KSR: a MAPK scaffold of the Ras pathway?

The Ras pathway plays a critical role in the transmission of many growth and developmental signals, and a major route by which Ras transmits signals is through the sequential activation of the cytoplasmic kinases Raf, MEK and MAP kinase (MAPK). Several years ago, large-scale genetic screens performed in Drosophila and C. elegans simultaneously led to the isolation of a new component of this pathway, which was termed Kinase Suppressor of Ras (KSR) because of its structural and genetic properties (Kornfeld et al., 1995; Sundaram and Han, 1995; Therrien et al., 1995). The discovery of KSR in these invertebrate model organisms was remarkable because it was the first time that genetic studies had clearly preceded biochemical approaches in identifying a novel component of the Ras pathway. However, although these genetic studies were instrumental in showing that KSR is a positive effector of Ras signaling that appeared to act either upstream of Raf or in a parallel pathway, they did not reveal the precise molecular mechanism by which KSR functions to transmit Ras signals. Naturally, the presence of a putative kinase domain in KSR led to widespread speculation that KSR might act as another kinase in the pathway. Surprisingly, however, biochemical experiments have been inconclusive in demonstrating that KSR actually has intrinsic kinase activity, and a growing body of evidence now indicates that a primary function of KSR is as a scaffolding protein that coordinates the assembly of a multiprotein complex containing MAPK and its upstream regulators. Here, I will examine the evidence supporting the role of KSR as a MAPK scaffold of the Ras pathway.

Deciphering the role of KSR as a signal transducer begins with an examination of the physical properties of the KSR proteins. The KSR members constitute a novel protein family that displays remarkable overall structural similarity to proteins of the Raf family (Fig. 1). All known KSR members contain five conserved domains (Therrien et al., 1995): a 40-residue region unique to KSR proteins (CA1), a proline-rich region (CA2), a cysteine-rich zinc-finger domain (CA3), a serine/threonine-rich region (CA4), and a putative kinase domain (CA5). In both Raf and KSR, the kinase domain occupies the C-terminal half of the protein, and the smaller conserved domains are found in the N-terminal region. The KSR CA3 and CA4 domains bear some resemblance to domains located at analogous positions within the Raf proteins; however, KSR lacks the Ras-binding domain that the Raf family members possess, in agreement with biochemical evidence indicating that KSR does not bind Ras directly. Although little is known about the function of the KSR CA1 and CA2 domains, the CA3 domain appears to be involved in regulating the subcellular localization of KSR (see below, Michaud et al., 1997), and the CA4 domain serves as the docking site for MAPK (Jacobs et al., 1999).

In addition to sequence homologies and domain structure, there are numerous other similarities between KSR and Raf. Like Raf genes, KSR genes are found in invertebrates and mammals, but no obvious homolog is present in yeast. Also, the subcellular localization of KSR is dynamic and changes in response to signaling events, as does that of Raf. In resting cells, KSR is found predominantly in the cytosol, but, upon Ras activation, a fraction of KSR translocates to the plasma membrane and is found in a high-molecular-weight complex (Xing et al., 1997; Michaud et al., 1997; Stewart et al., 1999). The CA3 domain is required for this process, since mutation of this domain prevents the relocalization of KSR (Michaud et al., 1997). KSR also resembles Raf in being a phosphoprotein whose phosphorylation state changes in response to signaling events; however, the types and positions of phosphorylation do not appear to be equivalent (Morrison and Cutler, 1997; Cacace et al., 1999; Volle et al., 1999).

A central problem with the apparent similarity between KSR and Raf is that, whereas Raf has been shown to be a direct effector of Ras that functions by phosphorylating MEK and activating the MAPK cascade, the function of KSR in similar events has been very difficult to demonstrate. As noted above, KSR contains a CA5 putative kinase domain, which immediately implicated KSR as a protein kinase. This notion was bolstered by the fact that many of the KSR-inactivating mutations identified in Drosophila and C. elegans occur in the CA5 domain. However, several observations have cast doubts on whether the CA5 domain does have enzymatic activity. In particular, although the CA5 region contains many of the hallmarks of a protein kinase, all of the mammalian KSR proteins contain an arginine residue at a position in the ATP-binding domain that is normally occupied by a lysine residue (Therrien et al., 1995). This conserved lysine is involved in the phosphotransfer reaction and is usually required for enzymatic activity. Ironically, the first KSR family members to be identified in flies and worms have the expected lysine residue at this position, contributing to the confusion about kinase function.

Moreover, the finding that the isolated catalytic domain of KSR acts as a dominant inhibitory protein also called into question the function of the CA5 domain (Therrien et al., 1996; Yu et al., 1997; Joneson et al., 1998). Because KSR was discovered as a positive effector of Ras signaling, it was predicted that deregulation of the kinase domain by removal of the N-terminal sequences would produce a constitutively active kinase that would enhance Ras-dependent signaling, as has been shown for the isolated catalytic domain of Raf. Thus, the findings that the KSR catalytic domain blocks Ras signaling and MAPK activation in mammalian cells, Xenopus oocytes and the Drosophila retina were surprising and suggested that either the function or regulation of KSR is more complex than originally anticipated. More recently, genetic studies in C. elegans have further undermined the idea that KSR acts as a kinase. In these experiments, KSR proteins containing mutations that would be expected to eliminate kinase activity rescued the KSR loss-of-function phenotype, indicating that KSR activity can be restored by kinase-independent mechanisms (Stewart et al., 1999).

If KSR is not a kinase, how does it facilitate Ras signaling? The notion that KSR functions as a scaffolding factor first emerged when KSR was found to interact with Raf-1, MEK1/2 and ERK1/2. Both MEK1 and MEK2 bind directly to the CA5 domain of KSR (Denouel-Galy et al., 1997; Yu et al., 1997) and this interaction appears to be essential for KSR function. MEK is stably associated with KSR in both quiescent and growth-factor-treated cells and mutations in the catalytic domain that inactivate the biological activity of KSR disrupt MEK binding (Stewart et al., 1999; Muller et al., 2000). ERK1/2 binding is also direct, mediated by an FxFP motif found in the CA4 domain (Jacobs et al., 1999). Significantly, unlike the interaction with MEK, the interaction with the ERKs is not constitutive but is induced upon Ras activation (Cacace et al., 1999; Muller et al., 2000). The interaction with Raf-1 occurs at the plasma membrane and is also Ras dependent, but appears to require other proteins, such as MEK (Therrien et al., 1996; Xing et al., 1997). Taken together, these observations suggest a model in which KSR localizes MEK to the plasma membrane in a Ras-dependent manner and promotes the assembly of a multiprotein complex that brings MEK into close proximity with its upsteam activator, Raf and downstream substrate, MAPK (Fig. 2A,B). In essence, KSR provides a scaffold that facilitates the phosphorylation reactions that are required for executing critical signal transduction steps downstream of Ras.

Further support for this scaffolding model has come from the finding that other proteins, including Hsp90, p50^cdc37, G protein γ subunits and 14-3-3 proteins also interact with KSR. Hsp90 and p50^cdc37 bind directly to the KSR CA5 domain and appear to be required for protein stability since pharmacological disruption of these interactions results in rapid KSR degradation (Stewart et al., 1999). The binding of KSR to the γ subunits of heterotrimeric G proteins is mediated by the CA3 domain of KSR, and this interaction may localize KSR to the plasma membrane in response to ligands that activate G protein-coupled receptors (Bell et al., 1999). 14-3-3 binding might also play a role in regulating the intracellular localization of KSR and in stabilizing the inactive and active conformations of KSR (Cacace et al., 1999; Xing et al., 1997). 14-3-3 proteins bind to two phosphoserine residues located on either side of the CA3 domain, and both sites are fully phosphorylated in quiescent cells. (Cacace et al., 1999). Because the CA3 domain is required for the translocation of KSR from the cytoplasm to the plasma membrane, binding of 14-3-3 proteins to these sites may keep KSR in an inactive state by masking the CA3 domain. When Ras is activated, the phosphorylation state of the 14-3-3-binding sites is reduced, which would result in the release of 14-3-3 proteins and perhaps exposure of the CA3 domain. Once at the membrane, 14-3-3 proteins might also play a role in stabilizing the membrane-bound signaling complex through interactions with KSR and Raf. Analyses of how these various proteins interact with KSR further support a model in which KSR transports MEK from the cytoplasm to the plasma membrane and organizes Raf-1, MEK and MAPK for efficient activation in response to membrane signaling events.

The scaffolding model also provides an attractive explanation for previously puzzling observations that the biological effects of KSR can vary dramatically, depending on the level of KSR protein expressed (Cacace et al., 1999). When KSR is expressed at low or near physiological levels, it functions as a positive effector of Ras signaling (Therrien et al., 1996; Cacace et al., 1999; Muller et al., 2000). However, when KSR is highly overexpressed, it inhibits Ras signaling in mammalian cells, Xenopus oocytes and the Drosophila retina (Denouel-Galy et al., 1997; Sugimoto et al., 1997; Yu et al., 1997; Joneson et al., 1998; Cacace et al., 1999). This is a property that KSR shares with other MAPK scaffolding factors, such as the JNK scaffolding protein JIP-3, which was initially identified as an inhibitor of JNK signaling because of similar overexpression effects (Dickens et al., 1997). For a scaffolding protein to function properly, the concentration of the scaffold would be expected to be closely titered to the level of the components to which it binds. When the level of the scaffolding protein drastically exceeds the level of its interacting components, nonfunctional complexes that lack some of the critical components would form, thereby inhibiting signal transmission. The scaffolding model also accounts for the dominant inhibitory effect of the isolated KSR catalytic domain. In this case, mutational studies have shown that KSR catalytic domain constructs unable to bind MEK no longer block Ras signaling (Muller et al., 2000), which indicates that the interaction with MEK is critical for the dominant-inhibitory activity of these protein fragments. Thus, if a key function of KSR is to shuttle MEK to the plasma membrane, the isolated catalytic domain would sequester MEK in the cytosol, since it does not contain the CA3 domain required for membrane localization. As a result, signal transduction from Ras to MAPK would be short-circuited, which is consistent with the experimental data (Fig. 2C).

A role for KSR as a MAPK scaffold does not necessarily preclude a kinase function for the protein. In contrast to the body of evidence against the kinase activity of KSR described above, evidence from one group suggests that KSR is a ceramide- and EGF-activated protein kinase that can phosphorylate Raf-1 on threonine 269 in vitro and thereby increase its kinase activity (Zhang et al., 1997, Xing and Kolesnick, 2001). However, attempts by other groups to corroborate these findings have been unsuccessful (Denouel-Galy et al., 1997; Yu et al., 1997; Michaud et al., 1997; Stewart et al., 1999; Muller et al., 2000). In addition, the phosphorylation of Raf-1 on threonine 269 has not been reported in vivo. The in vitro studies detecting Raf-1 phosphorylation may have been complicated by the association of KSR with several proteins that do possess kinase activity, such as Raf-1, MEK, MAPK and an unidentified kinase (Volle et al., 1999). Further, KSR constructs that lack the catalytic domain have been shown to augment Raf-1 activation in a kinase-independent manner (Michaud et al., 1997). One significant problem in determining whether KSR is a kinase is that although mammalian KSR proteins lack the conserved lysine residue that has long been considered essential for and diagnostic of a true protein kinase, a kinase that utilizes an alternative lysine residue for ATP binding has recently been identified (Xu et al., 2000). Thus, it is difficult to rule out kinase activity for the KSR CA5 domain on the basis of sequence considerations and available mutagenesis data, and the question of catalytic activity still stands. Has KSR evolved to be a non-functional pseudokinase? Or is KSR a protein kinase that perhaps has a very specific substrate specificity or unexpected activation mechanism that has not been elucidated? The answer to these questions will be resolved only with the conclusive identification of a KSR phosphorylation substrate.

In addition to the question of whether KSR is a protein kinase, there are other aspects of KSR function and regulation that remain to be answered. In particular, although KSR is clearly a conserved component of the Ras pathway, is it a protein required for all Ras signaling events or is its activity needed only under certain signaling conditions? Also, we do not know whether all the components of the KSR signaling complex been identified or whether the set of components will be the same in all cellular contexts. In addition, for other proteins of the MAPK cascade, such as Raf, MEK and MAPK, multiple mammalian genes have been identified. To date, however, only one mammalian KSR gene has been found, although it does give rise to different splice variants, which encode distinct isoforms (Muller et al., 2000). Therefore, do other mammalian KSR genes exist? Finally, what is the precise mechanism by which the CA3 domain mediates the translocation of KSR, and do other regulatory mechanisms influence KSR function? Clearly, these are all crucial questions that need to be addressed by further KSR studies.

A more global question regarding KSR function concerns the need for multiple MAPK scaffolding proteins in mammalian cells. In addition to KSR, both MP1 (MEK partner 1) and MEKK1 interact with components of the MAPK cascade. MP1 interacts with MEK1 and ERK1 (Schaeffer et al., 1998), and MEKK1 has been reported to associate with Raf-1, MEK1 and ERK2 (Karandikar et al., 2000). Do the different MAPK scaffolds function in distinct cellular events? Do they localize the MAPK components to different cellular compartments and/or substrates? Do they respond to different signaling inputs? These questions remain to be answered and are likely to be central to a full understanding of the pleiotropic effects of Ras signaling and MAPK activation.

In summary, despite early indications that KSR might function as a protein kinase, critical experimental validation of this idea has not been forthcoming. Instead, numerous biochemical and genetic lines of investigation have provided evidence that KSR serves as a scaffolding protein that facilitates the Ras-dependent assembly of the Ras pathway kinases Raf-1, MEK, and MAPK, thereby promoting their interactions and efficient signal transduction. Thus, while KSR was originally named ‘Kinase Suppressor of Ras’, ‘Kinase Scaffold of the Ras pathway’ may be a more accurate moniker for KSR.

I thank members of my laboratory and Mark E. Fortini for helpful comments on the manuscript.