does recruit-ment mean? If it means that Sos is first bound to theEGFR and then moves to Ras by two-dimensionaldiffusion, then why should this accelerate signaltransduction? Cytosolic Sos still requires the sameamount of time to reach the EGFR. Binding to EGFRwould slow down its diffusion unless Sos dissociatedagain, but then Sos would escape back to the cytosolrather than bind to Ras.To clarify the effect of membrane localization, weconsider two extremes. When two protein moleculesform a productive complex (i.e. transduce the signal)after each diffusive encounter, the signal-transductionprocess is ‘diffusion-limited’. If only a small fractionof the collisions leads to binding that lasts longenough to transfer the information, the signal trans-duction is ‘reaction-limited’. In this case, the reactionrate is controlled by the alignment of reactive patchesin the correct orientation or by the intrinsic chemicaltransformation rather than by the Brownian colli-sions of the molecules. The two protein molecules thenassociate and dissociate several times before signaltransduction takes place. We will now analyse theconsequences of the membrane translocation fordiffusion- and reaction-limited signal transduction.Does membrane localization enhance diffusion-limited signal transduction?Adam and Delbrück suggested that the reductionin dimensionality might enhance reaction rates be-tween solutes that bind to membranes and membrane-bound species9; the solutes should not get lost bywandering off into the bulk phase. The relevance andmagnitude of this enhancement has been studiedextensively in various biological systems10–12.Conservative estimates can be made of the timetaken by signal transduction proteins in the cellFORUMhypothesistrends in CELL BIOLOGY (Vol. 10) May 2000 0962-8924/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved. 173PII: S0962-8924(00)01741-4Why cytoplasmicsignalling proteinsshould be recruitedto cell membranesBoris N. Kholodenko, Jan B. Hoekand Hans V. WesterhoffIt has been suggested that localization of signal-transductionproteins close to the cell membrane causes an increase in their rateof encounter after activation. We maintain that such an increasein the first-encounter rate is too small to be responsible for trulyenhanced signal transduction. Instead, the function of membranelocalization is to increase the number (or average lifetime) ofcomplexes between cognate signal transduction proteins and henceincrease the extent of activation of downstream processes. This isachieved by concentrating the proteins in the small volume of thearea just below the plasma membrane. The signal-transductionchain is viewed simply as operating at low default intensity becauseone of its components is present at a low concentration. The steadysignalling level of the chain is enhanced 1000-fold by increasingthe concentration of that component. This occurs upon ‘piggyback’binding to a membrane protein, such as the activated receptor,initiating the signal-transduction chain. For the effect to occur, theprotein translocated to the membrane cannot be free but has toremain organized by being piggyback bound to a receptor, membranelipid(s) or scaffold. We discuss an important structural constraintimposed by this mechanism on signal transduction proteins thatmight also account for the presence of adaptor proteins.FORUMhypothesis174 trends in CELL BIOLOGY (Vol. 10) May 2000membrane or in the cytosol to encounter their firstpartner molecule by free diffusion13. A spherical cellwith a radius of 10 mm has a surface area of 1260 mm2.If it contains 10 000 copies of each signal transductionprotein in its membrane, then at 0.35 mm spacingthe protein molecules occupy the entire cell surface.Partners in signal transduction should then be ap-proximately L 5 0.25 mm apart. The average time forthem to meet a neighbour should be approximatelyL2/2D, where D is the lateral diffusion coefficient. Asthe membrane diffusion coefficient of the protein isapproximately 10–9–10–10cm2/s (Refs 14 and 15), itshould take about 0.3–3 s before the partner proteinshit each other when diffusing in two dimensions. Forthree-dimensional signal transduction with 10 000proteins per cytosol, the partner proteins will be onaverage 0.6 mm apart. Using a diffusion coefficient of10–8cm2/s for cytosol diffusion (see Ref. 16 and refer-ences therein), this leads to a time of L2/(3D) 5 0.1 s,which is faster, not slower, than the 0.3–3s for