This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Copyright Permission Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Pearson, G. Articles by Cobb, M. H. Search for Related Content PubMed PubMed Citation Articles by Pearson, G. Articles by Cobb, M. H.

Mitogen-Activated Protein (MAP) Kinase Pathways: Regulation and Physiological Functions¹

Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, Texas 75390

	Abstract

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
Mitogen-activated protein (MAP) kinases comprise a family of ubiquitous proline-directed, protein-serine/threonine kinases, which participate in signal transduction pathways that control intracellular events including acute responses to hormones and major developmental changes in organisms. MAP kinases lie in protein kinase cascades. This review discusses the regulation and functions of mammalian MAP kinases. Nonenzymatic mechanisms that impact MAP kinase functions and findings from gene disruption studies are highlighted. Particular emphasis is on ERK1/2.

I. Introduction

II. Overview of Regulation and Properties of MAP Kinases

III. MAP Kinases Are Activated by Phosphorylation Cascades

IV. Signal Integration and Specificity

V. Mammalian MAP Kinase Cascades

A. The ERK1 and ERK2 cascades

B. c-Jun N-terminal kinases/stress-activated protein kinases (JNK/SAPK)

C. p38 Pathways

D. Other MAP kinases

E. MEKKs, the first tier in the kinase cascade

VI. Activation of ERK1/2 and Other MAP Kinases from the Cell Surface

VII. Scaffolding and Its Role in Organization, Localization, and Specificity in MAP Kinase Cascades

A. Complexes predicted from studies in yeast

B. Protein associations in mammalian MAP kinase pathways

VIII. Regulation of MAP Kinase Localization

IX. Inactivation of MAP Kinases

X. Substrate Recognition and Stable Binding of Substrates to MAP Kinases

XI. Substrates of MAP Kinases

XII.

Biology of MAP Kinase Pathways

A. Development of inhibitors

XIII. Gene Disruption Experiments

A. The ERK1/2 pathway

B. The JNK/SAPK pathways

C. The p38 pathways

D. Other components of MAP kinase pathways

	I. Introduction

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
PROTEIN kinases and other messenger systems form highly interactive networks to achieve the integrated function of cells in an organism. To understand the signaling mechanism for any agent, its repertoire of signal transducers and their interactions within this network must be defined within the cellular context. This includes the production of second messengers, activation of protein kinases, and the subcellular distribution of these transducers to bring them into contact with appropriate targets. Within the repertoire of signaling molecules in the network is a family of protein kinase cascades known as mitogen-activated protein (MAP) kinase modules. These cascades contain at least three protein kinases in series that culminate in the activation of a multifunctional MAP kinase (1, 2, 3). MAP kinases are major components of pathways controlling embryogenesis, cell differentiation, cell proliferation, and cell death. This review contains a historical overview of the mammalian MAP kinases that have been studied to date, their regulatory cascades, and some of their functions. Current research on these pathways is described in detail, and emphasis is on nonenzymatic mechanisms and findings from gene disruption studies. Much of the review highlights work on extracellular signal-regulated kinases 1 and 2 (ERK1 and ERK2). Some mechanisms in yeast MAP kinase cascades that might offer insight into the mammalian pathways are also included.

	II. Overview of Regulation and Properties of MAP Kinases

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
Between 1989 and 1991 the sequences of the first MAP kinases, Kss1p and Fus3p in the pheromone response pathway of the budding yeast and the mammalian MAP kinases ERK1, ERK2 and ERK3, became available, revealing that these enzymes were members of a newly identified protein kinase family (4, 5, 6, 7, 8). The activities of ERK1 and ERK2 had been routinely measured with two substrates, myelin basic protein (MBP) and microtubule-associated protein-2 (MAP2); as a result, they had been called MBP and MAP2 kinases (9, 10). The MAP acronym was retained, but with a different meaning: the name mitogen-activated protein kinase was assigned to these enzymes to acknowledge the fact that they had first been detected as mitogen-stimulated tyrosine phosphoproteins in the early 1980s, during an intense search for tyrosine kinase substrates (11).

The concept that there were multiple MAP kinases with distinct regulation and functions arose from the description of additional pathways found initially in yeast, the high osmolarity glycerol (HOG) pathway containing the MAP kinase HOG1 and the cell wall pathway containing the kinase MPK1, and then in metazoans with the discovery of c-Jun N-terminal kinases/stress-activated protein kinases (JNK/SAPKs), p38 enzymes, and others discussed below (12, 13, 14, 15, 16, 17, 18). Extensive analyses of sequence relationships among these kinases have been published recently (19, 20, 21).

MAP kinases have some features in common with the cyclin-dependent kinases (cdks). These include an insert of unknown function between subdomains X and XI of the catalytic core and a preference for serine or threonine residues followed by proline in their substrates. Among the distinguishing features of the MAP kinases are activation directly by phosphorylation in the absence of a regulatory subunit, and usually two activating phosphorylation sites in the kinase activation loop, one a tyrosine and one a threonine, separated by a single, variable residue (Fig. 1 and Table 1). Kinases such as KKIALRE, for which cDNAs were first cloned as homologs of the cdk cdc2, KKIAMRE, and the nemo-like kinase NLK, identified by its similarity to Drosophila nemo, appear intermediate between the MAP kinase and cdk families and may function in a manner distinct from the majority of MAP kinases discussed in this review (22, 23, 24, 25). Analysis of the sequence of the Caenorhabditis elegans genome reveals 15 MAP kinase family members (26). Nearly 20 MAP kinases are now known in mammals and more are anticipated (Table 1).

View larger version (51K): [in this window] [in a new window]   Figure 1. Unphosphorylated structure of ERK2. ATP binds in the interior of the active site at the domain interface and protein substrates are bound on the surface. MAP kinase activity is controlled by phosphorylation of two residues, a tyrosine (185 ) and a threonine (183 ), that are in a surface loop known as the activation loop or phosphorylation lip. Phosphorylation of ERK2 or other MAP kinases on a single residue does not cause a substantial increase in activity, nor does replacement of the phosphorylation sites with acidic amino acids (77;445). This is probably because of the nature of the conformational changes that must occur upon phosphorylation. The sulfate ion that lies in the position occupied by phosphotyrosine in the active structure is shown. The aspartic acid residues (D316 and D319) in the proposed binding site for D domains are also indicated.

	III. MAP Kinases Are Activated by Phosphorylation Cascades

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

View larger version (25K): [in this window] [in a new window]   Figure 2. MAP kinase cascades. Enzyme cascades shown are described in the text.

Another feature of MAP kinase cascades derives in part from the dual phosphorylation of the MAP kinase by the MEK. In the case of ERK1/2, the kinases are phosphorylated on tyrosine before threonine is phosphorylated both in vitro and in cells (38, 39). The result of this nonprocessive phosphorylation is the establishment of a threshold (40, 41). The tyrosine-phosphorylated proteins are not active but must accumulate before phosphorylation of threonine. Once this accumulation threshold has been reached, the kinases are rapidly converted to the active state, as threonine is phosphorylated. It may be generally true that the MEK-MAP kinase step exists to enhance the cooperativity of activation of the MAP kinase and to allow modulation by other signaling events, in addition to or rather than amplifying the MEK signal.

MEKs are also activated by phosphorylation of two residues, either serine or threonine, in their activation loops (42, 43). At least in the case of MEK1, either phosphorylation alone significantly increases activity, in contrast to the effects of the phosphorylations on the MAP kinase. Nevertheless, activation of MEK also displays cooperativity at least in the Xenopus oocyte system as elucidated in detail by Ferrell and Machleder (44).

The MEK kinases (MEKKs) that activate MEKs are many and diverse. Enzymes with MEKK activity in metazoans include several relatives of the yeast MEKK Ste11p; several distant relatives of another yeast kinase Ste20p, which lies upstream of Ste11p; and Raf isoforms and Mos, which have no homologs in yeast (45, 46, 47). Few generalizations can yet be made about regulation of these MEKKs themselves, except that they may be subject to multiple regulatory inputs. Most, if not all, of these MEKKs are not abundant, suggesting that the MEKK-MEK step amplifies the signal emanating from a given MEKK.

	IV. Signal Integration and Specificity

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
Interactions among the cascades occur in numerous ways to integrate responses and moderate outputs. Abundant evidence demonstrates that MAP kinases have overlapping substrate specificities (1, 48, 49). The resulting activities of the substrates reflect the cumulative extent of phosphorylation on all regulatory sites, which may be shared among multiple protein kinases. MAP kinase cascades form complexes that facilitate their activation and impact their localization, specificity, and targets (50, 51, 52). Potential scaffold proteins and adaptor or linker molecules have been found for some of the pathways. Regulation of complex formation provides yet another site for cross-talk between signaling pathways. Several MEK family members contain sites that are phosphorylated by kinases in other pathways; these events may influence the ability of MEKs to interact in complexes, for instance (32, 53, 54). Integration may also occur early in the signaling pathway and at the top of the kinase module. Some MEKKs may regulate more than one MAP kinase cascade, and some cascades may be controlled by several, unrelated MEKKs.

	V. Mammalian MAP Kinase Cascades

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
A. The ERK1 and ERK2 cascades
ERK1 and ERK2 are proteins of 43 and 41 kDa that are nearly 85% identical overall, with much greater identity in the core regions involved in binding substrates (5, 7). The two phosphoacceptor sites, tyrosine and threonine, which are phosphorylated to activate the kinases, are separated by a glutamate residue in both ERK1 and ERK2 to give the motif TEY in the activation loop (55). Both are ubiquitously expressed, although their relative abundance in tissues is variable. For example, in many immune cells ERK2 is the predominant species, while in several cells of neuroendocrine origin they may be equally expressed. They are stimulated to some extent by a vast number of ligands and cellular perturbations, with some cell type specificity (1). In fibroblasts (the cell type in which the generalizations about their behavior and functions have been developed) they are activated by serum, growth factors, cytokines, certain stresses, ligands for G protein-coupled receptors (GPCRs), and transforming agents, to name a few. They are highly expressed in postmitotic neurons and other highly differentiated cells (7). In these cells they are often involved in adaptive responses such as long-term potentiation (56, 57, 58).

Recently an ERK1 splice variant, ERK1b, was found as an immunoreactive band that migrates more slowly than the ubiquitously expressed form of ERK1 (60). It is possible that ERK1b corresponds to the protein species originally named ERK4 (62). An alternatively spliced form of ERK2, lacking some residues from the N terminus, has also been reported; overexpression suggested that it was selectively membrane localized (59, 61). The three-dimensional structures of ERK2 in its unphosphorylated and phosphorylated states have been determined and reviewed elsewhere (63, 64, 65, 66, 67).

1. MEK1 and 2. ERK1 and ERK2 are activated by a pair of closely related MEKs, MEK1 and MEK2 (28, 29, 30, 68, 69, 70, 71). Both of these MEKs have been shown to fully activate ERK1/2 in vitro (72, 73). Upon dual phosphorylation, ERK1/2 activities increase by well over 1,000-fold to specific activities of 1–2 µmol/min/mg protein. The largest effect appears to be due to an increase in V_max; changes in K_m for substrates are small (74, 75). The stoichiometry of phosphorylation of ERK1/2 by MEK2 more readily approaches 2 mol phosphate/mol ERK than does phosphorylation by MEK1. Haystead and co-workers (76) purified a factor that enhances phosphorylation of ERKs by MEK1. The biological importance of this molecule remains uncertain. Replacement of the two ERK2 phosphorylation sites with acidic residues does not elevate the activity of the protein (77).

Phosphorylation of MEK1 on both sites has been reported to stimulate its activity by more than 7,000-fold; as noted above, phosphorylation of either site alone produces a significant increase in activity (42, 78). Both V_max and K_m values change; K_m decreases by nearly 100-fold. Substitution of the two sites of phosphorylation with acidic residues increases their activity; deletions in the N terminus increase activity even more. The combination of these two changes yields constitutive MEK1/2 mutants nearly as active as phosphorylated wild-type proteins (78). These MEK mutants, most often MEK1R4F, have been used in many systems to infer events associated exclusively with the ERK cascade (79, 80). It has been assumed, from lack of evidence to the contrary, that MEK1/2 have no other substrates. Although this may not be the case, at this time no other MEK1/2 substrates have been identified.

2. Raf isoforms. Of all the known MEKKs, Raf isoforms and Mos are perhaps the only ones that phosphorylate MEKs in a single cascade. These proteins appear to phosphorylate only two MEK family members, MEK1 and MEK2, placing these MEKKs exclusively in the ERK1/2 MAP kinase cascade (81, 82, 83).

The Raf family of protein kinases is composed of A-Raf, B-Raf, and Raf-1 (or c-Raf) (84, 85). Each isoform contains three conserved regions, termed CR1, CR2, and CR3. The first two conserved regions are located in the amino terminus and have been implicated in regulating the Raf catalytic domain, because their deletion creates a mutant of Raf-1 that either has constitutively high activity or can be activated in a Ras-independent manner (see below) (86, 87). The kinase domain is located in CR3. Raf-1 is ubiquitous; highest expression of B-Raf occurs in neuronal tissue and testis; and A-Raf appears to function primarily in urogenital tissue.

Most studies have focused on Raf-1. Raf-1 regulation is complex, involving protein-protein interactions, phosphorylation of tyrosine, threonine, and serine residues, and cellular localization (84). These multiple modes of regulation allow Raf-1 to fluctuate through a number of graded activity states. Raf exists as part of a multiprotein complex composed of Raf-1 or B-Raf, heat shock protein 90 (hsp90), p50, and an indeterminate number of 14–3-3 proteins (88, 89, 90, 91, 92, 93, 94, 95). 14–3-3 Appears to stabilize Raf-1 in both low and high activity conformations depending upon Raf phosphorylation state and interaction with other regulatory proteins such as GTPliganded Ras. 14–3-3 May also serve to regulate Raf-1 signaling specificity by recruiting Raf-1 to higher order protein complexes. Disruption of hsp90-p50 binding to Raf, through the use of pharmacological agents such as geldanamycin and dexamethasone and mutants of p50 that are deficient for hsp90 binding, disrupt Raf-dependent signaling to downstream effectors (92, 93, 96). Multiple lines of evidence indicate that geldanamycin’s effects are due to a depletion of Raf in the cell. Geldanamycin does not affect the ability of Raf to form complexes with an upstream activator Ras or reduce its specific activity upon epidermal growth factor (EGF) stimulation. Coexpression of p50 with Raf in Sf9 cells increases Raf activity and potentiates v-src activation of Raf (92). It is uncertain as yet whether p50 is an active regulator or whether it works passively in concert with hsp90 to stabilize Raf.

There are significant differences in regulation of Raf isoforms. One notable difference between Raf-1 and B-Raf is their differential regulation by the small G proteins Ras and Rap1a (97, 98, 99, 100, 101). Raf-1 is activated by H-, K-, and N-Ras. It has been suggested that proliferation in nontransformed cells may be controlled primarily by N-Ras, but most studies have employed H- or K-Ras (102). Although Raf-1 also interacts with Rap1a, the function of this interaction is uncertain, because no increase in activity is seen. On the other hand, B-Raf is activated by both Ras and Rap. In neuronal model systems such as PC12 cells, activation of B-Raf by Rap1 may be the dominant mechanism (Refs. 97, 100 ; G. Landreth, personal communication). This functional difference has been attributed to the cysteine-rich domains (CRDs) of these proteins. Swapping the Raf-1 and B-Raf CRDs allows for activation of Raf-1 by Rap1 and eliminates Rap activation of B-Raf (103).

The phosphorylation state of Raf-1 is influenced by multiple protein kinases, including Src, protein kinase C (PKC) family members, the p21 (Rac/Cdc42)-activated protein kinase PAK, and Akt (also called protein kinase B). The PAK and Src phosphorylation sites are located N-terminal to the catalytic domain at serine 338 and tyrosines 340 and 341, respectively (104, 105, 106). These sites have each been found to increase activity when phosphorylated and may do so in an interactive manner, depending on the signal context (105, 107, 108). The activation loop residues, serine 497 and 499, were the originally reported PKC phosphorylation sites (109); however, mutation of these sites has no discernible impact on Raf stimulation by serum (83). Wolfman and colleagues have recently found that PKC forms a stable complex with Raf-1 and phosphorylates serine 338 (Hamilton, M., M. K. Cathcart, and A. Wolfman, submitted), the same site as PAK (105, 110). Other serine 338 kinases have been proposed. Down-regulation of PKC blocks the phorbol ester activation of Raf-1 but has no effect on activation by EGF, one of many lines of evidence indicating multiple, independent mechanisms for activation of Raf-1.

Serine 259 is part of a putative 14–3-3 binding site (111, 112, 113). Phosphorylation of this serine may stimulate binding of 14–3-3 which, when bound to this region of Raf, has an inhibitory effect on Raf-1 activity. Mutation of this in vivo phosphorylation site to alanine creates an active mutant of Raf-1 (104, 114). Akt has been shown to phosphorylate serine 259 in MCF-7 breast cancer cells (115). Forced down-regulation of ERK1/2 in C2C12 cells cultured in serum can stimulate early stages of myotube differentiation (116). Akt may reduce Raf-1 activity in a number of contexts such as during C2C12 myoblast differentiation (117). The site of Raf-1 phosphorylation by Akt in C2C12 cells was not directly mapped. Instead, the authors show that in insulin-like growth factor I (IGF-I)-treated, postdifferentiated myotubes, there is reduced phosphorylation of serine 338 when a kinase active mutant of Akt is expressed. These methods of regulation are not mutually exclusive; however, Akt’s ability to inhibit Raf-1 activity may vary depending on cell type. It is also interesting to note that in C2C12 cells an Akt-Raf-1 association only occurs during differentiation and is dependent on Akt kinase activity whereas in MCF-7 breast cancer cells the association of the two proteins appears to be constitutive. Further study is required to reconcile these differences and determine the generality of Akt-mediated down-regulation of Raf-1 during physiological processes.

TC21, a Ras family member, was previously thought to use a Raf-1-independent mechanism to activate ERKs 1 and 2; however, both B-Raf and Raf-1 displayed increased kinase activity in TC21-transformed NIH 3T3 cells (118). Also, overexpressed TC21 coimmunoprecipitated with overexpressed Raf-1 or B-Raf; it interacts with the two isoforms in a directed two-hybrid assay; and disruption of the TC21-Raf-1 interaction abolished the ability of TC21 to transform cells.

The three-kinase cascade, so well defined for the ERK1/2 module, is more difficult to identify as a discrete unit for other MAP kinase cascades at the present time. This is in part due to the capacity of many MEKKs to phosphorylate many MEKs in vitro and to activate many MAP kinases when overexpressed. Thus, the other MEKKs that are currently known will be discussed as a group after the description of the MAP kinases and their probable MEKs. The MAP kinases and related enzymes are listed in Table 1. Those not mentioned below appear in the overview section.

B. c-Jun N-terminal kinases/stress-activated protein kinases (JNK/SAPK)
A form of JNK/SAPK was first purified as a 54-kDa MBP kinase from livers of cycloheximide-treated rats (119). Shortly thereafter, JNK/SAPKs of 46 and 54 kDa were purified by affinity adsorption to a c-Jun fusion protein (120). Isolation of cDNAs encoding these enzymes and subsequent analysis of their expression revealed three genes encoding proteins with 10 or more alternatively spliced forms (14, 15, 121). Within the core catalytic domains, JNK1/SAPK, JNK2/SAPK, and JNK3/SAPKß are more than 85% identical. Based on mutagenesis studies, JNK/SAPKs are activated upon phosphorylation of two sites, a tyrosine and threonine, like other MAP kinases (15). In all JNK/SAPKs these residues are separated by a proline residue to give the motif TPY in the activation loop. They are activated by cytokines, certain ligands for GPCRs, agents that interfere with DNA and protein synthesis, many other stresses, and to some extent by serum, growth factors, and transforming agents. The alternatively spliced forms and their properties have been reviewed in detail elsewhere (121A ).

1. MKK4 and MKK7. Two MEK family members, MKK4 (SEK1, MEK4, JNKK1, SKK1) and MKK7 (MEK7, JNKK2, SKK4), have been implicated in JNK/SAPK pathways. Both were identified initially by cDNA cloning strategies rather than by purification (122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132). One approach identified MKK7 as a two-hybrid binding partner of MEK1, although the significance of their association is not known. Unlike MKK4, MKK7 will rescue a lethal mutation in a Drosophila MEK (hemipterous) that is required for dorsal closure (126). Both MKK4 and MKK7 have the ability to phosphorylate p38 family members in vitro and when overexpressed, although JNK/SAPKs are the preferred substrates (133). JNK/SAPK activation is impaired in cells from animals in which the MKK4 gene was disrupted, but changes in p38 activation have been found that are dependent on cell type (Ref. 134 ; see below). The prevailing view that MKK4 acts exclusively in JNK/SAPK cascades remains an open question. JNK/SAPKs are still activated by certain stimuli in MKK4 -/- cells, consistent with the conclusion that MKK7 is also linked to JNK/SAPK cascades. In vitro MKK4 preferentially phosphorylates the tyrosine residue in the TPY activation loop motif of JNK/SAPKs, and MKK7 preferentially phosphorylates the threonine residue. Based on these specificity differences, it has been suggested that these kinases cooperate to activate JNK/SAPKs, perhaps allowing for signal integration (135, 136). Results also indicate that phosphorylation of threonine may be most important for activity changes of JNK3 (136).

C. p38 Pathways
p38 was discovered independently in three contexts. It was found as a tyrosine phosphoprotein present in extracts of cells treated with inflammatory cytokines (17); as the target of a pyridinyl imidazole drug that blocked production of tumor necrosis factor- (TNF) and as such was called cytokine-suppressive antiinflammatory drug-binding protein or CSBP (16); and as a reactivating kinase for MAP kinase-activated protein (MAPKAP) kinase-2 (18). Cloning strategies rather than biological approaches were used to identify the other three genes that encode members of the p38 subfamily: p38ß (or p38–2), p38 (ERK6 or SAPK3), and p38 (SAPK4) (137, 138, 139, 140, 141, 142, 143). All of these kinases contain the sequence TGY in their activation loops. A splice variant of p38ß lacks the eight-amino acid insertion unique to ß. p38 And ßisoforms are sensitive to pyridinyl imidazole inhibitors, but - and -isoforms are resistant to these drugs (141, 142). A variety of agents including cytokines, hormones, GPCRs, osmotic and heat shock, and other stresses activate p38 family members. In some contexts p38 family members have apparently opposite actions (144, 145).

A sixth protein, Mxi, is a splice variant of p38 in which the last 80 residues have been replaced by a novel 17-residue C terminus (146). Mxi was isolated from a two-hybrid screen with the c-Myc binding partner Max. Both Mxi and p38 bind Myc. The change in the C-terminal residues confers unique properties on Mxi. Unlike p38, Mxi is activated not only by stresses but also by growth factors (146A ). In contrast to p38, Mxi is relatively insensitive to pyridinyl imidazole compounds; Mxi also displays a reduced affinity for p38 substrates. Crespo and colleagues showed that deletion of the 80 C-terminal residues from p38 yielded a mutant with properties similar to Mxi. An explanation may be proposed for these findings from the crystal structures of MAP kinases (64, 147, 148, 149). In these enzymes the C-terminal residues, deleted in Mxi, make intimate contacts with the N-terminal domain of the kinase catalytic core. These contacts undoubtedly influence the interaction with ATP and other compounds that bind in the ATP pocket, such as pyridinyl imidazoles.

Two MEK family members, MEK3 and MEK6, have high activity toward p38 MAP kinases (123, 150, 151, 152). MEK3 appears to favor phosphorylation of p38 and p38ß isoforms, while MEK6 phosphorylates all p38 family members well (150). Both will also phosphorylate JNK/SAPK isoforms. MEK6 phosphorylates p38/ERK2 chimeras, and NLK (see below) in vitro, suggesting that it has a broader specificity than other MEKs (153, 154). The physiological implications of this broader specificity are not clear at this time.

D. Other MAP kinases
1. ERK3 isoforms. cDNAs encoding rat ERK3 were isolated from a library using a probe derived from ERK1 (7). A human cDNA predicted a second ERK3-like kinase, also 63 kDa, about 75% identical to ERK3 (59). These kinases are nearly 50% identical to ERK1 and ERK2 in the core catalytic domain, and both contain C-terminal extensions of approximately 200 residues. For the purposes of discussing them here, the first of these will be designated as ERK3 and the second as ERK3ß. Subsequently, Flier and colleagues isolated a human cDNA that predicted a 97-kDa protein 100% identical to ERK3 over their shared lengths but lacking a stop codon and longer by nearly 300 residues (155). Immunoblotting with antibodies specific for ERK3 revealed proteins of 63, 95, and 160 kDa in multiple rat tissues and several cell lines, consistent with multiple species predicted by the cDNAs (156). A clone encoding a 100-kDa form of ERK3 was recently isolated by Meloche and colleagues (157) from mouse and a single genomic locus was mapped. Database analysis indicates that there may be several loci encoding ERK3-like molecules. Genes encoding ERK3 homologs have not been found in the genomes of yeast or nematodes, suggesting that ERK3 and ß may have arisen from a relatively late gene duplication (26, 158).

Despite the similarity to ERK1/2, ERK3 and -ß have some features that are different from other family members. The phosphorylation site motif in the activation loop of ERK3 isoforms has a single phosphoacceptor site, serine189 in ERK3 in the sequence SEG. Glycine replaces the usual tyrosine phosphorylation site found in most other MAP kinases. ERK3 autophosphorylates, but data for other ERK3 substrates are weak (156). Several MAP kinases are largely cytoplasmic in unstimulated cells and translocate to the nucleus when cells are stimulated. In contrast, ERK3 is highly concentrated in the nucleus under all conditions examined (156) but the mechanism is unknown; ERK3 lacks a consensus nuclear localization sequence. A kinase that binds to and phosphorylates ERK3 on serine189 has been described but its molecular identity is unknown (159). This activity phosphorylates ERK3 but not other MAP kinases.

2. ERK5. ERK5 was identified independently by two groups. One used a two-hybrid screen with an upstream activator MEK5 as the bait; the other used a degenerate PCR strategy to clone novel MAP kinases (160, 161). Thus, the putative upstream activator MEK5 was found ahead of this MAP kinase. Among the most intriguing features of ERK5 is its size, 816 amino acids, due to a stretch of approximately 400 amino acids C-terminal to the kinase domain. When comparing the primary sequence of the catalytic domain of ERK5 to other mammalian MAP kinases, it appears to be most like ERK2. The 400-residue C terminus, however, neither displays sequence similarity to any known proteins nor has a known function, although it contains 10 consensus sites for MAP kinase phosphorylation. These phosphorylation sites may be autophosphorylated, consistent with the dramatic increase in autophosphorylation ERK5 displays when it is in a high activity state (162). Whether autophosphorylation plays an integral role in ERK5 function within the cell remains to be seen. The C terminus also contains a potential cytoskeletal targeting motif; however, there is no evidence supporting this putative function (160).

In mammals, ERK5 is ubiquitously expressed. Like the other MAP kinases, ERK5 activity is regulated by a wide variety of proliferative and cell-stressing agents. The proliferative stimuli include serum, EGF, nerve growth factor (NGF), lysophosphatidic acid (LPA), and phorbol ester (163, 164, 165). The ability of these agonists to activate ERK5 is Ras-dependent in some cell types; EGF activation of ERK5 requires MEKK3 activity in HeLa cells (Refs. 163, 164, 166 ; see below). The stress stimuli include sorbitol, H₂O₂, UV irradiation, vascular shear stress, and ischemia (164, 165, 167, 168, 169). These stimuli may sometimes exert their activity through Src (170). Cellular requirements for ERK5 activity have been better defined in proliferation models. Dominant negative forms of ERK5 can inhibit EGF-stimulated proliferation and RafBXB-stimulated focus formation in 3T3 cells (163, 171).

English and colleagues (162) examined the regulation of the catalytic domain through truncation of its C terminus. The ERK5 catalytic domain is activated by V¹²Ras and an active mutant of MEK5, MEK5DD (the two sites of activating phosphorylation are replaced with acidic residues), as determined by an increase in activity toward substrates. In vitro, the ERK5 catalytic domain expressed in bacteria is phosphorylated by immunoprecipitated MEK5DD on its TEY motif and displays an increased activity toward substrate, consistent with the behavior of the majority of MAP kinase family members, which are only slightly larger than a core catalytic domain (Pearson, G., and M. H. Cobb, unpublished). Coexpression of ERK5 with MEK5DD in cells increases ERK5 activity. The kinase domain displays the expected specificity of activation in that other MEK family members such as MEK1 fail to phosphorylate it in vitro or increase its activity when coexpressed in 293 cells.

ERK5 can affect cellular activity through phosphorylation of the MADS box transcription factors, myocyte enhancer factor 2A and C (MEF2A and C), and the ETS-like transcription factor SAP1a (164, 165, 172). The ability of ERK5 to activate MEF2 isoforms appears to allow it to positively regulate intracellular concentrations of c-Jun (172). Additional downstream effectors are likely to exist.

MEK5 is upstream of ERK5. MEK5 was identified by two groups using cDNA cloning strategies (160, 173). There are multiple splice variants including 50-kDa - and 40-kDa ß-isoforms. MEK5 is particulate and primarily expressed in liver and brain; the ubiquitously expressed ß-isoform is cytosolic. The only known substrate of MEK5 is ERK5; thus, effects of MEK5 have been attributed to its ability to activate ERK5.

According to primary sequence alignment, MEK5 is most closely related to MEKs 1 and 2. Perhaps as a consequence of this relationship, it is also inhibited by PD98059 and U0126, two compounds that have been considered highly selective inhibitors of MEK1 and MEK2 (Ref. 164 ; see below). At low concentrations, the effects of these inhibitors may be primarily on MEK1/2, since the K_i for MEK5 is significantly higher.

In spite of the similarity to MEK1/2, MEK5 is not phosphorylated or activated by Raf-1 (162). Although Raf-1 is unable to increase MEK5 activity, MEK5 is intimately involved in Raf-1 signaling. Kinase-defective MEK5, MEK5KM, can inhibit RafBXB-stimulated focus formation in 3T3 cells, whereas a constitutively active form of MEK5, MEK5DD, can synergize with RafBXB to form foci (171). MEK5DD cannot stimulate focus formation when expressed alone. MEK5KM can also inhibit focus formation induced by the Cot protooncogene product, also known as Tpl-2, and coexpression of Tpl-2 with MEK5 increases the phosphoserine content of MEK5 (174). Direct effects of Tpl-2 on MEK5 activity have not been demonstrated. The only MEK5 kinase identified thus far is MEKK3 (166).

3. ERK7. A cDNA encoding ERK7 was isolated by Rosner and colleagues (175). ERK7 is a 61-kDa MAP kinase with a TEY motif in the activation loop, like ERK1, ERK2, and ERK5. ERK7 is not activated by stimuli that activate ERK2 or the stress-responsive kinases, but appears to be constitutively activity in serum-starved cells. A role in growth inhibition has been proposed for ERK7. Its long C terminus has been suggested to be required for the localization and high basal activity of this protein. A cDNA encoding the protein CLIC3 was isolated using the tail of ERK7 as bait in a yeast two-hybrid screen (176). CLIC3 is related to human intracellular chloride channel proteins.

4. NLK. NLK was identified by Erikson’s group (24) as a mammalian relative of Drosophila nemo. This kinase has properties that place it between the MAP kinases and the cdks. Although it is nearly 45% identical to ERK2, the dual phosphorylation motif TXY in the activation loop is absent, and instead a single phosphorylation site in the sequence TQE, most similar to the cdks, is present. Nevertheless, NLK appears to lie in a MAP kinase cascade that negatively regulates Wnt signaling (154, 177). Studies in C. elegans have demonstrated that an NLK homolog lit-1 is activated by the MEKK Mom-4. Mom-4 is a homolog of TAK1, described below as an MEKK for the p38 MAP kinase module. In transfected cells TAK1 can enhance the activity of cotransfected NLK. Although a MEK specific for NLK has not been reported, NLK is activated in vitro by MKK6, a TAK substrate. Thus, it is possible that TAK1 and MKK6 may be normal cellular regulators of both p38 and NLK.

5. MOK. MOK has approximately 30% identity to members of the MAP kinase family and equivalent identity to the cdk family (178). Strikingly, however, MOK contains the TEY motif in its activation loop that is typical of MAP kinases. It has been shown to be activated by okadaic acid and phorbol ester, suggesting that it may be controlled by a kinase cascade. Its relatives include male germ cell-associated kinase (MAK) and the MAK-related kinase, MRK (25, 179, 180).

E. MEKKs, the first tier in the kinase cascade
A specific MEKK enzyme may regulate either a single or multiple MEKs depending upon the enzymatic specificity of the MEKK, the cellular and subcellular distribution of the signaling components, the formation of protein complexes, and the activating stimuli. Consequently, significant differences in both the magnitude and kinetics of MAP kinase activation may occur in response to a given agent under different circumstances. Many kinases acting at the MEKK level have been identified, adding to the complexity of unraveling signaling mechanisms. There is no apparent similarity among these proteins outside of their kinase catalytic domains. The relative contribution of each MEKK to the activation of individual MAP kinases, with the possible exception of Raf in the ERK1/2 module, is unclear.

Aside from Raf isoforms, the first of these to be isolated was the 195-kDa protein MEKK1. It is one of a family of molecules most closely related to the yeast kinase Ste11p, all of which contain C-terminal kinase domains and N-terminal regions of variable length (45). In their catalytic domains, MEKK2 and MEKK3, each approximately 70 kDa, and MEKK4, about 150 kDa, are nearly 50% identical to MEKK1 (181, 182, 183, 184).

The other enzymes with MEKK activity mentioned next are less similar with identities to MEKK1 generally in the 30–40% range. The following MEKK level kinases activate JNK/SAPKs when overexpressed or by in vitro reconstitution with MEKs: MEKKs(1, 2, 3, 4) (181, 182, 183, 184), MAP three kinase (MTK1) (181, 183, 184, 185, 186), Tpl-2/Cot (187), dual leucine zipper kinase (DLK) (188), mixed lineage kinase MLK2/MST (189), MLK3/PTK-1/SPRK (190, 191), transforming growth factor-ß (TGFß)-activated kinase (TAK1) (192), apoptosis signal-regulating kinases (ASK1)/MAPKKK5 (193, 194) and ASK2/MAPKKK6 (195), and thousand and one amino acid kinases 1,2 (TAOs1, 2) (196, 197). Of these, MEKKs(1, 2, 3) and Tpl-2 can also activate the ERK1/2 pathway (187); MEKK3 and Tpl-2 also activate the ERK5 pathway (172, 174); and TAK1, ASK1, TAOs1/2, and MTK1 also activate the p38 pathway (194, 196, 197).

Unraveling the relationships of these MEKKs to the MAP kinases they activate has been a daunting task. Identification of the intrinsic enzymatic specificities, the distribution, and the phenotypes of animals and cells with these MEKK genes disrupted should begin to help decipher their cascade specificity and their functions. The function of MEKK1, the first of these enzymes isolated, is still in question. It has been implicated in activation of JNK/SAPK, ERK, and p38 MAP kinase pathways, as noted above, and in the activation of nuclear factor-B (NF-B) (198, 199). In vitro MEKK1 phosphorylates MEKs 1, 2, 3, 4, 6, and 7 (73, 200, 201, 202, 203). However, despite the fact that the recombinant protein phosphorylates MEKs 1 and 2 on the same sites as Raf-1, it does so poorly relative to the phosphorylation of MEK4 in the JNK/SAPK pathway, consistent with the finding that signaling to JNK/SAPKs is most affected in cells lacking MEKK1 (Refs. 204, 205, 206 ; see below).

Although the classical MAPK module is a three-tiered kinase cascade, a fourth kinase may act directly upstream as an activator of the MEKKs. This was discussed earlier for Raf. Kinases implicated in JNK/SAPK activation at the MEKK kinase level include PAKs 1–4 (207, 208, 209), germinal center kinase (GCK) (210, 211), GCK-related kinase (KHS/GCKR) (212), GCK-like kinase (213, 214), hematopoietic progenitor kinase 1 (HPK1) (215), and Nck-interacting kinase (NIK) (216).

Both small G proteins and heterotrimeric G proteins can activate MAP kinase cascades as discussed in more detail for ERK1/2 below (217). Activation of JNK/SAPKs and p38 in response to interleukin (IL)-1ß, muscarine, bradykinin, and heterotrimeric G protein ß subunit complexes may be mediated by Rho family members Rac and Cdc42 (209, 218, 219, 220).

	VI. Activation of ERK1/2 and Other MAP Kinases from the Cell Surface

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
Perhaps the most well defined signaling pathway from the cell membrane to ERK1 and ERK2 is that used by receptor tyrosine kinases (reviewed in Refs. 221, 222). Stimulation of these receptors by the appropriate ligand results in an increase in receptor catalytic activity and subsequent autophosphorylation on tyrosine residues. Phosphorylation of these receptors results in the formation of multiprotein complexes whose organization dictates further downstream signaling events. Quite often one of these functions is the activation of the monomeric G protein Ras. This is achieved by the recruitment of adaptor proteins, such as Shc and Grb2, to the receptor through interactions between their SH2 domains and phosphotyrosine residues. The guanine nucleotide exchange factor (GEF) Son of Sevenless (Sos) then becomes engaged with the complex and induces Ras to exchange GDP for GTP. GTP-liganded Ras is capable of directly interacting with a number of effectors, including Raf isoforms, of which the best characterized is Raf-1. As discussed before, Ras binding to Raf may result in conformational changes in Raf that increase its kinase activity or simply provide the proper environment for Raf-1 signaling (223, 224, 225, 226, 227, 228). Localization of Raf to the plasma membrane may also allow protein kinases such as Src, PKC, and PAK to further modify Raf to increase its activity (105, 106, 109, 110, 228). The increase in Raf activity is subsequently transduced through the MEK-ERK module.

Signaling to ERKs by GPCRs also involves modulation of Raf activity; however, the mechanisms employed by these receptors are widely varied. The existence of multiple classes of G proteins, the ability of some receptors to activate more than one class of G protein, and cell type-specific mechanisms contribute to the diversity. For clarity, only the general trends observed for a few specific classes of G proteins will be discussed. There are several more detailed recent reviews (1, 67).

Signals transmitted from receptors through Gs are particularly diverse, consistent with the variety of effects on ERK activity evoked by elevation of cAMP concentration. cAMP-dependent protein kinase (PKA) has been reported to reduce Raf-1 activity through direct phosphorylation of serine 43 and serine 621 in some situations (229, 230, 231, 232). On the other hand, PKA can also phosphorylate Rap1a, which may positively influence ERK through activation of B-Raf in cells of neuronal origin (97, 233). Activation of Rap1a by a cAMP-binding Rap1 GEF, or by some other means, has been suggested to inhibit ERK activity through Rap1a-dependent sequestration of Raf-1 (99, 234). The particular effect of Rap1a activation on ERK may be determined by the expression level of B-Raf. Lefkowitz and colleagues (235) have reported that isoproterenol treatment of 293 cells overexpressing ß2- adrenergic receptors stimulates a PKA-dependent switch of receptor coupling form Gs to Gi, and that ERK activation is through the Gi pathway.

In Gi-dependent ERK activation, free ß-subunits may be the active signal transducers, reminiscent of their role in the yeast mating response. This is evidenced by overexpression studies showing ß-subunits are sufficient to activate ERKs and a ß sequestering peptide reduces ERK stimulation by Gi-coupled receptors (236, 237). In one proposed model, ß stimulates a Src family kinase activity in a PI-3 kinase dependent manner (238). The Src family kinase may then phosphorylate a tyrosine kinase receptor, PYK2, or focal adhesion kinase (FAK), to create SH2 domain binding motifs (239, 240, 241). Then, analogous to the signaling mechanism used by receptor tyrosine kinases described above, a Shc-, Grb2-, and Sos-containing complex is formed at the membrane to activate Ras and, in turn, Raf-1. ERK activation in cell types where PI-3 kinase expression is low may be dependent on alternative means to activate Src or PYK2 (239, 242).

Gq activation of ERK2 is often a PKC-dependent process, which may be Ras-dependent or independent (240, 243, 244, 245, 246, 247). The Gq effector is PLCß, which generates inositol triphosphate (IP₃) and diacylglycerol (DAG) through the cleavage of phosphatidylinositol 4,5-bisphosphate (PIP₂). Some isoforms of PKC are activated by DAG and the intracellular Ca released as a result of IP₃ production. PKC may then regulate Raf through direct phosphorylation, although this mechanism has not been fully characterized (109). As noted above, Wolfman and colleagues have found that PKC phosphorylates Raf-1 and increases its activity (Hamilton, M., M. K. Cathcart, and A. Wolfman, submitted). In PC12 cells, stimulation of Gq by receptors results in PYK2 and Shc phosphorylation. In Rat-1 cells phosphorylation of the EGF receptor, Neu, and Shc increases after treatment with endothelin. Thus, G proteins appear able to access multiple tyrosine kinases to activate the ERK pathway.

Overexpression of G12 activates Ras in 293 cells, although a strong link to the ERK pathway has not been made (245). GTPase-deficient mutants of G12 and G13 can stimulate focus formation in 3T3 cells, perhaps suggesting that some subset of the MAPK family is involved. Slight increases in JNK and p38 and -ß activity are seen when coexpressed with G13. JNK activity is also increased by G12 (248). Based on blocking experiments, Rac may be an intermediate (220).

	VII. Scaffolding and Its Role in Organization, Localization, and Specificity in MAP Kinase Cascades

Top Abstract I. Introduction II. Overview of Regulation... III. MAP Kinases Are... IV. Signal Integration and... V. Mammalian MAP Kinase... VI. Activation of ERK1/2... VII. Scaffolding and Its... VIII. Regulation of MAP... IX. Inactivation of MAP... X. Substrate Recognition and... XI. Substrates of MAP... XII. Biology of MAP... XIII. Gene Disruption... References

 
A. Complexes predicted from studies in yeast
1. The scaffold Ste5p. The first scaffolding protein identified that binds the kinase components of a MAP kinase pathway was the S. cerevisiae protein Ste5p. Mutants lacking Ste5p, as suggested by its name, are sterile. They fail to progress through the pheromone-induced mating pathway.

Two-hybrid studies from several laboratories revealed that Ste5p interacts with the three protein kinases of the MAP kinase module, either Fus3p or the similar MAP kinase, Kss1p, the MEK, Ste7p, and the MEKK, Ste11p (50, 51, 249). These results supported earlier biochemical studies with overexpressed protein, which also showed that Ste5p is a Fus3p substrate (250). Deletion analysis indicated that the binding sites for these kinases on Ste5p are distinct, suggesting that a multiprotein complex can form (50, 51, 249). Phosphorylation and activity states affect association of the kinases with Ste5p (251).

Epistasis analysis is consistent with the idea that Ste5p has an important function at more than one step of the cascade (252). Four properties of Ste5p have been discovered that are likely to be keys to its function. As discussed above, the first is its capacity to bind the components of a MAP kinase module. Second is its ability to interact with the upstream signal transducers. These signaling intermediates include the heterotrimeric G protein that is activated by pheromone binding to its receptor. In this pathway the ß-subunits (Ste4p and Ste18p) transduce the signal and do so in a manner that requires Ste5p (253); this is consistent with the finding that, when overexpressed, the -subunit inhibits the pheromone signal (254, 255). The interaction of Ste5p with the Gß-subunit is essential for activation of the MEKK Ste11p (255, 256). A close parallel exists in mammalian MAP kinase modules, which can be regulated by the Gi family through ß-subunit interactions. Third, Ste5p forms oligomers; these may promote complex activation (256, 257). Finally, Ste5p may also be an essential feature of the mechanism of localization of the kinases in the complex, because it must localize to the plasma membrane for cascade activation, yet its entry and exit from the nucleus are also required for pheromone-induced signaling (258).

2. Pbs2p. The formation of protein complexes may determine the regulation and functions of the associated MAP kinases. This idea was strongly suggested by findings in a second yeast MAP kinase module, which is part of a homeostatic response to osmotic shock (12, 259, 260). The HOG pathway contains the MAP kinase Hog1p, a relative of mammalian p38 (12). The MEK upstream in the pathway is Pbs2p (12, 260). Two osmosensors can activate the pathway through one of three different MEKKs—Ste11p, Ssk2p, or Ssk22p (261). A transmembrane osmosensor, Sho1p, activates Ste11p, the same MEKK that works in the pheromone response pathway (260). Thus, the mating and osmotic stress pathways share a common MEKK. When the osmotic pathway is activated, Ste11p binds to Pbs2p, which apparently scaffolds the MAP kinase module of the HOG pathway. Pbs2p binds Sho1p, Ste11p, and Hog1p (260). Ste5p is either absent or present in very low concentrations in diploid cells. Its presence may be required for recognition of Ste7p by Ste11p. Its absence may be an important factor in the specificity of Ste11p for Pbs2p, both the MEK and the scaffold, rather than Ste7p. Thus, the binding partners of Ste11p seem to determine the signals it transmits. This sort of mechanism may well hold in mammalian MAP kinase cascades.

B. Protein associations in mammalian MAP kinase pathways
Although the roles of Ste5p may not yet have been fully elucidated, the fact that Ste5p is required for the function of the MAP kinase module of the pheromone response pathway focused attention on the importance of assembly of cascade complexes. Furthermore, the control of specificity of Ste11p that appears to be exerted by its binding to either Ste5p or Pbs2p indicates that signal reception and transmission can be channeled by the formation of protein complexes. Another apparently essential function of Ste5p is its ability to move and become appropriately localized within cells.

Extrapolating from these findings in yeast, we expect that scaffold proteins have one or more key functions: 1) they may organize MAP kinase cascades for the efficient serial activation of the components; 2) they may restrict signal reception by recognizing signals from only a subset of possible receptor systems; 3) they may restrict the specificity of signal transmission by interacting with a limited repertoire of potential components of MAP kinase cascades; and 4) they may determine the output signal not only as a consequence of selectivity among MAP kinases, but also by localizing the cascade to selected sites of action, e.g., the transcription machinery, the microtubule cytoskeleton, etc.

While the inherent enzymatic specificity of Raf isoforms and MEK1/2 may be sufficient to account for their selectivity for ERK1/2 in cells, some of the mammalian MEKKs and MEKs implicated in the stress pathways appear less specific in vitro and when overexpressed in cells. For example, overexpression of Tpl-2 has been linked to the activation of at least five MAP kinase pathways, and MKK6 phosphorylates at least seven different MAP kinases in vitro. This apparent lack of enzymatic selectivity suggests that the assembly of these enzymes in complexes may restrict their actions to the MAP kinase or kinases in the complex and thereby determine their output signal. As a result of these considerations, the search for scaffolds for MAP kinase cascades has been intense.

1. Protein-protein interactions in the ERK1/2 cascade. Evidence from binding studies, cloning, and the behavior of mutant MEKs suggests that several protein-protein interactions are required for intracellular signal transmission through the ERK1/2 pathway. These interactions have proposed functions that lead to the localization of the kinases for signal reception, movement of the kinases to sites of action, substrate specificity and recognition, and temporal control of kinase activation. Several of these are described next.

2. Raf-1 forms complexes with Ras and MEK1. Wolfman and colleagues (223) showed that pull-down assays could be used to isolate Ras-Raf-1 complexes. MEK1 was also present in these complexes by virtue of a tight interaction with Raf-1, which can be demonstrated by coimmunoprecipitation. Raf-1 has been the subject of most studies in part because it is ubiquitous. Other Raf isoforms may display distinct properties. Less is known about binding interactions of Raf isoforms with MEK2.

3. Binding domains on MEK1/MEK2. MEK1 and MEK2 display one unique feature and one feature conserved in other MEK family members: both are required for efficient activation of their downstream MAP kinases in cells. The conserved feature is a stable binding site for MAP kinases, specifically ERK1/2, which is located at the N terminus of MEK1 and 2 in a short basic region. This sequence has all the hallmarks of a MAP kinase substrate-docking domain known as the D domain. An extensive list and examination of the presence of this domain in many proteins were presented by Nishida and colleagues (262) and others (263, 264) (see below). This docking site on MEK1 is not only required for ERK2 activation in vitro but is also necessary for its activation of ERK2 in cells. Several types of experiments support this conclusion. A MEK1 deletion mutant lacking N-terminal sequence including the docking domain interferes with activation of ERK2 by EGF (263). When introduced into cells, an N-terminal peptide derived from MEK1, which contains the docking site, inhibits progress through the cell cycle (265). Anthrax lethal factor cleaves the D domain from MEK and inhibits ERK activation (266). Using mutagenesis and deletion analysis, a binding site on ERK1/2 for this D domain has been localized to a pair of aspartate residues in the C-terminus of ERK2, just outside the catalytic core (Refs. 262, 267 ; Fig. 1 and Table 2). Additional sites of interaction on ERK2 have also been proposed (153, 263, 268).

4. MP-1. MP-1, a protein of approximately 13 kDa, was identified by Weber and colleagues (272) in a two-hybrid screen with MEK1. Deletion of the MEK1 proline-rich insert eliminates binding, suggesting that the insert is the primary site of interaction between MEK1 and MP-1. It has been suggested that MP-1 is a scaffold that enhances formation of protein complexes, because it also binds to ERK1. Interestingly, it binds much less well to ERK2, indicating an unexpected selectivity between these two very similar MAP kinases. Cellular studies demonstrated that MP-1 increases the activation of ERK1, consistent with the interpretation that it binds both MEK1 and ERK1 (272). Because of its small size, MP-1 is unlikely to be a functional equivalent of Ste5p. However, MP-1 may be one unit of a modular scaffolding system that may facilitate the formation of a smorgasbord of complexes with minor differences in protein composition.

5. Grb10. Nantel and co-workers (273) have shown that the proline-rich insert of MEK1 binds to Grb10. Grb10 is usually viewed as an adapter molecule. It was originally isolated in a screen for proteins that bound to the tyrosine-phosphorylated, C-terminal domain of the EGF receptor. Grb10-MEK1 complexes have been identified in association with mitochondria and may be involved in cell survival signals that can be generated by this pathway (274). MEK1 also binds to Grb2 (A. Dang and M. H. Cobb, unpublished data), a common adapter that links receptors to the Ras GEF Sos. The significance of these associations is unknown.

6. Kinase suppressor of Ras (KSR). Eye development in Drosophila and vulval induction in the nematode C. elegans have proven to be valuable systems in which to discover Ras signaling mechanisms and components of the ERK1/2 MAP kinase signaling cascade using genetics. In each system the MAP kinase cascade is regulated by a receptor tyrosine kinase—Sevenless in flies and the EGF receptor in worms. Each works through Ras to control cell fate. To identify molecules that were required for the function of Ras, mutants with impaired Ras signaling without effect on Raf or downstream molecules were sought using these two systems (275, 276, 277). Kinase suppressor of Ras or KSR resulted from these screens and was found to act in numerous tyrosine kinase pathways. KSR, like Raf, has an N-terminal cysteine-rich region and a C-terminal kinase domain. Also in common with Raf, KSR homologs have been found in numerous animal species but not in yeast (158).

Substantial evidence indicates that KSR acts as a scaffold to bind the kinases of the ERK1/2 MAP kinase module (278, 279, 280, 281, 282). On the other hand, there is little evidence to indicate that it is a protein kinase. It has strong primary sequence similarity to the protein kinase family, but has arginine in place of the lysine in kinase subdomain II that is required for catalysis (283). There is also little to suggest that its functions depend on protein kinase activity, although its kinase domain is required for binding to both Raf-1 and MEK1, and mutation of the above mentioned arginine impairs its function (278, 280, 281, 282). The autophosphorylating activity of KSR, the only reported evidence of its catalytic function, is most likely due to the association with MEK and ERK (D. K. Morrison, personal communication), raising further questions about its protein kinase activity.

The ability of the catalytic domain to bind to MEK1 is essential for the function of KSR (282). The CRD of KSR binds to ERK2 (278, 279, 280). One function of KSR, like Ste5p, may be to localize the MAP kinase module at the membrane to be activated by transmembrane cues. Also similar to Ste5p, KSR binds to -subunits of heterotrimeric G proteins, suggesting that KSR may have roles in signaling by G protein-coupled as well as tyrosine kinase receptors (284).

7. Raf kinase inhibitor protein (RKIP). A Raf-1-interacting protein, named RKIP, was isolated from a two-hybrid screen using Raf-1 as bait (285). As suggested by its name, RKIP inhibits the phosphorylation and activation of MEK by Raf-1. RKIP appears to disrupt the formation of Raf-MEK complexes. RKIP binds directly to Raf-1, MEK, and ERK as assessed by in vitro binding and coimmunoprecipitation from cell lysates, apparently preventing their productive interactions. Based on overexpression studies and the use of antisense RNA and inhibitory antibodies, it was concluded that RKIP functions physiologically to shut off the activation of the ERK1/2 module. It is possible that RKIP may have other functions, e.g., as a scaffold that promotes activation of the cascade under a select group of circumstances or to localize the cascade to a specialized organelle. Although there are no data supporting this idea currently, both JNK inhibitory proteins (JIPs, discussed below) and the inhibitor protein for PKA (PKI) were originally identified as inhibitors and are now believed to have additional functions. JIP is apparently a Ste5p-like scaffold and PKI terminates the nuclear activity of PKA by forming a complex that promotes the export of the catalytic subunit of PKA from the nucleus (286).

8. YopJ. Orth et al. (287) have identified a virulence factor from the bacterial pathogen Yersinia pestis that binds to multiple MEKs so that host signaling responses can be usurped or interrupted. YopJ blocks phosphorylation and activation of MEKs and thereby inhibits ERKs, JNK/SAPKs, p38 MAP kinases, and other signaling pathways. Among the consequences are prevention of cytokine biosynthesis and promotion of apoptosis. YopJ-related proteins exist in some other bacterial pathogens, but mammalian homologs of YopJ have not been reported.

9. STYX. Dixon’s group (288, 289) also identified a tyrosine phosphatase-related molecule, STYX, which lacks the cysteine required for phosphatase catalytic activity. When cysteine was introduced into the appropriate position in the molecule, it displayed phosphatase activity toward ERK1/2. STYX bound tightly to ERK1/2, suggesting that it may act as an inhibitor either of ERK1/2 activity or their dephosphorylation.

10. Sur-8. Sur-8 was identified as a loss-of-function mutation that can suppress the multivulval phenotype in C. elegans in the presence of an activated let-60 (ras) gene (290, 291). Loss of Sur-8 function in a wild-type genetic background produced no observable phenotype. When worms with mutated Sur-8 were crossed with worms deficient in either mpk-1 (ERK1/2 ortholog) or ksr-1, vulval induction was severely compromised. Ectopic expression of wild-type Sur-8 enhanced the multivulval phenotype caused by an activated let-60 mutation and also increased Raf-1 activity. Epistatic analysis in worms placed Sur-8 at the same level of the pathway or downstream of Ras. This was consistent with two-hybrid experiments that showed that Sur-8 interacts with Ras mutants. Overexpressed Sur-8 coimmunoprecipitated with complexes of Ras and Raf. Point mutations (cysteine 260 tyrosine, glutamate 457 lysine) in Sur-8 reduced its association with Ras and Raf-1 and its ability to enhance Raf kinase activity.

11. Connector enhancer of KSR (CNK). To identify molecules that modified the function of KSR in the Sevenless/photoreceptor system, Rubin and colleagues (292, 293) created a line of flies expressing only the putative catalytic domain of KSR, which they named KDN.