Mixed lineage kinases (MLKs): a role in dendritic cells, inflammation and immunity?

Abstract

This review summarizes current knowledge about the mixed lineage kinases (MLKs) and explores their potential role in inflammation and immunity. MLKs were identified initially as signalling molecules in the nervous system. They were also shown to play a role in the cell cycle. Further studies documented three groups of MLKs, and showed that they may be activated via the c-Jun NH₂ terminal kinase (JNK) pathway, and by Rho GTPases. The biochemistry of the MLKs has been investigated in considerable detail. Homodimerization and heterodimerization can occur, and both autophosphorylation and autoinhibition are seen. The interaction between MLKs and JNK interacting protein (JIP) scaffolds, and the resultant effects on mitogen activated protein kinases, have been identified. Clearly, there is some redundancy within the MLK pathway(s), since mice which lack the MLK3 molecule are not abnormal. However, using a combination of biochemical analysis and pharmacological inhibitors, several recent studies in vitro have suggested that MLKs are not only expressed in cells of the immune system (as well as in the nervous system), but also may be implicated selectively in the signalling pathway that follows on toll-like receptor ligation in innate sentinel cells, such as the dendritic cell.

It has become clear that the early stages in the response to injury are not merely a non-specific preamble to subsequent adaptive immune responses. Rather, the inflammatory process can itself assume many different guises, and the precise nature of this early ‘innate immunity’ is important in determining subsequent outcome. First propounded by the late Charles Janeway (Janeway 1989), it is now widely accepted that ‘pattern recognition receptors’ (PRR) (in contrast with ‘specific recognition receptors’) are a starting cellular trigger point in innate immunity. The different toll-like receptors (TLR) are molecules that have been conserved during evolution and are an important example of these PRRs, which have been investigated in considerable detail over the past few years (Medzhitov et al. 1997; Hallman et al. 2001; Janeway & Medzhitov 2002; Takeda et al. 2003), and have been shown to be implicated in this start-up phase of immunoregulation.

The recognition that dendritic cells (DC) may be an important component at the interface between innate and adaptive immunity (Ibrahim et al. 1995; Hemmi & Akira 2005; Steinman & Hemmi 2006) where these receptors may play their role is also widely acknowledged. These DC have been known for many years as the main antigen presenting cells (APC), and therefore what controls the DC at the innate immune level, as judged by activation, maturation, migration and survival, has very important consequences not only for defence against infection, but also for tolerance and auto-immune outcomes.

In parallel with these advances, the combination of molecular, genomic and proteomic approaches has generated a vast amount of new information about intracellular signalling pathways, and in particular about those that are implicated potentially in this part of the immune response. Much previous work in this field has floundered, probably because the signalling cascade operates in such a rapid, transient and synergistic, fashion. The new studies have led to a much more logical approach, providing a sound basis for the definition of families of molecules that have been regarded previously as disparate. Furthermore, they have facilitated the possibility of looking at selected members of these families in the context of the DC, where our understanding of signalling has hitherto been rudimentary.

An illustrative example of this increase in our understanding of signalling pathways is that of the so-called ‘mixed lineage kinases’ (MLK). The early studies of these kinases led to a complex nomenclature, and to the suggestion that these molecules were implicated in signalling events, especially in the nervous system (Mata et al. 1996; Sakuma et al. 1997), but also (more ubiquitously) in the cell cycle. However, our recent studies, starting with the use of an MLK inhibitor (Handley et al. 2005), raised the rather unexpected possibility that MLKs may also play a distinctive role in the immune system, since they appear to be involved in DC survival, maturation and morphogenesis. Therefore, in this review, after summarising what is known already about MLKs from a biochemical and functional perspective, we have attempted to integrate the background with our recent findings, and to suggest possible pathways whereby selected MLKs may be implicated in the way that DC behave during inflammation, and hence in DC–driven immunoregulation.

Mixed lineage kinases

The MLKs are a family of serine/threonine kinases all of which act as mitogen activated protein kinase kinase kinases (MAP3Ks) (Gallo & Johnson 2002). The name ‘Mixed Lineage Kinases’ derives from the fact that of the 11 conserved subdomains found in all protein kinases, domains 1–8 of the MLK proteins resemble serine/threonine kinases whereas regions 9, 10 and 11 share sequence similarity with those of tyrosine kinases, such as the Fibroblast Growth Factor Receptor and Src (Gallo & Johnson 2002).

Role of MLKs in MAPK pathway

Early experiments into the role of MLK proteins were performed with recombinant proteins or by overexpression in cell lines. Perhaps because of this the data is often contradictory. For example Merritt et al. (1999) showed that one of the MLK proteins, Dual Leucine Zipper bearing kinases (DLK), activates MKK7 and not MKK4, whereas Hirai et al. (1997) showed that DLK activates MKK4. Six different reports can be summarized by the statement that in vitro MLK proteins can activate both p38 and the c-Jun NH₂ terminal kinase (JNK), via any of the relevant intermediary proteins, MKKs 3, 4, 6 or 7 (Rana et al. 1996; Tibbles et al. 1996; Hirai et al. 1997; Cuenda & Dorow 1998; Hirai et al. 1998; Merritt et al. 1999). Like the early experiments it describes, this summary fails to take into account subtleties of physiological regulation. It is probable that all MLK proteins activate JNK, and all act via MKK7. MKK4-mediated activation of JNK occurs to a lesser degree, although the regulation of this pathway is not clear. MLK proteins can also activate p38 and this is via MKK3 phosphorylation.

The role of MLK proteins in extracellular signal regulated MAP kinase (ERK) activation was also confused. Overexpression of MLK3 has been shown to induce moderate activation of the MAP kinase or ERK kinase (MEK1)–ERK protein kinase pathway. This activation was prevented by the MEK-specific inhibitor PD98059 or by expression of a kinase-deficient MEK1 mutant (Hartkamp et al. 1999). In contrast to this, however, Shen et al. demonstrate that although MLK3 can phosphorylate and activate MEK-1 directly in vitro and can induce MEK phosphorylation in COS-7 cells, this induction of MEK phosphorylation does not result in ERK activation in vivo (Shen et al. 2003). However, silencing of mlk3 by RNAi suppressed mitogen and cytokine activation not only of JNK and p38, but also of ERK (Chadee & Kyriakis 2004b).

It therefore appeared as though the initial picture of MLKs as comparatively selective regulators of the JNK group of MAPKs was no longer valid, and that MLKs may in fact have a more general upstream role than was postulated initially.

Types of MLK

Eight mammalian MLKs have been identified, categorized into three subfamilies on the basis of domain organization and sequence similarity:

1. MLKs; MLK1-4 (Dorow et al. 1993; Ing et al. 1994).

2. Dual Leucine Zipper-Bearing Kinases (DLKs); Dual Leucine Zipper bearing kinases (DLKs) and Leucine Zipper bearing kinase (LZK) (Holzman et al. 1994; Sakuma et al. 1997).

3. Zipper Sterile-α-Motif Kinases (ZAKs); ZAKα and ZAKβ (Liu et al. 2000; Gotoh et al. 2001).

MLK2, MLK3, DLK and LZK (Dorow et al. 1993; Ing et al. 1994; Sakuma et al. 1997) are the only MLKs whose biochemistry has been studied in any detail. This is probably because they are either widely expressed (MLK3 and LZK), or are brain restricted (MLK2 and DLK), and hence offer potential therapeutic targets for neurodegenerative disorders, such as Parkinson's disease.

All eight MLKs contain a leucine zipper domain. These consist of one or two α-helices with predominantly non-aromatic lipophilic amino acids occupying one face, which facilitates protein multimerization via hydrophobic interactions. The DLK subfamily is identified by possession of two leucine zippers separated by a 31 amino acid spacer region. DLK and LZK share a high degree of homology.

The most recently identified MLK subfamily have been termed ‘ZAKs’ because the first identified member of the family, ZAKα, contains a sterile-α-motif (SAM): an independent globular domain of about 70 amino acids found in many signal transduction proteins. Biochemical and structural studies suggest that SAM domains mediate protein dimerization. The second ZAK protein, ZAKβ, has a much smaller C-terminal domain than ZAKα and as a result lacks the SAM. The ZAK proteins are only loosely related to the other six MLKs, sharing sequence homology in the kinase catalytic region and bearing a small leucine zipper domain (Gotoh et al. 2001).

MLK1-4 share a high degree of homology in the N-terminal 500 or so amino acids, which incorporate several functional domains: an SH₃ domain, a kinase catalytic domain, the leucine zipper, a proline-rich region, and the Cdc42/Rac interactive binding (CRIB) domain which mediates binding to GTP-bound cdc42 and rac. In contrast to the amino terminal portion of the proteins, the carboxyl termini share little homology and vary greatly in size, possibly to mediate specificity and orchestrate interactions with various other proteins. These kinases will be the main focus of this review.

Although MLK1 was identified originally in epithelial cell lines (Dorow et al. 1993), and MLK3 was found initially in human thymus (Ing et al. 1994), most studies on this topic have been on neuronal cell, and in the context of the nervous system. Therefore, the role of MLKs in the nervous system will be discussed first before reviewing the role within immunity and inflammation.

Activation of MLKs

Phosphorylation is an important mechanism in the regulation of protein kinases, in general, and there is much evidence suggesting that the activity of MLK proteins is also controlled in this manner. For example, HPK has been shown to bind to, and phosphorylate MLK3 (Kiefer et al. 1996). MLKs have also been shown to be phosphorylated in vivo, and furthermore co-expression with a constitutively active cdc42 mutant (V12cdc42) alters the phosphorylation pattern and increases the in vitro catalytic activity of MLK3 (Bock et al. 2000). In contrast, phosphorylation of MLK3 by Akt1 has been shown to have a negative regulatory effect (Barthwal et al. 2003).

Phosphopeptide mapping and mass spectrometry analysis has identified 12 potential MLK3 phosphorylation sites in total, most of which cluster at the C terminus (Vacratsis et al. 2002). Although these sites may not all be phosphorylated in vivo, this result emphasizes the importance of phosphorylation in MLK regulation. The motif T T X X S (residues 277–281 of MLK3) is found in the putative activatory region of all MLKs, and mutagenesis studies support thr²⁷⁷ and ser²⁸¹ as positive regulatory phosphorylation sites (Leung & Lassam 2001; Durkin et al. 2004).

MLK activation by JNK and ERK

It has been suggested that there is a close functional association between the MLKs and the JNK MAPK. A positive feedback was proposed in early studies, where an MLK mutant lacking the C-terminal domain was shown to be sufficient to activate JNK; but the resulting apoptotic response was abrogated (Sakuma et al. 1997). One of the potential MLK3 phosphorylation sites conforms to the consensus sequence for MAPK (including JNK) phosphorylation sites (P X S/T P), which would also be consistent with JNK phosphorylation of MLK3 as a feedback mechanism (Vacratsis et al. 2002).

In contrast, recent data has suggested the existence of a negative feedback loop involving JNK (Schachter et al. 2006). MLK3 can be indeed phosphorylated by JNK; and then a JNK mediated inhibition step inactivates MLK3. Removing the JNK inhibitor that was used for this study reverses the process. Furthermore, Schachter et al. also showed that the hypophosphorylated MLK3 redistributed into a Triton insoluble fraction, and suggest that this could represent an association either with cytoskeletal elements, such as merlin, the neurofibromatosis 2 gene product (Chadee et al. 2006) or with lipid rafts. If the inactive MLK3 resembles other signalling molecules, then it is likely to be found in the caveolae-enriched microdomain component of the rafts.

In these studies, the ERK pathway was explored, as well as the JNK mediated events (Chadee et al. 2006). A similar feedback loop has been shown to exist in this pathway. The MAPK Raf is phosphorylated by ERK in response to insulin and growth factors (Lee et al. 1992). Although MLK3 has been shown to be required for B-Raf activation, these studies suggested that the MLK3 role is required for the maintenance of the B-Raf/Raf-1 complex rather than the earlier step. Merlin–MLK3 association will disrupt the complex and thus inhibit ERK (Chadee et al. 2002).

With respect to other MLKs, it has also been shown that JNK can phosphorylate the C-terminal region of MLK2, although no specific phosphorylation sites were determined (Sakuma et al. 1997). An association with JNK has also been demonstrated: transfection of dominant-negative JNK inhibits MLK2 phosphorylation in response to anisomycin-induced JNK activation (Sakuma et al. 1997).

MLK activation by Rho family GTPases

The Ras-related Rho family of GTPases are small monomeric G proteins that function as a simple biochemical switch by cycling between two conformational states. Active GTPases recognize and activate target proteins until their GTPase activity results in GTP hydrolysis and auto-inactivation. Their activity is in turn activated by GEFs (guanine nucleotide exchange factors), and terminated by GAPs (GTPase-activating proteins). The Rho family in mammals represents 23 gene products including members of the rac and cdc42 subfamilies. Collectively, these proteins have been implicated in a huge number of cellular functions, including maintenance of morphology (Tapon & Hall 1997), motility (Aepfelbacher et al. 1994), adhesion (Nobes & Hall 1995; Braga et al. 1997), cell division (Dutartre et al. 1996) and proliferation (Olson et al. 1995), smooth muscle contraction (Hirata et al. 1992), and vesicular transport (Lamaze et al. 1996; Murphy et al. 1996). Many receptors have been implicated in the activation of Rho GTPases, including receptors for chemoattractants, cytokines, phagocytic particle uptake, growth factors and recently TLRs (Bokoch 2005).

The CRIB (cdc42 and rac Interactive Binding region), characteristic of the MLK subfamily of proteins, was first identified by Burbelo et al. (1995) in the PAK-1 MAP3K. A functionally diverse repertoire of proteins possess CRIB domains and hence act as cdc42 and rac effector proteins. These include kinases such as the PAK and MLK families, but also proline-rich and SH3-containing adaptor proteins. The number and range of potential effectors for cdc42 and rac is perhaps unsurprising given the many functions with which the G-proteins have been associated.

Activation of MLK3 by cdc42 and rac has been shown to be a pathway for JNK activation by these G-proteins (Teramoto et al. 1996). Two cdc42-inducible phosphorylation sites in MLK3 have been identified subsequently although it is unclear how cdc42 induces MLK3 phosphorylation since it lacks kinase activity, and other kinases must be involved.

The p21-cdc42/rac-activated kinase (PAK) family of MAP3K are the model CRIB-containing kinases, and thus the mechanism of PAK-cdc42/rac binding may provide some indication about the mode of MLK activation. The CRIB domains of MLK and PAK differ slightly (Burbelo et al. 1995), and therefore might differ in function. However, mutation of the MLK3 CRIB disrupts the ability of this protein to bind cdc42, indicating that this is indeed a functional CRIB motif.

Wen MLK3 is in the resting, unphosphorylated state it coprecipitates with cdc42 (Bock et al. 2000). However, following MLK3 phosphorylation, cdc42 is no longer detectable. This suggests that upon activation by cdc42, MLK3 has a reduced affinity for the GTPase. In an exactly analogous fashion, the affinity of PAK-2 for cdc42 decreases following activation (Manser et al. 1995). Furthermore, the residues phosphorylated in MLK3 in response to cdc42 stimulation are exactly the same as those of PAK-2 (Bock et al. 2000), as judged by tryptic phosphopeptides.

Cdc42 is geranylgeranylated (Maltese & Sheridan 1990) resulting in its localisation to cytoskeletal and cellular membrane elements (Erickson et al. 1996); it is therefore possible that cdc42 is responsible for recruitment of MLK3 to the vicinity of an activating kinase resulting in phosphorylation and hence activation. This mechanism is reminiscent of the activation of another MAP3K, Raf, which on being activated by ras, clusters at the plasma membrane and becomes a target for other membrane kinases (Stokoe et al. 1994). The transition between cytoplasm and membrane/cytoskeletal association is therefore emerging as a key element of MLK regulation..

Dimerization and autophosphorylation

In addition to phosphorylation by activatory kinases, there is evidence that MLK phosphorylation can also occur via dimerization and autophosphorylation (Leung & Lassam 2001). Indeed, the most striking feature of all MLK proteins is their ability to dimerize – suggesting that this is an important feature of their catalytic cycle. In addition to the leucine zipper domain characteristic of MLK proteins, all contain a second dimerization mechanism: MLK1-3 possess both an SH3 domain and its substrate proline-rich region, the DLKs have a second leucine zipper region and the ZAKs a sterile α motif, however, the significance of this is not as yet clear.

The importance of leucine zipper dimerization to MLK activity was first suggested by the observation that MLK3 forms homodimers that are stabilized by disulphide bonds. The leucine-zipper region is both necessary and sufficient for this homodimerization (Leung & Lassam 1998). The precise mechanism by which leucine zipper-mediated dimers are reinforced by disulphide linkages, and the importance of these covalent bonds to the catalytic mechanism, is not understood.

MLK3 proteins autophosphorylate following homodimerization (Leung & Lassam 2001), via a mechanism analogous to that of receptor tyrosine kinases (with which MLKs share some sequence homology). Dimerization is necessary for autophosphorylation because kinase-dead MLK3 can homodimerize but not autophosphorylate, whereas MLK3 minus the leucine zipper region lacks autophosphorylation activity. JNK is not activated by the MLK3 leucine-zipper deficient mutant suggesting that homodimerization and autophosphorylation are critical steps in the catalytic mechanism of MLK3 (Leung & Lassam 1998).

Some of these effects may have been because of the large protein structural alteration induced by domain deletion. To overcome this problem, using site directed mutagenesis, proline residues were introduced into the leucine zipper of MLK3 to disrupt the α-helical conformation with minimal protein structure disruption (Vacratsis & Gallo 2000). They found that cdc42 was still able to activate this MLK3 mutant suggesting that dimerization is not required for MLK3 activation by cdc42. Furthermore, the MLK3 mutant retains a high autophosphorylation capacity (Vacratsis & Gallo 2000). This is intriguing because although it supports the idea that monomeric MLK3 binds to active cdc42, in this scenario dimerization cannot subsequently occur, and yet autophosphorylation still results. Coexpression of cdc42 and MLK3 leads to a substantial increase in MLK dimerization suggesting that cdc42 catalyses MLK3 dimerization (Leung & Lassam 1998). Taken together, these studies suggest that by binding to membrane-localized cdc42, the local concentration of MLK3 is increased sufficiently to allow autophosphorylation to occur in the absence of zipper-mediated dimerization and disulphide reinforcement. Dimerization is therefore not a prerequisite for autophosphorylation, but presumably increases the probability of fruitful MLK–MLK interaction.

The idea that cdc42 and rac increase the local concentration of MLK2 is also supported by the observation that when overexpressed, MLK2 is constitutively active (Nagata et al. 1998). Presumably in this situation the local concentration does not need amplification by membrane localization for homodimerization and autophosphorylation to occur. In support of this interpretation, this constitutive MLK activation is unaffected by cotransfection with constitutively active or dominant negative cdc42 or rac (Nagata et al. 1998).

Role of the MLK SH3 domain and autoinhibition

As well as the CRIB domain, MLK1-4 are also different from other MLKs in possessing an N-terminal SH3 domain, an independently folding domain of about 60 amino acids which functions to bind proline-rich polypeptides in a consensus sequence P X X X P preceded or followed by several basic amino acids.

The SH3 region of MLK3 has been shown to bind Haematopoietic Progenitor Kinase 1 (HPK-1) (Leung & Lassam 2001), although the physiological significance of this is unclear (Kiefer et al. 1996). It was recently shown that the SH3 domain in MLK3 binds to a non-consensus SH3 binding site situated between the leucine-zipper and the CRIB motifs. Point mutations introduced into this sequence or the SH3 domain increased the activity of the protein. Hence MLK3 is capable of negative regulation of its own activity (Zhang & Gallo 2001). This was shown to be a specific autoregulation and not an inhibition by other SH3-containing proteins because an MLK3 with a mutated SH3 domain bound with much higher avidity to a GST-SH3 reporter. This suggests that in wild-type MLK3, intramolecular SH3 binding out-competes the reporter protein (Zhang & Gallo 2001).

A similar mechanism of auto-inhibition is observed in the Src family, where an intramolecular SH3-mediated association prevents ATP binding to the catalytic domain of the tyrosine kinase (Williams et al. 1998). Interestingly, in Src, the SH3 binds to the same non-consensus sequence as in MLK3. It is probable that this sequence induces a low affinity interaction because a more stable binding would prevent auto-inhibition. MLKs1, 2 and 4 also posses this binding sequence suggesting that all members of the MLK subfamily are regulated in this manner.

The mechanism of release of MLK auto-inhibition has not been elucidated. One speculation was that higher affinity ligands for the SH3 domain, such as HPK-1, might induce release of the intramolecular association (Zhang & Gallo 2001). An alternative mechanism becomes apparent however, when one considers that the functional motifs of MLK3 lying between the SH3 and its binding region are the CRIB and leucine-zipper domains. Therefore, this intramolecular bond prevents MLK3 from dimerizing or binding to cdc42 or rac. Association of the CRIB domain with a GTP-bound G-protein would therefore displace the intramolecular SH3 bond and permit homodimerization via the leucine-zipper and autophosphorylation. This model fits neatly with the observations outlined above regarding a model of cdc42 and rac increasing the local concentration of MLK3 before dimerization.

In this modified scenario, cdc42 serves not only to increase local concentration, but also to release the MLK auto-inhibition. The concentration of activated MLK is therefore increased sufficiently such that MLK3 proteins can dimerize and autophosphorylate. Perhaps upon release of auto-inhibition, adjacent MLK proteins dimerize via intermolecular SH3 bonds before hydrophobic zipper interactions. This would make sense to align the proteins because the catalytic kinase domain is in the N-terminus and the proposed site of phosphorylation is in the C-terminal lobe. The proteins must therefore bind in opposing orientations, and the leucine-zipper is incapable of determining a directional binding such as this.

A dimerization and autophophorylation model is also a feature of DLK, which forms homodimers, indicating an identical mechanism of autophosphorylation (Mata et al. 1996). Unlike MLK3 however, DLK does not contain a CRIB domain and hence cannot bind to the membrane-bound GTP forms of cdc42 and rac. Furthermore, DLK does not possess an SH3 binding region indicating that it is not capable of auto-inhibition, and therefore DLK does not require a G-protein interaction to activate it (Zhang & Gallo 2001). DLK dimerization does not require DLK phosphorylation or catalytic activity (Mata et al. 1996) and a protein corresponding to the DLK leucine-zipper domain only binds readily to itself or to wild-type DLK (Nihalani et al. 2000). These observations suggest that the DLK–DLK interaction is of high affinity, and DLK is probably capable of spontaneous homodimerization. The stimulus and regulation of DLK activity remains unclear, although it has been proposed that DLK is regulated by a sequestering protein.

Homodimerization, heterodimerization and autophosphorylation

With the auto-inhibition model in mind, one is forced to consider the question; if cdc42 is sufficient to induce autophosphorylation and activation of MLK monomers, then why is MLK dimerization important?

This question was addressed in studies which examined the ability of a leucine zipper-deficient MLK3 mutant to activate MKK4, requiring phosphorylation on both serine and threonine residues in a critical TXY activation motif. Mutation of either serine or threonine to a non-phosphorylatable amino acid prevents MKK4 activation of JNK. Monomeric mutant MLK3 could phosphorylate on the serine but not the threonine residue (Vacratsis & Gallo 2000). The reason for MLK dimerization is therefore that it is essential for activation of the substrate MAP2K, but the mechanism for this step is also unclear.

The importance of dimerization in the mechanism of MLK catalysis lead several authors to suggest that the ability of MLK proteins to heterodimerize would allow a few MLK proteins to induce MAPK activation in response to a wide selection of upstream signals (Leung & Lassam 1998). In support of the heterodimerization theory, coprecipitation of MLK2 and MLK3 was reported in this study. Further evidence is provided by the observation that there are interactions between MLK3 and DLK in 293T cells, and that kinase-dead DLK suppressed MLK3 activation of JNK, although the mechanism of this interaction was not investigated (Tanaka & Hanafusa 1998). The authors interpret this as DLK lying downstream of MLK3 in the JNK activation pathway, but an alternative and more logical hypothesis, in the light of the heterodimerization model, is that DLK and MLK3 bind via their leucine zippers. The catalytically inactive member would obviously be incapable of either autophosphorylation or substrate phosphorylation, resulting in termination of signal transduction.

However, this view is not universally accepted. It has been suggested that DLK exclusively forms homodimers, and does not dimerize with the closely related LZK or more distantly related MLK3 proteins (Nihalani et al. 2000). The inability of MLK proteins to heterodimerize may be important in maintaining signalling specificity and preventing cross-talk by limiting interactions between different MLK members in the same sub-cellular microenvironment.

JIP scaffolds

Research into MLK protein biology advanced considerably with the discovery of a mammalian MAPK scaffold protein, which was named JNK interacting protein-1 (JIP-1) (Whitmarsh et al. 1998).

The JIP-1 is a homologue of scaffolds found in yeast such as Ste5p and Pbs2p, which coordinate components of the pheromone and osmoregulatory response pathways, respectively.

Four JIP scaffolds have been now identified (Morrison & Davis 2003). JIP-1 and -2 (also called IB-1 and -2) share sequence homology, as do JIP-3 (also called JSAP-1) and JIP-4. The two functionally related pairs bear no structural similarities. The JIP proteins show different patterns of expression and slightly different MLK affinities, although all bind MKK7 and JNK. JIP-2 is unique in that it also binds MKK3 and p38 and has a higher affinity for p38 than JNK (Schoorlemmer & Goldfarb 2001; Buchsbaum et al. 2002). JIP-3 and JIP-4 bind MKK4 as well as MKK7 (Ito et al. 1999; Lee et al. 2002), and JIP-3 can also bind MEKK-1. MEKK-1 is a MAP3K which activates JNK (Yujiri et al. 1998) and has also been shown to activate ERK when overexpressed (Minden et al. 1994). MEKK-1 has also been shown to induce activation of the NF-κB transcription factor.

The scaffold proteins play a multi-regulatory role. Firstly, they provide a mechanism for regulated activation of JNK by facilitating the sequential phosphorylation of the MAPK pathway, and secondly they insulate the pathway to prevent cross-talk with other MAPK systems transducing unrelated messages in the cytosol. JIP scaffolds have also been shown to bind phosphatases and therefore negatively regulate signal transduction (Willoughby et al. 2003). It is also thought that JIP scaffolds regulate the signalling pathways via subcellular localisation (see below).

JIP-1 and JIP-2 have been shown to bind to MLK3 and DLK, but not to the other JNK activating MAP2K, MKK4. It has since been shown that other MLK family members such as MLK2 and LZK also bind JIP-1 and JIP-2, but MAP3Ks of other families such as MEKKs, Raf, MTK-1, PAKs, ASK-1, etc. do not bind. This suggested that a role of the JIPs is to provide specificity of MLK family members for JNK activation (Ikeda et al. 2001a; Nihalani et al. 2001).

This presence of a MAPK scaffold may explain the contrasting data regarding MLK protein heterodimerization. Studies examining such heterodimerization were based primarily on co-precipitating different MLK family members. However, the discovery of JIPs and the observation that JIP-1 can multimerize leads to the possibility that different MLKs are bound on each JIP scaffold and in this case co-precipitation would result in detection of different MLK family members. If considered together with the observation that DLK does not dimerize with the closely related LZK protein (Nihalani et al. 2000) it seems unlikely that heterodimerization occurs between any of the MLKs. It would also be consistent with the observation (Tanaka & Hanafusa 1998) that transfected kinase-dead DLK out-titrates active MLK3, perhaps by out-competing for binding to the JIP scaffold, thus inhibiting MLK3-induced JNK activation.

The identification of JIP scaffolds has led to studies which explore the mechanism of scaffold specificity. Using truncated constructs, the N-terminal domain of DLK has been shown to bind to JIP-1 and is necessary for JNK activation (Nihalani et al. 2001). The analogous region on the closely related LZK protein was also important for binding to this scaffold (Ikeda et al. 2001b). A region of 14 amino acids (entirely conserved between DLK and LZK) in the C-terminal portion of the protein was essential for the activation of MKK4 but not MKK7. This is particularly interesting because like MKK7, MKK4 specifically activates JNK, but unlike MKK7, MKK4 does not bind to all of the JIP scaffolds and MKK4-binding scaffolds are not as fully understood. A picture is therefore beginning to emerge of an MLK protein containing distinct functional regions, such as a JIP-binding domain in the N-terminus, a leucine zipper dimerization domain, and specific discrete regions containing consensus-binding sequences for activation of different substrate kinases.

JIPs and MAPK activation

The physiological importance of JIP proteins has been demonstrated in several studies. For example, JIP is required for cytokine-induced apoptosis (Haefliger et al. 2003), and overexpression of the JNK binding domain of JIP-1 prevents apoptosis in sympathetic neurones (Harding et al. 2001).

Before the discovery of JIP scaffolds it was believed that MAPK were activated by a three-tier phosphorylation cascade to allow signal amplification. In other words, phosphorylation induces kinase activation and each active enzyme could phosphorylate many substrate proteins. Mathematical models based on initial observations regarding levels of activated proteins appeared to support this amplification idea (Levchenko et al. 2000). However, while the model of MAPK enzymes bound to a scaffold protein in a ‘phospho-relay’ assembly is attractive because signal specificity is highly regulated, the ability to amplify the signal is lost because amplification requires free diffusion of component kinases.

Recent work investigating the nature of the interaction between MAPK pathway proteins and the JIP scaffolds may help to reconcile this conflict. Binding determinants have been identified mediating interactions at all levels of the pathway (Xu et al. 2001). JNK, MKK7 and MLKs bind to distinct regions of JIP-1 (Whitmarsh et al. 1998; Xu et al. 2001). Interestingly, studies using JIP-1 and DLK have suggested that the scaffold not only anchors the signalling proteins in the optimal orientation, but is also important in regulating the activity of the MAPK signal transduction.

Much of the work on this topic recently has been based on DLK rather than the MLKs. Using a mutagenesis approach, binding sites within MKK7 and DLK were identified that were involved in mediating interaction with JIP-1. Surprisingly, however, it was also found that the leucine-zipper of DLK binds to a region in the N-terminal portion of MKK7. This suggests that dimerization (and hence activation) of DLK, and binding to MKK7 may be mutually exclusive.

Dual Leucine Zipper-Bearing Kinase and JNK do not bind JIP simultaneously. Stimulation results in recruitment of JNK to the scaffold protein (Kelkar et al. 2000; Nihalani et al. 2001), and JNK recruitment coincides with decreased affinity of JIP for DLK. In the resting state, DLK is bound to JIP, and given the observation that in this state, MKK7 protein prevents DLK dimerization and hence activation, DLK is therefore held in a monomeric, unphosphorylated and catalytically inactive state.

A model has been proposed whereby appropriate stimulation induces an increase in the affinity of JNK for JIP. This induces a decrease in the affinity of the scaffold for DLK, presumably by some subtle structural alteration of JIP. Upon release, DLK spontaneously homodimerizes and becomes activated. The activated DLK dimer can phosphorylate MKK7 and induce MAPK signal transduction (Nihalani et al. 2001).

The model solves two key problems. Firstly, it suggests that JIP is the sequestrating protein responsible for preventing spontaneous DLK activation. Secondly, and more important, according to this scenario, activated DLK is not bound to JIP irreversibly and hence can phosphorylate many MKK7 proteins. Thus, although the scaffold provides specificity, signal amplification can still occur (Levchenko et al. 2000).

Other lines of evidence also support the model. JIP binding to all three module components simultaneously has not been observed. Decreased JIP–DLK binding is seen in the presence of excess JNK or conditions that favour JNK activation. The affinity of activated DLK for MKK7 is sufficient for interaction to occur without stabilization by a scaffold (Mata et al. 1996). However, it is unclear how JNK receives the initial signal to bind to JIP. One possibility is that JNK activation occurs via an alternative MKK4-dependent system and the role of MKK7 and DLK is to amplify JNK activation. Another is that there may be a reverse mechanism: a stimulus inducing DLK release from JIP would have the dual effect of allowing spontaneous DLK activation, and would increase the affinity of the JIP scaffold for JNK in readiness for MAPK signal transduction.

LZK shares a high degree of homology with DLK; it has also been shown to bind JIP-1 (Ikeda et al. 2001a), and is probably also regulated in this manner. some features are consistent, but the mechanism probably does not apply universally to all MLK family members, and in particular to MLK3, which can regulate its own activity negatively and remain associated with JIP-1 following stress (Kim et al. 2002).

In addition to JIP, there are other MLK-binding scaffold proteins that have been investigated in much less detail. Plenty Of SH3s (POSH) is thought to bind GTP-bound rac and mediate pro-apoptotic signalling via the JNK pathway (Tapon et al. 1998), because JNK-induced apoptosis is suppressed by POSH antisense oligonucleotides or siRNA (Xu et al. 2001). The POSH complex is regulated negatively by Akt-2, which induces complex dissociation (Figueroa et al. 2003). This is interesting because Akt-1 reduces the affinity of JNK for JIP-1 (Kim et al. 2002), suggesting that the negative regulatory effect of the Akt family on the JNK pathway may be mediated by disruption of scaffold complexes.

Subcellular localization of MLKs

There have been several studies investigating the subcellular localization of MLK proteins, using chiefly neuronal cells. Nagata et al. (1998) showed that MLK2 co-localizes with dually phosphorylated JNK to discrete microtubule-associated structures. Before the discovery of JIP scaffolds they also identified that MLK2 colocalizes with two members of the kinesin superfamily of motor proteins. Later all four JIP proteins were shown to be cargoes for the motor protein Kinesin-1 (Bowman et al. 2000; Whitmarsh et al. 2001) and this protein probably controls subcellular localization of MAPK activation. This was corroborated by intracellular staining for JIP-1 and JIP-2 proteins, which accumulated close to the plasma membrane in cytoplasmic projections that extend from the cell surface (Yasuda et al. 1999). Possibly the role of the kinesin motor is to localize the JIP-MAPK pathway cargo to a site of activation close to the cell membrane. Activated JNK phosphorylates its substrate c-jun in the nucleus, so following MAPK signal transduction, kinesin may carry deactivated JNK signalling components towards the plus end of microtubules, from the nucleus to the cell periphery, to re-prime the cells for signal transduction. JIP proteins may therefore mediate complex regulation of MAPK pathways via subcellular localization and transport.

This model was supported by the observation that there is co-localization of DLK to ‘dotted structures’ along microtubules (Hirai et al. 2002). These were distributed along the length of the microtubules with many present at the tips of the microtubules towards the cell periphery. MLK3 has also been reported to accumulate along microtubules, but primarily towards the minus ends at the centrosome (Swenson et al. 2003). Thus MLK3 may associate with a selection of different motor proteins, implying that there may be cell – type specific patterns of subcellular localization.

It has been suggested that different MLK proteins not only bind microtubules for localization, but regulate selective microtubule-associated cell processes. For example, MLK2 binds to the microtubule-associated GTPase, dynamin, which controls synaptic vesicle transmission (Rasmussen et al. 1998). while MLK2 can interact with clathrin and regulate clathrin-coated vesicle transport (Akbarzadeh et al. 2002).

The DLK has also been implicated in control of microtubule organization (Hirai et al. 2002). Before DLK transfection, Cos-1 cells possessed radially organized microtubules. These were disrupted completely upon DLK overexpression. Although DLK did not induce depolymerization of tubulin, the microtubules became organized in a random fashion or accumulated at the cell periphery. Cells expressing a kinase-deficient DLK mutant did not show this altered microtubule pattern, and microfilament organization and formation of focal contacts were not affected by DLK overexpression. In the developing mouse cortex neuronal DLK expression was observed in a ring of cells in which migration is arrested, suggesting that expression of DLK is developmentally regulated, and that hence the DLK–JNK MAPK pathway regulates neural cell migration during cortex development (Hirai et al. 2002).

MLKs and cell cycle regulation

Another potential MLK role is in cell cycle and proliferation. In a search for human proteins that are homologous to an essential fungal cell cycle G2/M transition protein NIMA a region of MLK3 was identified (Swenson et al. 2003). MLK3 phosphorylation was observed at this stage of the cell cycle, although JNK activity was unaltered, adding support to the idea that MLK3 has JNK-independent functions. In addition the subcellular localization of MLK3 is altered transiently during G2/M, localizing to the centrosomal region and inducing microtubule depolymerization. The microtubule effects were not observed after expression of kinase-deficient MLK3, and silencing of MLK3 expression using siRNA resulted in increased sensitivity to the microtubule stabilizing agent taxol, suggesting that endogenous MLK3 plays a physiological role in destabilizing microtubules during M phase entry.

In support of the possible role of MLK proteins in promoting cell proliferation, apparently overexpression of MLK3 induces growth of NIH 3T3 cells on soft agar (Hartkamp et al. 1999), and silencing of mlk3 by RNAi prevents both serum-stimulated cell proliferation and the proliferation of tumour cells (Chadee & Kyriakis 2004a). Inhibition of MLK proteins using the CEP11004 inhibitor (see below) blocked mitotic progression, probably by disruption of microtubule formation and spindle assembly (Cha et al. 2005). The tubulin-binding region of the N-terminus of MLK2 may also contribute to spindle formation (Rasmussen et al. 1997).

MLK3-/- mice

One obvious way to explore the role of MLKs further is to generate MLK deficient mice, and monitor their phenotype and function (Brancho et al. 2005). Remarkably, despite the above evidence linking MLK proteins with several different fundamental aspects of cell biology, the knockout mouse was viable and healthy. The only signalling phenotype observed was a defect in the ability of embryonic fibroblasts to activate JNK activation in response to TNFα. Clearly, a unique important regulatory role for MLK3 is not consistent with a phenotypically normal knockout mouse. This discrepancy may be due in part to functional reconstitution of MLK3 by other MLK proteins. But it is notable that several of the most recent reports are based upon MLK3 gene silencing using RNAi technology, and hence the in vitro results are difficult to reconcile with those of the in vivo knockout study.

MLK inhibitors

The importance of the MLK family in activating JNK and the role of this MAPK in neuronal cell death programmes and many other degenerative diseases presents MLKs as a potential therapeutic target for drug development. A semi-synthetic MLK inhibitor, CEP1347, is currently in clinical trials for Parkinson's disease and has also shown potential therapeutic effects in models of Alzheimer's disease and hearing loss (Cheng et al. 2002).

CEP 1347 was developed following observations of the neurotrophic and survival promoting effects of K252a, a small molecule naturally produced by the actinomycetes group of gram-positive bacteria (Kase et al. 1986). In initial reports K252a was observed to act as a relatively non-selective inhibitor of serine/threonine kinases, but later was a potent inhibitor of the receptor tyrosine kinase TrkA. Acting as a competitive agonist that binds to the ATP binding site, it blocks the autophosphorylation of TrkA with an IC₅₀ of 3 nM (Angeles et al. 1998) in response to neurotrophic ligands (Nye et al. 1992). Interestingly, however, despite inhibiting the action of neurotrophins, K252a was shown to possess its own ‘neurotrophic’ activity. For example, neurones deprived of NGF were shown to survive for at least 2 weeks in culture in the presence of K252a (Borasio 1990).

The drug was subsequently shown to exhibit these variable properties when present at different levels. At high concentrations, K252a inhibits neurotrophins, whereas at lower concentrations, K252a itself promotes survival. Compounds based on K252a were therefore synthesized to enhance the neurotrophic effects, while decreasing the inhibition of survival-promoting neurotrophins (Murakata et al. 2002). The 3,9-bis-[ethylthio(methyl)]-substituted K252a was found to have the most potent neurotrophic effects and was subsequently named CEP1347 or KT-7515 (Murakata et al. 2002). CEP 1347 is not cytotoxic above 200 nM, as is K252a, and does not possess the non-selective serine/threonine kinase inhibition properties of K252a.

However, the true molecular target mediating the survival-promoting effects of CEP1347 was not known. In NGF-deprived neurones, CEP1347 was shown to prevent JNK activation (Xu et al. 2001), while having no inhibitory effects on other MAPKS (Maroney et al. 1999). The possibility of direct JNK inhibition by CEP1347 was excluded because MEKK-1-mediated JNK activation in CHO cells was insensitive to CEP1347 addition.

The ability of upstream regulators of JNK to induce c-jun phosphorylation in the presence or absence of CEP1347 was assayed by in vitro kinase assay (Xu et al. 2001). Direct activators of JNK were excluded because no inhibition of MAP2K-induced JNK activation was observed by addition of CEP1347.

Rac and cdc42 have been shown to signal NGF-deprivation-induced neuronal apoptosis (Bazenet et al. 1998) making them a candidate for CEP action. The G-proteins induced a 20-fold increase in JNK activity, and this activation was blocked by CEP1347, although no interaction between rac/cdc42 and CEP1347 could be demonstrated, indicating that CEP1347 acts downstream of these proteins.

MLK proteins were shown to be activated by rac and cdc42 (Mota et al. 2001; Xu et al. 2001) and induce JNK activation, suggesting that these could be the targets of CEP1347 action, and indeed the drug was shown to prevent MLK3-induced activation of JNK in CHO cells. CEP1347 was later shown to be highly specific for the MLK family, and inhibit these proteins with high potency (Murakata et al. 2002).

The specificity of CEP1347 for MLK has been confirmed by genetic studies. For example cellular apoptosis induced by MLK transfection can be suppressed by CEP1347 treatment (Xu et al. 2001), and dominant negative MLKs and CEP1347 suppress death induced by constitutively active rac1 and cdc42 (Xu et al. 2001).

There are fewer published data using another indolocarbazole analogue of CEP1347 and CEP11004 (Murakata et al. 2002). This inhibitor wasused to prevent apoptotic death in dopamine neurons of the substantia nigra induced by 6-hydroxydopamine (Ganguly et al. 2004), and has also been explored in the developing heart (Tenhunen et al. 2004). Our own study, where CEP11004 was used initially primarily as an upstream inhibitor of JNK activation (Handley et al. 2005) suggested the concept of a wider role for MLK3 within the immune system as well.

With respect to the role outlined above of MLKs in microtubule organization, one of the earliest papers regarding the initial biochemical characterization of the K252a compound reported that not only did the drug mimic neurotrophins and promote survival, but also it could mimic growth factors and induce morphological change including neurite outgrowth (Maroney et al. 1998). Such cellular alteration has not been mentioned again in any of the inhibitor studies. However, in retrospect, an alternative explanation is that in the cells being studied, ciliary ganglion neurones, MLK proteins are involved in promoting microtubule instability. Inhibiting MLK therefore would induce microtubule polymerization giving the appearance of outgrowth. Such an interpretation would be consistent with other observations of MLK-treated cells; for example, the loss of microtubule organization in DLK-transfected neurones (Hirai et al. 2002), and the role of MLK3 in destabilization of microtubules during prometaphase (Swenson et al. 2003).

In summary, the inhibitor studies demonstrate that MLK proteins, initially thought to be specific activators of the JNK pathway, may in fact be involved in various cellular processes. The MLK-specific inhibitors provide an important and powerful tool for understanding MLK biology, not only in the nervous system, but also in other physiological and pathological processes, where there are few reported studies.

MLK and T-cell activation

Although most studies of MLK3 have used neuronal cells, MLK3 was identified first in human thymus. It has become clear that MLK3 may play a role in NF-κB activation in T cells, thus indicating how one signal-transducing protein can induce activation of both transcription factors required for expression of a particular gene; in this case, c-jun (via JNK) and NF-κB transcription factors, and the IL-2 gene (Hehner et al. 2000). MLK3 is implicated via its effect on IκB, the protein that sequestrates and controls NF-κB activity. IκB neutralization is signalled by phosphorylation by an Iκ Kinase complex (IKC) containing IκB kinases, IKKα and IKKβ. These kinases are activated by IKKKs, all of which are also MAP3Ks and are attached to the IKC. MLK3 functions as an IKKK, and its activity is potentiated by the expression of other IKKKs, such as MEKK-1 (Hehner et al. 2000). However, this MLK3-induced NF-κB activation was shown to be stimulus-specific, since CD3/CD28-induced NF-κB has been shown to be MLK3-dependent, whereas TNF-α and IL-1 activation of NF-κB is independent of MLK3 activity. Since rac and cdc42 are capable of activating NF-κB, the identification of MLK3 as an IKKK suggests a mechanism for this activation (Perona et al. 1997).

MLK, monocytes and macrophages

Following the studies of spatial regulation of MLK proteins, it has recently been reported that JIP-1 associates with the TLR signalling adaptor Myd88 (Sun & Ding 2006) in macorphages, and mediates interaction of MLK3 with this signalling complex. Previous studies have shown that JIP3 also associates with TLR2, TLR4 and TLR9 in a mouse macrophage cell line (Matsuguchi et al. 2003), suggesting this scaffold may also play a role in recruiting other MLKs to this signalling complex.

Inhibitor studies have confirmed that MLK3 does indeed play a role in signalling in cells of this lineage (Sui et al. 2006). The model system used exposed rat neurones to HIV-1 gene products (Tat and gp120) that are known to cause neurotoxicity. CEP1347 not only enhanced survival of these cells, but also inhibited activation of monocytes and macrophages. The findings were interpreted as suggesting that Tat and gp120 compromise MLK3 function in these cells, and that overall MLK3 may be a target for intervention in HIV-1 associated dementia.

MLK and DC

MLKs 2, 3, 4 (but not MLK1) as well as DLK have been shown to be expressed in DC (M.E. Handley et al. unpublished data) and these MLKs are all susceptible to the CEP inhibitors, as outlined above. In fact, the MLK inhibitor, and the part MLK plays in the JNK pathway, was the starting point for our studies of the role of MLK in non-neuronal cells.

The relationship between TLRs and MLKs in DC was explored in more detail using three stimuli: lipopolysaccharide (LPS), a known TLR4 ligand (Qureshi et al. 1999) (which has several non-signalling ‘accessory’ receptors such as CD14); poly (I:C), a synthetic double stranded RNA mimic, which activates TLR3 (Alexopoulou et al. 2001); and zymosan, a component of yeast cell walls, primarily composed of beta-glucans, which activates DC via a heterodimer of TLR2 and TLR6 (Frasnelli et al. 2005), as well as through the Dectin 1 glucan binding lectin (Rogers et al. 2005).

At the protein level polyclonal antibodies specific for total- and phosphorylated-MLK3 were used to confirm expression and to show that MLK3 phosphorylation occurs in response to LPS, and is maximal after 45 min, i.e. more rapidly than maximal phosphorylation of either p38 or JNK. The expression of MLK2 and DLK proteins was unexpected, as both isoforms were thought previously to be neuronally restricted. One plausible explanation for this is that these proteins are in some way involved in the regulation of cellular morphology, as has previously been suggested. DCs share with neurones the ability to extend long projections, or dendrites, and, indeed, when the DCs were first characterized they were thought to be a type of nerve cell. Therefore, the role of MLK2 and DLK in formation and regulation of these projections is of considerable interest, and may provide a important link between DC activation and migration that has not been documented previously.

There were two consistent bands of phosphorylated protein, possibly representing the phosphorylated and hyperphosphorylated forms (Swenson et al. 2003). Increased phosphorylation was seen in response to LPS and poly-I:C, although not in response to zymosan. The effect of DC–TLR stimulation upon the reintensity of the two bands was variable. In response to LPS, the lower band is of a higher intensity than the upper band, both in response to a time course of stimulation, and at all LPS concentrations. In response to poly (I:C), however, the two bands are of equivalent intensity at all concentrations. Little, if any, MLK3 phosphorylation was observed with any concentration of zymosan.

When signalling molecules downstream of MLKs were examined, again the TLR stimuli had differential effects. DC preincubated with CEP showed activation of the ERK pathway occurred in response to LPS, poly (I:C), or zymosan, and degradation of IκBα was not prevented. Thus MLK does not seem to regulate the canonical pathway of NF-κB activation in response to TLR ligation in DC. In response to both LPS and poly (I:C), CEP also inhibited the activation of p38 and JNK, but in response to zymosan, very little effect on p38 activation was observed. Furthermore, upregulation of all three classic DC markers (CD83, CD86 and HLA-DR) induced by LPS or poly (I:C) was blocked by the MLK inhibitor, but MLK inhibition had no effect on the upregulation of surface markers stimulated by zymosan. CEP inhibitors had no effect on baseline cytokine secretion (TNF-α, IL6, IL-10 and IL-12), whereas p38 inhibition blocked secretion of all cytokines. However, CEP blocked cytokine secretion in response to LPS and poly (I:C), but had no inhibitory effect on cytokine secreted in response to zymosan. In the LPS stimulation p38 and MLK inhibitors had similar effects, but in response to poly (I:C), CEP had a greater quantitative effect than SB for all three cytokines.

MLKs, DC and innate immunity

Taken together with the previous work in neuronal cells, and the known biochemical details of the MLKs as outlined above, these studies suggest that the MLKs may be an important, albeit hitherto unexplored, component of the signalling pathway in DC.

A link between at least one member of the MLK family, MLK3 and TLR signalling has been demonstrated clearly. MLK3 is phosphorylated in response TLR3 and TLR4 ligation but TLR2/6 ligation has only a very slight transient effect. This phosphorylation in response to TLR ligation, was not blocked by the MLK inhibitor CEP suggesting that an additional kinase is involved linking the ligation of TLR to MLK activation rather than simply an autophosphorylation event. However, an alternative explanation is that CEP functions in such a way that autophosphorylation of MLK proteins is not prevented, whereas the subsequent phosphorylation of substrate proteins is inhibited. Although possibly one of the two bands observed when immunoblotting for phosphorylated MLK3 represents a non-MLK3-specific target of the antiserum, the 12 potential phosphorylation sites in MLK3 (Vacratsis et al. 2002), suggest that phosphorylation leads to an activatory loop that generates a hyperphosphorylated MLK3 species possessing a slightly different electrophoretic mobility.

An attractive hypothesis is that these differences may reflect the utilization of different intracellular adaptor proteins by the TLRs. Mice deficient for different adaptor proteins show that TLR3 and TLR4, but not TLR2, possess a strong signalling requirement for the TIR-domain-containing adaptor inducing IFN-β (TRIF) protein (Yamamoto et al. 2003), and by analogy these studies suggest that TRIF-dependent MAPK activation may operate via MLK.

An alternative is that the different MLK dependency is a function of the presence of a second receptor for zymosan in addition to TLR2/6 (Rogers et al. 2005). Dectin-1 is a C-type lectin, leads to IL2 and IL10 production via syk tyrosine kinase-induced intracellular signalling. This signalling pathway has not been examined specifically in DC, but clearly possible by using this parallel pathway zymosan is able to by-pass MLK.

Post-MLK, in further support of the observations regarding the utilization of different signalling pathways by TLR3 and TLR2/6, the different stimuli have been shown previously to induce TNFα and IL6 by different mechanisms. While LPS-induced TNFα and IL6 production in macrophages is inhibited by overexpression of IκBα and thus NF-κB inhibition, the zymosan-induced production of these cytokines is not, although both stimuli induce significant amounts of both cytokines (Bondeson et al. 1999).

The involvement of the MLK family in the regulation of DC maturation is a novel observation. Effectively, MLKs act as MAP3Ks, inducing the phosphorylation and activation of the downstream kinases p38 and JNK and this is in turn translated into a distinctive surface and cytokine phenotype. Although this may prove to be a simplistic view, as there may well be some MLK functional redundancy, making detailed analysis of the MLK role more difficult than was expected [for example, in the recent reports of the MLK3-deficient mouse, there are far fewer phenotypic aberrations than one might expect given the number and variety of putative MLK3 functions (Brancho et al. 2005)], it does add a new dimension to signalling which needs to be explored further.

Dissecting the in vivo relevance of any innate immunity, DC–MLK is thus an important topic in itself, albeit likely to prove difficult to unravel. However, if different TLRs utilize different downstream MLKs, and the nature of the TLR and its agonist is a factor in determination of the T-cell polarization of the resulting immune response, then the TLR–MLK linkage might also influence polarity of outcome into adaptive immunity. For example, TLR2 ligands have been implicated in the selective induction of Th2 responses (Agrawal et al. 2003), and zymosan induces production of more IL-10 than TNFα, whereas both LPS and poly (I:C) induce secretion of more proinflammatory cytokines, such as TNFα than IL10 (Qi et al. 2003).

It is thus tempting to speculate that the differential role for MLK proteins in the responses to LPS and poly (I:C) vs. zymosan is in some way responsible for the DC to induce an effector T-cell bias. Further studies of the MLK pathway, in particular using a larger selection of stimuli (including known Th2-inducing ligands such as SEA or Pam-3-cys), and using a panel of variably efficient APC as controls, would help to confirm this hypothesis.