Determination of IKKε substrate specificity
IKKε has recently been identified as a breast cancer oncogene. However, the mechanism by which IKKε contributes to cell transformation is not known. Since IKKε is a serine/threonine kinase, we sought to determine whether the kinase activity of IKKε is necessary for its oncogenic activity. Wild-type IKKε or kinase-dead IKKε K38A was introduced into NIH-3T3 cells (). We found that cells expressing WT IKKε, but not IKKε K38A, exhibited robust anchorage-independent colony growth (). These results confirm that the kinase activity of IKKε is necessary for its activity as an oncogene.
Identification of the optimal IKKε phosphorylation motif
While several IKKε substrates involved in interferon signaling pathways have been identified, direct substrates required for the oncogenic functions of IKKε have not been reported. To address this question, we utilized proteomic and bioinformatic approaches to perform an unbiased screen for likely IKKε substrates. Specifically, we used a positional scanning peptide library assay recently developed in this laboratory to identify the optimal phosphorylation motif for IKKε (Hutti et al., 2004
; Turk et al., 2006
). This assay employs 198 biotinylated peptide libraries. Each library has a 1:1 mixture of serine and threonine fixed at the central position, and has one additional position fixed to one of the 20 naturally-occurring amino acids. All other positions contain a degenerate mixture of amino acids (excluding serine, threonine, and cysteine). Phosphothreonine and Phosphotyrosine were included at the fixed positions in order to facilitate the identification of kinases which have a requirement for priming phosphorylation events. Using recombinant GST-IKKε purified from HEK-293T cells, we performed kinase assays simultaneously on all 198 peptide libraries in solution using γ-32
P-ATP. Biotinylated peptides were captured with a streptavidin-coated membrane and the relative preference for each amino acid at each position was determined by the relative level of radiolabel incorporation into the corresponding peptides.
We found that IKKε exhibits strong sequence selectivity at multiple positions surrounding the phosphorylation site. IKKε strongly prefers substrates with a hydrophobic residue at the +1 position relative to the phosphorylation site (). This kinase also exhibits strong sequence selectivity for aromatic residues at the -2 position, and for bulky hydrophobic residues at the +3 position (). Kinase-dead IKKε K38A does not exhibit selectivity at these positions, confirming that peptides were not being phosphorylated by a contaminating kinase ().
To validate this motif, we generated a consensus peptide substrate, IKKε-Tide, and measured the phosphorylation of peptides bearing individual alanine substitutions. By comparing these results to those obtained using the consensus sequence we confirmed that substitution of amino acids at the +1 or -2 positions resulted in a decrease in the efficiency of peptide phosphorylation (). This motif has similarities to the sequence surrounding the autophosphorylation site in the activation loop of IKKε (DEKFVS172
VYGTE) due to the aromatic residue at -2 and aliphatic at +1, although this site is predicted to be suboptimal due to the lack of an aromatic residue at +3 (Peters et al., 2000
IKKε has been shown to phosphorylate one of the essential serines on IκBα in vitro
, but in vivo
phosphorylation of IκBα by this kinase has not been described (Peters et al., 2000
). Therefore, the role of IKKε in IκBα phosphorylation and degradation remains unclear. To examine this question, we performed kinase assays using a peptide substrate corresponding to the sequence surrounding Ser32 and Ser36 of IκBα. This peptide contains two potential phosphorylation sites, but neither site is within a sequence context that matches the optimal motif for IKKε. We found that this peptide was a poor in vitro
substrate for IKKε when compared with the optimal peptide determined from the peptide library screen (). In contrast, when recombinant GST-IKKβ was used to phosphorylate the same set of peptides, the IκBα peptide was phosphorylated by IKKβ much more efficiently than IKKε-Tide (Figure S1
). These observations suggest that IκBα is unlikely to be an important physiological substrate of IKKε. We recently demonstrated that, like IKKε, IKKβ prefers aromatic residues at the -2 position and hydrophobic residues at the +1 position (Hutti et al., 2007
). However, the phosphorylation motifs for these kinases differ at the -4, -5, and +3 positions. Taken together, these observations demonstrate that while the substrate specificities of IKKβ and the related kinase IKKε have overlapping characteristics, the optimal substrate peptides for these kinases differ in substantial ways and therefore can be predicted to have different (though possibly overlapping) in vivo
Prediction of IKKε substrates
Spot intensities from the peptide library screen were then quantified (Table S1
) and converted into a matrix which could be used with the bioinformatic search engine Scansite. Scansite (http://scansite.mit.edu
) allows proteome-wide searches for sites which best match the data provided by the input matrix (Obenauer et al., 2003
; Yaffe et al., 2001
). shows top-scoring candidate IKKε substrates obtained following the Scansite analysis, all of which scored in the top 0.05% of sites in the SwissProt database. Interestingly, a large number of predicted IKKε substrates are known to be involved in inflammatory and/or oncogenic signaling pathways. Of these potential substrates, the deubiquitinating enzyme CYLD was of particular interest, as it has been shown to have roles as both an inflammatory mediator and tumor suppressor, functions that could be downstream of IKKε (Bignell et al., 2000
). Our bioinformatic analysis predicted that CYLD is likely to be phosphorylated by IKKε at Ser418.
Candidate IKKε substrates identified by Scansite
CYLD is phosphorylated by IKKε at Ser418
To further facilitate the identification of novel IKKε substrates we raised antibodies against a collection of phosphopeptides biased towards the optimal IKKε phosphorylation motif We verified that these antibodies recognize known IKKε substrates in a kinase-dependent manner (data not shown). These antibodies were then used to determine whether CYLD is phosphorylated at a site matching the IKKε phosphorylation motif HEK-293T cells were cotransfected with Myc-epitope tagged CYLD (Myc-CYLD) and either GST-IKKε WT or kinase-dead IKKε K38A. CYLD was immunoprecipitated via its Myc tag and immunoblotted with the anti-IKKε phospho-substrate antibody. shows that the phospho-substrate antibody blotted CYLD which had been transfected with WT IKKε, but not IKKε K38A. CYLD treated with calf-intestinal phosphatase (CIP) following cotransfection with IKKε was no longer recognized by the phospho-substrate antibody, confirming that the IKKε phospho-substrate antibody specifically recognizes phosphorylated CYLD ().
CYLD co-immunoprecipitates with and is phosphorylated by IKKε
In order to determine whether IKKε can directly phosphorylate CYLD, an in vitro kinase assay was performed. Wild-type GST-IKKε or GST-IKKε K38A was purified from HEK-293T cells. Myc-CYLD was separately transfected into HEK-293T cells and immunoprecipitated. When the CYLD immunoprecipitate was incubated in an in vitro kinase assay with WT IKKε, strong phosphorylation of CYLD was observed (). This phosphorylation was not observed in the presence of IKKε K38A.
To determine whether IKKε and CYLD physically interact, Myc-CYLD was cotransfected into HEK-293T cells expressing GST- IKKε WT or K38A. CYLD was immunoprecipitated via its Myc tag and these immune complexes were blotted with an anti-GST antibody to identify IKKε. In CYLD immune complexes we identified both WT and kinase-dead IKKε (). Moreover, when we performed the reciprocal experiment we found that Myc-CYLD was also observed in the IKKε precipitates ().
While CYLD Ser418 was predicted by Scansite to be the optimal site for IKKε phosphorylation (ENRFHS418
LPFSL), two additional serines within the CYLD sequence were potential, though less optimal, IKKε phosphorylation sites (DSRFAS547
LQPVS and KKIFPS772
LELNI). Therefore, we used mass spectrometry to determine which residue(s) of CYLD is phosphorylated in vivo
. Myc-CYLD and GST-IKKε were cotransfected into HEK-293T cells. CYLD was immunoprecipitated, subjected to SDS-PAGE, and Coomassie stained (Figure S2
). The band corresponding to Myc-CYLD was excised, digested with trypsin or chymotrypsin, and subjected to microcapillary LC/MS/MS. A phosphopeptide consistent with phosphorylation at Ser418 was identified. This analysis confirmed that CYLD is phosphorylated at Ser418, the same site predicted by our bioinformatic analysis (). Ser418 is evolutionarily conserved in all sequenced mammals, as well as Gallus gallus
and Xenopus tropicalis
(). In addition, the -2F, +1L, and +3F relative to Ser418 (which correspond to the IKKε phosphorylation motif) are also conserved, providing further evidence for the evolutionary importance of this phosphorylation site.
CYLD is phosphorylated by IKKε at Ser418
To further verify that Ser418 is the critical site phosphorylated by IKKε, we created Myc-epitope tagged mutants of CYLD at Ser418 (S418A), Ser547 (S547A), and Ser772 (S772A). We introduced wild-type CYLD or each of these mutants into HEK-293T cells alone or with GST-IKKε and isolated CYLD immune complexes. Mutation of either Ser547 or Ser772 did not affect recognition of CYLD by the IKKε phospho-substrate antibody (). However, Myc-CYLD S418A was no longer recognized by the IKKε phospho-substrate antibody when cotransfected with IKKε (). In addition, when we introduced Myc-CYLD or Myc-CYLD S418A with IKKε into HEK-293T cells and performed an immunoprecipitation using the IKKε phospho-substrate antibody, we found that wild-type Myc-CYLD, but not Myc-CYLD S418A, was efficiently immunoprecipitated by the IKKε phospho-substrate antibody (). Finally, endogenous CYLD was immunoprecipitated from IKKε-transformed NIH-3T3 cells utilized in . Immunoblotting with the IKKε phospho-substrate antibody revealed that CYLD is indeed phosphorylated in cells transformed by IKKε, but not in cells expressing IKKε K38A (). Together, these data confirm that CYLD Ser418 is phosphorylated in vivo in the presence of IKKε.
It has been reported that Ser418 of CYLD is a substrate of the canonical IKK family members, IKKβ and IKKα (Reiley et al., 2005
). Therefore, in order to determine which IKK family member(s) can most efficiently phosphorylate CYLD, we cotransfected Myc-CYLD with individual IKK family members fused to GST. CYLD was immunoprecipitated using an anti-Myc antibody and blotted using the anti-IKKε phospho-substrate antibody. Figure S4
shows that CYLD was efficiently phosphorylated in cells transfected with IKKε. In contrast, CYLD is phosphorylated inefficiently in cells transfected with IKKα and IKKβ, even when greater than 5× more kinase is present (Figure S3
). While it is therefore possible for any of these IKK family members to phosphorylate Ser418 of CYLD, CYLD is a much better substrate for IKKε than for IKKα or IKKβ. Our analysis of IKKε and IKKβ peptide substrate specificities also demonstrated that these kinases should be expected to have largely different, though possibly overlapping, substrate pools ( and S1
IKKε-mediated phosphorylation of CYLD at S418 facilitates transformation
Since CYLD is a known tumor suppressor and IKKε is a newly-discovered oncoprotein (Bignell et al., 2000
; Boehm et al., 2007
; Massoumi et al., 2006
), we hypothesized that phosphorylation of CYLD by IKKε might play a role in the regulation of IKKε-mediated cell transformation. In we show that forced expression of IKKε induces a tumorigenic phenotype in NIH-3T3 cells. To determine whether suppression of CYLD also induced cell transformation in this experimental model, we suppressed murine CYLD using two distinct short hairpin RNAs (shRNA), each of which strongly suppresses endogenous levels of CYLD (). Suppression of CYLD induced substantial anchorage-independent growth of NIH-3T3 cells when compared to control NIH-3T3 cells ().
Suppression of CYLD protein expression is sufficient to induce transformation
To determine whether CYLD phosphorylation is required for IKKε-mediated transformation, we generated IKKε-transformed NIH-3T3 cells that stably express wild-type CYLD, CYLD S418A, or CYLD S772A (). We found that Flag-epitope tagged IKKε-expressing cells exhibited robust anchorage-independent growth that was 4-fold above that observed in control NIH-3T3 cells or cells expressing WT CYLD or CYLD S772A (). In contrast, in tumorigenic cells expressing Flag-epitope-tagged IKKε, co-expression of CYLD S418A suppressed anchorage-independent growth (). A similar result was observed in vivo when we evaluated the contribution of CYLD phosphorylation to IKKε-induced tumorigenicity. Introduction of IKKε-transformed NIH-3T3 cells into immunodeficient animals yielded tumor formation at a high penetrance (, data not shown). While expression of WT CYLD or CYLD S772A failed to significantly alter tumorigenicity, expression of CYLD S418A led to statistically-significantly smaller tumors (p<0.05) (). Taken together, these findings demonstrate that phosphorylation of CYLD by IKKε at serine 418 is necessary for IKKε to fully induce transformation.
CYLD phosphorylation at serine 418 is important for IKKε-mediated transformation
Phosphorylation of CYLD at Ser418 decreases CYLD activity
CYLD is a deubiquitinating enzyme which removes Lys63-linked ubiquitin chains from a large number of inflammatory mediators including TRAF2, TRAF6, and NEMO, as well as the pro-proliferation transcription factor BCL-3 (Brummelkamp et al., 2003
; Kovalenko et al., 2003
; Massoumi et al., 2006
; Trompouki et al., 2003
). Removal of ubiquitin chains by CYLD inactivates these substrates, inhibiting inflammatory signaling and cell cycle progression. Cell transformation induced by loss of CYLD is hypothesized to occur following the accumulation of a variety of ubiquitinated species, which leads to both increased cell proliferation and the increased transcription of NF-κB-regulated anti-apoptotic factors. Therefore, we sought to determine the effect of CYLD phosphorylation on ubiquitination of known CYLD substrates.
TRAF2 is a critical NF-κB mediator ubiquitinated via Lys63-linked ubiquitin chains following stimulation with TNFα, and CYLD removes these ubiquitin chains to prevent uncontrolled TNFα-induced inflammation. To determine the effect of CYLD phosphorylation on TRAF2 ubiquitination, HEK-293T cells were transfected with Myc-epitope tagged TRAF2 in combination with GFP-WT CYLD or CYLD S418A and GST-IKKε. In order to examine only changes in Lys63-linked ubiquitin chains, cells were also transfected with an HA-tagged ubiquitin variant which has all lysines mutated to arginine except for Lys63 (HA-Ub-K63). Cells were stimulated with TNFα for 10 minutes prior to lysis and CYLD-mediated deubiquitination of TRAF2 was assessed by immunoprecipitation. As expected, CYLD efficiently removed Lys63-linked ubiquitin chains from TRAF2 (). Interestingly, CYLD-mediated deubiquitination of TRAF2 was blocked in the presence of IKKε. In constrast, cotransfection of CYLD S418A with IKKε did not block deubiquitination of TRAF2 (). These data demonstrate that phosphorylated CYLD exhibits less deubiquitinating activity than unphosphorylated CYLD. Thus, IKKε regulates the deubiquitination of TRAF2 by phosphorylating CYLD.
Phosphorylation of CYLD at Ser418 decreases CYLD activity
The IKK regulatory subunit NEMO undergoes Lys63-specific ubiquitination at Lys285 when coexpressed with RIP2, resulting in increased NF-κB activity (Abbott et al., 2004
; Abbott et al., 2007
). CYLD efficiently removes these RIP2-induced ubiquitin chains from NEMO (Abbott et al., 2004
). To determine whether phosphorylation of CYLD affects RIP2-induced NEMO ubiquitination, HEK-293T cells were cotransfected with Myc-NEMO, OMNI-RIP2, HA-Ubiquitin, GFP-CYLD, and GST-IKKε. As expected, cotransfection of RIP2 with NEMO increased NEMO ubiquitination, and this ubiquitination was suppressed by CYLD (). However, cotransfection of IKKε with CYLD resulted in the stabilization of NEMO ubiquitination, suggesting that phosphorylation of CYLD decreases its deubiquitinase activity. As in , cotransfection of CYLD S418A with IKKε failed to stabilize NEMO ubiquitination (). Together, these observations demonstrate that IKKε-mediated phosphorylation of CYLD at Ser418 inhibits CYLD deubiquitinase activity.
We next sought to determine what effect this phosphorylation exerts on NF-κB activity. We have shown previously that the gene encoding IKKε, IKBKE
, is amplified and overexpressed in MCF-7 cells (Boehm et al., 2007
). These cells were transiently transfected with Myc-TRAF2 alone or in combination with Myc-CYLD, Myc-CYLD S418A, or Myc-CYLD S772A and we performed a NF-κB-luciferase reporter assay. In cells expressing TRAF2 alone, we found substantial (14.9 fold) activation of the NF-κB reporter (, data not shown). Cotransfection with a sub-saturating amount of WT CYLD or CYLD S772A resulted in an approximately 50% decrease in TRAF2-induced NF-κB activation. However, co-transfection with CYLD S418A resulted in a statistically significantly larger decrease in NF-κB reporter activity (p<0.05) (). These results demonstrate that phosphorylation of CYLD at Ser418 suppresses CYLD activity, resulting in increased NF-κB transcriptional activation.
As further validation, we assessed whether CYLD phosphorylation modulates NF-κB activity in the context of transformation. NF-κB luciferase reporter activity was measured in IKKε-transformed NIH-3T3 cells overexpressing WT or mutant CYLD (S418A and S772A). IKKε-transformed cells exhibited strong NF-κB activation in comparison to control cells, and introduction of WT CYLD and S772A failed to induce a substantial change in NF-κB activity. As observed in MCF-7 cells, however, CYLD S418A more strongly inhibited IKKε-induced NF-κB activation. These findings indicate that IKKε-mediated phosphorylation of CYLD at serine 418 plays a role in the observed activation of NF-κB. As the decrease in NF-κB activation observed in the presence of CYLD S418A was not as robust as the decrease in transformation that was observed in , it is likely that NF-κB is not the only pathway affected by phosphorylation of CYLD. However, these findings suggest that the regulation of CYLD phosphorylation by IKKε contributes to the NF-κB activity necessary for IKKε-mediated cell transformation.
To further confirm that cell transformation mediated by CYLD suppression is NF-κB dependent, we utilized the NIH-3T3 cells expressing shRNAs targeting murine CYLD which were described in . When a non-phosphorylatable IκBα mutant (IκBα S32,36A), which acts as an NF-κB “superrepressor,” was stably expressed in these cells, we found a dramatic reduction in anchorage-independent colony growth (). Thus, suppression of CYLD induces NF-κB-dependent cell transformation similar to that induced by overexpression of IKKε.