Protein kinase C-associated kinase can activate NFkappaB in both a kinase-dependent and a kinase-independent manner.

Protein kinase C-associated kinase (PKK, also known as RIP4/DIK) activates NFkappaB when overexpressed in cell lines and is required for keratinocyte differentiation in vivo. However, very little is understood about the factors upstream of PKK or how PKK activates NFkappaB. Here we show that certain catalytically inactive mutants of PKK can activate NFkappaB, although to a lesser degree than wild type PKK. The deletion of specific domains of wild type PKK diminishes the ability of this enzyme to activate NFkappaB; the same deletions made on a catalytically inactive PKK background completely ablate NFkappaB activation. PKK may be phosphorylated by two specific mitogen-activated protein kinase kinase kinases, MEKK2 and MEKK3, and this interaction may in part be mediated through a critical activation loop residue, Thr184. Catalytically inactive PKK mutants that block phorbol ester-induced NFkappaB activation do not interfere with, but unexpectedly enhance, the activation of NFkappaB by these two mitogen-activated protein kinase kinase kinases. Taken together, these data indicate that PKK may function in both a kinase-dependent as well as a kinase-independent manner to activate NFkappaB.

Protein kinase C-associated kinase (PKK, 1 also known as RIP4/DIK) is an ankyrin repeat domain containing serine/threonine kinase that can activate NFB when expressed in cell lines (1,2) and is required for keratinocyte differentiation in vivo (3). PKK was originally identified on the basis of its association with PKC␤ (4) and PKC␦ (5) in yeast two-hybrid screens. Although PKC␤ can phosphorylate PKK, there is no evidence for the catalytic activation of PKK by this or other PKCs (4).
The kinase domain of PKK shares a high degree of homology with the catalytic domains of members of the receptor-interact-ing protein (RIP) family of protein kinases (1,2). RIP is a death domain containing serine/threonine kinase that was first described as a consequence of its association with TNFR1 (tumor necrosis factor-␣ receptor 1) and FAS (CD95) (6,7) and has since been shown to be recruited by TRADD (TNFR1-associated death domain protein) (6 -8) and TRAF2 (TNF receptor associated factor 2) (8) following TNF-␣ signaling. RIP has also been shown to associate with other TNFR family members such as RAIDD and DR3 (9,10). Two other kinases, RIP2 and RIP3, were designated on the basis of the homology of their kinase domains to the corresponding domain of RIP, although other segments of these proteins are highly divergent. RIP2 (RICK/ CARDIAK) is a caspase-associated recruitment domain-containing kinase that associates with TNFR1 and the TRAF1, TRAF5, and TRAF6 adaptors (11). It has been implicated in Toll-like receptor signaling and has been shown to be important for both innate and adaptive immune responses (12,13). The C terminus of RIP3 has no homology to any known functional domain, although this segment is critical for RIP3 to interact with and negatively regulate RIP (14).
RIP, RIP2, and RIP3 have all been shown to both activate NFB and to induce apoptosis in a kinase-independent manner when overexpressed (7,8,11,15,16), although RIP3 has also been shown to inhibit NFB induction in some cases (14,17). Following TNFR stimulation, both RIP and RIP2 have been shown to recruit IKK-␥ and to consequently contribute to the activation of the IKK complex, suggesting that these RIP family members function as scaffold-like molecules rather than as protein kinases when they activate NFB (18 -20). As with other RIP kinases, PKK can activate NFB when overexpressed in cell lines, although the catalytic activity of PKK has been considered to be required for NFB activation (1,2). In addition, PKK can activate NFB in IKK-␥-deficient cell lines (1), further suggesting that it induces the activation of NFB by a distinct mechanism from that employed by other RIP kinases.
Given our limited understanding of how PKK is activated and how it activates NFB, we wished to further explore these issues biochemically. There is an appropriately positioned SXXXS motif in the activation segment of the catalytic loop of murine and human PKK (in kinase subdomains VII and VIII) that is identical to a motif in members of the MAP2K family that is phosphorylated by MAP3Ks (21). Although we have established that these serines are not required for the catalytic activity of PKK, we have identified two MAP3Ks, MEKK2 and MEKK3, that can phosphorylate PKK in a specific manner. Surprisingly, we have also determined that certain catalytically inactive mutants of PKK can activate NFB, whereas other enzymatically compromised mutants cannot. Catalytically inactive mutants of PKK that are incapable of activating NFB can block PMA-induced NFB activation. These mutants do not abrogate but surprisingly enhance MEKK2-and MEKK3-induced activation of NFB. Taken together, these data support a model in which PKK can participate in the activation of NFB in both a kinase-dependent and a kinaseindependent manner.

MATERIALS AND METHODS
Plasmids and Cell Lines-FLAG-tagged PKK has been described previously (4). Plasmids for c-Raf, ASK1, MEKK1, and MEKK4 were kindly provided by Dr. John Kyriakis. Plasmids for MEKK2 and MEKK3 were kindly provided by Dr. Michael Karin and Dr. Gary Johnson. PCR-based site-directed mutagenesis was used to generate FLAG-tagged versions of all PKK mutants. 293T cells were maintained as described previously (4). U2OS cell lines stably expressing M96G-PKK were generated by co-transfection of pCMV-FLAG-M96G-PKK in conjunction with pBabePURO and maintained in complete medium supplemented with puromycin (1 g/ml).
Assays for Kinase Activity and Phosphorylation-293T cells were transiently transfected, and assays were performed as described previously (4). Briefly, 293T cells were transfected using the calcium phosphate method. After 48 h the cells were lysed in 1% Triton X-100, and the post-mitochondrial supernatant was immunoprecipitated using anti-FLAG (M2 monoclonal; Sigma) antibodies. Equal amounts of each immunoprecipitate were analyzed using an anti-FLAG immunoblot assay and an in vitro kinase assay. Histone H1 (Sigma) was used as a substrate in in vitro kinase assays.
Luciferase Assays-The luciferase assay was performed using the dual luciferase reporter assay system (Promega). The pBIIxLuc NFB reporter construct, containing four NFB-binding sites upstream of a firefly luciferase reporter gene, was kindly provided by Dr. Sankar Ghosh (22). A Renilla luciferase reporter construct, pRL-TK (Promega), was used to normalize for transfection.

RESULTS
Thr 184 Is a Critical Residue within the Activation Loop of PKK-A notable feature of the activation loop of PKK is an appropriately positioned SXXXS motif (Fig. 1A), characteristically seen in MAP2Ks, which is typically phosphorylated by MAP3Ks when MAP2Ks are activated. Mutation of these serine residues typically results in a catalytically inactive kinase, as has been experimentally demonstrated for a number of MAP2Ks including members of the MEK, SEK, and IB kinase families (21,(23)(24)(25). In an analogous fashion, the serine residues (Ser 173 and Ser 177 ; Fig. 1A) within this motif in PKK were mutated to alanines to assess their relevance to the catalytic activity of this kinase. Constructs encoding FLAG-tagged wild type and mutant PKKs were independently transfected into 293T cells. The lysates from these transient transfections were then immunoprecipitated with anti-FLAG antibodies. Half of each immunoprecipitate was used in an in vitro kinase assay employing histone H1 as a substrate; the other half was examined using an anti-FLAG immunoblot assay (to assess protein levels). The catalytic activity of PKK was measured by examining the autophosphorylation of PKK and the phosphorylation of histone H1 in an in vitro kinase assay. We have previously shown that PKK migrates as two major species: a hyperphosphorylated 110-kDa band and an underphosphorylated 97-kDa band when FLAG-tagged PKK is over expressed in 293T cells, and we have also shown that the catalytic activity of PKK is required for it to become hyperphosphorylated (4). Therefore, a third criterion for assessing catalytic activity is the decreased altered migration of PKK on a Western blot. Interestingly, the mutation of Ser 173 , Ser 177 , or both to alanines does not affect the catalytic activity of PKK (Fig. 1, B and C, compare lanes 2-4 with wild type PKK, lane 1) as assessed by any of these criteria. Furthermore, the mutation of these serines to glutamate residues does not appreciably increase the catalytic activity of PKK (data not shown), although similar mutants frequently, but not invariably, contribute to the activation of many MAP2Ks (21). Taken together, these data strongly sug-gest that the SXXXS motif is not of relevance for the catalytic activation of PKK.
We subsequently mutated additional serines and threonines that were proximal to the SXXXS motif (Fig. 1A). The S171A mutation in concert with the S173A/S177A mutation does not have any affect on the catalytic activity of PKK as measured in our assays (Fig. 1, B and C, lane 5). Thr 184 , which is conserved across all species (Fig. 1A), was mutated to an alanine, either alone or in conjunction with the S171A/S173A/S177A triple mutation. This latter combination, including T184A, is referred to as quadruple mutant (QM). Both T184A-PKK and QM-PKK exhibit markedly compromised catalytic activity as measured by autophosphorylation as well as the ability of PKK to phosphorylate histone H1 in an in vitro kinase assay. In addition, the T184A and QM mutants migrate largely as 97-kDa species, further suggesting that their catalytic activity has been compromised ( Fig. 1, B and C, lanes 6 and 7). Although the mutation of Thr 184 to a glutamic acid (T184E) did not restore the catalytic activity of PKK (Fig. 1, B and C, lanes 8), these data suggest that Thr 184 is a critical residue within the activation loop of PKK with respect to catalytic activity.
MEKK2 and MEKK3 Can Phosphorylate PKK-As stated above, SXXXS motifs are characteristically found in the activation loops of MAP2Ks, and serine residues in these motifs are typically phosphorylated by MAP3Ks. Based on this knowl-

FIG. 1. Thr 184 is a critical residue within the catalytic loop of PKK.
A, alignment of the catalytic loop of PKK from three different species: mouse (GenBank TM accession number AF302127), human (GenBank TM accession number AJ278016), and zebrafish (GenBank TM accession number AF487541). Diamonds signify, in order, Ser 171 , Ser 173 , Ser 177 , and Thr 184 . B, in vitro kinase assay performed with mutant PKKs as described in text. Each FLAG-tagged mutant in a pCMV5 expression vector was transiently transfected as described. The lysate was subjected to an anti-FLAG immunoprecipitation. Half the lysate was used in the in vitro kinase assay; the other half was used for an anti-FLAG Western blot (in panel C) (4). Histone H1 was used as a substrate in the in vitro kinase assay. The asterisk represents an as yet unidentified protein. C, anti-FLAG Western blot of half the immunoprecipitate from B. Equal amounts of each immunoprecipitate were loaded onto an 8% SDS-PAGE gel. The immunoblots were probed with M2 monoclonal antibody (Sigma). The 110-kDa band represents a hyperphosphorylated form, and the 97-kDa band represents a hypophosphorylated form (4). edge, we surveyed a panel of MAP3Ks as to whether or not they could phosphorylate PKK. Because wild type PKK has been shown to be able to presumably phosphorylate itself in in vitro kinase assays (Fig. 1), a previously described catalytically inactive mutant of PKK, K51R-PKK (4), was used in this assay (see Fig. 4). The MAP3Ks surveyed included MEKK1, MEKK2, MEKK3, MEKK4, ASK-1, c-Raf, and NIK, but only MEKK2 and MEKK3 were found to be able to phosphorylate PKK when co-transfected in 293T cells. This was demonstrated by the phosphorylation of K51R-PKK in an in vitro kinase assay and the decreased mobility of K51R-PKK on an anti-FLAG Western blot ( Fig. 2A and data not shown).
We initially predicted that PKK would be phosphorylated by MAP3Ks on the serine residues within the SXXXS motif. However, we have already shown that these serines, when mutated to alanines, have no affect on the catalytic activity of PKK. Therefore, as expected in light of the data from Fig. 1, MEKK3  (Fig. 2B) and MEKK2 (data not shown) were still able to phosphorylate PKK when the SXXXS motif was mutated to AXXXA. This was assessed by the phosphorylation of PKK in an in vitro kinase assay and the altered migration of this kinase in an anti-FLAG Western blot. Some information regarding the sites that may be critical for the interaction of PKK with MEKK2 and MEKK3 came from the analysis of the quadruple mutant, QM-PKK (S171A/S173A/S177A/T184A). When MEKK3 (Fig. 2C) or MEKK2 (data not shown) is co-transfected with QM-PKK into 293T cells, a portion of QM-PKK becomes phosphorylated as evidenced by the appearance of a more slowly migrating PKK species on an anti-FLAG Western blot. However, neither MEKK3 nor MEKK2 can phosphorylate QM-PKK in an in vitro kinase assay, whereas they can phosphorylate K51R-PKK. In addition, the proportion of the slower migrating QM-PKK species in cells co-transfected with MEKK2 or MEKK3 is less than what is seen with K51R-PKK (co-transfected with MEKK3 or MEKK2) on an anti-FLAG Western blot (compare Fig. 2, B, lane 2, and C, lane 4). These results suggest that Thr 184 may participate in the interaction between PKK and MEKK2/MEKK3. Some Catalytically Inactive Mutants of PKK Can Activate NFB-It has recently been reported that PKK can activate NFB when overexpressed in cell lines. In contrast to RIP, RIP2, and RIP3, this activity can be abolished when a point mutation (K51R or D143A) is made within the predicted ATPbinding pocket to catalytically inactivate this kinase (1, 2). Moreover, these catalytically inactive mutants can function in a dominant negative manner, blocking the induction of NFB by a variety of different stimuli, including phorbol ester. Surprisingly, when the catalytically inactive mutants, K51R-PKK, QM-PKK ( Figs. 1 and 2), and T184A-PKK ( Fig. 1) were cotransfected with an NFB reporter plasmid into 293T cells, we were able to detect a significant amount of NFB activation, albeit to a lesser degree than with wild type PKK (Fig. 3).
To further investigate whether the catalytic activity is in fact dispensable for PKK to activate NFB, we generated three additional catalytically inactive mutants: M96G-PKK, D143A-PKK, and D143N-PKK. M96G-PKK represents a mutation within the predicted ATP-binding pocket that is thought to enlarge the pocket so that specially engineered ATP analogs may bind. Analogous mutations render some kinases unable to bind ATP and thus catalytically inactivate them (26). D143A- FLAG-tagged K51R-PKK was co-transfected with a panel of MAP3Ks into 293T cells; similar assays to those described in the legend to Fig. 1 were performed. B, S173A/S177A-PKK on a K51R mutant background can be phosphorylated by MEKK3. FLAG-tagged K51R PKK or triple mutant FLAG-tagged K51R/S173A/S177A PKK was co-transfected into 293T cells with vector DNA (pCMV5) or with MEKK3. An in vitro kinase assay (upper panel) and an anti-FLAG immunoblot assay (lower panel) were performed on immunoprecipitates, in a manner similar to that described in the legend to Fig. 1. C, FLAG-tagged versions of S173A/S177K PKK and quadruple mutant S171A/S173A/S177A/T184A PKK were co-transfected with vector or with MEKK3 into 293 T cells. An in vitro kinase assay (upper panel) and an anti-FLAG immunoblot assay (lower panel) were performed on immunoprecipitates.
PKK is an ATP-binding pocket mutation that has previously been shown not to activate NFB (1). D143N-PKK represents a more conservative mutation of the aspartate residue similar to one that has been shown to catalytically inactivate other kinases (33). In contrast to the catalytically inactive kinases described above, when these mutants were co-transfected with an NFB reporter construct into 293T cells, no NFB activation was detected (Fig. 3). Therefore, we have identified two groups of catalytically inactive PKK mutants. One group (K51R-PKK, QM-PKK, and T184A-PKK) can activate an NFB reporter construct when expressed in cell lines, whereas another group (M96G-PKK, D143A-PKK, and D143N-PKK), cannot.
To determine whether there was a difference between these two groups of PKK mutants that might account for their distinct abilities to activate NFB, we closely examined the catalytic activities of each of the mutants. All six of the catalytically inactive mutants and wild type PKK were individually expressed in 293T cells and immunoprecipitated by virtue of their FLAG tag. As before, half the immunoprecipitate was used in an in vitro kinase assay, with the other half being utilized in an anti-FLAG Western blot. All of the catalytically inactive mutants were unable to phosphorylate histone H1 to any appreciable degree in an in vitro kinase assay (Fig. 4). In addition, all six mutants migrate primarily as a faster 97-kDa species, which is consistent with an underphosphorylated, catalytically inactive form. The only consistent difference that we have seen between the two groups is the degree of phosphorylation of these mutants in an in vitro kinase assay (Fig. 4, arrow). The PKK mutants that are able to activate NFB in this assay (K51R-PKK, QM-PKK, and T184A-PKK; Fig. 4, lanes 2-4,  arrow) appear to be phosphorylated to a greater extent than the mutants that cannot activate NFB (M96G-PKK, D143A-PKK, and D143N-PKK; Fig. 4, lanes 5-7, arrow). This difference is not a result of a significant variation in protein levels, as can be seen from the Western blot (Fig. 4). Therefore, the specific phosphorylation seen may be accounted for either as a consequence of a small amount of catalytic activity (and there-fore autophosphorylation) that these mutants still retain, the ability of these mutants to co-immunoprecipitate endogenous factors that can phosphorylate PKK, or both. If these mutants do have some residual kinase activity, then this would support the view that minimal catalytic activity of PKK is necessary for NFB activation by this enzyme. On the other hand, if these mutants are able to associate with endogenous kinases/regulators that phosphorylate PKK (including endogenous PKK itself), which in turn facilitates the activation of NFB, then these data suggest that PKK may also function in a kinaseindependent scaffold-like manner to activate NFB. These possibilities are explored further below. Regardless, these data suggest that the catalytic activity of PKK contributes in a major way to its ability to activate NFB. Certain mutations that disrupt ATP-binding also decrease the ability of PKK to activate NFB, clearly illustrating the importance of a kinasedependent step in this pathway.
Full-length PKK Is Necessary for the Optimal Activation of NFB-We wished to further explore why K51R-PKK, generally considered to be an inactive kinase, can activate NFB, and to assess what domains of PKK are necessary for NFB activation. To address these issues, we created a series of truncation and deletion mutants of wild type PKK based on predicted domain structures. In addition, we introduced the K51R substitution into each of these truncation mutants (Fig.  5A). All truncation/deletion mutants that were catalytically active could activate an NFB reporter construct when expressed in 293T cells, albeit to a lesser level than wild type PKK (Fig. 5B). Deletion of four ankyrin repeats alone or of the intermediate domain also compromised the ability of PKK to activate NFB. These results suggest that the entire ankyrin repeat domain and the intermediate domain are necessary for the optimal activation of NFB.
Intriguingly, all of the truncation and deletion mutants surveyed that also contained the K51R mutation were unable to activate NFB in our assay (Fig. 5B). We have already seen that these truncations significantly decrease the ability of wild type PKK to activate NFB. These results suggest that the K51R mutation does indeed compromise the catalytic activity of this kinase. Because full-length K51R-PKK has only a fraction of the NFB activating ability of wild type PKK, these assays further reveal the essential role of the ankyrin repeat region and the intermediate domain of PKK in NFB activation.  Fig. 1. The lower panel is an anti-FLAG Western blot, similar to that seen in Fig. 1. The asterisk represents unidentified proteins also seen in mock transfected lysates. The arrow shows the specific, phosphorylated, PKK band.

PKK Mutants That Cannot Activate NFB Act as Dominant
Negative Mutants and Block PMA-induced Activation of NFB-It has been recently reported that catalytically inactive mutants of PKK can block PMA-induced activation of NFB (1,2). We wished to determine whether specific domains of PKK were required in a catalytically inactive context to compromise PMA induced activation of NFB. As expected, full-length M96G-PKK, which we have shown is catalytically inactive and cannot activate NFB (Figs. 3 and 4), significantly blocks PMAinduced NFB activation (Fig. 6). We then examined whether the truncation mutants K51R/442-PKK and K51R/304-PKK can also act in a dominant negative manner. As can be seen in Fig. 6, both of these mutants block PMA-induced activation of NFB. This is an interesting observation given that the catalytically active forms of these mutants, 304-PKK and 442-PKK, are both able to activate NFB (Fig. 5). Although it might have been predicted that inactive kinases containing the ankyrin repeat or intermediate domains or both might have been required to compete with endogenous PKK, it is clear that the catalytically compromised kinase domain of PKK alone suffices to abrogate PMA-mediated activation of NFB.

Dominant Negative PKK Does Not Block the NFB Activating
Ability of MEKK2 and MEKK3-Aside from PMA-induced activation of NFB, we were interested in identifying other pathways in which PKK may contribute to NFB activation. Transient co-transfection-based luciferase assays for NFB activation, in which a dominant negative form of PKK is used in conjunction with other expression constructs, have been used to examine whether PKK lies downstream of specific receptors or signaling molecules. In our hands these kinds of transient assays give highly variable results most likely because of varying expression levels and transfection efficiencies. We therefore generated U2OS cell lines that stably express M96G-PKK and chose three stable transfectants that expressed increasing amounts of mutant PKK (Fig. 7A) to examine whether specific MAP3Ks may be functionally linked to PKK.
We have already shown that two specific MAP3Ks, MEKK2 and MEKK3, can phosphorylate PKK (Fig. 2). These two kinases have been reported to activate NFB (27,28). We reasoned that if PKK were required for the activation of NFB by MEKK2 or MEKK3, then M96G-PKK would be expected to circle, ankyrin repeat domain. B, PKK and the various mutants described in the text (1 g) were co-transfected with pBI-IxLuc (0.1 g) and pRL-TK (0.02 g) into 293T cells. In these luciferase assays, all of the values have been normalized to the mock (empty vector) transfected lysate, and each assay was performed in triplicate. The error bars denote one standard deviation in this experiment, similar to Fig. 3. Wild type PKK has been shown on a separate graph for reference. Note that the scales of the two graphs are different. C, equal volumes of each of the above lysates were loaded on an 8% SDS-PAGE gel and subjected to an anti-FLAG Western blot to determine protein levels, similar to Fig. 3. Certain truncation mutants, such as PKK 442, reproducibly present as multiple bands, presumably reflecting multiple phosphorylation states. IMD, intermediate domain.
compromise the activation of NFB by these MAP3Ks in our stable U2OS transfectants. It is clear from Fig. 7B that MEKK2 and MEKK3 do not activate NFB in a PKK-dependent manner, implying that PKK does not functionally lie downstream of these MAP3Ks. Surprisingly, the ability of MEKK2 and MEKK3 to activate NFB was enhanced in U2OS cells expressing intermediate levels of M96G-PKK (p Ͻ 0.05) as well as in cells expressing higher levels of M96G-PKK (p Ͻ 0.01; Fig. 7B). In contrast, M96G-PKK did not significantly influence the activities of MEKK1, which activates NFB (Ref. 29 and Fig. 7B) but does not phosphorylate PKK ( Fig. 2A and data not shown), or MEKK4, which neither phosphorylates PKK (data not shown) nor activates NFB (Ref. 30 and Fig. 7B).
To test the possibility that M96G-PKK contributes to MEKK2-and MEKK3-mediated activation of NFB as a result of being phosphorylated by these MAP3Ks, we initially asked whether MEKK2 or MEKK3 can phosphorylate M96G-PKK. Surprisingly, MEKK2 (data not shown) and MEKK3 (Fig. 7C) are unable to phosphorylate M96G-PKK as measured by in vitro kinase assay and Western blot. Although we do not understand the relevance of the interaction between PKK and these MAP3Ks, these data lend some support to the view that PKK may participate in the activation of NFB in a kinaseindependent manner.

DISCUSSION
Very little is understood about the mechanisms by which PKK is activated or about the process by which PKK activates NFB. We have presented evidence that Thr 184 , located within the activation loop, is critical with respect to the catalytic activity of PKK. In addition, we have shown that specific MAP3Ks, MEKK2 and MEKK3, can phosphorylate PKK, and this phosphorylation event is influenced by Thr 184 . Every domain of PKK, including the kinase domain, the intermediate domain, and the ankyrin repeat region, contributes to its ability to activate NFB. Finally, we have provided evidence that supports a model that PKK can act in both a kinase-dependent and a kinase-independent manner to activate NFB, and this is discussed here.
The role for the catalytic activity of PKK in the activation of NFB is most evident when we consider that wild type PKK activates NFB to levels that are significantly higher than those induced by catalytically inactive mutants. Identification of proteins phosphorylated by PKK may be crucial to obtaining a better understanding of this pathway. Perhaps this kinasedependent step requires the autophosphorylation of PKK and/or the phosphorylation of a downstream target. Indeed, other RIP-like kinases have been reported to phosphorylate factors involved in cell signaling events. RIP has been shown to phosphorylate and catalytically activate MEKK1 (31), and RIP2 has been shown to phosphorylate and catalytically activate ERK2, thereby activating AP-1 (32). In an analogous fashion, PKK may phosphorylate and thereby activate a downstream factor that can activate NFB.
More surprising are the results that suggest PKK may participate in the activation of NFB in a kinase-independent manner. The observation that K51R-PKK, QM-PKK, and T184A-PKK, three presumably catalytically inactive mutants, can activate NFB when overexpressed in cell lines was unexpected, given that it has been reported that the catalytic activity of PKK is absolutely required for the activation of NFB (1,2). The question obviously arose as to whether or not these mutants are truly catalytically inactive. Although we cannot rule out the possibility that QM-PKK and T184A-PKK retain a small amount of catalytic activity, there is considerable evidence to suggest that K51R-PKK is catalytically inactive. The mutation of this critical lysine within the ATP-binding domain has been well characterized as one that ablates the ability of a kinase to associate with ATP (33)(34)(35)(36). Additional evidence comes from the analysis of truncation mutants that carry the K51R mutation. All of the catalytically active truncations of wild type PKK can activate NFB to some degree in our hands, whereas truncation mutants that carry the K51R mutation act as dominant negative inhibitors, blocking PMA-induced NFB activation. If the K51R mutation retained some catalytic activity, it would have been expected that these truncation mutants would not have acted in a dominant negative fashion. We therefore believe that many pieces of evidence suggest that K51R-PKK is catalytically inactive and that these data illustrate that PKK can activate NFB in a kinase-independent manner.
How then do K51R-PKK, QM-PKK, and T184A-PKK activate NFB in a kinase-independent manner? Some evidence comes from in vitro kinase studies, where we have observed that these PKK mutants become slightly phosphorylated. It could be that phosphorylated PKK can recruit other proteins that contribute to the activation of NFB. As stated earlier, RIP and RIP2 are thought to recruit IKK-␥ and assemble the IKK complex in a kinase-independent manner (18 -20). Because PKK can activate NFB in IKK-␥-deficient cells (1), it is possible that PKK recruits IKK-␤, or some other unidentified protein in the IKK complex, to activate NFB. In addition, PKK may possibly recruit other kinases, such as MEKK2 or MEKK3, to the IKK complex to facilitate the activation of NFB. Why three additional catalytically inactive PKK mutants, M96G-PKK, D143A-PKK, and D143N-PKK, cannot activate NFB when expressed at high levels is still unclear. It may be that these mutations may contribute to a more global distortion of PKK, blocking their association with other factors, possibly other kinases. This is evidenced by the observation that these mutants are not phosphorylated in an in vitro kinase assay. Functional studies avoiding the use of PKK overexpression may reveal further insights into this pathway.
We do not fully understand why K51R-PKK activates NFB in our studies, whereas it has been shown to act as a dominant FIG. 7. M96G-PKK does not block the NFB activating activity of MEKK2 and MEKK3. A, anti-FLAG Western blot of equivalent amounts (30 and 60 g in alternate lanes) of lysates from three different U2OS cell lines stably expressing M96G-PKK. Lanes 1, 2, and 3 refer to three different transfectants. The cell lysates were prepared as described under "Materials and Methods." B, luciferase assay for NFB. 1 g of MEKK1, MEKK2, MEKK3, and MEKK4 were co-transfected with pBIIxLuc (0.1 g) and a pRL-TK (0.02 g) into either wild type (wt) U2OS cells or U2OS cells stably expressing M96G-PKK (1, 2, and 3 refer to the three clones categorized above). Statistical significance was assessed by a paired t test. *, p Ͻ 0.5; **, p Ͻ 0.01. C, MEKK3 or control empty vector were co-transfected with PKK, M96G-PKK, or K51R-PKK into 293T cells. The upper panel depicts an in vitro kinase assay. The lower panel shows an anti-FLAG Western blot, similar to Fig. 1 (B and C). negative mutation by others (2). One possible explanation is that we used mouse PKK with a K51R mutation in our studies, whereas the other study used a human K51R-PKK mutant. There are a number of small structural differences between human and murine PKK, some of which might contribute to species-specific differences in PKK function. A more detailed analysis might provide further insights regarding NFB activation by PKK.
We have also provided evidence that PKK is phosphorylated by two MAP3Ks, MEKK2 and MEKK3. This phosphorylation appears to be specific, because other MAP3Ks were tested and were not found to phosphorylate PKK, although it should be noted that this panel was far from exhaustive, and there still may be other MAP3Ks that can phosphorylate PKK. Ser 173 and Ser 177 , the serines within the SXXXS motif are not phosphorylated by MEKK2 and MEKK3. We have shown that Thr 184 may participate in the interaction between MEKK2/3 and PKK. However the significance of the phosphorylation of PKK by MAP3Ks remains unclear. Interestingly, RIP2 has been reported to be phosphorylated by the MAP3K, c-Raf, even though it does not contain the canonical SXXXS motif; this interaction is important for the activation of ERK2 (32). Although the interaction between MEKK2 and MEKK3 with PKK may have parallels when one considers other RIP kinases and MAP3Ks, the biological significance of this interaction remains to be determined.
It is clear from these data that PKK can activate NFB in both a kinase-dependent and a kinase-independent manner. The relative contributions of each of these mechanisms, which may not be mutually exclusive, remains to be ascertained. Further studies involving PKK-deficient cells may prove useful in this regard.