Introduction

Protein kinases confer novel identity and, often, novel functionality to their substrates by adding phosphate moieties to specific sites, thus playing key regulatory roles in a diverse array of signaling systems. One such example is the action of Inhibitor of κB Kinase-complex (IKK-complex), which is composed of two catalytic subunits, IKK1 (also known as IKKα) and IKK2 (or IKKβ), and an essential regulatory scaffolding protein, NEMO (IKKγ). IKK-complex precisely triggers induction of NF-κB family transcription factors in metazoans in response to inflammatory or pathogenic signals (Hoffmann and Baltimore, 2006)(DiDonato et al., 1997; Rothwarf et al., 1998; Zandi et al., 1997). In resting cells, activity of IKK2 is maintained at a low basal level (Ghosh and Karin, 2002; Hacker and Karin, 2006; Hinz and Scheidereit, 2014; Liu et al., 2012). In response to inflammatory or pathogenic signaling cues, the IKK2 subunit becomes activated via phosphorylation at two serine residues within its activation loop and subsequently phosphorylates two critical serine residues (S32 and S36) near the N-terminus of IκBα (Inhibitor of NF-κB α), marking the inhibitor protein for ubiquitin-dependent 26 S proteasome-mediated degradation (Figure 1A). Signal-responsive phosphorylation of IκBα refers to phosphorylation of its S32 and S36 residues by the IKK-activity. Degradation of IκBα liberates NF-κB, which then translocates to the nucleus to execute its gene expression program (Hayden and Ghosh, 2008; Hinz and Scheidereit, 2014; Karin and Ben-Neriah, 2000; Scheidereit, 2006). Mutation of IκBα residues S32 and S36 to phospho-ablative alanine converts it into a super-repressor of NF-κB that is incapable of being degraded in response to NF-κB signaling (Brown et al., 1995; Lin et al., 1995), highlighting the need for exquisite specificity of IKK2 in phosphorylating these two serines (DiDonato et al., 1997). The IKK2 subunit fails to phosphorylate IκBα when S32 and S36 are mutated even to threonine, another phosphorylatable residue. Despite its critical role in the regulation of NF-κB via high fidelity phosphorylation specificity toward IκBα, IKK2 behaves with functional pleiotropy toward other substrates and in other contexts (Antonia et al., 2021; Schröfelbauer et al., 2012; Schrofelbauer and Hoffmann, 2011).

Figure 1:

IKK2 is a multidomain protein consisting of a kinase domain (KD) in association with a ubiquitin-like domain (ULD) followed in sequence by a scaffold dimerization domain (SDD) and a C-terminal region that contains the NEMO-binding domain (NBD) (Figure 1B). As mentioned previously, the hitherto enigmatic activation of IKK2 from its inactive state is manifested in the phosphorylation status of two serine residues, S177 and S181 (S176 and S180 for IKK1), located within in the activation loop (AL) of the KD (Huse and Kuriyan, 2002; Liu et al., 2013; Polley et al., 2013; Xu et al., 2011). Phospho-mimetic substitution of these two serines to glutamates (S177E, S181E; henceforth EE) renders IKK2 constitutively active. It has been demonstrated that the upstream kinase TAK1 primes IKK2 through phosphorylation at S177, leading to its autophosphorylation at S181 and full activation in cells (Zhang et al., 2014). Alternatively, it has also been shown that increased oligomerization upon association with NEMO and linear or Lys63-linked poly-ubiquitin chains, or due to a high concentration of IKK2, enables trans autophosphorylation of IKK2 at S177 and S181 (Chen, 2012; Du et al., 2022; Ea et al., 2006; Polley et al., 2013). NEMO appears not only to aid in activation of the IKK-complex, but also to direct IKK activity towards IκBα bound to NF-κB (NF-κB:IκBα complex) (Schröfelbauer et al., 2012). The SDD of IKK2 and the C-terminal segment (residues 410-756, collectively) are also required for specific phosphorylation of IκBα at S32 and S36. A shortened version of IKK2 with deletion of these regions failed to retain the above-mentioned exquisite specificity of IKK2 but phosphorylated IκBα at serine and threonine residues within its C-terminal PEST (residues 281-302) region (Shaul et al., 2008). Interestingly, this C-terminal PEST region of IκBα, which is rich in proline, glutamic acid, serine, and threonine residues, is phosphorylated in vivo by other kinases that are not capable of phosphorylating IκBα at S32 and S36 (Barroga et al., 1995; Tergaonkar et al., 2003). Despite its central role in the regulation of transcription factor NF-κB, the connection or interdependence of IκBα phosphorylation at the PEST region and the signal responsive S32/S36 sites remains elusive.

In this study, we employed structure-based biochemical and cell-based experimental approaches to explore the basis of IKK2 specificity towards S32/S36 of IκBα that is required for rapid induction of NF-κB transcriptional activity in response to canonical signaling. Biochemical analysis of catalytically competent IKK2 revealed that it is capable of autophosphorylation at tyrosine residues in addition to autophosphorylation of its own serines or phosphorylation of serines in its substrate IκBα. Therefore, IKK2 is a dual-specificity protein kinase. IKK2 tyrosine autophosphorylation depends upon prior phosphorylation of activation loop S177/S181 and plays a significant role in phosphorylation of signal responsive S32/S36 of IκBα. The phosphorylation of IκBα at S32 was severely compromised upon substitution of Y169 of IKK2 to a phospho-ablative phenylalanine. Additionally, signal-responsive phosphorylation of IκBα upon TNF-α treatment in MEF cells reconstituted with IKK2-Y169F was severely diminished in comparison to wild type IKK2 (IKK2-WT). We also observed that an IKK2 bearing the K44M mutation resulted in a loss of S177 and S181 as well as tyrosine autophosphorylation activities and consequent loss of S32/S36 phosphorylation in IκBα, confirming a critical function of this lysine residue. Interestingly, phosphorylation of residues within the C-terminal PEST region of IκBα was retained in IKK2-K44M. Finally, we observed an intriguing activity of IKK2, when fully autophosphorylated at its serine and tyrosine residues, to phosphorylate IκBα in the absence of an exogenous supply of ATP, if ADP is present in the reaction. This result suggests the possibility of a unique, and likely transient, autophosphorylated form of IKK2 that can serve to relay phosphoryl group(s) specifically to S32/S36 of IκBα. Such a mechanism contrasts with the conventional transfer of γ-phosphate groups directly from ATP to the substrate observed in eukaryotic protein kinases.

Results

IKK2 undergoes autophosphorylation at uncharacterized sites

Signal-induced phosphorylation of activation loop residues Ser177 and Ser181 is the hallmark signature of IKK2 catalytic activity (Figure 1B). In contrast, hyperphosphorylation of other serine residues located within the flexible C-terminal region of IKK2, spanning amino acids 701 to 756 (Figure 1B), has been reported to down-regulate IKK2 activity in cells (Delhase et al., 1999). Ectopically overexpressed IKK2 has been observed to autophosphorylate the two serines of the activation loop in trans even in the absence of an activating physiological signal {Polley, 2013 #27}. We observed that purified recombinant full length IKK2 is autophosphorylated when incubated with ATP in the absence of NEMO in a time-dependent manner, and likely at multiple sites as indicated by the diffuse nature of the slower migrating phospho-IKK2 band (Figure 1C). The autophosphorylated IKK2 produced a sharper band in the presence of NEMO – possibly reflecting enhanced specificity of phosphorylation (Figure 1D, left panel) – although NEMO did not alter the efficiency of phosphorylation for its bona fide substrate IκBα (Figure 1D, right panel). An in vitro kinase assay with a K44M mutant of IKK2 that lacks both its autophosphorylation and IκBα phosphorylation activities along with the native sequence IKK2 using γ-³²P-ATP confirmed that phosphorylation of IKK2 was self-catalyzed and not due to a spurious contaminating kinase (Figure 1 - figure supplement 1A). K44 is an ATP-contacting residue that is conserved across STYKs, including IKK2, and its mutation to methionine typically reduces or eliminates kinase activity. An in vitro kinase assay showing inhibition of specific phosphorylation of S32/36 on IκBα and IKK2-autophosphorylation performed in the presence of an IKK-specific ATP-competitive inhibitor, Calbiochem Inhibitor VII, further validates that phosphorylation of IKK2 is due to its own activity (Figure 1E, Figure 1 - figure supplement 1B). In these experiments, the reaction mixtures of kinase and kinase:substrate were incubated with the inhibitor for 30 minutes prior to addition of the phosphate donor, γ-³²P-ATP or cold ATP. Purity of recombinant IKK2 and IKK2 K44M proteins is shown in Figure 1 - figure supplement 1C, and an LC-MS/MS analysis of the IKK2 K44M is shown in Figure 1 - figure supplement 1D.

Next, we performed kinase assays with variants of IKK2 protein with activation loop S177 and S181 mutated to phospho-mimetic glutamate resulting in the constitutively active IKK2-EE mutant form (Zandi et al., 1998) and with the C-terminal serine-rich region truncated to various extents. The shortest construct, Δ664EE, which lacked the entire flexible C-terminal region (670-756), was subjected to limited proteolysis using trypsin to further eliminate flexible N- and C-terminal residues and subsequently purified for in vitro kinase assays. These deletion IKK2-EE constructs displayed both autophosphorylation and substrate IκBα phosphorylation activities (Figure 1F). The observed autophosphorylation of IKK2 lacking serines of the activation loop and the C-terminal segment suggested the existence of hitherto uncharacterized sites of autophosphorylation in IKK2.

Autophosphorylation of IKK2 reveals its dual specificity

We were intrigued as to whether the observed autophosphorylation activity of IKK2 might involve other residues in addition to the previously characterized serines, i.e., the possibility of IKK2 being a dual specificity kinase. Indeed, we observed by western blot with anti-phosphotyrosine antibodies the unexpected de novo autophosphorylation of IKK2 at tyrosine residues (Figure 2A, middle panel) in addition to the previously reported autophosphorylation of activation loop serines (Figure 2A, upper panel) when full length native sequence IKK2 was incubated with Mg²⁺-ATP for various amounts of time. Increased autophosphorylation of activation loop S177 and S181 over time reflects nonhomogeneous and incomplete phosphorylation of activation loop S177 and S181 in the IKK2 obtained from recombinant baculovirus-infected Sf9 insect cells. To address the possibility of non-specificity by the pan-phosphotyrosine antibody (used in Figure 2A), we further performed the experiment with an alternate anti-phosphotyrosine specific antibody that confirmed detection of de novo tyrosine phosphorylation was independent of the source of the antibody used (Figure 2B). ATP-dependent tyrosine autophosphorylation was not observed in the presence of an IKK-specific inhibitor, Inhibitor VII (Figure 2C), and with the kinase inactivating IKK2 K44M mutant (Figure 2D). Additionally, autophosphorylation assays performed in the presence of urea revealed tyrosine autophosphorylation to be undetectable at urea concentrations greater than 1M (Figure 2 – figure supplement 1A). More sensitive assays using γ-³²P-ATP recapitulated elimination of pan-autophosphorylation and IκBα substrate phosphorylation of IKK2 above 1M urea (Figure 2 – figure supplement 1B). These results demonstrate the dual specificity of recombinant IKK2 in phosphorylating serine residues on itself and on substrates, as well as tyrosine residues on itself. The fact that activities of IKK2 towards S32 and S36 of IκBα and its own tyrosine residue(s) were affected similarly by both general (AMPPNP and Staurosporine) and IKK-specific (MLN-120B, TPCA, Inhibitor VII) (Figure 2 – figure supplement 1C) kinase inhibitors, indicates that both activities occur in the same active site. Using various truncated versions of IKK2, we further observed that the C-terminal NBD of IKK2 is not required for its autocatalytic dual specificity, and in phosphorylating IκBα at S32 and S36 (Figure 2E) in vitro. Semi-quantitative assessment of phospho-Ser and phospho-Tyr residues on a constitutively active NBD-deficient IKK2 (IKK2Δ664EE) was performed by western blot with monoclonal anti-phosphoserine and anti-phosphotyrosine antibodies, which revealed de novo autophosphorylation at tyrosine residues but not at serines (Figure 2F). The phospho-ablative S177A/S181A (AA; non-phosphorylatable activation loop) mutant of IKK2, that is incapable of phosphorylating S32 and S36 of IκBα (Figure 2 – figure supplement 1D), also lacked tyrosine autophosphorylation activity (Figure 2G) in an in vitro kinase assay suggesting phosphorylation of activation loop S177/S181 to be critical for dual specificity of IKK2.

Figure 2:

Dual-specific autophosphorylation is critical for function of IKK2

Previous reports suggested that phosphorylation of IKK2 on Y169, Y188, and/or Y199 is critical for its function (Darwech et al., 2010; Meyer et al., 2013). We performed mass spectrometric analyses on in vitro autophosphorylated IKK2 and identified Y169 to be the prominent tyrosine autophosphorylation site (Figure 3 – figure supplement 1A). To explore the possible relevance of phosphorylation at tyrosine residues within the activation loop to kinase catalytic activity, we analyzed the experimentally determined x-ray crystal structures of IKK2 monomers within reported models of human IKK2 dimers (PDB ID: 4KIK, 4E3C) (Liu et al., 2013; Polley et al., 2013). In the active state conformers, the position and conformation of Tyr169 appears to be well positioned for accepting a phosphate from ATP intramolecularly (Figure 3 – figure supplement 1B). 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 3 – figure supplement 1B and 1C). Interestingly, the position occupied by Y169 in IKK2 primary sequence corresponds to the classic DFG+1 (DLG in cases of IKK2 and IKK1) (Figure 3A & B). The DFG+1 position has been reported to be critical in defining the substrate specificity of a kinase (Chen et al., 2014).

Figure 3:

We primarily focused our attention to the previously described K44M mutant and the newly generated Y169F mutant of IKK2, since these residues are located within the vicinity of the enzyme active site. We observed that the IKK2 K44M mutant fails to undergo autophosphorylation of tyrosine (Figure 2D) and the activation loop serines (Figure 3 – figure supplement 1D). Surprisingly, IKK2 K44M displayed a very weak (compared to that of WT IKK2) autophosphorylation activity but a more robust phosphorylation of full length IκBα, albeit to a much lesser degree than native sequence IKK2, in a radioactive in vitro kinase assay (Figure 3C). Interestingly, we observed a lack of phosphorylation of IκBα at S32 and S36 by IKK2 K44M (Figure 3D). This suggests that IKK2 K44M phosphorylates IκBα at residues within its C-terminal PEST, consistent with the observed phosphorylation of IκBα (Figure 3C & D). Moreover, IKK2 K44M phosphorylated super-repressor IκBα S32A/S36A more efficiently than native sequence IκBα, in sharp contrast to IKK2 of native sequence (Figure 3 – figure supplement 1E). This suggests that IKK2 K44M retains significant kinase activity towards IκBα, though it is incapable of specific IκBα phosphorylation at S32,S36. LC-MS/MS analysis with 20 μg of Sf9-derived IKK2 K44M protein used in our studies did not indicate the obvious presence of any contaminating kinase. It is noteworthy that an equivalent K to M mutant of Erk2 retained ∼5 % of its catalytic activity (Robbins et al., 1993).

Next, we assessed the activity of a tyrosine-phosphoablative Y169F mutant of IKK2 in an in vitro kinase assay and observed a severe reduction in both autophosphorylation and IκBα phosphorylation activities (Figure 3E). We measured its phosphorylation activity towards native sequence IκBα as well as IκBα with either S32 or S36 residues independently substituted to either phosphoablative alanine or phosphomimetic glutamic acid. IKK2 Y169F displayed drastically reduced levels of kinase activity towards S36A/E mutants of IκBα while the S32A/E mutants were minimally affected when compared against the native IκBα substrate (Figure 3F). The reduction in phosphorylation of S36A/E by IKK2 was less pronounced than that by IKK2 Y169F (Figure 3 – figure supplement 1F). These observations suggest that the tyrosine residue at position 169 (and possibly its phosphorylation) might contribute towards signal-responsive phosphorylation of IκBα.

We also measured signal-induced activation of IKK2 upon TNF-α treatment in ikk2^-/-mouse embryonic fibroblast (MEF) cells reconstituted with native sequence IKK2 and IKK2 Y169F. Catalytic activity is significantly reduced in IKK2 Y169F (Figure 3G). Nonetheless, phosphorylation of activation loop serines was only marginally defective for IKK2 Y169F as judged by anti-phosphoserine-177 western blot (Figure 3 – figure supplement 1G) hinting at the necessity, but not sufficiency, of activation loop phosphorylation for IKK2 to become fully active. Together, these results imply a possible interdependence of phosphorylation at Y169 of IKK2 and S32 of IκBα.

Structural analyses of IKK2 autophosphorylation

We next analyzed three states of IKK2: unphosphorylated (UnP-IKK2), phosphorylated at S177 and S181 (p-IKK2), and phosphorylated at Y169, S177, and S181 (P-IKK2) using computational approaches of molecular dynamic (MD) simulations of 200 ns scale and flexible molecular docking (see methods)^43–50. Phosphorylation at S177 and S181 increased folding stability of the kinase, which was further stabilized by phosphorylation at Y169 as evidenced by a gradual decrease in the total energy of the system (Figure 4A). Results of differential scanning calorimetry (DSC) in the presence of ADP vs. ATP also indicated a change of Tm from ∼40°C to ∼50°C i.e., reflecting a striking enhancement in stability upon autophosphorylation of IKK2 (Figure 4 – figure supplement 1A). These results correlate with observations of increased solubility of IKK2 upon autophosphorylation; a property found to be essential for its successful crystallization (Polley et al., 2013).

Figure 4:

Counterintuitively, we observed the highest RMSD for P-IKK2, followed by p-IKK2, and then UnP-IKK2 (Figure 4B). This result might be indicative of an increased internal motion despite phosphorylation-induced stabilization of IKK2 because of phosphorylation. The changes in total energy and RMSD values were consistent throughout the course of simulation even though the magnitude was small. We observed distinct structural alteration within the KD upon phosphorylation at S177 and S181 or S177,S181, and Y169 (Figure 4C, models shown separately in Figure 4 – figure supplement 1B). Analyses of the trajectory of RMSF values for the three states revealed different regions of the kinase domain with different RMSF values in different phosphorylated states (Figure 4 – figure supplement 1C). We observed that, while the glycine-rich loop and the αC-helix in p-IKK2 were positioned in a manner reminiscent of an active kinase, those in UnP and P-IKK2 were oriented differently (Figure 4C). In fact, helix αC was found to be distorted in P-IKK2 in a manner that would make it unlikely to support canonical phosphotransfer from ATP. The activation loop in these three states adopted conformations distinct from each other. The activation loop in the UnP model was splayed out and moved closer to the tip of the C-lobe, whereas the activation loop in p-IKK2 and P-IKK2 moved inwards and adopted more compact conformations closer to the N-lobe (Figure 4C). The formation of the proper R- and C-spines in p-IKK2 confirmed its active conformation (Figure 4D). The p-IKK2 KD exhibited additional features consistent with an active kinase: a salt bridge between K44 and E61 (K72 and E91 in PKA, respectively), and the ‘DFG in’ conformation (though the sequence is DLG in IKK2). In addition, a dynamic cross-correlation matrix (DCCM) or contact map of each structure suggests that specific phosphorylation events render distinct allosteric changes to the kinase, even at locations distant from the phosphorylation sites (Figure 4 – figure supplement 1D).

Next, we docked ATP onto IKK2 structures in all three states using LeDock and GOLD followed by rescoring with AutoDock Vina using the 0 ns (starting, S) and 200 ns (ending, E) structures of our MD simulation studies (see methods). The results indicated that ATP binding to P-IKK2 is relatively unfavorable (docking score in positive range) in both starting and ending structural models (Figure 4 – figure supplement 1E), whereas ATP binding to UnP-IKK2 and p-IKK2 is favorable. ADP binding is however favorable for all phosphorylated or unphosphorylated IKK2 structural models except for the starting conformation of P-IKK2 (Figure 4 – figure supplement 1E). We also extracted 50 intermediate structures from a 10 ns MD simulation run and calculated their binding free energies (ΔG) using the MM-PBSA (Molecular Mechanics Poisson-Boltzmann Surface Area) method. We again observed that ATP-binding for the P-IKK2 population is highly unfavourable in contrast to their high relative preference for ADP, whereas p-IKK2 displayed comparable preference for both ATP and ADP (Figure 4 – figure supplement 1F). In the ATP-docked structures, the ATP phosphate groups exhibited a pose (ΔG = -10.64 kcal/mol) very similar to that observed in the PKA structure (PDB ID: 1ATP) (Figure 4 – figure supplement 1G), and the terminal phosphate was in close proximity to the Y169-OH, consistent with the likelihood of autophosphorylation at Y169 (Figure 4E). Further analyses of the ATP-bound p-IKK2 structure confirmed that the presence of ATP enhanced formation of the R- and C-spines, as observed in active forms of kinases (Figure 4F). Superposition of the ATP from the p-IKK2 on P-IKK2 indicated that narrowing of canonical the ATP-binding pocket in P-IKK2 may lead to rejection of ATP from the canonical binding pocket, triggered by severe clashes between the glycine-rich loop and ATP (Figure 4G). However, this does not exclude the possibility that P-IKK2 might interact with ATP through alternative binding modes. Furthermore, the R-spine in P-IKK2 exhibited a discontinuity (Figure 4H), with M65 (L95 in PKA) located away from the other three conserved spine residues, and from the glycine-rich loop.

Binding of ATP to p-IKK2 could lead to autophosphorylation at Y169, thus generating P-IKK2 bound to ADP. The superposition of ADP (from the docked complex of p-IKK2 and ADP, ΔG = -9.88 kCal/mol, Figure 4I) onto the P-IKK2 structure confirmed that while P-IKK2 is unable to accommodate ATP in a manner similar to that observed in experimentally determined kinase structures, it could accommodate ADP into the ATP-binding cleft comfortably (Figure 4J). The affinities of ATP and ADP for p-IKK2 appears comparable as indicated by their respective binding energy/docking score values, like in other kinases (Becher et al, 2012). Thus, the cellular concentrations of ATP vs ADP could play an important role in kinase activity. We speculate that ADP-bound P-IKK2, containing various and perhaps transiently phosphorylated groups within the activation loop and in close proximity to the active site, could act as an intermediate to relay phosphates back to ATP or to substrates.

These observations of IKK2 structural propensities hint at the possibility of residues within the flexible activation loop undergoing context-sensitive autophosphorylation and possibly rendering IKK2 with novel phospho-transfer activities distinct from the conventional direct transfer of γ-phosphate from ATP to substrate.

Freshly autophosphorylated IKK2 phosphorylates IκBα even in absence of ATP

We observed phosphorylation of tyrosine residue(s) only upon ATP-treatment of purified IKK2 proteins, hinting at the likelihood of efficient cellular phosphotyrosine phosphatases (Eckhart et al., 1979). Surprisingly, crystal structures of Sf9-derived human IKK2 did not indicate phosphates on tyrosine residues even though recombinant IKK2 was treated with Mg²⁺-ATP as part of the purification (PDB ID: 4E3C) (Polley et al., 2013). It is unclear if the month-long period of crystallization at 18-20°C contributed to loss of the generally very stable phosphotyrosine residues. This raises the intriguing possibility that phosphotyrosine residue(s) in IKK2 might somehow serve as a transient sink of phosphate for its eventual transfer to ADP or serines in the IκBα substrate. To investigate this hypothesis, we first incubated purified IKK2 with limiting amounts of radiolabeled γ-³²P-ATP for autophosphorylation and subsequently removed unreacted ATP via two passes through desalting spin columns (40 kDa MWCO). The nucleotide-free autophosphorylated IKK2 (P-IKK2) was then incubated with substrate IκBα either in the absence or with the fresh addition of unlabeled ATP/ADP. Resolution of the reaction mixture on an SDS-PAGE and subsequent autoradiography indicated transfer of radiolabeled phosphate to IκBα (Figure 5A), which was significantly enhanced in the presence of cold ADP (Figure 5B). Addition of excess cold ATP did not reduce the extent of IκBα phosphorylation, strongly suggesting the transfer of ³²P-phosphate from ³²P-IKK2 to IκBα (Figure 5B) and consistent with an IKK2-mediated phosphate relay mechanism. It is unclear if the transfer of phosphate from the phospho-IKK2 intermediate to IκBα occurs with or without direct involvement of ADP.

Figure 5:

To further investigate the possibility of transfer of phosphate from the kinase (and not from ATP) to the substrate, we produced autophosphorylated IKK2 obtained by treatment of recombinant full length IKK2 with excess cold ATP and purified using size-exclusion chromatography on a Superdex200 column to remove any traces of unreacted ATP (Figure 5C). P-IKK2 eluted at SEC fractions corresponding to MW between 670 and 158 kDa, much earlier than free ATP (Figure 5D). This P-IKK2 was incubated with IκBα substrate either in the presence or absence of ADP for varying time periods. A similar reaction was set up in the presence of ATP instead of ADP as a positive control. We observed, by western blot with sequence-specific anti-phosphoserine antibody, efficient phosphorylation of IκBα S32 and S36 residues in the presence of ADP (10 and 50 μM) but not in its absence. As anticipated, a robust phosphorylation of IκBα was observed in the presence of 50 μM ATP (Figure 5E). The very low-level detection of IκBα phosphorylation by radioactive assay in the absence of ADP and complete lack of signal in the cold immunoblotting assay (compare lanes marked with vertical arrows in Figures 5B and 5E) may indicate higher sensitivity of the radioactive assay. We quantified the intensities of each band in Figure 5E, and the ratios of the normalized average intensities of the tyrosine phosphorylated IKK2 and S32/36-phosphorylated IκBα with their respective full-length protein controls were plotted as shown in Figure 5 – figure supplement 1A. Next, we followed the same experimental scheme as shown in Figure 5C and assessed IκBα phosphorylation by relay mechanism at a finer time interval. Phosphorylation of S32/S36 of IκBα increased with time as shown in Figure 5 – figure supplement 1B. The requirement for ADP in the observed phosphotransfer prompted us to consider if our ADP was contaminated with trace amounts of ATP. ESI-MS analysis of our 50 μM ADP solution did not detect any peak corresponding to ATP (Figure 5 – figure supplement 1C). Dependence of IκBα phosphorylation on ADP raises the possibility of infinitesimal reversibility generating ATP from ADP in the active site of the kinase. We could not detect the generation/presence of ATP in a setting similar to that described in Figure 5A & B by TLC-based analysis (Figure 5 – figure supplement 1D).

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

We next tested whether the observed phosphotransfer to IκBα is restricted only to the S32 and S36 residues required for NF-κB activation or also directed toward serine/threonine residues within C-terminal PEST domain of IκBα. Kinase assays performed with P-IKK2 and IκBα of native sequence or harboring S32A and S36A mutations (AA) in the presence of γ-³²P-ATP revealed a lack of phosphorylation of IκBα AA, though that protein retains the serines and threonines of the C-terminal PEST region (Figure 6A). Furthermore, a similar assay with IKK2-Y169F showed its inability to transfer phosphate to IκBα (Figure 6 – figure supplement 1). Taken together, this hints at transfer of phosphate exclusively to S32 and S36 of IκBα from Y169. We also observed no inhibitory effect on the phosphotransfer by AMPPNP, an unhydrolyzable analogue of ATP, irrespective of the presence of ADP. This suggests that diverse adenine nucleotides prompt phosphotransfer by the kinase (Figure 6A). It is noteworthy that a recent study reported that IKK2 phosphorylates IκBα sequentially 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).

Figure 6:

Taken together with our results, we propose that catalytically active IKK2, that is by definition phosphorylated at S177 and S181, can undergo autophosphorylation on tyrosine residues within its activation loop. Furthermore, the tyrosine autophosphorylated IKK2 could serve as a phosphoenzyme intermediate to transfer its phosphate group to S32 of IκBα, thereby enhancing the rate of phosphorylation at S36, possibly using ATP for the second site phosphorylation at S36 (Figure 6B).

Discussion

The phosphorylation of cognate substrates by protein kinases is regulated in a variety of ways (Cullati et al., 2022; Sang et al., 2022). IKK2, a kinase central to inflammation, phosphorylates serine residues of its primary substrate IκBα using γ-phosphate of ATP. In this study, we report several intriguing structural and biochemical properties of IKK2 that contribute to its substrate phosphorylation specificity and efficiency: 1) dual-specificity of IKK2 that entails autophosphorylation of its own tyrosine residues, in addition to known autophosphorylation of activation loop serines; 2) loss of specificity, but retention of diminished catalytic activity, upon disruption of a universally conserved salt bridge mediated by K44 in IKK2; and 3) relay of phosphate to substrate IκBα from autophosphorylated IKK2 (P-IKK2).

Dual specificity of IKK2

Our present analyses of IKK2 revealed a surprising property wherein IKK2 phosphorylated at activation loop serines S177/S181 could be further autophosphorylated by ATP at multiple sites, including at least one of its tyrosine residues, yielding a hyper-phosphorylated form (P-IKK2) (Figure 6B). This defines IKK2 as an autocatalytic dual-specificity kinase rather than simply a prototypical Ser/Thr kinase. Several members of the Ser/Thr kinase family have been reported to undergo tyrosine phosphorylation and it appears that many of these kinases employ diverse strategies for phosphorylation (Bhattacharyya et al., 2006; Ge et al., 2002; Lochhead, 2009; Lochhead et al., 2006, 2005; Sugiyama et al., 2019; Tigno-Aranjuez et al., 2010). Tyrosine residues within the activation loop of IKK2 are also reported to undergo signal-induced phosphorylation in cells, and this is proposed to be mediated by a tyrosine kinase (Meyer et al., 2013; Otero et al., 2008; Rieke et al., 2011). However, dual specificity is notably manifested by DYRK and GSK3β, even though function of tyrosine phosphorylation in DYRK/GSK3β does not appear to be related to that of IKK2.

The phosphorelay mechanism

The Ser/Thr/Tyr family of eukaryotic protein kinases transfer γ-phosphate of ATP directly to substrates, usually at S, T, and Y residues (Cohen, 2002; Fischer and Krebs, 1955; Hunter, 1991; Krebs and Fischer, 1955; Nolen et al., 2004). In contrast, histidine kinases found in prokaryotes and in some simple eukaryotes employ a mechanism in which the γ-phosphate of ATP is transferred first to an aspartate residue on a response regulator (RR) substrate through an internal phosphohistidine intermediate (Borkovich and Simon, 1990; Gao and Stock, 2009; Hunter, 2022; Kalagiri and Hunter, 2021; Laub et al., 2007; Robbins et al., 1993). Surprisingly, the hyper-phosphorylated P-IKK2 appears to display an ability to transfer phosphate(s) to substrate IκBα in the presence of ADP without requiring an exogenous supply of fresh ATP – hinting at the possibility of IKK2 acting as a phosphate sink capable of relaying phosphate from the phophoenzyme intermediate to the substrate (Figure 5B & 5E). Since this phosphotransfer to S32 and S36 of IκBα is observed with autophosphorylated, constitutively active IKK2 EE, the transferred phosphates are unlikely to derive from phosphorylated S177 or S181. Phosphorylated Y169, on the other hand, could be a suitable candidate to act as an intermediate and a source of transferable phosphate due to its conducive location at the DFG+1 (DLG in IKK2) position. Y169 seems particularly fit for this role as it is the only active site-proximal residue prominently autophosphorylated within the activation loop of IKK2 (Figure S3A). Tyrosine at this position is unique only to IKK1 and IKK2 and the plant kinase SNRK1, and is absent even in close mammalian orthologs such as IKKε and TBK1 (Larabi et al., 2013; Tu et al., 2013). Interestingly, a study has implicated the role of DFG+1 residue in differentiating kinase specificity towards Ser versus Thr residues (Chen et al., 2014). These observations hint at the possibility of Y169 being involved in phosphorylation selectivity of IκBα substrate. We observed a role for Y169 in phosphorylating S32 of IκBα using both in vitro and ex vivo experiments (Figure 3F & G); however, our data does not confirm if Y169 phosphorylation is the sole contributor to specific signal-responsive phosphorylation of IκBα. It is noteworthy that a S32I mutant of IκBα containing the phosphoablative mutation at S32 is associated with ectodermal dysplasia and T-cell immunodeficiency (Courtois et al., 2003; Mooster et al., 2015). Autophosphorylation in the kinase domain has been observed to regulate function in other kinases as well, e.g., autophosphorylation at T220 greatly influenced both the activity and substrate specificity of CK1 (Cullati et al., 2022). So far, IKK2 appears to be the only other metazoan protein kinase that derives its substrate specificity through the formation of a phosphoenzyme intermediate apart from MHCK of Dictyostelium discoideum (an atypical eukaryotic protein kinase) (Ye et al., 2010). The transfer of phosphate from Tyr169 to Ser32 of IκBα represents an intriguing and unprecedented mechanism among metazoan protein kinases. We propose a possible explanation for the process: the hydroxyl group of Tyr169 is optimally positioned to readily accept the γ-phosphate from ATP at the expense of minimal energy in presence or absence of the substrate IκBα. This relay of phosphate – from ATP to Tyr169 to Ser32 – might offer a more energetically favorable pathway than a direct transfer of phosphate from ATP to the substrate. Another intriguing feature of phosphorelay in IKK2 is its dependence on ADP (Figure 5B & E), which the generic kinase inhibitor AMPPNP (unhydrolyzable ATP analogue) failed to inhibit (Figure 6A). Taken together, we surmise that ADP or AMPPNP might act as an ATP surrogate in helping IKK2 adopt a conformational state commensurate with selective phosphorylation of S32 and S36 in IκBα. A similar phenomenon is reported for IRE1, where ADP was found to be a potent stimulator of its ribonuclease activity and AMPPNP was also reported as a somewhat effective potentiator (Lee et al., 2008; Sidrauski and Walter, 1997). We speculate that Ser32 cannot be efficiently phosphorylated by p-IKK2:ATP due to a deliberate chemical mismatch of the hydroxyl group of Ser32 within the active site. However, it aligns well with the P-IKK2:ADP intermediate. Indirect supporting evidence, as described above, underscores the existence of a unique phosphotransfer mechanism; however, detailed structural, biochemical and MD simulations studies are necessary to validate this model.

Signaling specificity and digital activation

The amplitude, duration, and kinetics of IKK activity correlate with the strength of input signal, level of IκBα degradation, and NF-κB activation (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. The primary target of IKK2 is IκBα which underlies the rapid activation of NF-κB, although multiple other substrates are targeted in vivo. This digital activation profile may be intrinsically linked to the phosphorelay process. The selection of specific IκBα residues for phosphorylation is unclear, although NEMO is reported to have a direct role (Schröfelbauer et al., 2012). We observe that P-IKK2 transfers phosphate(s) efficiently and specifically to S32 and S36 of IκBα, but not to other phosphorylatable sites, such as the serine and threonine residues within the IκBα C-terminal PEST region (Figure 5B, 5E, and 6A). We surmise that the transiency of various phosphorylated residues in P-IKK2 could underlie the activation-inactivation event during signal response. In this context, it will be worth investigating if functional regulation of IKK2 involves properties analogous to that of the two-component histidine kinase-effector systems (Bhate et al., 2015; Lamarche et al., 2008; Laub and Goulian, 2007). It is noteworthy that phosphate transfer directly from phosphohistidine is energetically far more favorable than from phosphotyrosine, except in enzymatic processes. The phosphotransfer from P-IKK2 to IκBα investigated and reported herein reflects a single turnover event in the absence of the phosphate donor, ATP (Figure 5B). Not every phosphorylated residue over the long span of the P-IKK2 is expected to have the ability to transfer its phosphate to IκBα and additionally, the P-IKK2 likely represents a pool with non-homogenous distribution of phosphorylated residues. This could be a reason for observance of the substoichometric phospho-relay.

The K44 conserved salt bridge is a determinant of specificity but not activity

We also observe a critical role of K44 in autophosphorylation of IKK2 and specific phosphorylation of S32 and S36 in IκBα. Interestingly, the IKK2 K44M mutant retained its non-specific kinase activity toward the C-terminal PEST region of IκBα. Along these lines, an engineered monomeric version of IKK2 (lacking its NBD and major portions of its SDD) was observed to retain specificity for the N-terminal serines of IκBα (Hauenstein et al., 2014), whereas a smaller version of IKK2 with only the KD and ULD lacked S32 and S36 specificity but retained the ability to phosphorylate IκBα within its C-terminal PEST region (Shaul et al., 2008). It is possible that a particular structural state of IKK2 or its interaction with a partner protein could direct IKK2 to phosphorylate specific sites within different substrates.

Conclusion

It is intriguing how the activation of NF-κB, through unique phosphorylation events catalyzed by IKK2, appears to be regulated by a complex and multi-layered fail-safe mechanism. In absence of upstream signals, the IKK-complex is unable to phosphorylate Ser 32/36 of IκBα efficiently thereby keeping a check on the aberrant and untimely activation of NF-κB. Upon encountering proinflammatory cues, cells need to activate NF-κB immediately, efficiently and specifically, 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 an activation loop tyrosine 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. Our discovery of an intriguing phosphotransfer reaction that helps accomplish the desired specificity could present the beginning of a new aspect in eukaryotic cell signaling by EPKs.

Methods summary

Protein purification, autophosphorylation assay, immunoprecipitation-coupled kinase assay, phosphotransfer assay and other experiments are described in detail in the supplementary document.

Data availability

All computational data supporting the findings of this study are available in a repository on Zenodo (DOI: 10.5281/zenodo.15309725).

Supplementary figures

Supplementary Figure 1.

Supplementary Figure 2.

Supplementary Figure 3.

Supplementary Figure 4.

Supplementary Figure 5.

Supplementary Figure 6.

Acknowledgements

We thank Tony Hunter (Salk Institute) and Kaushik Biswas (Bose Institute) for commenting on the work, and members of Polley lab 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 and California Metabolic Research Foundation support to Biochemistry research at SDSU to TH. 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].

Additional information

Author contribution

TB, GG, and SP designed experiments; PB, TB, GG, and SP performed biochemical experiments; PB prepared MS samples, PRT and SRC performed the MS analyses, AC and SC performed computational analyses; PB, TB, AC, TH, SC, GG, and SP analyzed data; PB, AC, SP prepared figures and images; SP wrote the manuscript; PB, TB, TH, and GG edited the manuscript.

Additional files

Materials and methods