096363u116 Cell, Vol. 100, 113–127, January 7, 2000, Copyright 2000 by Cell Press Signaling—2000 and Beyond Review a large family, and, indeed, with the recent addition ofTony Hunter The Salk Institute odorant receptors, this has become by far the largest receptor family, numbering over a thousand. Analysis10010 North Torrey Pines Road La Jolla, California 92037 with purified components showed that the liganded re- ceptor interacts directly with the heterotrimeric G pro- tein leading to the exchange of GTP for GDP bound toShips that pass in the night and speak the a subunit, which thereupon dissociates from theeach other in passing b/g subunits, allowing both the a.GTP and the b/g com-Only a signal shown and distant voice plexes to signal to downstream effectors, such as ade-in the darkness nylyl cyclase, again by direct interaction. Sustained sig-—Henry Wadsworth Longfellow, naling in response to a stimulus is generally undesirable,“The Theologian’s Tale” (1863) and therefore mechanisms for signal termination are required. The study of G proteins revealed that signalingThe Past is turned off when the a subunit hydrolyzes the boundWhen the history of signal transduction is written in GTP, either spontaneously or upon interaction with a2100, what will be remembered of the exciting advances GTPase activating protein (GAP), permitting the b/gthat occurred in the latter half of the twentieth century? complex to rebind.Everyone will surely have their own list of landmark find- Transmembrane Signaling by Phosphorylationings that have contributed to our current picture of signal During the 1980s and 1990s several other distinct princi-transduction. Undoubtedly there are seminal earlier ples of transmembrane signaling have been uncovered.events, but from my own perspective this era began The discovery of a new class of protein kinases associ-with the discovery in the mid-1950s that phosphorylation ated with the polyomavirus, v-Src and v-Abl viral trans-can reversibly alter the activity of an enzyme through forming proteins that phosphorylate tyrosine immedi-the combined action of a protein kinase and a protein ately suggested that tyrosine phosphorylation plays aphosphatase (Krebs and Beavo, 1979). At much the role in growth control (Hunter and Cooper, 1985). Tyro-same time the hormones adrenalin and glucagon were sine phosphorylation was quickly revealed as a majorfound to increase the level of intracellular 3959 cyclic mechanism of transmembrane signaling. The demon-AMP, and the concept of the second messenger was stration that EGF stimulated tyrosine phosphorylationborn. About 10 years later the cAMP-dependent protein- serine kinase (PKA) was isolated as a target for cAMP, EGF receptor in membrane preparations led to the iden- tification of a large family of ligand-stimulated receptorand through its pleiotropic substrate specificity PKA was shown to be responsible for many of the effects of protein-tyrosine kinases (PTKs). Receptor PTKs are all type I transmembrane proteins with a cytoplasmic do-cAMP. With these findings the field of signal transduc- tion was born. main that has intrinsic catalytic activity that is activated upon ligand binding. This established that intracellularSince these early days, progress in understanding mechanisms of signal transduction has been astonish- protein phosphorylation can be used as a direct means of transmembrane signal transduction. Another impor-ingly rapid. In the past 40 years many important themes and principles of signal transduction have emerged. The tant step forward in understanding how signals are transmitted across the membrane came with the discov-highly conserved nature of eukaryotic signaling path- ways has been revealed, and defects in signaling have ery that receptors lacking intrinsic catalytic activity can be coupled to nonreceptor PTKs via noncovalent asso-been found as the underlying basis of cancer and other human diseases. Progress has benefited tremendously ciation with the cytoplasmic domain of a receptor sub- unit, thus forming “binary” receptors (Neet and Hunter,from structural and genetic analysis, the development of analytical techniques and reagents, and the availabil- 1996). For instance, cytokine receptors use members of the JAK PTK family. The discovery of transmembraneity of pharmacological modulators of signaling. What are some of the major milestones? protein-serine kinases, first in plants and then as recep- tors for TGFb family cytokines in vertebrates, providedG Protein–Coupled Receptors The identification of G proteins and their role in activat- another example of the use of ligand-induced intracellu- lar protein phosphorylation as means of transmembraneing membrane bound adenylyl cyclase to synthesize cAMP in response to hormonal stimulation in the mid- signal transduction. The fact that receptor PTKs proved to be type I trans-1970s was another great step forward in understanding transmembrane signal transduction (Gilman, 1987). The membrane proteins raised the conundrum of how a the- oretically flexible protein with a single transmembranestudy of G proteins revealed the principle that hydrolysis of protein-bound GTP could act as a signaling switch, domain could propagate a signal across the membrane in response to ligand binding. The seminal finding wasand also brought us very near to the membrane receptor itself. The cloning of hormone receptors that are coupled that activation of receptor PTK cytoplasmic catalytic domains requires ligand-induced dimerization (Schles-to adenylyl cyclase as well as several neurotransmitter and drug receptors revealed that they all had a close singer, 1988). This juxtaposes the two catalytic domains allowing mutual transphosphorylation of residues in therelationship to rhodopsin, the seven transmembrane do- main G protein–coupled light receptor. This suggested activation loop of the catalytic domain, leading to enzy- matic activation, and autophosphorylation of tyrosinesthat serpentine G protein–coupled receptors would be Cell 114 Figure 1. Protein Modules and Signal Trans- duction This figure of a cell, showing how modular protein and lipid interaction domains are used in a variety of cell signaling pathways, has been modified from a concept provided by Tony Pawson. outside the catalytic domain, which are the key to down- With the identification of the SH2 domain came the stream signaling. Receptor protein-serine kinases are realization that most protein–protein interaction do- also activated by ligand-induced dimerization. Subse- mains involved in signal transduction are modular in quently, ligand activation of many types of surface re- nature (Figure 1). The study of signaling initiated by ceptor, including those that do not activate protein kinases receptor PTKs and other protein kinases has uncovered directly, has been found to involve oligomerization. a plethora of different protein interaction domains, rang- Protein Signaling Modules ing in size from 40 to 150 residues, used for inducible or The discovery of the SH2 domain as a means of recog- constitutive interactions involved in signaling (Pawson, nizing specific phosphorylated tyrosines in activated 1995). Examples are SH2, PTB, SH3, WW, FHA, SAM, receptor PTKs and receptor PTK targets was a break- LIM, PX, EH, EVH1, and PDZ domains. Like the SH2 through in understanding how activated PTKs propa- domain, most of these domains recognize short linear gate signals, since it revealed how the association of sequences from 4 to 10 amino acids in length, in some two proteins could be induced by phosphorylation thus cases requiring phosphorylation of a specific Ser/Thr propagating a signal (Pawson and Gish, 1992). Indeed, or Tyr within the recognition sequence, thus providing this discovery illustrated a totally new function for pro- inducible association. These domains fold indepen- tein phosphorylation, namely the regulation of protein– dently and their N and C termini protrude from one side protein association. Other types of phosphotyrosine of the domain with the interaction surface on the oppo- (P.Tyr)-binding domain have subsequently been found site side, allowing them to be assembled almost like (e.g., PTB domains), but SH2 domains are the most beads on a string (Kuriyan and Cowburn, 1997). prevalent type of P.Tyr-binding domain involved in sig- Another important step forward was the discovery naling downstream of activated receptor PTKs. SH2 do- that domains involved in membrane signaling can recog- mains bind in a sequence-specific fashion, recognizing nize molecules other than proteins. For instance, plecks- one or more residues in positions 1–6 C-terminal to the trin homology domains (PHD), found in many signaling P.Tyr. PTB domains recognize residues up to 5 away proteins, recognize specific phospholipids and thus on the N-terminal side of P.Tyr, but only a subset of allow inducible membrane association dependent on PTB domains (e.g., the Shc and IRS-1 PTBs) bind to their the formation of lipid second messengers (Ferguson et target proteins in a phosphorylation-dependent manner al., 1995). For example, the PHDs of the PDK1 and Akt/ (van der Geer and Pawson, 1995). PKB protein-serine kinases bind PIP3 and this promotesSH2 and PTB domain interactions are used as a means their membrane recruitment in response to PI-39 kinaseof recruiting target proteins to activated PTKs, thus per- activation and leads to activation of Akt/PKB by PDK1.mitting their phosphorylation, and also for translocation Ca21 elevation can regulate membrane association ofto the plasma membrane, where many effector proteins proteins via Ca21-dependent interaction of C2/CalB do-activated by receptor PTKs, such as phospholipase Cg mains with phospholipids.and PI-39 kinase, have their substrates (Schlessinger, Recently, phosphoserine (P.Ser)-binding domains have1994). SH2/PTB domains are present not only in proteins also been identified, and these too play roles in signalwith intrinsic enzymatic activity that can be regulated propagation, by facilitating phosphorylation-dependentby tyrosine phosphorylation, such as phospholipase Cg, protein–protein interactions (Yaffe and Cantley, 1999).but also in so-called adaptor proteins, such as Grb2, The first example was the phosphorylation-dependentthat bind to and thereby bring effector enzymes to the binding of the CBP coactivator protein to the CREBplasma membrane. Receptor-induced recruitment of transcription factor when phosphorylated at Ser133.proteins to the plasma membrane as a mechanism of Subsequently, other families of P.Ser/P.Thr-binding do-activating signaling pathways has emerged as a major theme in signaling. mains have been found. Review 115 Intracellular Signaling Pathways translocates to the nucleus. An important step in under- standing the cAMP pathway was the identification ofFollowing the identification of the transmembrane re- ceptors, and their proximal targets, intracellular signal- the CREB transcription factor, which binds to cAMP response elements in inducible genes, and the demon-ing pathways were the next to come on stage. The emer- gence of a large family of small monomeric G proteins, stration that it is a nuclear target for PKA. Once the C subunit enters the nucleus it phosphorylates CREB atheralded by Ras, was an important step in understand- ing transmembrane signaling. Ras is anchored via C-ter- Ser133, triggering binding of the CBP/p300 coactivator and transcription of cAMP-responsive genes. This find-minal lipid modifications to the inner face of the plasma membrane, and, like heterotrimeric G proteins, is acti- ing reinforced the notion that transcytoplasmic signaling pathways could use transcription factor phosphoryla-vated by GTP-GDP exchange catalyzed by GTP ex- change factors. Ras.GTP levels are increased upon tion as a regulatory mechanism. Many transcription fac- tors are now known to be directly regulated by phos-activation of many types of receptor. A key to under- standing how Ras is activated was the finding that the phorylation, through positive or negative control of nuclear import or export, DNA binding, or transactivationCdc25-related Sos protein has Ras.GTP exchange fac- tor activity, and that the Grb2 SH2/SH3 adaptor protein activity (Karin and Hunter, 1995). MAP Kinase Pathwaysis constitutively associated with Sos via its SH3 domains (Schlessinger, 1993). The Grb2/Sos complex is recruited Two separate lines of inquiry, namely analysis of protein- serine kinases activated by receptor PTKs and the studyto an activated receptor PTK upon binding of the Grb2 SH2 domain to an autophosphorylated tyrosine in the of transcription factor phosphorylation, converged to establish that the MAP kinase pathway is a major mech-appropriate sequence context, thus bringing the Sos catalytic domain into proximity with Ras at the plasma anism for controlling transcription in eukaryotes (Seger and Krebs, 1995). MAP kinase was originally discoveredmembrane and stimulating GTP exchange. Once acti- vated, Ras.GTP interacts with a series of effector pro- as an insulin-activated protein-serine kinase, and bio- chemical studies, reinforced by genetic analysis of theteins, including the Raf protein-serine kinase and PI-39 kinase, which initiate downstream signaling (Witting- pheromone response in budding yeast, showed that this pathway consists of a cascade of three protein kinases,hofer and Nassar, 1996). Analysis of Ras function also led to the discovery of the first GAP, which stimulates a MAP kinase kinase kinase (MAPKKK), a MAP kinase kinase (MAPKK), and a MAP kinase (MAPK) (WaskiewiczGTP hydrolysis and thus acts to terminate Ras signaling (McCormick, 1989). Other small G proteins also function and Cooper, 1995). These protein kinases are activated in series, such that the MAPKKK phosphorylates theas signaling switches that are activated by receptor- induced GTP exchange. For instance, the Rho family MAPKK at serines in its activation loop, which is thereby activated and phosphorylates the MAPK at a threoninesmall G proteins, Rho, Cdc42, and Rac, are all activated via receptor signaling pathways, and, like Ras, induce and tyrosine in its activation loop, leading to its activa- tion. Every eukaryotic organism has multiple MAPKdivers signaling pathways, which lead to gene expres- sion and, perhaps more importantly, cause dramatic pathways, which are largely separate from one another. The elucidation of the modular MAPK cascade with itsrearrangements in the actin cytoskeleton. Ras was the first signaling protein whose function was three consecutive protein kinases had immediate impli- cations for understanding signal amplification and switch-shown to be conserved from yeast to vertebrates, and this presaged the identification of many highly con- ing, and most importantly for nuclear signaling, because the terminal MAPK, once activated, can migrate into theserved signaling pathways throughout the eukaryotic kingdom. A prime example is the receptor PTK-Ras- nucleus, and there phosphorylate and activate tran- scription factors.MAP kinase pathway, where genetic and biochemical analysis has revealed that this pathway exists in essen- Nuclear Translocation of Transcription Factors Receptor-induced nuclear translocation of a latent cyto-tially identical form in species ranging from nematodes to vertebrates, with most of the components being func- plasmic transcription factor into the nucleus emerged as a different transcytoplasmic signaling principle. Thetionally interchangeable between organisms. Transcriptional Regulation by Surface Receptors NF-kB activation pathway provided the first example of a cytoplasmically sequestered transcription factorvia Transcytoplasmic Signaling Upon binding ligand, many receptors invoke gene ex- (Karin and Hunter, 1995). In this case, NF-kB is held in the cytoplasm through binding to an inhibitor protein,pression responses. Indeed, the discovery that growth factors induce de novo expression of a specific set of IkB, which masks the nuclear localization signal in the NF-kB heterodimer. Activating stimuli induce phosphor-genes independent of new protein synthesis led to the search for transcytoplasmic signaling pathways that ylation of IkB, at specific sites, which targets it for ubiqui- tin-mediated degradation, thus releasing active NF-kBregulate transcription (Karin and Hunter, 1995). In the past 10 years, there has been spectacular progress in to migrate into the nucleus. The simplest pathway of this sort is the JAK/STAT system, in which ligand bindingunderstanding how plasma membrane signals are trans- mitted to the nucleus. The first and simplest transcyto- to a cytokine receptor activates associated JAK family PTKs, which first phosphorylate receptor subunits andplasmic nuclear signaling mechanism to be identified was the PKA/CREB system (Montminy et al., 1990). In then specific STAT transcription factors, whereupon the STATs dimerize, migrate to the nucleus and activatethis pathway, cAMP, elevated in response to receptor stimulation, activates the R2C2 PKA holoenzyme local- transcription (Karin and Hunter, 1995). Other examples are activation of the Wnt/Frizzled pathway, which resultsized in the cytoplasm by binding to the regulatory (R) subunits, thus releasing the catalytic (C) subunit which in translocation of b-catenin/LEF1 transcription factor Cell 116 complexes into the nucleus, and activation of TGFb fam- 1989). The discovery of calmodulin as a major Ca21- ily receptors, which leads to phosphorylation of recep- sensing protein in the cells, and identification of protein tor-specific Smad transcription factors, which then as- targets for Ca21/calmodulin complexes, including a fam- semble with the common Smad4 subunit, translocate ily of Ca21/calmodulin activated protein kinases, has into the nucleus, and induce transcription of target provided us with an explanation for how Ca21 release genes. is translated into molecular consequences. Moreover, Signaling via Notch family receptors utilizes the novel the development of cell-permeant Ca21 sensor fluoro- principle of ligand-induced proteolytic cleavage releas- phores has afforded us with the most detailed spatio- ing the Notch cytoplasmic domain, which acts as a pro- temporal picture of signaling events occurring in the tein second messenger that migrates to the nucleus and cell. The intricate and dynamic real-time patterns of local regulates gene expression by binding into and con- and global Ca21 release and resorption and the propaga- verting the suppressor-of-hairless transcriptional re- tion of Ca21 waves and oscillations are surely the harbin- pressor into a transcriptional activator (Artavanis-Tsa- ger of how other signaling systems work at the subcellu- konas et al., 1999). lar level. Nuclear Receptors New Second Messengers The finding that lipid-soluble ligands, such as retinoic With cAMP, cGMP, and IP3 the concept of the second acid, estrogen, and other hormones, can traverse the messenger was well established. However, the discov- plasma membrane without utilizing surface receptors ery that NO is the endothelial cell–derived relaxing factor and induce cellular responses by binding to and activat- for vascular smooth muscle and that NO activates solu- ing members of a family of zinc finger transcription fac- ble cytoplasmic guanylyl cyclase to elevate cGMP re- tors revealed yet another nuclear signaling mechanism vealed, somewhat unexpectedly, that a gas could act (Mangelsdorf et al., 1995). These so-called nuclear re- as a second messenger (Murad, 1994). This finding re- ceptors prove to be a huge family that transduce tran- kindled interest in cGMP-mediated signaling, which is scriptional responses to an astonishing variety of chemi- mainly effected by the cGMP-dependent protein kinase, cals. In principle, the nuclear receptors represent the although there are other targets for cGMP. simplest possible mechanism for nuclear signaling, al- Signaling in Prokaryotes and Plants though analysis of how ligand binding activates recep- Although much of the emphasis in signal transduction tor-mediated transcription of target genes reveals a sur- has been on eukaryotes, prokaryotes should not be ne- prisingly complex machinery in which the unliganded glected in this litany of progress. For instance, an impor- receptor heterodimer acts as repressor as a result of tant new signaling principle was established with the its association with histone deacetylases, and ligand discovery of the two component histidyl-aspartyl phos- binding converts the receptor into an activator by trig- phorelay systems, comprised of a sensor “histidine” gering dissociation of histone deacetylases and recruit- protein kinase that phosphorylates a response regulator ment of histone acetylases. on an aspartate residue (Stock et al., 1990). These sys- Phospholipid- and Ion-Based Signaling tems allow prokaryotes to respond to a wide variety of The discovery in the 1960s of the phosphatidylinositol extracellular stimuli. However, somewhat surprisingly, (PI) cycle, which results in stimulus-induced turnover of two component systems are rare in eukaryotes, and PIP2, was the harbinger of the finding of a series of for the most part seem to have been superseded by phospholipid-derived second messengers, including di- conventional protein kinases. acylglycerol (DAG), IP3, and PI3,4,5P3, and more recently Likewise, although animals and fungi have received IP4 and IP6 (Liscovitch and Cantley, 1994). The identifica- the lion’s share of attention, plants have also taught us tion of protein kinase C as a DAG-regulated enzyme new principles in signal transduction. For instance, the provided another example of a second messenger-regu- first receptor protein-serine kinases were found in lated protein kinase. Phospholipase action on PIP2 not plants, and one type of leucine-rich receptor kinase may only generates DAG but also IP3, which acts as a second be directly regulated by plant steroid family ligands. messenger to release Ca21 from intracellular stores, thus Moreover, ethylene, an important plant hormone, was triggering a program of Ca21-activated events. The un- the first gas shown to induce cellular responses, and expected discovery of PI-39 kinases and the family of significant progress has recently been made in under- 39 phosphoinositides they generate led to the elucida- standing how ethylene signals.tion of a new phospholipid-based signaling system in which proteins are recruited via PIP3-binding PH do- The Futuremains to the membrane where they are activated (e.g., What does the future hold? There are many areas inthe Akt/PKB protein-serine kinase) (Kapeller and Cant- signaling where it is safe to predict that rapid progressley, 1994). will be made, but because of its preeminence as a mech-The regulated entry and exit of ions across the plasma anism of signal transduction, most attention will be paidmembrane, through ligand-gated, voltage-sensitive and to protein phosphorylation and its role in intracellularstretch-activated ion channels and via ion pumps, has signaling.important signaling functions, particularly in impulse Modular Protein Interaction Domains in Signalingpropagation in the nervous system and in muscle con- A central theme in receptor-initiated signaling is the usetractility. Receptor-induced changes in cytoplasmic and of modular protein–protein interaction domains, eithernuclear Ca21 levels as a result of release of membrane- for inducible or constitutive interactions (Pawson, 1995).bound Ca21 stores through IP3 or entry of extracellular Well over a hundred putative protein domains have beenCa21 through plasma membrane ion channels constitute an important signaling mechanism (Berridge and Irvine, defined by comparative sequence analysis (http://smart. Review 117 EMBL-Heidelberg.de). Although the function of many of signaling complexes. FHA domains bind pTXXD motifs these domains is known, many remain uncharacterized in proteins, such as Rad9p involved in DNA damage and a majority of them could prove to be new protein– responses. One group of WW domains bind P.Ser/ protein interaction domains utilized in signaling path- P.Thr.Pro motifs in mitotic phosphoproteins and in the ways. Partners for these domains can be identified by RNA polymerase II large subunit CTD. A subset of WD40 two-hybrid screens and affinity methods combined with domains and leucine-rich repeats can also recognize the use of degenerate peptide libraries to establish a P.Ser. Given the great potential and specificity of phos- binding consensus sequence. Several phospholipid in- phorylation-induced protein interaction as a signal trans- teraction domains, including PH and FYVE domains, are fer mechanism, it is a good bet that additional P.Ser/ already known that provide inducible membrane associ- P.Thr-binding domains will be discovered. ation in response to generation of lipid second messen- Induced Protein Proximity in Signaling gers (e.g., PIP3) or stabilize membrane association of Another overriding principle that has emerged from anal- signaling proteins via binding to constitutive membrane ysis of receptor-activated signaling pathways is that phospholipids. Given the importance of membrane re- signaling can be initiated and propagated using the prin- cruitment in signaling, additional lipid recognition do- ciple of induced protein proximity (Austin et al., 1994). mains seem certain to be discovered. Identification of As described above, ligand binding to receptor PTKs nonprotein targets for such domains should be aided initiates signaling by causing dimerization, which facili- by affinity chromatography combined with mass spec- tates transphosphorylation. Recruitment of proteins trometric analysis. with P.Tyr-binding domains to activated membrane- The finding that short linear motifs were sufficient localized receptors where they can be phosphorylated for recognition by protein interaction domains came as or act on membrane targets is another example. This something of a surprise, given the complex nature of mechanism elevates the local concentration of proteins the subunit interaction surfaces found in multisubunit at the membrane thereby increasing the number of pro- proteins. However, the fact that recognition only re- ductive signaling complexes. SH2 domain binding also quires such short sequences combined with the func- greatly increases the affinity of these proteins as sub- tional independence of these domains presumably strates for phosphorylation by activated receptor PTKs. facilitated the rapid genesis of new protein–protein A dramatic illustration of the importance of protein prox- interactions during evolution. Indeed, how easily these imity has been afforded by the use of artificial ligand- domains could be duplicated and used to evolve new binding domains fused to signaling proteins that can signaling proteins is illustrated by the experimental por- then be induced to oligomerize by membrane-permeant tability of these domains. For example, replacement of dimeric chemical ligands. In several cases, this has been the C terminus of the Sos Ras activator with the Grb2 found to trigger the activation of signaling pathways just SH2 domain yields a fusion protein that largely rescues as efficiently as extracellular ligands (Spencer et al., the defects in Grb22/2 ES cell differentiation (Cheng et 1993). al., 1998). The ability to make chimeric signaling proteins Induced protein proximity is used to activate signaling of this sort to test hypotheses about specific signaling pathways by many other receptor systems, particularly connections will be increasingly useful as an analytical the TNF, IL-1, Toll, and death receptors, where initial tool. signaling events are critically dependent on oligomeric Phosphorylation-Dependent Protein– protein–protein interactions. The subset of TNF family Protein Interaction receptors that promote cell death use homotypic inter- In just 10 years it has become apparent how widely used actions between death domains, death effector domains sequence-dependent P.Tyr recognition is for tyrosine and CARD domains to trigger an intracellular signaling phosphorylation-mediated signaling. To add to the well- cascade that involves sequential proteolytic activation established SH2 and PTB domains, novel P.Tyr-binding of a series of proenzymes in the caspase protease fam- domains have recently been identified in c-Cbl, IRS-2, ily. In death receptor-induced caspase activation, pro- and Gab1. Interestingly, structural analysis indicates tein proximity is used to accentuate the low basal that the P.Tyr-binding domain in c-Cbl is a variant SH2 catalytic activity of a procaspase leading to efficient domain (Meng et al., 1999), whereas the Gab1 P.Tyr- autoprocessing in the context of an oligomer (Salvesen binding domain appears to have a novel structure. Addi- and Dixit, 1999).tional P.Tyr-binding domains are likely to be identified, A new concept in protein proximity has emerged fromalthough they probably will not be large families. A dis- the study of signaling by receptor PTKs. Although ittinct type of P.Tyr-binding domain is exemplified by the is well established that ligand-induced dimerization is“anti-phosphatases”, which are proteins structurally re- required for activation of receptor PTKs, recent evi-lated to the PTP or dual-specificity phosphatase fami- dence suggests that dimerization per se may not belies, but which lack one or more of the residues essential sufficient for activation, and that there is an additionalfor catalytic activities, and therefore can bind but not requirement for the relative orientation of the two dimerhydrolyze P.Tyr-containing proteins (Hunter, 1998a). subunits to be such that the catalytic domains are cor-Although it has only recently been appreciated that rectly juxtaposed (Jiang and Hunter, 1999). In fact, thereP.Ser/Thr can also be recognized by modular binding is good evidence in some systems that receptor dimersdomains, three families of sequence-specific P.Ser/Thr- can exist in the absence of ligand, and that ligand-binding domains are already known (Yaffe and Cantley, induced conformational switches in preformed recep-1999). 14-3-3 proteins, which recognize RSXpSXP and tors are critical for activation of the catalytic domains.RXY/FXpSXP motifs, bind many protein kinases, and as a result of their dimeric nature can potentially assemble In this context, an understanding of the way in which Cell 118 Figure 2. Ribbon Diagram Showing the Struc- ture of the c-Src Protein-Tyrosine Kinase in Its Inactive State Bound to AMP-PNP This figure is reproduced with permission from Xu et al. (1999). multimeric ligands are presented to receptors will be Structural analysis of protein kinase and phosphatase catalytic domains bound to inhibitors and protein inter-increasingly important. Structural Analysis of Signaling Proteins action domains bound to ligands is now a routine step in the development of drugs that inhibit activity or blockMany of the most important advances in understanding signaling pathways initiated by tyrosine phosphorylation interactions, for therapeutic use in diseases where sig- naling is deregulated. The structures of protein kinaseshave come through the determination of the three- dimensional structures of PTKs and PTPs, their immedi- bound to protein (e.g., p16/Cdk6) and chemical (e.g., FGFR1 receptor PTK bound to SU5402 and PD173074)ate targets, including modular domains such as SH2, SH3, PTB and PH domains, and downstream components, inhibitors have provided insights into how inhibition is brought about and how selectivity is achieved, and thisincluding protein-serine kinases and phosphatases. Structural analysis of signaling proteins has already promises to accelerate the pace of development of spe- cific small molecule or peptide-based inhibitors (Mo-revealed new and unexpected principles of regulation. For instance, the structures of several protein kinases hammadi et al., 1997, 1998; Russo et al., 1998). More- over, based on inhibitor-bound protein kinase structureshave shown how the “activation” loop acts as a negative regulator of activity, by blocking ATP and/or substrate it is possible to make mutant protein kinases resistant to a specific inhibitor that can be expressed and usedbinding, and how rotation of the C helix in the N-terminal lobe is used as a regulatory principle. The structure of to pinpoint which responses are a direct consequence of that protein kinase (Eyers et al., 1999).the c-Src PTK in its inhibited P.Tyr527-phosphorylated state revealed that the SH2 and SH3 domains bound to Structures of most of the well-characterized protein and lipid interaction domains have been generated.P.Tyr527 and the SH2-catalytic domain linker, respec- tively, lie on the backside of the catalytic domain, instead Comparative analysis of multiple examples of these do- mains bound to their ligands has given us importantof blocking substrate access to the catalytic cleft as had been expected (Figure 2). In fact, c-Src activity is insights into ligand binding specificity. Indeed, analysis of SH2 domains bound to P.Tyr-containing peptides hasinhibited indirectly as a result of a conformational change propagated through the small N-terminal lobe made it possible to make designer point mutations that alter SH2 domain sequence-binding specificity in a pre-that leads to rotation of the C helix and distortion of the active site (Xu et al., 1999). The dimeric structure of dictable way. Likewise, a point mutation in the WW do- main can transform its ligand specificity (Espanel andcatalytic domain 1 of RPTPa led to the prediction that receptor PTP activity can be negatively regulated Sudol, 1999). This heralds the possibility of designing unique signaling pathways with customized signalingthrough a symmetrical D1 dimer interaction in which a helix-turn-helix motif blocks access to the catalytic cleft components. An area of structural understanding that has lagged(Bilwes et al., 1996). Likewise, the N-terminal SH2 do- main was found to be bound to the catalytic cleft of the behind is the molecular basis of functional regulation by protein phosphorylation. There are very few struc-Shp2 PTP thus blocking substrate access, an interaction that is reversed upon the binding of a P.Tyr-containing tures of proteins in both their phosphorylated and un- phosphorylated states. The structures of the phospho-protein ligand to the SH2 domain leading to exposure of the active site (Hof et al., 1998). The structures of Ras and dephospho-forms of glycogen phosphorylase and isocitrate dehydrogenase kinase show that addition ofand trimeric G proteins bound to GTP and GDP have revealed how GTP hydrolysis results in a conformational a phosphate can induce a local conformational change increasing substrate affinity or a steric block to substrateswitch that blocks binding of effector proteins. Undoubt- edly additional regulatory principles will be identified binding, respectively. Several protein kinases structures reveal how activation as a result of phosphorylation ofthrough structural analysis. Moreover, as molecular docking programs become more sophisticated it may be residues in the catalytic domain activation loop occurs through a local conformational change that allows ATPpossible to predict which signaling proteins can interact. Review 119 and protein substrate access to the active site. Struc- particular, the PP2A protein-serine phosphatase has a large number of regulatory subunits that dictate sub-tures of many more proteins in their phosphorylated and strate specificity and localization. There are severalunphosphorylated states will be needed to determine known inhibitor proteins for PP1 (e.g., inhibitor 1 andthe structural principles underlying functional regulation inhibitor 2), some of which require phosphorylation toby phosphorylation. be active, and there is some evidence for PTP inhibitorProtein Kinase and Phosphatase Regulation proteins. PTPs are also regulated by phosphorylationProtein kinases can be regulated by second messen- and by subcellular localization. There is a large familygers, by allosteric mechanisms, by intrasteric inhibition of receptor-like PTPs whose activity could theoreticallythrough pseudosubstrate sequences, by positive and be regulated by ligands. In some instances receptor-likenegative phosphorylation events, by regulatory subunit PTPs appear to be negatively regulated by dimerizationbinding, by inhibitor proteins and by subcellular localiza- (e.g., RPTPa and CD45) (Majeti et al., 1998), in a mirrortion through targeting domains and anchoring proteins. image fashion to the activation of receptor PTKs byWill new mechanisms of protein kinase and phospha- dimerization. The identification of ligands that regulatetase regulation emerge? Additional second messengers receptor PTP dimerization will be an important step for-and nonprotein regulators are likely to be added to the ward in understanding these enzymes.current list, which includes cAMP, cGMP, DAG, Ca21/ Spatiotemporal Aspects of Signalingcalmodulin, polyamines, double-stranded RNA, and An area that will be increasingly important is the spatio-double-stranded DNA ends. Sphingosine and ceramide temporal aspects of signaling by protein kinases andare potential second messengers that reportedly regu- phosphatases. Over the past few years, it has beenlate protein kinases; cyclic ADP-ribose and sphingosine appreciated that protein kinases and phosphatases and1-phosphate and other recently described second mes- their substrates are often discretely localized in the cell,sengers, such as IP6, may also prove to regulate protein and that this is critical for specificity of phosphorylation.kinases. New second messengers may be identified The finding that signaling components are highly orga-through mass spectrometric analysis of total low molec- nized at the plasma membrane, in the cytoplasm, and ular weight compounds in extracts of stimulated cells. in the nucleus suggests that there are discrete signaling Conversely, one should be aware that second messen- domains within the cell. Traditionally, the transduction gers known to regulate protein kinases might have other of signals was thought to involve freely diffusible enti- functions. For example, that most venerable of second ties, and this is indeed the case for second messengers messengers, cAMP, has recently been shown to regu- like cAMP. However, pathways that involve predomi- late a Rap1 guanine-nucleotide-exchange factor (de nantly protein components may signal in a semi–solid Rooij et al., 1998), suggesting that cAMP, like cGMP, state fashion with minimal free diffusion. Moreover, sig- has multiple intracellular targets. naling proteins may be directionally transported via fila- Protein mass spectrometry is leading to the identifica- ment-bound motor proteins or by vesicular transport. tion of new types of posttranslational modification, and This leads to the idea that signals can be channeled some of these modifications may be used to regulate through specific routes from the membrane to the nu- protein kinase and phosphatase activity. An attractive cleus and other destinations and that there may be spe- possibility is that nuclear protein kinases and phospha- cific cellular compartments that are competent to relay tases involved in transcription might be regulated by signals. This mechanism would be expected to greatly acetylation. Additionally, the HIPK2 protein-serine ki- enhance the efficiency and specificity of signal trans- nase is modified by SUMO-1 (Kim et al., 1999), sug- mission. gesting that modification by members of the ubiquitin Anchoring and Scaffolding Proteins. Anchoring pro- family might also regulate protein kinase function. teins are often used to localize protein kinases and phos- Intracellular protein inhibitors of protein kinases have phatases to their substrates in particular places in the been known for many years (e.g., PKI for PKA, p21, and cell (Pawson and Scott, 1997). This increases the effi- p16 family Cdk inhibitors). Generally, these inhibitors are ciency and specificity of substrate phosphorylation and highly specific for individual protein kinases, as would coordinates the actions of protein kinases and phospha- be expected given that they work on the principle of tases. The best studied anchoring proteins are the protein–protein interaction. New protein kinase inhibitor AKAPs that are used to localize PKA, via its R subunits, proteins continue to be reported, and increasingly these and its substrates to particular places in the cell. For are likely to be of physiological importance. example, AKAP18, a myristoylated protein is localized Protein Phosphatases and Signaling to the plasma membrane where it binds PKA and the In any process governed by protein phosphorylation, L-type Ca21 channel, a PKA target (Colledge and Scott, regulation of the protein phosphatase(s) is just as likely 1999). Anchoring proteins are also known for TGFb RI, to be important as regulation of the protein kinase where SARA is used to recruit Smad proteins and to (Hunter, 1995). Indeed, although initially regarded as localize the signaling complex to discrete membrane constitutively active enzymes that reversed the action regions via a FYVE domain (Tsukazaki et al., 1998), and of inducibly activated protein kinases, the regulation of for protein kinase C, where RACKs are used for mem- protein phosphatases is proving to be highly sophisti- brane localization. Undoubtedly, many more anchoring cated. Protein phosphatase activity can regulated by proteins await discovery. posttranslational modification (e.g., phosphorylation and A different means of bringing signaling components methylation), second messengers (e.g., Ca21/calmodulin together is through the formation of multiprotein com- plexes through protein–protein interaction domains. Inand PP2B), targeting subunits, and inhibitory proteins. In Cell 120 this regard, one important principle is the use of scaf- of the FKHR transcription factor by Akt/PKB results in nuclear export (Brunet et al., 1999). Vesicular proteinfolding proteins to assemble sequential components of a signaling pathway (Garrington and Johnson, 1999). trafficking may also be used as a mechanism of signal transduction. For instance, most ligands induce endocy-The best characterized example is Ste5, a scaffolding protein that interacts with the Ste11, Ste7, and Fus3p/ tosis of their receptors, and there is increasing evidence that receptors internalized in endocytic vesicles carryKss1p protein kinases in the mating pheromone MAPK pathway. Scaffolding proteins are known for the JNK/ out specialized signaling functions. Temporal Aspects. Historically, signaling has beenSAPK and ERK MAPK pathways as well. Individual mem- bers of a MAPK pathway can also serve as scaffolds studied in populations of cells, but it is clear that there is significant kinetic variation in the responses of individualby interacting with other protein kinases in the same pathway. An important property of many scaffolding cells to a stimulus. This underscores the importance of developing tools to study signaling at high resolutionproteins is their ability to dimerize, which allows phos- phorylation in trans between protein kinase molecules in single cells, as has been done for Ca21 signaling. Extensive use is already being made of GFP-fusion pro-bound to different subunits in the dimer. A deeper under- standing of how assemblies of signaling proteins are teins to study the real-time movement of signaling mole- cules such as protein kinases in single living cells inbuilt up is clearly important. Plasma Membrane Signaling Domains. Another higher response to extracellular stimuli. For instance, translo- cation to the plasma membrane of various isoforms oforder principle of signal localization is afforded by plasma membrane signaling domains. Membrane mi- PKC fused to GFP can be detected in response to recep- tor stimulation (Sakai et al., 1997; Oancea and Meyer,crodomains known as caveolae, marked by the pres- ence of an intracellular protein coat comprised of caveo- 1998). SH2-GFP fusion proteins are being used to study the localization of tyrosine phosphorylation eventslin, can be isolated in a detergent-insoluble cholesterol/ glycolipid-rich fraction (Kurzchalia and Parton, 1999). (Stauffer and Meyer, 1997), and PH domain–GFP fusions to study where PIP2 and PIP3 are generated in vivoBy virtue of this property, caveolae are found to be enriched for many membrane signaling proteins, usually (Stauffer et al., 1998). A better knowledge of the exact timing of the onset and termination of signaling andlocalized by GPI linkages or palmitoyl groups. These lipid rafts are estimated to contain 50–100 protein mole- the speed of movement of signals combined with their precise location will be a key to understanding signalingcules, and this could be a means of concentrating signal- ing proteins for more efficient interaction. Another exam- output. Signaling Networks and Signaling Specificityple of a discrete plasma membrane signaling domain are focal adhesions, which are organized membrane- One of the major challenges in understanding signaling networks is to elucidate how signaling specificity isassociated structures where clustered integrins bind to extracellular matrix proteins and to the actin cytoskele- achieved when many of the same core signaling path- ways are activated by receptors that elicit different cellu-ton, and which mediate integrin signaling through tyro- sine and serine phosphorylation. lar responses (Tan and Kim, 1999). For instance, why does activation of the PI-3 kinase pathway by the insulinYet another mechanism of organizing proteins on the inner face of the membrane is through the use of proteins receptor PTK lead to metabolic responses such as trans- location of the GLUT4 glucose transporter, whereas acti-with multiple PDZ domains (Fanning and Anderson, 1999). PDZ domains bind to short C-terminal E.S/T.X.V/ICOOH vation of PI-3 kinase by growth factor receptor PTKs in the same cell does not? Signaling specificity can inmotifs, and a single protein can have up to 20 PDZ domains. A good example of a submembraneous PDZ principle result from the activation of unique signaling pathways by a receptor. Alternatively, since cellular re-organizing protein is InaD, which contains 5 PDZ do- mains with different specificities and can also dimerize. sponses reflect an integration of outputs from all the pathways activated by a single receptor, signaling spec-InaD interacts with multiple components of the visual signaling system in photoreceptors in the Drosophila ificity can also result from a unique combination of sig- naling pathways activated by the receptor. Receptoreye, including rhodopsin, thus concentrating all the nec- essary components into complexes and thereby speed- PTKs use not only autophosphorylation site tyrosines but also phosphorylate-specific docking proteins (e.g.,ing up signal transmission. An understanding of how membrane proteins involved in signaling are organized IRS1/2 for the insulin receptor PTK and FRS2 for the FGF and NGF receptor PTKs) to diversify signaling andon the inner face of the membrane will be increasingly important. provide specificity. The exact location of ligand-acti- vated receptors on the cell surface may dictate whichSignaling Protein Translocation. Regulated protein lo- calization has emerged as a fundamental principle in pathways can be activated, thus using spatial separa- tion as another means of generating specificity. Cellsignaling, and spatial separation of proteins is com- monly used as a mechanism for preventing spontaneous type–specific nuclear responses to activation of a given receptor can be explained by the expression of differentsignal activation. In addition to local movement of pro- teins, exemplified by inducible membrane association repertoires of downstream signaling proteins including transcription factors, and by which active chromatin do-of cytoplasmic P.Tyr-binding domain proteins, longer range movement of activated signaling proteins within mains are accessible for transcriptional induction by transcytoplasmic signaling pathways in each cell type.the cell is essential, particularly for transcytoplasmic signal transmission, where phosphorylation triggers the The apparent redundancy in signaling pathways is another challenge in understanding specificity. Redun-translocation of proteins in and out of the nucleus. For instance, tyrosine phosphorylation of Stats induces their dancy is evident from the finding that mutation of individ- ual tyrosine phosphorylation sites in a receptor PTKdimerization and nuclear import, and phosphorylation Review 121 often does not abolish a specific response, or even af- through phosphorylation and dephosphorylation and also fect the spectrum of genes induced upon receptor acti- by protein degradation. Thus, ligand-induced receptor vation (Valius and Kazlauskas, 1993; Fambrough et al., PTK signaling can be negatively regulated by receptor 1999). This may be explained by the existence of more endocytosis, dephosphorylation, feedback serine phos- than one means of activating a critical pathway (e.g., phorylation, by binding of inhibitory proteins including the Ras MAP kinase pathway can be activated by several inhibitory ligands, and through phosphorylation-depen- distinct mechanisms). Where there are several closely dent ubiquitination and degradation. The recruitment related proteins in a family, there can also be functional of nonreceptor PTPs such as Shp1 either to receptors redundancy. A good example is the Src PTK family, themselves or to inhibitory receptors (e.g., KIRs) or in- where c-Src, Fyn, and c-Yes are commonly coex- hibitory signaling proteins (e.g., SIRPs) is one mecha- pressed, and where any one of the three can be sufficient nism to downregulate receptor PTK signaling (Moghal to propagate a signal requiring a Src family PTK (Kling- and Sternberg, 1999). Interestingly a number of SH2 hoffer et al., 1999). domain proteins, including Socs, Cbl, and Dok family Signal pathway cross-talk will become increasingly members have recently been shown to act as negative important for our understanding of signaling networks. regulators of PTK signaling. Negative regulation of sig- Cross-talk can occur between pathways activated by a naling may well turn out to be as important as positive single receptor, or more commonly by pathways acti- regulation in understanding the specificity of signaling vated by different receptors. Indeed, integration of cellu- networks. lar responses elicited by different receptors must occur Role of Protein Kinases and Phosphatases by cross-talk. There are already well-established links in Disease between G protein–coupled receptors and receptor Increasing numbers of human diseases are known to PTKs, and between integrin adhesion receptors and re- involve mutations, overexpression, or malfunctioning of ceptor PTKs, to name just two examples (Luttrell et al., protein kinases and phosphatases, and their regulators 1999; Moghal and Sternberg, 1999). Cross-talk can take and effectors. The realization that protein kinases might place at many levels from the membrane to the nucleus, play a direct role in disease came with the discovery and involve components that are in common between that the v-Src oncoprotein is a PTK. Subsequently, a two pathways, as well as positive and negative feedback plethora of diseases have been show to be due to muta- signals that can act at many steps in a pathway from tions that activate or inactivate PTKs and PTPs, or lead transcription factors to the receptors themselves. On to their misexpression and/or overexpression (Hunter, the other side of the coin, pathway insulation to prevent 1998b). At least 18 PTK genes have been identified as cross-talk is proving an increasingly important concept. oncogenes either in acutely transforming retroviruses For instance, scaffolding proteins that assemble protein or in human tumors. Mutations in PTKs are also involved kinase signaling cascades insulate related MAPK path- in other diseases. In particular, mutational inactivation ways (Whitmarsh and Davis, 1998). of nonreceptor PTKs is observed in several immunodefi- Another important concept in signaling specificity is ciency diseases. Inactivation of both copies of ZAP70 signal thresholds. In many systems a 2-fold decrease or JAK3 causes severe combined immunodeficiency, in the level of a signaling protein can be sufficient to and mutation of the X-linked BTK gene results in agam- abrogate signaling, and conversely a 2-fold increase can maglobulinemia. initiate signaling. Moreover, signaling thresholds may Given the importance of activated PTKs in cancer, be different in different cell types. Productive signaling one might have anticipated that PTP genes would be can occur with only a few hundred activated receptor found as tumor suppressor genes. So far this has not molecules per cell, and therefore there has to be signifi- proved to be the case. However, there has been recent cant signal amplification. In this context, one aspect of excitement over the finding that the PTEN/MMAC gene, receptor PTK signaling that is not well understood is which is mutated in a variety of sporadic cancers and in the fact that growth factor receptor signaling is needed the hereditary Cowden’s hamartoma cancer syndrome, for 6–8 hr before a cell is committed to respond, even encodes a member of the dual-specificity protein phos- though all the early signaling events that have been so phatase family. However, PTEN’s main function in intensively studied are over by 1–2 hr, and the receptor negative growth regulation appears to be as a 39 phos- itself and the signaling pathways have been downregu- phoinositide phosphatase, rather than as a proteinlated by degradation and negative feedback loops. It phosphatase (Maehama and Dixon, 1998).remains unclear exactly what signal(s) is sensed by the Many genetic diseases also result from mutations incell at later times. This also highlights the fact that most protein-serine kinases and phosphatases. For instance,studies of ligand-induced signal transduction use acute the Coffin-Lowry syndrome is due to inactivation of thestimulation with saturating doses of ligand, and often X-linked Rsk2 protein-serine kinase gene, and myotonicoverexpressed receptors and heterologous cell types, dystrophy is due to decreased levels of expression of theand one can question whether all of the responses that myotonic dystrophy protein-serine kinase. In addition,are detected are physiologically relevant. overexpression of the aurora2 protein-serine kinase isFinally, temporal aspects of signaling are also impor- implicated in colon carcinoma, and the Lats1 and Lkb1tant in defining cellular responses. For instance, tran- protein-serine kinases have both been identified as tu-sient activation of ERK MAPK by a receptor PTK in PC12 mor suppressors. Conversely, inactivating mutations incells fails to induce differentiation, whereas sustained the Pr65 PP2A regulatory subunit are found in lung andERK MAP kinase activation induces neurite outgrowth colon cancers.(Marshall, 1995). Signaling kinetics are governed by neg- ative feedback systems that downregulate signaling As molecular analysis of human disease proceeds, Cell 122 one can predict that many additional somatic and hered- kinases and phosphatases are likely to be approved for clinical use in the near future. In the next few years, weitary mutations with causal roles in disease will be found in protein kinases and phosphatases, and in other sig- can anticipate that the rational structure-based design and development of highly specific protein kinase andnaling proteins. These enzymes and proteins will be- come candidates for the development of therapeutic phosphatase inhibitors (and activators) will become rou- tine, and that drugs that intercede in phosphorylation-drugs. Clinical Implications. In general enzymes make good mediated signaling pathways will become a major class of drug.targets for drugs, and the widespread involvement of protein kinases and phosphatases in disease has led to Genomics and Phosphorylation Complete genome sequences have the potential to re-a massive effort to develop drugs that either activate or more usually inhibit individual protein kinases and veal the totality of protein kinases and phosphatases and phosphorylation-related signaling proteins expressedphosphatases. Drugs designed to target these enzymes fall into two categories—monoclonal antibodies or mod- by a single organism. However, a priori assignment of function to a gene product relies on its sequence beingified protein ligands, and small molecules. Significant success has been achieved with the development of related to a protein of known function, and it clear that not all of the types of proteins that mediate phosphoryla-small molecule inhibitors of a number of PTKs, and sev- eral PTK inhibitors are either in or are beginning to enter tion-dependent signaling have yet been described. Novel Protein Kinases and Phosphatases. Biochemi-clinical trials. One group of small molecule inhibitors in cancer therapy trials target the EGF receptor or EGF cal and genomic analysis has already uncovered novel types of protein kinases and phosphatases, distinctreceptor PTK family members. Another group of inhibi- tors is directed against the VEGF receptor PTKs that from the major protein-serine/tyrosine kinase superfam- ily and the known protein phosphatase families, and willare essential for tumor angiogenesis (Fong et al., 1999). One receptor PTK antagonist, Herceptin, a monoclonal undoubtedly continue to do so. For example, one newly described group of protein kinases (MHCK A, eEF-2antibody that blocks the function of the ErbB2 receptor PTK, is already on the market, and is being used as an kinase, NFK, etc.) have a conserved catalytic domain that is completely unrelated to that of the protein-serine/adjuvant breast cancer therapy in the significant fraction of breast carcinomas where ErbB2 is overexpressed. tyrosine kinase superfamily (Ryazanov et al., 1999). Re- cent sequence analysis has uncovered a set of microbialAnother disease being treated with a PTK inhibitor is chronic myelogenous leukemia (CML), where a chromo- “protein kinase” families (ABC1, piD261, RIO1 families, and aminoglycoside kinases) that are all very distantlysomal translocation results in the expression of a chime- ric Bcr-Abl protein that is a constitutively activated PTK related in sequence to the protein-serine/tyrosine kinase superfamily catalytic domain, which are also repre-(Sawyers and Druker, 1999). CGP57148, a Bcr-Abl PTK inhibitor, is proving very effective as a treatment for sented in budding yeast and C. elegans (Leonard et al., 1998). These may represent the evolutionary origin of theCML. There are also protein-serine kinase inhibitors in cancer clinical trials, including flavopiridol, a Cdk2 in- eukaryotic protein kinase superfamily. Histidine, lysine, and arginine can be phosphorylated in proteins; how-hibitor. A number of diseases are due to insufficient receptor ever, with the exception of a yeast histidine kinase that phosphorylates histone H4, little is known about thePTK signaling, including noninsulin-dependent diabetes and peripheral neuropathies. If it were possible to en- protein kinases that phosphorylate these residues, and this will be an important task for the future.hance signaling through the receptors in question, this could serve as a viable therapy. One way to do this New types of protein phosphatase will also emerge to add to the currently known families of protein-serine,would be to find inhibitors of the cognate PTP for a receptor PTK. One candidate emerges from the recent protein-tyrosine, and dual-specificity phosphatases. For instance, the recently described CTD phosphatase,exciting finding that the knockout of PTP1B nonreceptor PTP in the mouse results in insulin hypersensitivity, indi- which dephosphorylates the CTD of the large subunit of RNA polymerase II, is a new type of protein-serinecating that PTP1B is a major insulin receptor PTK phos- phatase (Elchebly et al., 1999). A PTP1B-specific inhibi- phosphatase (Cho et al., 1999). Moreover, like PTEN, some members of the PTP superfamily may have non-tor has just entered clinical trials for diabetes. Another candidate is the NGF receptor PTK, and a drug that protein phosphate ester substrates. Protein Kinase and Phosphatase Catalogs. One excit-enhances NGF receptor signaling is also about to start clinical trials. ing outcome of the ongoing genomic sequencing proj- ects is that complete catalogs of protein kinases andSmall molecule activators of receptor PTKs have also been developed. For instance, an activator of the insulin phosphatases will become available for increasingly complex eukaryotic organisms. We already know thatreceptor PTK has recently been reported, which could in principle act as an orally available insulin mimetic S. cerevisiae has 114 conventional protein kinase genes (but no bona fide PTKs) out of 6,217 genes (1.8%)(Zhang et al., 1999). Interestingly, this pseudodimeric molecule acts intracellularly, apparently by reorienting (Hunter and Plowman, 1997). The recently completed sequence of C. elegans reveals that the worm genomethe two subunits of the insulin receptor dimer into an active configuration in a redox-dependent manner. Like- encodes 400 protein kinase catalytic domains (92 are PTKs 5 23%) out of 19,099 genes (2.1%) (Plowman etwise, small molecule activators of the G-CSF (Tian et al., 1998) and erythropoietin (Wrighton et al., 1996; John- al., 1999). Based on the existence in public human EST databases of .650 distinct protein kinases (.98 areson et al., 1998) binary receptor PTKs have been de- veloped. PTKs 5 z16%) and extrapolation from C. elegans, the human genome is predicted to encode .1100 proteinThe first small molecule drugs that act on protein Review 123 kinases (z150 PTKs), assuming the human genome has genes are expressed in a particular cell, and sophisti- about 80,000 genes. Indeed, it is almost inevitable that cated bioinformatics analysis. The new discipline of pro- the ballyhooed millenary of protein kinases will be teomics, where protein modifications and interactions reached! Analysis of the C. elegans genome shows that are analyzed in a high throughput format to provide the number of protein phosphatases encoded is surpris- protein linkage maps and other information, will be of ingly high, being more than half the number of protein fundamental importance in this endeavor. Databases kinases. Indeed, there are hints that there can be one that catalog protein–protein and protein–small molecule to one relationships between protein kinases and phos- interactions will also be an important step in this direc- phatases (e.g., the Clr-1 receptor PTP and Egl-15 FGF tion. The availability of consensus sequences for induc- receptor PTK in C. elegans) (Kokel et al., 1998). ible and constitutive protein–protein interactions and for Comparative analysis of protein kinase and phospha- phosphorylation by particular protein kinases will also tases from different species will tell us more about the be useful in predicting signaling connections. evolution of different protein kinases. For instance, it is One of the major challenges will be to model protein– already clear from the lack of bona fide PTKs in the protein interactions in vivo, where protein concentra- yeasts and their presence in the simplest of multicellular tions are much higher than those used in vitro. High eukaryotes that protein-tyrosine phosphorylation evolved protein concentrations favor low affinity interactions, hand in hand with multicellularity, presumably in re- which are hard to measure in vitro and yet are likely sponse to a need for intercellular communication. In to be important in signal propagation. Indeed, many keeping with this idea, a majority of PTKs play a role in signaling assemblies may rely on a combination of multi- transmembrane signaling in response to ligands that ple low-affinity interactions for their existence. Knowl- bind to surface receptors. We will also learn which pro- edge of the concentration of a given protein at a particu- tein kinases have been conserved throughout evolution, lar location in the cell will be critical information to and which ones have a specialized function in a particu- obtain. Incorporation of kinetic and particularly spatial lar type of organism. aspects of signaling into models will also be a serious Such catalogs of protein kinases and phosphatases challenge; and given that several thousand different pro- used in conjunction with detailed cellular expression teins could be involved in signaling in a single cell, the patterns derived from expression array analysis will cir- computational difficulties will be immense. Nonetheless, cumscribe which protein kinases and phosphatases are successful in silico analysis of signaling pathways and present in a cell and can therefore be involved in any networks must be a major goal for the future. particular phosphorylation event. This information, in Protein Phosphorylation Methodology combination with the available repertoire of protein ki- Advances in methodology play a key role in progress in nase target proteins and algorithms for simulating sig- most fields of biology, and this has certainly been true naling networks, may allow predictions of signaling out- in the field of signal transduction in general and protein comes in response to individual stimuli or combinations phosphorylation in particular. New methods will surely of stimuli. also play a key role in future advances. Signal Network Modeling: E-Phosphorylation. Analo- Genetic Analysis. Our understanding of signaling gies have been drawn between cellular signaling net- pathways has benefited enormously from the use of works and electronic circuits. Indeed cellular signaling organisms where genetic analysis is feasible. The highly pathways have many of the attributes of electronic cir- conserved nature of most of the major signaling path- cuits. Individual proteins can act as amplifiers or ways in eukaryotes allows use of genetic information switches, and protein kinase cascades can act as serial obtained in one organism to predict the existence of amplifiers or switches. Signal pathways can have posi- pathways and specific components in other organisms. tive and negative feedback loops, and networks can be A good example is the elucidation of the MAPK path- built up out of multiple signaling pathways. ways where genetic analysis in yeast, C. elegans, and Even though cell signaling networks are multidimen- Drosophila was combined with biochemical analysis of sional rather than two dimensional, their properties have growth factor–stimulated protein kinases in mammalian encouraged attempts to develop predictive algorithms. cells to provide a complete picture of this pathway. Efforts have been made to model MAPK cascades and Genetic analysis of signaling systems will increasingly the regulation of cell cycle transitions, and informative be used to investigate the nature and function of signal-predictions have emerged. For instance, theoretical ing pathways in vivo. Gene disruption through homolo-analysis of the ERK MAPK cascade has shown that it gous recombination, selection of null mutants, or theacts as a switch to provide a very sharp activation curve powerful new interfering double-stranded RNA (RNAi)in response to increasing levels of stimulus, rather than “knockout” technology can be used to determineacting an amplifier (Huang and Ferrell, 1996). Ultimately, whether a protein is essential for a given response. Inhowever, useful prediction of signaling pathways and vertebrates, increasing use will be made of tissue-spe-networks will require a knowledge of all the players, cific conditional knockouts, to circumvent potential de-their kinetic properties, their interaction partners and fects in embryogenesis. Future genetic analysis of sig-mechanisms of positive and negative regulation, and naling pathways will rely more and more on makingtheir subcellular localizations and concentrations. With subtle germline mutations in organisms or cells (e.g.,the availability of complete genome sequences, efforts inactivating point mutations) so that the gene productare underway to define all of the proteins involved in is expressed at physiological levels at the correct timesignaling responses induced by specific receptors (e.g., and place, thus mitigating the potential problems withG protein coupled receptors). These efforts will be aided by the use of expression array analysis to define which using overexpressed proteins and heterologous cell Cell 124 types. In the phosphorylation arena, the function of indi- (FRET) to monitor the timing and location of an intramo- lecular interaction dependent upon protein kinase acti-vidual phosphorylation sites will be studied by mutating them to a nonphosphorylatable residue to block phos- vation or phosphorylation. For instance, the localization of activated protein kinase Ca has been imaged by ex-phorylation or an acidic residue that can mimic phos- phorylation. Such an analysis of the functions of individ- pressing a GFP-tagged PKCa and microinjecting Cy3- labeled anti-P.Thr250 antibodies, which recognize acti-ual tyrosine phosphorylation sites in the PDGF b receptor PTK has recently been carried out by making vated phosphorylated PKCa (Ng et al., 1999). Specific cell-permeant inhibitors of protein kinasesknockin mutations in the mouse germline (Heuchel et al., 1999). Finally, genetic screens for suppressors and and phosphatases, and other signaling proteins are ex- tremely powerful tools for analyzing signal transductionenhancers of sensitized conditional mutations, synthetic lethal screens, and analysis of modifier genes will remain processes in intact cells. Many such inhibitors are al- ready available and are extensively used. For instance,important tools in our armamentarium for analyzing sig- naling pathways. there are nearly 2000 papers reporting the use of the Parke-Davis MEK inhibitor PD98059, and over 1500Biochemical Analysis. Many new techniques have been developed in the protein phosphorylation arena. where wortmannin has been used as a PI-39 kinase inhib- itor. Phosphopeptides or other peptides that block pro-Traditionally, 32P labeling has been used to study protein phosphorylation both in vivo and in vitro, but techniques tein–protein interactions are also increasingly used to interdict signaling pathways. Conversely, rapid and re-that emphasize nonradioactive detection of phosphory- lation are becoming more and more prominent (e.g., versible activation of exogenously expressed protein kinases and phosphatases can be achieved by usingmeasurements of protein kinase activity using fluores- cence anisotropy with fluorescently tagged peptides chemical inducers of homo- or heterodimerization, or using 4-hydroxy tamoxifen to activate protein kinasesthat bind to phosphospecific antibodies upon phosphor- ylation). Anti-P.Tyr antibodies have been widely used to and phosphatases fused to the estrogen receptor hor- mone binding domain.detect tyrosine phosphorylation, but attempts to de- velop anti-P.Ser and anti-P.Thr antibodies for similar The ongoing efforts of the pharmaceutical industry to develop drugs that act as highly specific inhibitors (anduses have been less successful, because these antibod- ies tend to recognize other phosphate esters, and are activators) of protein kinases and phosphatases has already had significant benefits in basic research bygenerally of low affinity. Of much greater value have been antibodies directed against specific P.Ser- and providing inhibitors as research tools. In many cases a potent inhibitor of a target protein is developed, but forP.Thr-containing peptide sequences, corresponding to known or suspected sites of phosphorylation. Phospho- a variety of reasons is not taken forward into clinic trials or fails in trials. Such inhibitors as well as early deriva-specific antibodies against single or closely spaced sites will be increasingly useful tools for studying serine/ tives of successful drugs are perfectly suited for analyti- cal research. In the coming years, we can expect to seethreonine and tyrosine phosphorylation of individual proteins by immunoprecipitation, immunoblotting, and increasing use of compounds from the pharmaceutical industry as tools for investigating signaling processes.immunofluorescence staining. Analysis of histidine, ly- sine, and arginine phosphorylation would be greatly fa- Protein Kinase and Phosphatase Substrates. Protein kinases recognize their substrates in part through thecilitated by the development of antibodies against these phosphoamino acids. primary sequence surrounding the phosphorylatable residue. An emerging theme is that protein kinases alsoThe mapping of phosphorylation sites is a critical endeavour, which traditionally has been done using recognize substrates via secondary docking sites dis- tinct from primary phosphorylation sites (Holland and32P-labeling either in vivo or in vitro. However, within the past few years mass spectrometry has emerged as the Cooper, 1999). This concept is well developed for PTKs where SH2 and PTB domains are used to recruit sub-best method for identifying phosphorylation sites, and, in the future, phosphorylation site mapping and mea- strates to specific P.Tyr residues in activated receptor PTK dimers, triggering their phosphorylation. Protein-surements of phosphorylation stoichiometry will be ac- complished routinely using mass spectrometry. serine kinases also recognize substrates through dock- ing sites. For instance, JNK1/2 MAPKs interact specifi-Single Cell Analysis. It is imperative that we devise better methods to study the kinetics and intracellular cally with a short LXL motif in the d region of c-Jun via a loop in their C-terminal lobe, and ERK MAPKslocalization of signaling processes in single living cells, so that spatiotemporal aspects of signaling can be de- recognize substrates via at least two distinct motifs (e.g., FXFP and LAQRRXXXXL/I) (Kallunki et al., 1994; Yangtermined. New assays to determine in real time where in the cell a protein kinase is active using fluorescent et al., 1998; Gavin and Nebreda, 1999; Jacobs et al., 1999; Smith et al., 1999). In addition, the G1 cyclin/reporter proteins will be important (e.g., using engi- neered GFP derivatives with a grafted protein kinase Cdks select substrates through interaction of the cyclin subunit via a ZRXL motif (Adams et al., 1996), and theconsensus sequence that exhibit fluorescence changes upon phosphorylation by a specific protein kinase). The L45 loop in the N-terminal lobe of the TGFb type I recep- tor catalytic domain interacts with a short sequencelocation and kinetics of specific phosphorylation events can be monitored simultaneously in living cells using in the Smad2/3 MH2 domain (Chen et al., 1998). The emerging picture is that a variety of surface loops onexpressed or microinjected fluorescent reporter pro- teins (e.g., natural or artificial P.Ser or P.Tyr binding the catalytic domain as well as sites elsewhere in a protein kinase can be used for selection of substrates.proteins and phosphospecific antibodies). Better yet, one can use fluorescence resonance energy transfer Protein kinase catalytic domains interact with many Review 125 types of protein including activators, inhibitors, tar- sequencing will probably be expressed and function in specific neurons. From a technical standpoint a majorgeting proteins, and substrates. An important challenge for the future will be to understand the structural bases requirement is the development of assays that allow temporal analysis of signaling events at high spatial res-for these multiple protein kinase catalytic domain inter- actions. olution in single cells. Also better methods for rapid detection of phosphorylation, phosphorylation site map-One major problem in the protein phosphorylation field is the difficulty in identifying true substrates for ping, and protein kinase and phosphate substrate identi- fication are essential. The evolution of theoretical meth-individual protein kinases and phosphatases in vivo. Specific protein kinase inhibitors can be useful, but it ods for analyzing signaling networks and sophisticated protein interaction databases to support modeling ef-is hard to prove that they are truly selective in vivo. A new substrate detection method has recently been forts is another obvious need. From a practical stand- point, one can foresee that the accumulating knowledgedeveloped in which the ATP-binding site of a protein kinase is “enlarged” by mutation to accept an N6-modi- about signaling networks and the proteins involved will permit development of potent and specific pharmaco-fied ATP analog, which can be used as a phosphate donor for the mutant protein kinase when expressed in logical modulators of signaling that can be used thera- peutically. Finally, we should not be so arrogant as tovivo but which cannot be used by other normal cellular protein kinase (Liu et al., 1998). This method has been think that we already know all the possible principles of signaling, and we should certainly expect totally newvalidated with members of the Src PTK family, and ef- forts are currently being made to extend this to other types of signaling systems to be uncovered. PTKs and to protein-serine kinases. The analysis of degenerate oriented peptide libraries Acknowledgments for sequences that can be phosphorylated by purified protein kinases continues to be a valuable method for The important findings in the history of signal transduction are ade- quately covered in many reviews, and I have therefore cited reviewsdeducing primary sequence consensuses for newly that discuss the seminal papers. Space constraints have unfortu-identified protein kinases (Songyang et al., 1994). Such nately meant I have had to omit many pertinent citations in theconsensus sequences, combined with compilations of forward-looking part of this review. I thank Tony Pawson for provid- phosphorylation sites in known targets, can be used to ing the basis for Figure 1. screen sequence databases for potential substrates. Physiological targets have already been identified by References this strategy in a number of instances. Since many pro- tein kinases have secondary docking sites on their sub- Adams, P.D., Sellers, W.R., Sharma, S.K., Wu, A.D., Nalin, C.M., and strates, database searches for potential substrates that Kaelin, W.G. (1996). Identification of a cyclin-cdk2 recognition motif present in substrates and p21-like cyclin-dependent kinase inhibi-contain both consensus phosphorylation site and dock- tors. Mol. Cell. Biol. 16, 6623–6633.ing site sequences should have greatly increased Artavanis-Tsakonas, S., Rand, M.D., and Lake, R.J. (1999). Notchchances of finding physiological targets. signaling: cell fate control and signal integration in development.The identification of physiological substrates for pro- Science 284, 770–776. tein phosphatases is equally difficult. Cell permeant Austin, D.J., Crabtree, G.R., and Schreiber, S.L. (1994). Proximityphosphatase inhibitors (e.g., okadaic acid) can be used versus allostery: the role of regulated protein dimerization in biology. as a method of identifying substrate proteins through Chem. Biol. 1, 131–136. increased in vivo phosphorylation. For PTPs catalytically Berridge, M.J., and Irvine, R.F. (1989). Inositol phosphates and cell impaired “substrate-trapping” mutant PTPs that bind signaling. Nature 341, 197–205. stably to their phosphoprotein substrates have success- Bilwes, A.M., den Hertog, J., Hunter, T., and Noel, J.P. (1996). Struc- fully been used to isolate substrates (Flint et al., 1997). tural basis for inhibition of receptor protein-tyrosine phosphatase-a by dimerization. Nature 382, 555–559.For identification of a PTP for a known phosphoprotein, in gel phosphatase assays are proving useful. Like pro- Brunet, A., Bonni, A., Zigmond, M.J., Lin, M.Z., Juo, P., Hu, L.S., Anderson, M.J., Arden, K.C., Blenis, J., and Greenberg, M.E. (1999).tein kinases, protein phosphatases appear to have sec- Akt promotes cell survival by phosphorylating and inhibiting a fork-ondary docking sites on their substrates, recognized head transcription factor. Cell 96, 857–868.either by associated targeting subunits or protein inter- Chen, Y.G., Hata, A., Lo, R.S., Wotton, D., Shi, Y., Pavletich, N., andaction domains on the catalytic subunit itself, and these Massague, J. (1998). Determinants of specificity in TGF-b signal interactions can be used to identify potential substrates. transduction. Genes Dev. 12, 2144–2152. Cheng, A.M., Saxton, T.M., Sakai, R., Kulkarni, S., Mbamalu, G., Futurescope Vogel, W., Tortorice, C.G., Cardiff, R.D., Cross, J.C., Muller, W.J., and Pawson, T. (1998). Mammalian Grb2 regulates multiple stepsPhosphorylation touches on most aspects of cell physi- in embryonic development and malignant transformation. Cell 95,ology. Which areas are most likely to be important in the 793–803.near future? Emergent areas are transcriptional control, Cho, H., Kim, T.K., Mancebo, H., Lane, W.S., Flores, O., and Rein-apoptosis, phosphorylation-dependent protein degra- berg, D. (1999). A protein phosphatase functions to recycle RNAdation, phosphorylation-dependent nuclear import and polymerase II. Genes Dev. 13, 1540–1552. export, cytoskeletal regulation, and checkpoint signal- Colledge, M., and Scott, J.D. (1999). AKAPs: from structure to func- ing. However, it is in the function of the vertebrate central tion. Trends Cell Biol. 9, 216–221. nervous system where studies of phosphorylation seem de Rooij, J., Zwartkruis, F.J., Verheijen, M.H., Cool, R.H., Nijman, likely to have the greatest impact. The majority of protein S.M., Wittinghofer, A., and Bos, J.L. (1998). Epac is a Rap1 guanine- kinases are expressed in the brain, and many of the nucleotide-exchange factor directly activated by cyclic AMP. Nature 396, 474–477.novel vertebrate protein kinases revealed by genomic Cell 126 Elchebly, M., Payette, P., Michaliszyn, E., Cromlish, W., Collins, S., Johnson, D.L., Farrell, F.X., Barbone, F.P., McMahon, F.J., Tullai, J., Hoey, K., Livnah, O., Wrighton, N.C., Middleton, S.A., Loughney,Loy, A.L., Normandin, D., Cheng, A., Himms-Hagen, J., Chan, C.C., et al. (1999). Increased insulin sensitivity and obesity resistance in D.A., et al. (1998). Identification of a 13 amino acid peptide mimetic of erythropoietin and description of amino acids critical for the mimeticmice lacking the protein tyrosine phosphatase-1B gene. Science 283, 1544–1548. activity of EMP1. Biochemistry 37, 3699–3710. Kallunki, T., Su, B., Tsigelny, I., Sluss, H.K., Derijard, B., Moore,Espanel, X., and Sudol, M. (1999). A single point mutation in a group I WW domain shifts its specificity to that of group II WW domains. G., Davis, R., and Karin, M. (1994). JNK2 contains a specificity- determining region responsible for efficient c-Jun binding and phos-J. Biol. Chem. 274, 17284–17289. phorylation. Genes Dev. 8, 2996–3007.Eyers, P.A., van den IJssel, P., Quinlan, R.A., Goedert, M., and Co- hen, P. (1999). Use of a drug-resistant mutant of stress-activated Kapeller, R., and Cantley, L.C. (1994). Phosphatidylinositol 3-kinase. Bioessays 16, 565–576.protein kinase 2a/p38 to validate the in vivo specificity of SB 203580. FEBS Lett. 451, 191–196. Karin, M., and Hunter, T. (1995). Transcriptional control by protein phosphorylation: signal transmission from the cell surface to theFambrough, D., McClure, K., Kazlauskas, A., and Lander, E.S. (1999). Diverse signaling pathways activated by growth factor receptors nucleus. Curr. Biol. 5, 747–757. induce broadly overlapping, rather than independent, sets of genes. Kim, Y.H., Choi, C.Y., and Kim, Y. (1999). Covalent modification Cell 97, 727–741. of the homeodomain-interacting protein kinase 2 (HIPK2) by the ubiquitin-like protein SUMO-1. Proc. Natl. Acad. Sci. USA 96, 12350–Fanning, A.S., and Anderson, J.M. (1999). Protein modules as orga- nizers of membrane structure. Curr. Opin. Cell Biol. 11, 432–439. 12355. Klinghoffer, R.A., Sachsenmaier, C., Cooper, J.A., and Soriano, P.Ferguson, K.M., Lemmon, M.A., Sigler, P.B., and Schlessinger, J. (1995). Scratching the surface with the PH domain. Nat. Struct. Biol. (1999). Src family kinases are required for integrin but not PDGFR signal transduction. EMBO J. 18, 2459–2471.2, 715–718. Flint, A.J., Tiganis, T., Barford, D., and Tonks, N. (1997). Development Kokel, M., Borland, C.Z., DeLong, L., Horvitz, H.R., and Stern, M.J. (1998). clr-1 encodes a receptor tyrosine phosphatase that nega-of “substrate-trapping” mutants to identify physiological substrates of protein tyrosine phosphatases. Proc. Natl. Acad. Sci. USA 94, tively regulates an FGF receptor signaling pathway in Caenorhab- ditis elegans. Genes Dev. 12, 1425–1437.1680–1685. Fong, T.A., Shawver, L.K., Sun, L., Tang, C., App, H., Powell, T.J., Krebs, E.G., and Beavo, J.A. (1979). Phosphorylation-dephosphory- lation of enzymes. Annu. Rev. Biochem. 48, 923–959.Kim, Y.H., Schreck, R., Wang, X., Risau, W., et al. (1999). SU5416 is a potent and selective inhibitor of the vascular endothelial growth Kuriyan, J., and Cowburn, D. (1997). Modular peptide recognition factor receptor (Flk-1/KDR) that inhibits tyrosine kinase catalysis, domains in eukaryotic signaling. Annu. Rev. Biophys. Biomol. Struct. tumor vascularization, and growth of multiple tumor types. Cancer 26, 259–288. Res. 59, 99–106. Kurzchalia, T.V., and Parton, R.G. (1999). Membrane microdomains Garrington, T.P., and Johnson, G.L. (1999). Organization and regula- and caveolae. Curr. Opin. Cell. Biol. 11, 424–431. tion of mitogen-activated protein kinase signaling pathways. Curr. Leonard, C.J., Aravind, L., and Koonin, E.V. (1998). Novel families Opin. Cell Biol. 11, 211–218. of putative protein kinases in bacteria and archaea: evolution of the Gavin, A.C., and Nebreda, A.R. (1999). A MAP kinase docking site “eukaryotic” protein kinase superfamily. Genome Res. 8, 1038–1047. is required for phosphorylation and activation of p90(rsk)/MAPKAP Liscovitch, M., and Cantley, L.C. (1994). Lipid second messengers. kinase-1. Curr. Biol. 9, 281–284. Cell 77, 329–334. Gilman, A.G. (1987). G proteins: transducers of receptor-generated Liu, Y., Shah, K., Yang, F., Witucki, L., and Shokat, K.M. (1998). signals. Annu. Rev. Biochem. 56, 615–649. Engineering Src family protein kinases with unnatural nucleotide Heuchel, R., Berg, A., Tallquist, M., Ahlen, K., Reed, R.K., Rubin, K., specificity. Chem. Biol. 5, 91–101. Claesson-Welsh, L., Heldin, C.H., and Soriano, P. (1999). Platelet- Luttrell, L.M., Daaka, Y., and Lefkowitz, R.J. (1999). Regulation of derived growth factor beta receptor regulates interstitial fluid ho- tyrosine kinase cascades by G-protein-coupled receptors. Curr. meostasis through phosphatidylinositol-39 kinase signaling. Proc. Opin. Cell Biol. 11, 177–183. Natl. Acad. Sci. USA 96, 11410–11415. Maehama, T., and Dixon, J.E. (1998). The tumor suppressor, PTEN/ Hof, P., Pluskey, S., Dhe-Paganon, S., Eck, M.J., and Shoelson, S.E. MMAC1, dephosphorylates the lipid second messenger, phosphati- (1998). Crystal structure of the tyrosine phosphatase SHP-2. Cell dylinositol 3,4,5-trisphosphate. J. Biol. Chem. 273, 13375–13378. 92, 441–450. Majeti, R., Bilwes, A.M., Noel, J.P., Hunter, T., and Weiss, A. (1998). Holland, P.M., and Cooper, J.A. (1999). Protein modification: docking Dimerization-induced inhibition of receptor protein tyrosine phos- sites for kinases. Curr. Biol. 9, R329–R331. phatase function through an inhibitory wedge. Science 279, 88–91. Huang, C.Y., and Ferrell, J.E. (1996). Ultrasensitivity in the mitogen- Mangelsdorf, D.J., Thummel, C., Beato, M., Herrlich, P., Schutz, G., activated protein kinase cascade. Proc. Natl. Acad. Sci. USA 93, Umesono, K., Blumberg, B., Kastner, P., Mark, M., Chambon, P., and 10078–10083. Evans, R.M. (1995). The nuclear receptor superfamily: the second Hunter, T. (1995). Protein kinases and phosphatases: the yin and decade. Cell 83, 835–839. yang of protein phosphorylation and signaling. Cell 80, 225–236. Marshall, C.J. (1995). Specificity of receptor tyrosine kinase signal- Hunter, T. (1998a). Anti-phosphatases take the stage. Nat. Genet. ing: transient versus sustained extracellular signal-regulated kinase 18, 303–305. activation. Cell 80, 179–185. Hunter, T. (1998b). The Croonian lecture 1997. The phosphorylation McCormick, F. (1989). Ras GTPase activating protein: signal trans- of proteins on tyrosine: its role in cell growth and disease. Philos. mitter and signal terminator. Cell 56, 5–8. Trans. R. Soc. Lond. B Biol. Sci. 353, 583–605. Meng, W., Sawasdikosol, S., Burakoff, S.J., and Eck, M.J. (1999). Hunter, T., and Cooper, J.A. (1985). Protein-tyrosine kinases. Annu. Structure of the amino-terminal domain of Cbl complexed to its Rev. Biochem. 54, 897–930. binding site on ZAP-70 kinase. Nature 398, 84–90. Hunter, T., and Plowman, G.D. (1997). The protein kinases of budding Moghal, N., and Sternberg, P.W. (1999). Multiple positive and nega- yeast: six score and more. Trends Biochem. Sci. 22, 18–22. tive regulators of signaling by the EGF-receptor. Curr. Opin. Cell. Biol. 11, 190–196.Jacobs, D., Glossip, D., Xing, H., Muslin, A.J., and Kornfeld, K. (1999). Multiple docking sites on substrate proteins form a modular system Mohammadi, M., McMahon, G., Sun, L., Tang, C., Hirth, P., Yeh, that mediates recognition by ERK MAP kinase. Genes Dev. 13, B.K., Hubbard, S.R., and Schlessinger, J. (1997). Structures of the 163–175. tyrosine kinase domain of fibroblast growth factor receptor in com- plex with inhibitors. Science 276, 955–960.Jiang, G., and Hunter, T. (1999). Receptor activation: when a dimer is not enough. Curr. Biol. 9, R568–R571. Mohammadi, M., Froum, S., Hamby, J.M., Schroeder, M.C., Panek, Review 127 R.L., Lu, G.H., Eliseenkova, A.V., Green, D., Schlessinger, J., and Tan, P.B., and Kim, S.K. (1999). Signaling specificity: the RTK/RAS/ Hubbard, S.R. (1998). Crystal structure of an angiogenesis inhibitor MAP kinase pathway in metazoans. Trends Genet. 15, 145–149. bound to the FGF receptor tyrosine kinase domain. EMBO J. 17, Tian, S.S., Lamb, P., King, A.G., Miller, S.G., Kessler, L., Luengo, 5896–5904. J. I., Averill, L., Johnson, R.K., Gleason, J.G., Pelus, L.M., et al. Montminy, M.R., Gonzalez, G.A., and Yamamoto, K.K. (1990). Regu- (1998). A small, nonpeptidyl mimic of granulocyte-colony-stimulat- lation of cAMP-inducible genes by CREB. Trends Neurosci. 13, ing factor. Science 281, 257–259. 184–188. Tsukazaki, T., Chiang, T.A., Davison, A.F., Attisano, L., and Wrana, Murad, F. (1994). Regulation of cytosolic guanylyl cyclase by nitric J.L. (1998). SARA, a FYVE domain protein that recruits Smad2 to oxide: the NO-cyclic GMP signal transduction system. Adv. Pharma- the TGFb receptor. Cell 95, 779–791. col. 26, 19–33. Valius, M., and Kazlauskas, A. (1993). Phospholipase C-g 1 and Neet, K., and Hunter, T. (1996). Vertebrate non-receptor protein- phosphatidylinositol 3 kinase are the downstream mediators of the tyrosine kinase families. Genes Cells 1, 147–169. PDGF receptor’s mitogenic signal. Cell 73, 321–334. Ng, T., Squire, A., Hansra, G., Bornancin, F., Prevostel, C., Hanby, van der Geer, P., and Pawson, T. (1995). The PTB domain: a new A., Harris, W., Barnes, D., Schmidt, S., Mellor, H., Bastiaens, P.I., protein module implicated in signal transduction. Trends Biochem. and Parker, P.J. (1999). Imaging protein kinase Ca activation in cells. Sci. 20, 277–280. Science 283, 2085–2089. Waskiewicz, A.J., and Cooper, J.A. (1995). Mitogen and stress re- Oancea, E., and Meyer, T. (1998). Protein kinase C as a molecular sponse pathways: MAP kinase cascades and phosphatase regula- machine for decoding calcium and diacylglycerol signals. Cell 95, tion in mammals and yeast. Curr. Opin. Cell. Biol. 7, 798–805. 307–318. Whitmarsh, A.J., and Davis, R.J. (1998). Structural organization of Pawson, T. (1995). Protein modules and signaling networks. Nature MAP-kinase signaling modules by scaffold proteins in yeast and 373, 573–580. mammals. Trends Biochem. Sci. 23, 481–485. Pawson, T., and Gish, G.D. (1992). SH2 and SH3 domains: from Wittinghofer, A., and Nassar, N. (1996). How Ras-related proteins structure to function. Cell 71, 359–362. talk to their effectors. Trends Biochem. Sci. 21, 488–491. Pawson, T., and Scott, J.D. (1997). Signaling through scaffold, an- Wrighton, N.C., Farrell, F.X., Chang, R., Kashyap, A.K., Barbone, choring, and adaptor proteins. Science 278, 2075–2080. F.P., Mulcahy, L.S., Johnson, D.L., Barrett, R.W., Jolliffe, L.K., and Plowman, G.D., Sudarsanam, S., Bingham, J., Whyte, D., and Hunter, Dower, W.J. (1996). Small peptides as potent mimetics of the protein T. (1999). The protein kinases of Caenorhabditis elegans: a model hormone erythropoietin. Science 273, 458–464. for signal transduction in multicellular organisms. Proc. Natl. Acad. Xu, W., Doshi, A., Lei, M., Eck, M.J., and Harrison, S.C. (1999). CrystalSci. USA 96, 13603–13610. structures of c-Src reveal features of its autoinhibitory mechanism. Russo, A.A., Tong, L., Lee, J.O., Jeffrey, P.D., and Pavletich, N.P. Mol. Cell 3, 629–638. (1998). Structural basis for inhibition of the cyclin-dependent kinase Yaffe, M.B., and Cantley, L.C. (1999). Signal transduction. GrabbingCdk6 by the tumour suppressor p16INK4a. Nature 395, 237–243. phosphoproteins. Nature 402, 30–31. Ryazanov, A.G., Pavur, K.S., and Dorovkov, M.V. (1999). Alpha-kinases: Yang, S.H., Whitmarsh, A.J., Davis, R.J., and Sharrocks, A.D. (1998).a new class of protein kinases with a novel catalytic domain. Curr. Differential targeting of MAP kinases to the ETS-domain transcrip-Biol. 9, R43–R45. tion factor Elk-1. EMBO J. 17, 1740–1749. Sakai, N., Sasaki, K., Ikegaki, N., Shirai, Y., Ono, Y., and Saito, Zhang, B., Salituro, G., Szalkowski, D., Li, Z., Zhang, Y., Royo, I.,N. (1997). Direct visualization of the translocation of the gamma- Vilella, D., Diez, M.T., Pelaez, F., Ruby, C., et al. (1999). Discoverysubspecies of protein kinase C in living cells using fusion proteins of a small molecule insulin mimetic with antidiabetic activity in mice.with green fluorescent protein. J. Cell Biol. 139, 1465–1476. Science 284, 974–977.Salvesen, G.S., and Dixit, V.M. (1999). Caspase activation: the in- duced-proximity model. Proc. Natl. Acad. Sci. USA 96, 10964–10967. Sawyers, C.L., and Druker, B. (1999). Tyrosine kinase inhibitors in chronic myeloid leukemia. Cancer J. Sci. Am. 5, 63–69. Schlessinger, J. (1988). Signal transduction by allosteric receptor oligomerization. Trends Biochem. Sci. 13, 443–447. Schlessinger, J. (1993). How receptor tyrosine kinases activate Ras. Trends Biochem. Sci. 18, 273–275. Schlessinger, J. (1994). SH2/SH3 signaling proteins. Curr. Opin. Genet. Dev. 4, 25–30. Seger, R., and Krebs, E.G. (1995). The MAPK signaling cascade. FASEB J. 9, 726–735. Smith, J.A., Poteet-Smith, C.E., Malarkey, K., and Sturgill, T.W. (1999). Identification of an extracellular signal-regulated kinase (ERK) docking site in ribosomal S6 kinase, a sequence critical for activation by ERK in vivo. J. Biol. Chem. 274, 2893–2898. Songyang, Z., Blechner, S., Hoagland, N., Hoekstra, M.F., Piwnica- Worms, H., and Cantley, L.C. (1994). Use of an oriented peptide library to determine the optimal substrates of protein kinases. Curr. Biol. 4, 973–982. Spencer, D.M., Wandless, T.J., Schreiber, S.L., and Crabtree, G.R. (1993). Controlling signal transduction with synthetic ligands. Sci- ence 262, 1019–1024. Stauffer, T.P., and Meyer, T. (1997). Compartmentalized IgE recep- tor-mediated signal transduction in living cells. J. Cell Biol. 139, 1447–1454. Stauffer, T.P., Ahn, S., and Meyer, T. (1998). Receptor-induced tran- sient reduction in plasma membrane PtdIns(4,5)P2 concentration monitored in living cells. Curr. Biol. 8, 343–346. Stock, J.B., Stock, A.M., and Mottonen, J.M. (1990). Signal transduc- tion in bacteria. Nature 344, 395–400.