Introduction

Protein kinases bestow novel identity to their substrates by adding phosphate moieties at the site of modification that often play key regulatory roles in a diverse array of signaling events. For example, inflammatory or pathogenic signals trigger precisely regulated induction of the transcription factor NF-κB in metazoans(Hoffmann and Baltimore, 2006) that depends upon the upstream activation of the Inhibitor of κB Kinase-complex (IKK-complex), comprising two catalytic subunits, IKK1 (also known as IKKα) and IKK2 (or IKKβ), and an essential regulatory scaffolding protein, NEMO (IKKγ)(DiDonato et al., 1997; Rothwarf et al., 1998; Zandi et al., 1997). IKK2 activity is maintained at a very low basal level in resting cells(Ghosh and Karin, 2002; Hacker and Karin, 2006; Hinz and Scheidereit, 2014; Liu et al., 2012). Upon activation, IKK2 phosphorylates the N-terminus of IκBα (Inhibitor of κBα) protein specifically at two Serine residues (S32 and S36) marking it for ubiquitinylation-mediated proteasomal degradation (Figure 1A). Signal-responsive phosphorylation of IκBα refers to the phosphorylation of S32 and S36 by the IKK-activity. Degradation of IκBα liberates NF-κB to execute its gene expression program (Hayden and Ghosh, 2008; Hinz and Scheidereit, 2014; Karin and Ben-Neriah, 2000; Scheidereit, 2006). Mutation of both S32 and S36 to alanine residues converts IκBα into a non-phosphorylatable super-repressor of NF-κB(Brown et al., 1995; Lin et al., 1995). Furthermore, IKK possesses exquisite specificity for these two serines, and it fails to phosphorylate IκBα when S32 and S36 are mutated even to another phosphorylatable residue, threonine(DiDonato et al., 1997).

Figure 1:

IKK2 is an essential Ser/Thr kinase that ensures the activation of a particular signaling cascade - NF-κB, with high fidelity despite its reported pleiotropy in other contexts(Antonia et al., 2021; Schröfelbauer et al., 2012; Schrofelbauer and Hoffmann, 2011). IKK2 is a multidomain protein consisting of a kinase domain (KD) followed by a Ubiquitin-like domain (ULD) and the scaffold dimerization domain (SDD) (Figure 1B). IKK2 toggles between inactive and active states based on the phosphorylation status of two serine residues (S177 and S181; S176 and S180 in the case of IKK1) within its activation loop (activation loop) in the KD(Huse and Kuriyan, 2002; Liu et al., 2013; Polley et al., 2013; Xu et al., 2011). Replacement of these two serines with phospho-mimetic glutamate residues (S177E, S181E; henceforth EE) renders the kinase constitutively active. IKK2 is reported to be primed through phosphorylation at S177 by the upstream kinase TAK1 that leads to auto-phosphorylation at S181, thus rendering the kinase fully active in cells(Zhang et al., 2014). However, oligomerization upon association with NEMO, in the presence of linear or Lys-63 linked poly-ubiquitin (Ub) chains or at a high concentration of IKK2, also enables trans auto-phosphorylation at S177 and S181(Chen, 2012; Du et al., 2022; Ea et al., 2006; Polley et al., 2013). It was also shown that NEMO not only helps in the activation of the IKK-complex, but also ensures that the IKK-activity is well directed towards IκBα in the inhibitor bound transcriptionally inactive NF-κB (NF-κB: IκBα complex)(Schröfelbauer et al., 2012). Activated IKK2 phosphorylates IκBα specifically at residues S32/S36, for which the C-terminal SDD of IKK2 (residue range 645-756) is critical – a shorter construct of IKK2 lacking the NBD and SDD displayed CK2-like activity that did not retain the exquisite specificity of IKK2 to phosphorylate the S32/36 of IκBα (Shaul et al., 2008). Other sites of IκBα e.g., the residues in the C-terminal PEST region that contains proline (P), glutamic acid (E), serine (S) and threonine (T) residues, but not S32/S36, can be phosphorylated by various kinases(Barroga et al., 1995; Tergaonkar et al., 2003). The role of IκBα-phosphorylation at sites other than the signal responsive S32/S36 phosphorylation in NF-κB activation and signaling is elusive. These findings underscore the ability of IKK2 to function specifically in order to guarantee rapid NF-κB activation in canonical signaling. It is however, still not fully understood how IKK2 achieves its exquisite specificity towards S32/S36 of IκBα to mark it for proteasomal degradation that liberates active NF-κB dimers.

Here we report an in-depth analysis of the catalytic features of IKK2 that revealed that IKK2 undergoes multisite autophosphorylation and can phosphorylate its own tyrosine residue(s) in addition to its own serines or the serines of its substrate IκBα, i.e., IKK2 possesses dual specificity. This phosphorylation at tyrosine residue(s) is autocatalytic. This phosphorylation at tyrosine residues is dependent upon the prior activation of IKK2 via activation loop S177/S181 phosphorylation, and is critical to phosphorylation of signal responsive S32/S36 of IκBα. Mutation of Y169 of IKK2 to phospho-ablative phenylalanine severely compromised phosphorylation of IκBα at S32. Mutation of Y188 to phenylalanine rendered the kinase practically inactive in vitro. IKK2 activity towards signal responsive S32/S36 of IκBα was severely compromised in MEF cells reconstituted with these two mutant IKK2s in response to treatment with TNFα. We also observed that mutation of the conserved ATP-interacting residue K44 in IKK2 to methionine leads to a loss of tyrosine as well as S177,S181 auto-phosphorylation activity of the kinase along with its ability to phosphorylate S32/S36 of IκBα, thereby clearly suggesting a critical role of this Lysine residue in regulating IKK2’s kinase activity for tyrosine as well as serine phosphorylations. Interestingly, IKK2 K44M retained non-specific kinase activity referring to phosphorylation of IκBα at residues other than S32 and S36. Furthermore, we observed that, after auto-phosphorylation at tyrosine, IKK2 can phosphorylate IκBα in the absence of exogenous supply of ATP, as long as ADP is present in the reaction. Overall, our results reveal unique, and likely transient, autophosphorylated intermediates of IKK2 that appear to be critical in relaying phosphoryl group(s) specifically to S32/S36 of IκBα rather than following the conventional mode of phosphate group transfer directly from ATP to the substrate seen in eukaryotic protein kinases. IKK2 thus ensures fidelity of NF-κB activation. Recent reports in other kinases indicate that multisite (auto)phosphorylation is a major determinant in regulating function of respective kinases and regulation exerted by those kinases. Our study adds another layer of regulation and critical nature of such autophosphorylation that plays a significant biological role in metazoans.

Results

IKK2 undergoes auto-phosphorylation at uncharacterized sites

Signal-induced phosphorylation of activation loop Ser177/Ser181 marks the general activation of IKK2 (Figure 1B) from an unphosphorylated inactive state. Phosphorylation of several other serines located within the flexible C-terminal region of IKK2, spanning amino acid 701 to the end, has also been reported (Figure 1B)(Delhase et al., 1999). Hyperphosphorylation of these C-terminal sites down-regulates IKK2 activity in cells. It was also reported earlier that, ectopically overexpressed IKK2 is capable of autophosphorylating these two serines in the activation loop in trans even in the absence of an activating physiological signal. Indeed, recombinantly purified FL IKK2WT is capable of autophosphorylation in vitro when incubated with γ³²P-ATP that increases with the time of incubation (Figure 1C). The spread-out of slower migrating phospho-IKK2 band likely indicates phosphorylation of the protein at multiple sites. We also observed that autophosphorylation of IKK2 occurred irrespective of the presence of its cognate binding partner NEMO and bonafide substrate IκBα (Figure 1D). To confirm that this phosphorylation of IKK2 was indeed autophosphorylation and not due to an unrelated kinase copurified with IKK2, we generated a K44M mutant of full-length IKK2 where the ATP-binding invariant lysine residue conserved in Ser/Thr/Tyr kinases was mutated to methionine. This lysine residue anchors ATP in the respective binding pocket of the kinase by forming hydrogen bonds with the α- and β-phosphate of ATP that is critical for productive phosphorylation reaction in protein kinases. Mutating this K to M is expected to greatly reduce or eliminate the kinase activity. An in vitro kinase assay using this mutant along with the WT kinase using γ³²P-ATP as the phosphor-donor confirmed that the IKK2 K44M lacked both IκBα (substrate) and autophosphorylation activities (Figure S1A). That these phosphorylations were indeed autophosphorylation was further supported by another in vitro radioactive kinase assay performed in the presence or absence of an IKK-specific ATP-competitive inhibitor, Calbiochem Inhibitor VII (Figure 1E) where the kinase and kinase-substrate reactions were incubated for 30 minutes with the inhibitor prior to adding the phosphor-donor, γ³²P-ATP. A similar assay was performed with cold ATP and specific phosphorylation of S32/36 on IκBα was monitored by immunoblotting in presence and absence of the same inhibitor (Figure S1B). Purification status of recombinant FL IKK2 WT and FL IKK2 K44M proteins used in this study are shown in Figure S1C.

Next, we used different versions of the IKK2 protein where all previously known phosphorylation sites were lacking: Ser177 and 181 were mutated to phosphor-mimetic Glu (EE; this version of IKK2 is known to be constitutively active (Zandi et al., 1998)), and the C-terminal serine-rich region was truncated to different extents. The shortest construct,

Δ664EE that entirely lacked the flexible C-terminus (670-756) was further subjected to limited proteolysis using trypsin to eliminate residues, if any, from the flexible ends of the enzymes and this trypsin-treated IKK2 was purified prior to in vitro kinase assays. Each of these constructs was active in phosphorylating itself and the substrate IκBα (Figure 1F). The occurrence of autophosphorylation despite the absence of activation loop serines and serines within the C-terminus strongly suggested the presence of hitherto uncharacterized sites of autophosphorylation in IKK2.

Auto-phosphorylation of IKK2 reveals its dual specificity

IKK2 is known as a bonafide Ser/Thr kinase that undergoes autophosphorylation. In the previous section, we found that IKK2’s autophosphorylation activity might extend beyond the previously characterized site(s). We asked if it is a dual specificity kinase. To test if Tyr residues can undergo auto-phosphorylation in addition to Ser, IKK2 incubated with Mg²⁺-ATP for various time periods was probed with antibodies that specifically recognize activation loop serines of IKK2 (Figure 2A, upper panel) or phospho-tyrosine residues (Figure 2A, middle panel). Unexpectedly, we observed de novo auto-phosphorylation of IKK2 on tyrosine residues along with the anticipated auto-phosphorylation of activation loop serines. It may be noted that autophosphorylation activation loop-S177/181 also increased with time, indicating inhomogeneous and incomplete phosphorylation of activation loop-S177/181 in the recombinant IKK2 when purified from Sf9 cells. We tested another phospho-tyrosine specific antibody to further substantiate this de novo tyrosine autophosphorylation of IKK2. It confirmed that detection of de novo phosphotyrosine on IKK2 was independent of the source of the antibody used (Figure 2B). The tyrosine auto-phosphorylation was abolished in the presence of an IKK-specific inhibitor, Inhibitor VII (Figure 2C). A similar assay was performed to monitor the phospho-tyrosine level in IKK2 as a function of urea concentration. This revealed the detrimental effects of urea on tyrosine autophosphorylation of IKK2 at concentrations beyond 1M (Figure S2A). A more sensitive assay with γ³²P-ATP as the phospho-donor also indicated that urea concentrations beyond 1M inhibited both auto-phosphorylation (pan-phosphorylation) and substrate (IκBα) phosphorylation (Figure S2B). Furthermore, the IKK2-K44M mutant (kinase inactivating) did not display any tyrosine autophosphorylation when incubated with ATP unlike WT IKK2 (Figure 2D). These results confirm the novel tyrosine phosphorylation ability of IKK2, demonstrating its dual specificity to both its own Ser and Tyr residues. We also observed that activity of IKK2 towards S32 and S36 of IκBα and its own tyrosine residue(s) were affected similarly when the kinase assays were performed in the presence of various kinase inhibitors, both general (AMPPNP and Staurosporine) and IKK-specific (MLN-120B, TPCA, Inhibitor VII) (Figure S2C). These observations confirm that canonical activity of IKK2 towards S32 and S36 of IκBα, and towards its own tyrosine residues are mediated by the same active site.

Figure 2:

Next, we checked the requirement of the C-terminal NEMO binding domain (NBD) for autocatalytic dual specificity. Constructs of various truncations (lacking NBD) were capable of phosphorylating itself on tyrosine residue(s), and phosphorylating IκBα at signal-responsive S32 and S36 (Figure 2E), indicating that the NBD domain is not required for either its ability to phosphorylate S32 and S36, as well as autophosphorylating itself on tyrosine. Next, we assessed the autocatalysis at Ser and Tyr by IKK2 using a constitutively active NBD-truncated version of IKK2 (IKK2^Δ664EE). As anticipated, the protein was recognized by monoclonal phospho-serine antibody with or without fresh ATP-treatment; however, phospho-tyrosine antibody recognized it strongly only upon fresh ATP-treatment (Figure 2F) further confirming dispensability of the NBD in de novo tyrosine autophosphorylation of IKK2.

Next, we asked if phosphorylation of activation loop S177/S181 was necessary for this dual specificity of IKK2. To this end, we purified the S177A, S181A (AA; non-phosphorylatable activation loop) phospho-ablative mutant of IKK2, and checked its tyrosine autophosphorylation activity. We found that the IKK2AA mutant lacked tyrosine autophosphorylation activity (Figure 2G) in an in vitro kinase assay. It suggests that activation loop S177/S181 serine phosphorylation is essential in conducting dual specificity in IKK2. Indeed, the IKK2 AA mutant was incapable of phosphorylating S32/36 of IκBα as well (Figure S2D) in an in vitro kinase assay.

IKK2 autophosphorylation is critical to specificity of IκBα phosphorylation

As mentioned earlier, it was long known that activation loop-serine phosphorylation was necessary for IKK-activation; we have now established that IKK2 phosphorylates itself on tyrosine residue(s) in addition to activation loop-serines. We wanted to further understand the significance of IKK2 autophosphorylation in the context of IκBα S32/S36 phosphorylation. It is known that disruption of a universal salt bridge mediated by the ATP-binding Lys (K44 in IKK2) is detrimental for kinase activity. However, it is not known if disruption of this salt bridge has any bearing on substrate specificity of eukaryotic protein kinases (EPKs). During our investigation, we noted unusual behavior in the IKK2-K44M mutant. We had noticed earlier that IKK2-K44M was incapable of de novo autophosphorylation on tyrosine residues (Figure 2D). It also failed to phosphorylate itself on the activation loop Ser177 and 181 (Figure 3A). Interestingly, IKK2-K44M phosphorylated full-length WT IκBα robustly albeit much less efficiency than WT IKK2 (Figure 3B) in a radioactive in vitro kinase assay. IKK2-K44M also showed a minor auto-phosphorylation activity compared to WT IKK2 in the absence of substrate (Figure 3B). We performed a similar kinase assay using full-length WT IκBα where cold ATP was used as the phosphate donor, and phosphorylation status of IκBα at S32, S36 was monitored by immunoblotting using an antibody that specifically recognizes IκBα when phosphorylated at those two N-terminal serines. It revealed that IKK2-K44M is unable to phosphorylate S32/S36 of IκBα (Figure 3C). Since radioactive kinase assays report pan-phosphorylation of a substrate (Figure 3B), it possibly reflects phosphorylation of the PEST region of FL IκBα. We additionally compared IKK2-K44M with WT IKK2 using radioactive kinase assay with both FL IκBα WT and FL IκBαAA substrates in the absence or presence of Inhibitor VII. As anticipated, FL IκBα was phosphorylated much more efficiently than FL IκBαAA by WT IKK2, and phosphorylation was practically abolished in the presence of 10 μM Inhibitor VII (Figure 3D). In contrast, IKK2-K44M phosphorylated FL IκBαAA more efficiently than FL IκBα, and reduction in the presence of inhibitor VII was strikingly less (Figure 3D). It indicates that IKK2-K44M, despite retaining non-specific kinase activity towards the PEST region, lost phosphorylation activity towards S32, S36 of the bona fide substrate IκBα. We subjected 100ug of FL IKK2 K44M protein used in the above studies to LC-MS/MS analysis, which did not detect any contaminating kinase from Sf9 cells. It is noteworthy that an equivalent K to M mutation in Erk2 also did not abolish the basal activity, but rather retained ∼5 % of it(Robbins et al., 1993). The above set of experiments confirm that K44 residue is not only crucial for both IKK2’s well-established serine phosphorylation as well as newly-found tyrosine autophosphorylation activities, as well as its specificity for phosphorylating its bona fide substrate IκBα at S32/36. Hence, these results strongly suggest that IKK2-autophosphorylation is critical for its substrate specificity.

Figure 3:

Tyrosine phosphorylation is critically important for IKK2 activity

A number of earlier studies indicated that phosphorylation of IKK2 on Y169, Y188 and/or Y199 to be critical for its function(Darwech et al., 2010; Meyer et al., 2013). Our mass spectrometric analyses identified multiple tyrosine residues to be phosphorylated, e.g., Y169, Y188, Y199, Y325 (data not shown). We compared available structural models of human IKK2 (PDB ID: 4KIK, 4E3C) to investigate whether tyrosine phosphorylation within the activation loop could be supported based on its structural features(Liu et al., 2013; Polley et al., 2013). The structure of IKK2 dimer is asymmetric in that the activation loop of one monomer is largely disordered while that of the other monomer is mostly ordered (PDB ID: 4KIK)(Liu et al., 2013). In the active monomers S177 and S181 are clearly observed. In these active state conformers, the position and conformation of Tyr169 appears to be well-poised for accepting a phosphate from ATP (Figure S4A) intramolecularly. A superposition of IKK2 with a pseudo-substrate-bound PKA shows that the hydroxyl of Tyr169 in IKK2 projects toward the γ-phosphate of ATP similarly to the serine hydroxyl on the PKA inhibitor pseudo-substrate(Knighton et al., 1991; Nolen et al., 2004) (Figure S4A & B). Interestingly, Y169 is the DFG+1 (DLG in case of IKK2 and IKK1) residue while Y188 (equivalent to Y204 of PKA) is located at the end of activation loop prior to the conserved APE motif (Figure 4A and 4B). It was previously established that the DFG+1 position is crucial for defining the substrate specificity of a kinase(Chen et al., 2014). Y188 counterpart Y204 of PKA is reported to allosterically modulate PKA activity by altering its conformational entropy(Ahuja et al., 2017; Yang et al., 2005). We further focused our attention to Y169 and Y188 as they are located in the activation loop of the kinase.

Figure 4:

To test the effect of phosphorylation at Y169 and Y188, we generated Y169F and Y188F mutants of IKK2. Radioactive in vitro kinase assays using recombinant purified IKK2^FL-Y188F and IKK2^FL-Y169F mutants revealed a drastic reduction in the activity of IKK2 (Figure 4C). Since the DFG+1 residue in kinases (Y169 in IKK2) is known to dictate its specificity, we created different mutants of the IκBα substrate where S32 and S36 were singly and independently altered to either alanine (phosphoablative) or glutamic acid (phosphomimetic) residues. Kinase activity of IKK2-Y169F was monitored in an in vitro radioactive kinase assay using WT as well as mutant IκBα. Mutation of S36 to either alanine or glutamate drastically reduced the phosphorylation signal of IκBα while that for the S32A/E mutants were minimally affected (Figure 4D). This suggests that the presence of Tyr (and possibly its phosphorylation) is critically important for signal-responsive phosphorylation of IκBα.

Next, we tested the consequence of Y169F and Y188F mutants on the pro-inflammatory signal-induced IKK activity in cells. We reconstituted IKK2 knock-out mouse embryonic fibroblast (ikk2^-/- MEF) cells with WT, Y169F and Y188F mutants of IKK2, and measured signal-induced activation of IKK2 upon TNF-α treatment. IKK2 activity is significantly reduced in Y169F and nearly abolished in Y188F (Figure 4E). Interestingly, phosphorylation of activation loop serines of the less active Y169F mutant was only marginally defective (Fig. 4F). This suggests that phosphorylation of the activation loop serines at 177 and 181 is necessary but not sufficient for IKK2 to become fully active. Together, these results suggest an unreported communication between the phosphorylation of Y169 of IKK2 and that of S32 of IκBα.

Structural analyses of IKK2 autophosphorylation

To get a more comprehensive picture, it is necessary to analyze at least three phosphorylated states of IKK2: unphosphorylated (UnP-IKK2), phosphorylated at S177 and S181 (p-IKK2), and phosphorylated at Y169, S177 and S181 (P-IKK2). We took advantage of computational approaches using molecular dynamic (MD) simulations and flexible molecular docking (see methods).^43–50 MD simulation was performed for 200ns for each of these three states of IKK2, unphosphorylated and phosphorylated. Phosphorylation at S177, S181 increased the stability of the kinase, which was further increased by the phosphorylation at Y169 as evidenced by a gradual decrease in the total energy of system (Figure 5A). A striking enhancement in stability of IKK2 upon this auto-phosphorylation is reflected in the change of Tm from ∼ 40 °C to ∼ 50 °C in differential scanning calorimetry (DSC) scans in the presence of ADP and ATP (Figure S4A). This behavior supports our observation in MD simulation, and tallies with increased solubility of IKK2 observed after auto-phosphorylation that was essential for its successful crystallization (Polley et al., 2013)

Figure 5:

Interestingly, RMSD values increased with more phosphorylation (Figure 5B) despite phosphorylation-induced stabilization, P-IKK2 showed the highest RMSD followed by p-IKK2 and UnP-IKK2, indicating increased internal motion in the kinase as a result of phosphorylation. It may be noted that, the changes in total energy and RMSD are consistent throughout the course of simulation though the magnitude is small. We found that, phosphorylation at S177, S181 and S177, S181, Y169 induced distinct structural alteration in the KD when compared with the unphosphorylated version (Figure 5C, models shown separately in Figure S4B). Careful analyses of the trajectory of RMSF values for these three states revealed that, different regions of the kinase domain had different RMSF values in different phosphorylated states (Figure S4C). We observed that, while the glycine-rich loop and the αC-helix in p-IKK2 were situated in a manner reminiscent of an active kinase (Figure S4D), those in UnP and P-IKK2 were placed very differently. In fact, helix αC was found to be distorted in P-IKK2, that is unlikely to support canonical phosphotransfer from ATP. Activation segment/loop (activation loop) in these three differently phosphorylated states adopted conformations distinct from each other. activation loop in the UnP model was displayed out and moved closer to the tip of the C-lobe, whereas activation loop in p-IKK2 and P-IKK2 moved inwards and adopted conformations closer to the N-lobe. The formation of the proper R- and C-spines in p-IKK2 confirmed its active form (Figure 5D). p-IKK2 KD also exhibited other canonical features of an active kinase, viz., a salt bridge between K44 and E61 (K72, E91 in PKA, respectively), ‘DFG in’ (DLG here) conformation. Next, we docked ATP onto the p-IKK2 structure (ΔG = -10.64 kcal/mol) using a flexible docking protocol (see methods). The ATP and its phosphate groups exhibited a pose (Figure S4D) very similar to that in the PKA structure (PDB ID: 1ATP), and the terminal phosphate was in close proximity to Y169-OH making autophosphorylation of Y169 highly plausible (Figure 5E). Further analyses of the ATP-bound p-IKK2 structure confirm that presence of ATP helped the R and C-spine form the desired continuum seen in the active form of a kinase (Figure 5F). Superposition of the ATP from the p-IKK2 on P-IKK2 indicated that the narrowing of ATP-binding pocket in P-IKK2 may lead to rejection of ATP emanating from severe clashes between the glycine-rich loop and ATP (Figure 5G). Furthermore, the R-spine in P-IKK2 was not continuous (Figure 5H), wherein M65 (L95 in PKA) of the four conserved residues were located at a position far away from the rest of three residues that constitute the R-spine, and from the glycine rich loop. In addition, dynamic cross-correlation matrix (DCCM) or contact map of each structure suggests that specific phosphorylation events render the kinase display distinct allosteric changes at locations far from the phosphorylation sites themselves (Figure S4E).

ATP binding to p-IKK2 leads to autophosphorylation at Y169 generating P-IKK2 in an ADP bound state. Interestingly, the superposition of ADP (from the docked complex of p-IKK2 and ADP, ΔG = -9.88 kCal/mol, Figure 5I) onto the P-IKK2 structure confirmed that while P-IKK2 was unable to accommodate ATP in a manner similar to that in other known experimentally derived kinase structures, it could comfortably accommodate ADP into the ATP-binding cleft (Figure 5J). It may be noted that, p-IKK2 has higher affinity for ATP than ADP as indicated by their respective binding energy values. We speculate that ADP-bound P-IKK2 is an intermediate step where presence of as many phosphate groups in the activation loop and ADP in such a close proximity persuades the kinase to adopt a conformation that can withstand this high density of negative charges albeit risking the integrity of an active fold, and making some of those de novo phosphate groups transient in nature to return back to the p-IKK2 state to readily bind to ATP that is abound in the reaction.

These structural observations suggest that it is possible for the residues within the flexible activation segment to undergo auto-phosphorylation and possibly render the kinase with a novel phospho-transfer activity distinct from the conventional transfer of γ-phosphate from ATP to the substrate.

Freshly auto-phosphorylated IKK2 phosphorylates IκBα even in absence of ATP

We observed phosphorylation of tyrosine residue(s) only upon fresh ATP-treatment, indicating that tyrosines of purified IKK2 loses its phosphate group during the course of purification (Eckhart et al., 1979). It may be noted that, we did not observe any density of phosphates on any of these residues in the human IKK2 crystal structure derived from ATP-treated protein sample (Polley et al., 2013). Of note, human IKK2 crystals grown with autophosphorylated IKK2 (PDB ID 4E3C) took about a month at 18-20°C before they were subjected to X-ray diffraction. This suggested an intriguing possibility that the phosphate linked to Tyr residue(s) may serve as a temporary phosphate-sink and could eventually be transferred to substrate IκBα. To test this possibility, we first allowed purified IKK2 to undergo auto-phosphorylation by incubating the excess kinase with limiting amounts of radiolabeled γ-³²P-ATP followed by two passes through desalting spin columns (40 kDa MW cut-off) to remove unreacted ATP. Nucleotide-free auto-phosphorylated IKK2 (P-IKK2) was incubated with IκBα either in absence of nucleotide or with newly added unlabeled nucleotides, and phosphorylation status of the substrate was subsequently monitored by autoradiography after resolving the reaction mixture on an SDS-PAGE (Figure 6A). Transfer of radiolabeled phosphate to IκBα was observed and the transfer was significantly enhanced in the presence of cold ADP (Figure 6B). Phosphorylation of IκBα did not significantly decrease in the presence of a high concentration of cold ATP. That addition of excess cold ATP did not reduce the extent of IκBα phosphorylation-signal on the autoradiograph strongly suggests the transfer of ³²P-phosphate from ³²P-IKK2 to IκBα (Figure 6B) by a relay mechanism. This transfer perhaps occurs through a phosphoenzyme intermediate with or without involving ADP.

Figure 6:

We undertook a different strategy to further confirm transfer of phosphate from the kinase (and not ATP) to the substrate. We auto-phosphorylated IKK2^FL-WT with excess cold ATP and removed the unreacted ATP from the reaction mixture using size-exclusion chromatography on a Superdex200 column (Figure 6C). P-IKK2 eluted at fractions corresponding to MW between 670 and 158 kDa, i.e., much larger than free ATP (Figure 6D). This P-IKK2 protein was incubated with FL IκBα either in presence or absence of ADP for different time periods. A similar reaction in presence of ATP instead of ADP was performed as a positive control. Phosphorylation of S32 and S36 (detected by the phospho-serine antibody specific to S32 and S36 of IκBα) was efficient in presence of 10 and 50 μM ADP but absent when ADP was absent in the reaction mixture. Phosphorylation of IκBα in the presence of 50 μM ATP was significantly more robust as anticipated (Figure 6E). The minor detection of IκBα phosphorylation in absence of ADP in the radioactive assay and lack of it in the cold immunoblotting assay (compare lanes marked with vertical arrows in Figures 6B and 6F) may indicate higher sensitivity of the radioactive assay. The need of ADP in phosphotransfer prompted us to consider whether the ADP was contaminated with a trace amount of ATP. ESI-MS analysis of 50 μM ADP did not detect any peak corresponding to ATP (Figure S5A). Dependence of IκBα phosphorylation on ADP could raise another possibility if microscopic reversibility generated ATP from ADP in the active site of the kinase. To this end, we employed TLC-based analysis to detect generation/presence of ATP in a similar setting as described in Fig. 5A-B which revealed that no detectable ATP was present in the system, i.e., no detectable ATP was retained in the protein after passing it through the spin column twice, nor did any γP³²-ATP was generated when incubated with cold ADP (Figure S5B).

Phosphotransfer from P-IKK2 is critical to fidelity and specificity of IκBα phosphorylation

We next tested if phosphotransfer is restricted only to the N-terminal signal responsive S32/S36 or to serines/threonines of the PEST domain as well. Radioactive kinase assay as described earlier was performed with full-length WT and mutant S32A, S36A (AA) IκBα as the phosphate recipient. Phosphate group(s) of P-IKK2 was not transferred to full-length IκBα AA which retains the serines and threonines within its PEST domain (Figure 7A), indicating phosphotransfer only to the N-terminal serines of IκBα that are S32 and S36. We also tested the effect of AMPPNP, an unhydrolyzable analogue of ATP, on the phosphotransfer, and found no inhibitory effect irrespective of the presence of ADP. This suggests adenine nucleotides poising the kinase for phosphotransfer (Figure 7A). A recent report proposed that IKK2 phosphorylates IκBα at S32 followed by S36 following a single binding event and that phosphorylation of S32 increases the phosphorylation rate of S36(Stephenson et al., 2022). Based on our results and other evidence, we propose that IKK2 phosphorylated at S177/S181 undergoes autophosphorylation on tyrosine residues in its activation loop area that serves as a phosphoenzyme intermediate to transfer phosphate group from IKK2 to S32 that enhances the rate of second site (S36) phosphorylation (Figure 7B). For the second site phosphorylation ATP serves as the phosphate donor.

Figure 7:

Discussion

Phosphorylation of substrates by protein kinases is regulated in various ways(Cullati et al., 2022; Sang et al., 2022). IKK2, a kinase central to inflammation, phosphorylates serine residues of its primary substrate IκBα with γ-phosphate of ATP. Here, we report multiple intriguing properties of IKK2 that provide specificity to its substrate phosphorylation. These include: 1) dual specificity that entails autophosphorylation of tyrosine residues of IKK2, in addition to known auto-phosphorylation at activation loop serines; 2) relay of phosphate to substrate IκBα from auto-phosphorylated IKK2 (P-IKK2) and 3) loss of specificity of the kinase due to disruption of a universally conserved salt bridge (mediated by K44).

Ser/Thr/Tyr family of eukaryotic protein kinases transfer γ-phosphate of ATP directly to substrate, in general to S, T and Y residues(Cohen, 2002; Fischer and Krebs, 1955; Hunter, 1991; Krebs and Fischer, 1955; Nolen et al., 2004). Histidine kinases (in prokaryotes and in some lower eukaryotes) however shuttles the γ-phosphate to Asp residue on a response regulator (RR) substrate through a phospho-His intermediate(Borkovich and Simon, 1990; Gao and Stock, 2009; Hunter, 2022; Kalagiri and Hunter, 2021; Laub et al., 2007; Robbins et al., 1993). Our present analyses of eukaryotic protein kinase IKK2 revealed its surprising property in which IKK2 phosphorylated at activation loop-serines (S177/S181) could be further auto-phosphorylated by ATP at multiple other residues yielding another form (P-IKK2) (Figure 7B). This marks IKK2 as an autocatalytic dual-specificity kinase rather than a prototype S/T kinase. This multisite phosphorylated form of IKK2 (P-IKK2) could transfer phosphate(s) to substrate IκBα in presence of ADP without requiring exogenous supply of fresh ATP, hinting at an unusual property of IKK2 acting possibly as a phosphate-sink. It is possible that phospho-tyrosines residues of IKK2 act as this sink. The Y169 located at the DFG+1 (DLG in IKK2) position is unique only to IKK1 and IKK2, and is absent even in close mammalian homologues such as IKKε and TBK1. Interestingly, neither IKKε nor TBK1 can phosphorylate IκBα at S32 and S36, in vitro or in vivo(Larabi et al., 2013; Tu et al., 2013). A recent report also indicated the role of DFG+1 residue in differentiating Ser versus Thr residues(Chen et al., 2014). These observations hinted to an involvement of Y169 in substrate selectivity. Indeed, our experimental data suggests a pivotal role of Y169 in phosphorylating S32 of IκBα both in vitro and ex vivo (Figure 4D & 4E). It is worthwhile to note that, a phosphoablative mutation at S32 in IκBα to Ile (S32I) is associated with ectodermal dysplasia and T-cell immunodeficiency in patients(Courtois et al., 2003; Mooster et al., 2015). More interestingly, the placement of Y169 in the kinase active site makes it amenable to a phosphorylation-dephosphorylation cycle, suggesting its possible role both in phosphotransfer and substrate specificity. Mutation of Y188, a conserved tyrosine, to non-phosphorylatable phenylalanine abolished the kinase activity of IKK2 in vitro as well as in MEFs upon TNF-treatment (Figure 4C & 4E). The placement of Y188 indicates that its mutation will be detrimental to the structural fold of kinase, thus it is unclear if or how Y188 plays a role in the phosphotransfer. Since phosphotransfer to S32, S36 of IκBα is observed with auto-phosphorylated IKK2 EE, transferred phosphates are not from S177 or S181. Kinase domain autophosphorylation is known to regulate function of other kinases as well, e.g., autophosphorylation at T220 greatly influenced both the activity and substrate specificity of CK1(Cullati et al., 2022).

The strength of input signal, level of IκBα degradation, and NF-κB activation correlate with the amplitude, duration, and kinetics of IKK activity(Behar and Hoffmann, 2013; Cheong et al., 2006). This digital (all-or-none) activation profile of NF-κB appears to be due to rapid activation and inactivation of IKK2. IKK targets multiple substrates in vivo; however, the rapid activation of NF-κB is regulated primarily by IκBα. How IKK specifically selects IκBα for phosphorylation is unclear, although NEMO is reported to direct the kinase activity of IKK2 specifically towards IκBα(Schröfelbauer et al., 2012). We observe that P-IKK2 could transfer phosphate(s) specifically to the S32/S36 of IκBα, and not to other sites (such as the PEST sequence of IκBα) (Figure 6B, 6E and 7A). We surmise if transiency of multisite phosphorylated IKK2 (P-IKK2) underlie this activation-inactivation events during signal response. It is possible that regulatory system of IKK2 is somewhat analogous to the two-component histidine kinase-effector systems in prokaryotes that show a threshold activation response(Bhate et al., 2015; Lamarche et al., 2008; Laub and Goulian, 2007). It is noteworthy that, transfer of a phosphate moiety directly from pHis is energetically much more probable than that from a pTyr, unless it is an enzymatic process which is likely the case for IKK2:IκBα system.

Tyrosines in activation loop of IKK2 are previously reported to undergo signal-induced phosphorylation in cells. These modifications were proposed to be mediated by a tyrosine kinase although there is no evidence of a specific tyrosine kinase(Meyer et al., 2013; Otero et al., 2008; Rieke et al., 2011). Several members of the Ser/Thr kinase family undergo tyrosine phosphorylation and these kinases employ diverse mechanisms for auto-phosphorylation(Bhattacharyya et al., 2006; Ge et al., 2002; Lochhead, 2009; Lochhead et al., 2006, 2005; Tigno-Aranjuez et al., 2010).

The phosphotransfer from P-IKK2 to IκBα monitored here is detection of a single turnover event in absence of the phosphate donor, ATP in the bulk (Figure 6B). Phosphorylation on IKK2 occurs at multiple sites, and not every phosphorylated residue in P-IKK2 expected to transfer its phosphate to IκBα. There is more than one type of phosphorylated residues in P-IKK2 distributed over the length of the kinase, arguably with differential stability wherein only a fraction of the total pool can be relayed. In addition, it is likely that the phosphorylated IKK2 in the bulk harbors a non-homogenous pool where differently phosphorylated species of the kinase are present, and are indistinguishable in our assay schemes. Another intriguing feature of phosphorelay in IKK2 is IKK2’s dependence on ADP (Figure 6B & 6E). It was also observed that, the generic kinase inhibitor AMPPNP (unhydrolyzable ATP analogue) failed to inhibit phosphor-transfer from IKK2 (Figure 7A) and IκBα phosphorylation (Figure S2C). Taken together, we surmise that presence of ADP or AMP-PNP help IKK2 attain a conformational state poised for phosphorylating S32/36 of IκBα even in absence of fresh exogenous ATP. Similar phenomenon was reported for IRE1, where ADP was found to be a more potent stimulator of its ribonuclease activity and that AMPPNP could also potentiate this activity in IRE1(Lee et al., 2008; Sidrauski and Walter, 1997).The list of Ser/Thr kinase family proteins that undergo tyrosine phosphorylation is growing that employ diverse mechanisms for auto-phosphorylation(Bhattacharyya et al., 2006; Ge et al., 2002; Lochhead, 2009; Lochhead et al., 2006, 2005; Tigno-Aranjuez et al., 2010). The most notable example of this dual specificity are DYRK and GSK3β. However, the role of Tyr-phosphorylation in DYRK/GSK3β is not related to that in IKK2, nor does it render DYRKs/ GSK3β capable of relaying phosphates to the substrates. To the best of our knowledge the only other reported example of an EPK, though atypical, transferring phosphate through a phosphoenzyme intermediate is MHCK of Dictyostelium discoideum(Ye et al., 2010).

Our study reveals tyrosine auto-phosphorylation in IKK2 and indicates that consequences of Ser and Tyr autophosphorylations in dual-specific IKK2 are distinct. Moreover, the crucial tyrosine residues are positioned in such a manner that some that are amenable to auto-phosphorylation are essential for substrate phosphorylation. Most surprisingly, the tyrosine-linked phosphate groups can be transferred by a relay mechanism to specific serines in IκBα, an event reminiscing phospho-relay in two-component systems prevalent in prokaryotes and some lower eukaryotes. Our study also reveals the role of K44 of IKK2 in auto-phosphorylation and thus in specific phosphorylation of S32, S36 of IκBα, although the mutant kinase retains its non-specific kinase activity. It may be noted that, while a monomeric version of IKK2 (devoid of NBD and large portion of the dimeric interface in the SDD) still retained its specificity for the N-terminal serines of IκBα(Hauenstein et al., 2014), while a shorter construct containing only the KD and ULD lacked this specificity that phosphorylated the PEST domain like CK2(Shaul et al., 2008). It requires further investigation to determine if a structural state of IKK2, perhaps through a different modification or through binding to another protein, allow it to phosphorylate substrates relatively non-specifically.

In all, it is intriguing how the gateway to the activation of NF-κB is regulated through unique phosphorylation events in IKK2 (Figure 7C). It appears that a multi-layered fail-safe mechanism is devised to tightly control NF-κB activation. Firstly, in absence of upstream signals IKK-complex is unable to phosphorylate Ser 32,36 of IκBα efficiently thereby keeping check on the aberrant and untimely activation of NF-κB. Upon encountering proinflammatory cues, cells need to activate NF-κB immediately and efficiently, i.e., IKK needs to be activated and phosphorylate Ser 32,36 of IκBα. NEMO warrants IKK-activation and ensures that the IKK-complex specifically chooses IκBα from a large pool of substrates of IKK2. IKK2 is designed in such a manner that it phosphorylates itself at activation loop-tyrosines when activated, such that phosphate group(s) can be relayed directly to Ser32,36 of IκBα with great fidelity, thus leaving little chance of a misconstrued signaling event thereby confirming NF-κB activation (Figure 7C). Our discovery might also present the beginning of a new aspect in eukaryotic cell-signaling by EPKs and provide a foundation for mechanistic studies of an intriguing phospho-transfer reaction that helps accomplish the desired specificity. To the best of our knowledge, it is the first report, apart from MHCK of Dictyostelium discoideum (an atypical EPK), to suggest that a metazoan EPK confirms its specificity through the formation of a phosphoenzyme intermediate.

Methods summary

Protein purification, auto-phosphorylation assay, immunoprecipitation coupled kinase assay, phosphotransfer assay and other experiments have been described in details in the supplementary document.

Acknowledgements

We thank Tony Hunter (Salk Institute) and Kaushik Biswas (BI) for commenting on the work. Authors also thank members of Polley lab (BI) and Ghosh Lab (UCSD) for their continuous support. This work was supported by an Intermediate Fellowship from the DBT Wellcome Trust India Alliance (IA/I/15/1/501852) & intramural funding from Bose Institute to SP, NIH grant to GG (AI163327) and American Cancer Society grant RSG-08-287-01-GMC to TH. Biochemistry research at SDSU is supported in part by the California Metabolic Research Foundation. PB was supported by the graduate research fellowship from Bose Institute. SC acknowledges CSIR-IICB for infrastructure support. AC acknowledges ICMR for post-doctoral Research Associateship [BMI/11(55)2022].

Author contribution

SP, GG and TB designed experiments; PB, SP, TB & GG performed experiments. Computational analyses were performed by AC & SC. SP, GG, TB, PB, TH, SC and AC analyzed the results. SP wrote the manuscript with the help of PB, TB and AC in preparing figures and images. GG, TB, PB and TH edited/modified the manuscript. All authors commented on the manuscript.

Ethics declarations

Competing interests: The authors declare no competing financial interests.