Molecular Basis of Tank-Binding Kinase 1 Activation by Transautophosphorylation

Molecular basis of Tank-binding kinase 1 activation by transautophosphorylation

Xiaolei Maa,1, Elizabeth Helgasonb,1, Qui T. Phungc, Clifford L. Quanb, Rekha S. Iyera, Michelle W. Leec, Krista K. Bowmana, Melissa A. Starovasnika, and Erin C. Dueberb,2

Departments of aStructural Biology, bEarly Discovery Biochemistry, and cProtein Chemistry, Genentech, South San Francisco, CA 94080

Edited by Tony Hunter, Salk Institute for Biological Studies, La Jolla, CA, and approved April 25, 2012 (received for review December 30, 2011)

Tank-binding kinase (TBK)1 plays a central role in innate immunity: it the C-terminal scaffolding/dimerization domain (SDD), a do- serves as an integrator of multiple signals induced by receptor- main arrangement that appears to be shared among the IKK mediated pathogen detection and as a modulator of IFN levels. family of kinases (3). Deletion or mutation of the ULD in TBK1 Efforts to better understand the biology of this key immunological or IKKε severely impairs kinase activation and substrate phos- factor have intensified recently as growing evidence implicates phorylation in cells (22, 23). Furthermore, the integrity of the aberrant TBK1 activity in a variety of autoimmune diseases and ULD in IKKβ is not only required for kinase activity (24) but was fi cancers. Nevertheless, key molecular details of TBK1 regulation and shown also to confer substrate speci city in conjunction with the β substrate selection remain unanswered. Here, structures of phos- adjacent SDD (25). Recent crystal structures of the IKK phorylated and unphosphorylated human TBK1 kinase and ubiq- homodimer demonstrate that the ULD and SDD form a joint, uitin-like domains, combined with biochemical studies, indicate three-way interface with the KD within each protomer of the a molecular mechanism of activation via transautophosphorylation. dimer, suggesting that the ULD and SDD serve to buttress the These TBK1 structures are consistent with the tripartite architecture KD as well as to contribute additional binding surfaces that observed recently for the related kinase IKKβ, but domain contribu- properly orient substrates (25). Here we report the crystal structure of the kinase and ubiq- tions toward target recognition appear to differ for the two uitin-like domains (KU) of TBK1 in complex with a potent small- enzymes. In particular, both TBK1 autoactivation and substrate spec- molecule inhibitor, BX795. This structure unexpectedly reveals ificity are likely driven by signal-dependent colocalization events.

an activation loop-swapped TBK1 conformation: S172 from one BIOCHEMISTRY protomer is located in close proximity to the active site of the kinase activation | crystallography neighboring protomer, providing a snapshot of a potential transautoactivation reaction intermediate. Biochemical analyses nvading bacteria and viruses possess distinct molecular sig- further demonstrate that the kinase domain alone is sufficient to Inatures that are recognized by conserved host receptors, thereby fully autoactivate and is capable of phosphorylating both mac- triggering the assembly of signaling complexes that activate the romolecular and peptide substrates. A high-resolution structure inhibitor of κB kinase (IKK) family of kinases (1–3). The canonical of monophosphorylated TBK1 KD reveals that S172 phosphor- IKKs (IKKα and IKKβ) in turn induce NF-κB–dependent gene ylation reorganizes the activation segment into a conventional transcription via phosphorylation of the inhibitory κBα (IκBα) configuration that is compatible with polypeptide substrate protein. This modification marks IκBα for K48-linked poly- binding. Taken together, these results offer insights into the ubiquitination and subsequent proteasomal degradation, resulting structural basis of TBK1 transautophosphorylation, highlight the in the release of free NF-κB to up-regulate expression of proin- structural transitions accompanying TBK1 activation, and sup- flammatory cytokines (4). By contrast, activated IKK-related port a model in which TBK1 recruitment to discrete signaling kinases [Tank-binding kinase (TBK)1 and IKKε] directly phos- complexes induces TBK1 activation through proximity. phorylate IFN regulatory factors 3 and 7 (IRF3 and IRF7). Phosphorylation promotes the dimerization and nuclear trans- Results location of these transcription factors that stimulate production of Domain Organization of TBK1 Is Consistent with the IKKβ Tripartite type I interferons (IFNs) (1, 3, 5). Recent studies have identified Architecture. To understand how the KD and ULD are organized an additional role for TBK1 in the xenophagic elimination of within TBK1, we crystallized a fragment of human TBK1 con- bacteria (6–9) and better-defined how cross-talk within the IKK taining both domains and a catalytic residue mutation (KU ; family regulates innate immune response (10). D135N residues Q2–E385) in the presence of an inhibitor, BX795 (26, Under pathological conditions, IKK-mediated pathways can 27). The structure was solved to 2.6-Å resolution by single- also be activated inappropriately by endogenous signals, contrib- fl wavelength anomalous dispersion using a selenomethionine- uting to in ammatory disorders and oncogenesis (11, 12). Whereas A canonical IKKs have long been recognized as bridges between substituted sample (Fig. S1 and Table S1). Both molecules in the chronic inflammation and cancer, IKK-related kinases more re- asymmetric unit demonstrate a tandem arrangement of the kinase cently have also been implicated in cell transformation and tumor and ubiquitin-like domains, which possess the typical bilobe progression (13). TBK1 has been of particular interest, given its identification both as an activator of the oncogenic AKT kinase (14–18) and as an essential factor in KRAS-driven cancers (19). Author contributions: X.M., E.H., Q.T.P., and E.C.D. designed research; X.M., E.H., Q.T.P., R.S.I., TBK1 activity is regulated by phosphorylation on S172 within M.W.L., K.K.B., and E.C.D. performed research; E.H., C.L.Q., R.S.I., and K.K.B. contributed new reagents/analytic tools; X.M., E.H., Q.T.P., M.W.L., M.A.S., and E.C.D. analyzed data; and X.M., the classical kinase activation loop. Serine-to-alanine substitution E.H., M.A.S., and E.C.D. wrote the paper. at this position abolishes TBK1 activity, whereas the phospho- Conflict of interest statement: All authors are employees of Genentech. mimetic mutation S172E partially restores activity to within ∼200- fold of the wild-type kinase (20). Genetic and pharmacological This article is a PNAS Direct Submission. inhibition studies have indicated that TBK1 can be activated by Freely available online through the PNAS open access option. IKKβ, as well as by apparent autophosphorylation (10). Addi- Data deposition: Atomic coordinates and structure factors reported in this paper have tional posttranslational modifications of TBK1 lysine residues by been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 4EUT and 4EUU). K63-linked polyubiquitin chains have been shown to promote 1X.M. and E.H. contributed equally to this work. production of IFNs in viral infections (21). 2To whom correspondence should be addressed. E-mail: [email protected]. TBK1 contains a predicted ubiquitin-like domain (ULD) (22) This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. that is located between the N-terminal kinase domain (KD) and 1073/pnas.1121552109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1121552109 PNAS Early Edition | 1of6 Downloaded by guest on September 30, 2021 kinase and ubiquitin folds, respectively (Fig. 1A). In this config- occupied by the αAL-helix of the neighboring protomer in the uration, the ULD abuts the C-terminal lobe (C lobe) of the KD. TBK1 structure (Fig. S4B). Thus, although the unphosphorylated The conventional hydrophobic surface patch of the ULD, which KUD135N structure is endowed with many active kinase traits, the often serves as a protein–protein interaction surface in the case of domain-swapped form is not compatible with substrate binding ubiquitin, faces away from the kinase domain. and, therefore, would not act on a polypeptide substrate. However, A very similar domain organization is observed for IKKβ (25). S172, located in the αAL of the exchanged TBK1 activation loop, is Superposition of the TBK1 KUD135N fragment onto the full- only 6 Å away from the P0 site of the PKA substrate and less than length IKKβ structure (via alignment of the kinase domains) 11 Å from the catalytic D135 residue. Elevated temperature factors appears to poise both TBK1 domains for interaction with the in the αAL-helix underscore the dynamic nature of the TBK1 ac- SDD (Fig. 1B). Residues identified in the KD–SDD and ULD– tivation loop (Fig. S4C) and suggest that a local reorganization of SDD interfaces of the IKKβ structure show moderate conserva- this region, which would allow phosphotransfer to S172, may be tion across IKK-family members and map to comparable surfaces plausible. B of the TBK1 KUD135N fragment structure (Fig. 1 and Fig. S2); The extended interface created by this KUD135N loop swapping however, the orientation of the ULD relative to the KD differs is comparable in size and composition to bona fide protein–pro- slightly in the two kinase structures (Fig. 1A, Inset). In the full- tein complexes, burying 2,570 Å2 of surface area per molecule. β ∼ β length IKK structure, an 30° rotation in the IKK ULD relative However, despite this extensive interaction surface, KUD135N β ’ to that of TBK1 allows the IKK ULD s hydrophobic patch to was found to be monomeric in solution (Fig. 2B). KUD135N, and A B bind to the adjacent SDD (Fig. S3 and ). Similarly, an insertion a shorter construct containing just the kinase domain (KDD135N; into the KD fold specific to IKK-family members (residues 241– Q2–R308), both behaved as monomers by size-exclusion chro- 265 in TBK1) also appears to rearrange in the presence of the matography (SEC). By contrast, full-length TBK1 (FLD135N;Q2– SDD (Fig. S3 C and D). Despite these minor differences, the L729) containing the C-terminal SDD eluted as a dimer. Multi- TBK1 KUD135N fragment structure appears to be compatible with angle light scattering (MALS) confirmed the dimeric or mono- the tripartite domain arrangement observed for full-length IKKβ. meric status of each sample. Thus, the activation loop-swapped interactions in the KUD135N structure do not appear to mediate TBK1 KUD135N Structure Adopts an Activation Loop-Swapped formation of stable dimers, but may instead constitute more Conformation. Although the KD of the TBK1 KUD135N structure transient self-associations. generally assumes the typical kinase fold, the activation segment adopts a unique conformation in this unphosphorylated crystal TBK1 Can Catalyze Autophosphorylation in Trans. To probe whether form. Residues encompassing the entire activation loop (L164– the placement of the activation loop within the substrate-binding G199) surprisingly extend away from the kinase core of each cleft of the KUD135N protomers observed in the structure could KUD135N molecule in the asymmetric unit and associate intimately also occur in solution, we tested whether active TBK1 could with the C lobe of the neighboring KUD135N molecule (Fig. 2A). catalyze S172 phosphorylation. To this end, we expressed and Whereas residues 191–192 and 192–198 are disordered in mole- purified three distinct TBK1 constructs that each contained the cules A and B, respectively, the electron density over the ex- wild-type aspartic acid residue at position 135: FLWT,KUWT, changed region is very clear from L164–R191 in both protomers, and KDWT. Each of these TBK1 constructs used the same do- supporting the domain-swap interpretation (Fig. S1B). main boundaries as their equivalent D135N variant (FLD135N, The fully formed α-helix at the activation-loop C terminus KUD135N,KDD135N) and showed similar SEC elution profiles as (αEF) docks onto a surface cleft between two α-helices from the their “kinase-dead” counterparts (Fig. S5 A and B). opposite protomer via the HPD tripeptide motif (canonically the FLWT,KUWT, and KDWT stocks treated with ATP before SDS/ “APE” motif). The placement of this motif on the C lobe, albeit in PAGE separation displayed a marked increase in S172 phos- trans, mirrors intramolecular interactions typically observed in phorylation by anti-pS172 immunoblotting compared with stocks A active kinase structures. Moreover, the KUD135N structure dem- incubated without ATP (Fig. 3 ). Quantitation of these samples onstrates several other features that are characteristic of active by LC-MS revealed that ∼6% of the purified FLWT stock con- kinases (Fig. S4A). tained pS172, whereas ∼98% was phosphorylated after a 20-min At the activation-segment N terminus, a helix forms between incubation with ATP (Fig. 3B). By comparison, purified KUWT D167 and L173 (αAL). Superposition of the activation loop-ex- and KDWT stocks were ∼32% S172-phosphorylated and also changed KUD13N structure with the protein kinase A (PKA)– achieved near-complete modification after the same ATP treat- peptide inhibitor complex structure (28) (based on alignment of ment (Fig. 3B, Inset). the kinase domains) reveals that the P+1, +2, and +3 sites inferred To determine whether TBK1 autophosphorylation could oc- from the PKA complex (where P0 is the site of modification) are cur in trans, active TBK1 constructs were added to the purified

Fig. 1. TBK1 KUD135N structure is compatible with IKKβ domain organization. (A) A cartoon of TBK1 KUD135N.An alignment of the TBK1 and IKKβ kinase domains (Inset) shows the difference in relative ULD orientation. (B)

Substituting the TBK1 KUD135N structure for the equivalent domains of the IKKβ protomer structure (Left) suggests a comparable juxtaposition of these domains

to the SDD for the two proteins. KUD135N placement is based on the alignment shown in A (Inset). Residues in IKKβ’s KD and ULD that contact the SDD are highlighted in purple in a surface representation of the IKKβ KD and ULD (Upper Right). Equivalent residues are mapped

(purple) on the surface of the TBK1 KUD135N structure (Lower Right).

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1121552109 Ma et al. Downloaded by guest on September 30, 2021 Fig. 2. Activation-loop swapping

and solution behavior of KUD135N. AB (A)ThetwoKUD135N molecules in the asymmetric unit, one shown in cyan surface and the other in orange cartoon, adopt a “loop- swapped” conformation. In the foreground, S172 of the orange

KUD135N lies near the active site of the cyan KUD135N (D135N, yellow; BX795, sticks; disordered residues 192–198, orange dots). The re- verse interaction is observed in the background. (B)SEC-MALSanaly-

sisofFLD135N (black; molecular massmonomer =84kDa),KUD135N (blue; molecular massmonomer =44kDa),andKDD135N (red; molecular massmonomer = 35 kDa). Normalized A280 traces of the TBK1 variants are plotted (dotted lines) as a function of elution time from an S200 column coupled to a MALS detector. Calculated molecular masses are plotted (solid lines) for each eluted peak, with the average apparent molecular masses listed.

D135N mutants and pS172 status was monitored by immuno- KDD135N, again bound to BX795, to a resolution of 1.8 Å using blotting (Fig. 3C). The kinase-dead D135N variants, which molecular replacement (Figs. S1C and S7 A–C and Table S1). showed negligible activity against a commercial peptide substrate The unphosphorylated and phosphorylated TBK1 KD structures, (Fig. S5C), were unable to mediate S172 phosphorylation, excluding the activation loops, can be superimposed with an rms whereas strong pS172 signal was observed upon the addition deviation of 0.58 Å (for 270 Cα atoms). By contrast, the activa- of preactivated FLWT,KUWT,orKDWT. While transphos- tion loop undergoes a dramatic conformational change upon phorylation with prephosphorylated wild-type TBK1 constructs phosphorylation (Fig. 4 A and B). In the monophosphorylated occurred rapidly, equivalent reactions using fully dephosphory- KDD135N structure, the well-ordered pS172 docks onto its own

lated KUWT showed a substantial lag phase in the kinetics of kinase domain, shifting 18 Å from its position in the loop-swap- BIOCHEMISTRY KDD135N (and KUWT) S172 modiﬁcation (Fig. 3D). Similar ped KUD135N structure to form a network of intramolecular reactions using a KU variant that cannot be activated by phos- contacts (Fig. S7D). phorylation (KUS172A; Fig. S5C) were even further retarded, The substantial refolding of the activation loop is hinged at with transphosphorylation of KUD135N only observable at longer residues L164 and G199, with remarkably few phosphorylation- time points (Fig. 3D). Further comparisons of prephosphory- induced conformational changes in the TBK1 KD outside this A B lated and dephosphorylated KUWT activities using a commercial region (Fig. 4 and ). All of the other structural features in- peptide substrate recapitulated the lag-phase behavior observed dicative of an active conformation present in unphosphorylated for transphosphorylation by pS172 immunoblotting; however, TBK1 KUD135N are preserved in the pS172-KDD135N structure; the lag was shortened with increasing concentrations of however, the HPD motif is now docked onto the KD in an dephosphorylated kinase (Fig. S6). intramolecular interface. More importantly, the backbone of the P+1 binding pocket now adopts a similar contour as PKA in the Structure of the S172-Phosphorylated TBK1 KD. To understand the PKA–peptide inhibitor complex, and thus appears competent to structural changes that accompany TBK1 activation, we phos- bind polypeptide substrate. Closer inspection of the site suggests phorylated TBK1 KDD135N in vitro using activated FLWT and that TBK1 would strongly favor a hydrophobic residue at the P+1 determined the crystal structure of the monophosphorylated position (Fig. S7 E and F).

Fig. 3. TBK1 autoactivation via S172 phosphorylation. FLWT,KUWT, and KDWT were analyzed by (A) anti-pS172 immunoblotting and (B) LC-MS before and after incubation with ATP. Extracted ion chromatograms for S172 and pS172 peptides are shown for TBK1 FLWT, with quantitation of all samples summarized (mean ± SD error bars, n =3)(Inset). (C) Transautoactivation of TBK1 was probed using combinations of wild type and kinase-dead (D135N) variants and anti-

pS172 Western blots. (D) A time course of KDD135N phosphorylation by preactivated (pre-phos) or dephosphorylated (de-phos) KUWT was evaluated by pS172 immunoblotting. The KUS172A-mediated transphosphorylation of KUD135N is observable at longer time points.

Ma et al. PNAS Early Edition | 3of6 Downloaded by guest on September 30, 2021 K ABwithin experimental error of the m of FLWT. The truncated variants also displayed only minor differences in Vmax compared with the full-length kinase. In agreement with this detailed kinetic evaluation, FLWT,KUWT, and KDWT also displayed comparable activities toward a commercial peptide substrate over a wide range of kinase concentrations (Fig. S9). Importantly, all comparisons of TBK1 constructs with IRF3, AKT, and peptide substrates were performed using TBK1 stocks that were preactivated with ATP, so as to remove TBK1 autophosphorylation events from the analyses. Discussion Although both the sheer number and complexity of TBK1-mediated signaling pathways appear to increase with continued research (21, 23), a clear understanding of the molecular determinants of TBK1 activation and substrate specificity has not been achieved. To Fig. 4. Conformation of the TBK1 activation loop changes upon S172 address this issue, we sought to evaluate how the discrete domains fi modi cation. Comparison of the (A) unphosphorylated KUD135N and (B) within TBK1 contribute to the global architecture and enzymatic monophosphorylated KDD135N kinase domain structures reveals a dramatic fi – activity of the kinase using a panel of puri ed, recombinant TBK1 reorganization of the activation loop (residues 160 180; orange) upon S172 constructs. phosphorylation (spheres). The HPD motif switches from an intermolecular In our studies, all wild-type TBK1 constructs were found to to an intramolecular docking state in the presence of pS172, whereas the in trans position of the DFG motif remains unchanged. autoactivate in the presence of ATP and could act to phosphorylate kinase-dead D135N variants. Further analysis of TBK1 autoactivation revealed a substantial lag phase in TBK1 ac- Determinants of TBK1 Catalytic Activity and Substrate Recognition tivation kinetics when the TBK1 starting material was dephos- Appear to Be Confined to the Kinase Domain. To assess domain phorylated; however, this slow phase could be shortened by contributions to TBK1 kinase activity, the ability of each active increasing the concentration of dephosphorylated TBK1, or abol- construct to phosphorylate putative peptide substrates was eval- ished by prephosphorylating the wild-type TBK1 variant in the reaction. Finally, transphosphorylation reactions catalyzed by an uated. Here, activated FLWT,KUWT, and KDWT stocks were each incubated with a panel of biotinylated peptide substrates, in- S172A variant, which itself cannot be activated, demonstrated very cluding a peptide encompassing the S172 activation site, in the slow TBK1D135N phosphorylation kinetics. Taken together, these data indicate a biphasic reaction pathway for TBK1 autophos- presence of 32P-labeled ATP. Reaction mixtures were blotted phorylation consisting of an initial phase with slow kinetics followed onto a streptavidin membrane, and peptide phosphorylation was by a second phase characterized by rapid kinetics. We propose that visualized by autoradiography (Fig. 5). The activation-loop pep- the activation loop-swapped conformation observed in the KU tide displayed robust phosphorylation signal that exceeded that D135N fi structure likely constitutes a meaningful self-activation intermediate of peptides containing established IRF3 or AKT modi cation for the slow, initial phase of TBK1 autophosphorylation. sites (15, 29). In accordance with previous studies using this Indeed, activation-segment swapping has also been observed methodology, the TBK1 constructs were also active against in crystal structures from other kinase families (e.g., DAP3K, a peptide sequence originating from the Dead-box helicase SLK, LOK, and CHK2), where transient dimerization facilitates DDX3X (30); however, they did not mediate phosphorylation of transautophosphorylation despite the fact that the activation the negative control peptide, GSK3-like. Conversely, the AKT segments constitute non–consensus-site sequences for these fi control showed no discernable modi cation of the TBK1-, IRF3-, kinases (32–34). In each loop-swapped kinase structure, the or AKT-derived peptides, but did phosphorylate its proven αEF-helix from the neighboring molecule is anchored to the C GSK3-like peptide substrate (31). Finally, all TBK1 constructs lobe via the APE motif (HPD in TBK1) at the C terminus of the showed comparable activities across a given substrate, and no activation loop, and the residue targeted for phosphorylation is phosphorylation was detected in the absence of kinase. TBK1- in close proximity to the catalytic aspartate residue. Analytical catalyzed phosphorylation of recombinant IRF3 and AKT protein ultracentrifugation and cross-linking studies have confirmed the substrates, as evaluated by phospho-specific immunoblotting and presence of minor dimer populations for the majority of these LC-MS, gave similar results (Fig. S8). kinases; however, the transient nature of these interactions is Additionally, full enzymatic profiles were generated for each of reflected in the predominantly monomeric behavior of these the active TBK1 constructs using IRF3 as the substrate (Fig. 6). proteins in solution (32–34). In the case of TBK1, no such as- Fits of the data to the Michaelis–Menten equation yielded the sociation was observed in solution even at 100 μM kinase. As enzymatic parameters Vmax and Km for each variant. Loss of the prolonged dimerization via the activation loops would exclude SDD alone, or in combination with the ULD, essentially had no binding and phosphorylation of other substrates, it is perhaps not effect on IRF3 binding, as Km values for KUWT and KDWT were surprising that these self-interactions are relatively short-lived.

1 2 3 4 5 6 7 8 9 10 peptide # sequence source ] ] ] ] ] ] ] ] ] ]

] 172 ] 1 DDEQFVS LYGTEE TBK1 TBK1 + +

2 RVGGASS386LENTVD IRF3

TBK1 FLWT + -

3 DLHIS396NS398HPLSLT IRF3 ] ] 4 HPLS402LT404S405DQYKA IRF3

TBK1 KUWT + +

AKT - + 5 KDEVAHT195LTENRV AKT ] 6 DGATMKT308FCGTPE AKT TBK1 KDWT 7 LGPEAKS378LLSGLL AKT

+ 1.5 nM Kinase (n=3) 8 PHFPQFS473YSASST AKT ] no Kinase control AKT - 9 RKQYPIS269LVLAP DDX3X 10 KRPRAASFAE GSK3-like

Fig. 5. Each TBK1 construct phosphorylates peptide substrates. Biotinylated peptides were incubated with FLWT,KUWT, and KDWT constructs in the presence of [γ-32P]ATP and blotted onto a streptavidin membrane. Phosphorylation was imaged by autoradiography. Purchased, activated AKT, and an AKT substrate peptide (GSK3-like) were included as controls (key). Established sites of phosphorylation are colored red in the peptide sequences.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1121552109 Ma et al. Downloaded by guest on September 30, 2021 1 and KDWT were active kinases. Whereas the shorter constructs were less stable than FL TBK1 (Fig. S5D), the TBK1 variants all 0.8 showed comparable activities in the context of both macromo- 0.6 lecular and peptide substrates, indicating that the SDD and ULD domains do not strongly influence the catalysis or substrate spec- 0.4 ificity of purified TBK1 in vitro. Instead, these features appear to FLWT be maintained wholly within the KD of TBK1, a result that is in 0.2 KUWT opposition to previous studies with IKK-family kinases (22–25). initial velocity (a.u./min) KDWT α β ε fl 0 For IKK , IKK , and IKK , this may simply re ect differential 0 20 40 60 80 100 120 140 160 domain functionalities that evolved in addition to their apparent [IRF3] (μM) common structural roles. In the case of TBK1, one possible ex- fi μ planation for this disparity is our nding that production of well- FLWT: Vmax = 0.8 ± 0.1 a.u./min; Km = 71 ± 19 M behaved KDWT and KUWT fragments required coexpression of μ KUWT: Vmax = 1.0 ± 0.1 a.u./min; Km = 53 ± 7 M a phosphatase enzyme, seemingly to limit autophosphorylation and activity during overexpression and purification. Vmax = 1.0 ± 0.1 a.u./min; Km = 55 ± 9 μM KDWT: Although the TBK1 SDD did not impact substrate specificity, as was observed for IKKβ, it did maintain a similar structural role Fig. 6. TBK1 constructs show comparable kinase activities. A detailed kinetic analysis of preactivated FL ,KU , and KD activities was conducted in mediating dimerization. Indeed, our results suggest that TBK1 WT WT WT likely adopts the same overall tripartite architecture observed for using the macromolecular substrate IRF3. Initial rates of IRF3 phosphoryla- β tion were averaged and plotted (mean ± SD error bars; n = 3) as a function protomers in the full-length IKK dimer, although the exact fold

of IRF3 concentration. Enzymatic parameters Vmax and Km were determined of the TBK1 SDD and the molecular details of the protomer and by fitting the data to the Michaelis–Menten equation, with errors reported dimer assembly surfaces likely differ given the sequence di- for these values derived from the fits. vergence in the SDD domains of these two kinases. Certainly, differential sequence conservation of SDD–SDD dimerization residues (as defined by the IKKβ structure) in the canonical Furthermore, once phosphorylated, the TBK1 activation loop IKKs versus the IKK-related kinases appears to segregate the completely reorganizes, with pS172 making a number of intra- two subfamilies, consistent with established homodimer and molecular contacts with the kinase domain. Such highly coor- heterodimer pairings within the family (23, 36).

dinated interactions likely reduce the mobility of the loop com- Placement of the observed TBK1 KUD135N activation-loop BIOCHEMISTRY pared with its unmodified state and can only be achieved by interactions in the context of a full-length, IKKβ-like dimer sup- a phosphorylated residue, perhaps explaining why the TBK1 ports a model of TBK1 transautoactivation in which phosphory- S172E phosphomimetic mutant shows decreased activity com- lation across two separate dimers is possible but phosphorylation pared with the wild-type enzyme (20). The conformational change of protomers within a dimer pair is prohibited. As demonstrated nucleated by the phosphorylation of S172 further propagates to by the IKKβ structure, protomer dimerization would place the the rest of the activation loop, resulting in intramolecular docking TBK1 kinase active sites on opposite faces of the molecule, of the HPD motif onto the kinase C lobe and completion of the making it unlikely for activation to occur within the dimer (Fig. P+1 pocket, thus providing the full complement of interactions 7A). Instead, we hypothesize that transautoactivation occurs required to facilitate phosphotransfer to polypeptide substrates when multiple TBK1 dimers are recruited to signaling complexes, (23, 35). Accordingly, pS172-activated TBK1 can quickly catalyze thus enabling activation-loop swapping of locally clustered TBK1 the modification of other TBK1 molecules, likely constituting the molecules that results in S172 phosphorylation (Fig. 7B) (23). In second rapid phase observed in TBK1 autoactivation reactions. In this manner, dimeric TBK1 is maintained in an inactive state until this regard, TBK1 differs from the other loop-swapped kinases recruited to a signaling platform where it can undergo autoacti- discussed as it readily phosphorylates its activation-segment se- vation due to high local concentration or activation via a different quence as a classical substrate, not just when presented as an effector kinase that has been localized to the same molecular activation loop-swapped substrate. scaffold. Consistent with this model of TBK1 regulation, KDWT Although previous domain deletion/truncation studies in- and KUWT constructs accrued much higher levels of S172 phos- dicated that TBK1 kinase activity was dependent on the presence phorylation during expression/purification than FLWT TBK1, of both the ULD and SDD (22, 23), we found that purified KUWT despite the fact that these shorter constructs were coexpressed

A B “off” TBK1 activation loops modeled onto IKKβ dimer KD

ULD 90° activation loop SDD phosphorylation signaling (scaffolding) active PO4 kinase

trans- PO4 autoactivation IKKβ protomer 1 IKKβ protomer 2 TBK1 activation loop more transautoactivation

Fig. 7. Proposed mechanism of TBK1 transautoactivation. (A) The extended conformation of the TBK1 KUD135N activation loop is shown superimposed onto two views of the IKKβ dimer to illustrate the separation of the kinase active sites in the dimer. Assuming that TBK1 adopts a similar dimeric form via the SDD, we propose (B) a model of TBK1 autoactivation in which a tripartite arrangement of domains within each protomer orients the kinase active sites away from one another. Signaling events lead to scaffolding of TBK1 and local clustering of TBK1 molecules that promotes interdimer interactions and phosphorylation of S172 in the activation loop. Additional transautoactivation events lead to fully activated TBK1 dimers.

Ma et al. PNAS Early Edition | 5of6 Downloaded by guest on September 30, 2021 with a phosphatase enzyme to limit their phosphorylation and complexes. In this way, different stimuli can exact specific displayed overall lower expression levels than FLWT. In a similar downstream effects even though they each use TBK1 activity to vein, recent studies found that overexpression of scaffolding-de- mediate signaling. Recent systematic proteomic analyses of TBK1 fective TBK1 mutants led to TBK1 activation, whereas no acti- interactions are consistent with this model of TBK1 regulation vation was observed when these mutants were expressed at (21, 23). Although these studies revealed a wide array of TBK1- endogenous levels. In the latter expression regime, activation was binding partners, corroborating the growing body of literature only observed for scaffolding-competent variants when an up- that links TBK1 to a variety of discrete cellular pathways, TBK1 stream stimulus was applied to the cells (23). Together, these was found to form complexes with distinct scaffolding partners fi ndings support a model in which the tripartite arrangement of that were mutually exclusive and appeared to localize TBK1 to the KD, ULD, and SDD, observed for IKKβ and inferred for “ ” different subcellular compartments (23). Given the ever-growing TBK1, lock these dimeric kinases in an off state until they are number of TBK1 interaction partners and substrates, localized appropriately scaffolded and phosphorylated in response to spe- fi fi kinase activation and substrate modi cation could be an effective ci c stimuli. In fact, similar strategies of autoinhibition and lo- strategy to limit TBK1 activity to a specific pathway. Insights calization of activity have been observed for other kinases, such as gained from the BX795-bound TBK1 structures presented here PKA, albeit via disparate molecular mechanisms (37, 38). may potentially aid the design of more selective TBK1 inhibitors, Structural analysis of the activated pS172-KDD135N suggests that TBK1 would likely prefer substrates in which a hydrophobic which would serve as important tools for further dissecting the side chain immediately follows the residue being modified, as intricacies of TBK1-mediated signaling events. was previously reported (30). Accordingly, many of the sub- Methods strates phosphorylated by TBK1 in this study housed hydrophobic residues at the P+1 position, with leucine being the most TBK1 variants were expressed in insect cells and purified to homogeneity. prevalent. However, some peptide substrates containing leucine Dephosphorylated stocks were treated with phosphatase during purification in the P+1 site were not phosphorylated by TBK1, whereas other whereas prephosphorylated stocks were incubated with ATP just prior to use. seemingly nonideal side chains, including serine and even Concentrated TBK1 stocks were incubated with BX795 inhibitor and crys- aspartic acid, appeared to be well-tolerated at this position. The tallized by hanging drop vapor diffusion method. Detailed methods are weak TBK1 consensus sequence combined with the wide variety provided in SI Methods. of substrates phosphorylated by TBK1 instead suggests a model ’ of substrate selection that depends more on colocalization ACKNOWLEDGMENTS. We thank Karen O Rourke and Maia Vinogradova fi for technical assistance; Andrea Cochran, Jacob Corn, and Sarah Hymowitz strategies than local sequence speci city of target sites. for many helpful discussions; and the staff at the Advanced Light Source and Thus, we propose that both TBK1 activation and substrate Stanford Synchrotron Radiation Lightsource for their assistance with sample specificity are likely driven by recruitment to discrete signaling handling and data collection.

1. Kawai T, Akira S (2010) The role of pattern-recognition receptors in innate immunity: 20. Kishore N, et al. (2002) IKK-i and TBK-1 are enzymatically distinct from the homolo- Update on Toll-like receptors. Nat Immunol 11:373–384. gous enzyme IKK-2: Comparative analysis of recombinant human IKK-i, TBK-1, and 2. Kawasaki T, Kawai T, Akira S (2011) Recognition of nucleic acids by pattern-recog- IKK-2. J Biol Chem 277:13840–13847. nition receptors and its relevance in autoimmunity. Immunol Rev 243(1):61–73. 21. Li S, Wang L, Berman M, Kong Y-Y, Dorf ME (2011) Mapping a dynamic innate im- 3. Häcker H, Karin M (2006) Regulation and function of IKK and IKK-related kinases. Sci munity protein interaction network regulating type I interferon production. Immu- – STKE 2006(357):re13. nity 35:426 440. 4. Hayden MS, Ghosh S (2008) Shared principles in NF-κB signaling. Cell 132:344–362. 22. Ikeda F, et al. (2007) Involvement of the ubiquitin-like domain of TBK1/IKK-i kinases in – 5. Chau TL, et al. (2008) Are the IKKs and IKK-related kinases TBK1 and IKK-ε similarly regulation of IFN-inducible genes. EMBO J 26:3451 3462. activated? Trends Biochem Sci 33(4):171–180. 23. Goncalves A, et al. (2011) Functional dissection of the TBK1 molecular network. PLoS 6. Thurston TLM, Ryzhakov G, Bloor S, von Muhlinen N, Randow F (2009) The TBK1 One 6:e23971. 24. May MJ, Larsen SE, Shim JH, Madge LA, Ghosh S (2004) A novel ubiquitin-like domain adaptor and autophagy receptor NDP52 restricts the proliferation of ubiquitin-coated in IκB kinase β is required for functional activity of the kinase. J Biol Chem 279: bacteria. Nat Immunol 10:1215–1221. 45528–45539. 7. Gleason CE, Ordureau A, Gourlay R, Arthur JSC, Cohen P (2011) Polyubiquitin binding 25. Xu G, et al. (2011) Crystal structure of inhibitor of κB kinase β. Nature 472:325–330. to optineurin is required for optimal activation of TANK-binding kinase 1 and pro- 26. Feldman RI, et al. (2005) Novel small molecule inhibitors of 3-phosphoinositide-de- β – duction of interferon . J Biol Chem 286:35663 35674. pendent kinase-1. J Biol Chem 280:19867–19874. 8. Wild P, et al. (2011) Phosphorylation of the autophagy receptor optineurin restricts 27. Bain J, et al. (2007) The selectivity of protein kinase inhibitors: A further update. Bi- Salmonella growth. Science 333:228–233. ochem J 408:297–315. 9. Randow F (2011) How cells deploy ubiquitin and autophagy to defend their cytosol 28. Zheng J, et al. (1993) 2.2 Å reﬁned crystal structure of the catalytic subunit of cAMP- from bacterial invasion. Autophagy 7:304–309. dependent protein kinase complexed with MnATP and a peptide inhibitor. Acta 10. Clark K, et al. (2011) Novel cross-talk within the IKK family controls innate immunity. Crystallogr D Biol Crystallogr 49:362–365. Biochem J 434(1):93–104. 29. Panne D, McWhirter SM, Maniatis T, Harrison SC (2007) Interferon regulatory factor 3 11. Hammaker D, Boyle DL, Firestein GS (2012) Synoviocyte innate immune responses: is regulated by a dual phosphorylation-dependent switch. J Biol Chem 282: TANK-binding kinase-1 as a potential therapeutic target in rheumatoid arthritis. 22816–22822. Rheumatology (Oxford) 51:610–618. 30. Soulat D, et al. (2008) The DEAD-box helicase DDX3X is a critical component of the 12. Clément J-F, Meloche S, Servant MJ (2008) The IKK-related kinases: From innate im- TANK-binding kinase 1-dependent innate immune response. EMBO J 27:2135–2146. munity to oncogenesis. Cell Res 18:889–899. 31. Alessi DR, Caudwell FB, Andjelkovic M, Hemmings BA, Cohen P (1996) Molecular basis ﬁ 13. Shen RR, Hahn WC (2011) Emerging roles for the non-canonical IKKs in cancer. On- for the substrate speci city of protein kinase B; comparison with MAPKAP kinase-1 – cogene 30:631–641. and p70 S6 kinase. FEBS Lett 399:333 338. 14. Wang D, et al. (2011) The leader proteinase of foot-and-mouth disease virus nega- 32. Pike ACW, et al. (2008) Activation segment dimerization: A mechanism for kinase autophosphorylation of non-consensus sites. EMBO J 27:704–714. tively regulates the type I interferon pathway by acting as a viral deubiquitinase. J 33. Oliver AW, et al. (2006) Trans-activation of the DNA-damage signalling protein kinase Virol 85:3758–3766. Chk2 by T-loop exchange. EMBO J 25:3179–3190. 15. Ou Y-H, et al. (2011) TBK1 directly engages Akt/PKB survival signaling to support 34. Oliver AW, Knapp S, Pearl LH (2007) Activation segment exchange: A common – oncogenic transformation. Mol Cell 41:458 470. mechanism of kinase autophosphorylation? Trends Biochem Sci 32:351–356. 16. Ng M-HJ, et al. (2011) MIP-T3 is a negative regulator of innate type I IFN response. J 35. Lee S-J, Cobb MH, Goldsmith EJ (2009) Crystal structure of domain-swapped STE20 – Immunol 187:6473 6482. OSR1 kinase domain. Protein Sci 18:304–313. 17. Joung SM, et al. (2011) Akt contributes to activation of the TRIF-dependent signaling 36. Scheidereit C (2006) IκB kinase complexes: Gateways to NF-κB activation and tran- pathways of TLRs by interacting with TANK-binding kinase 1. J Immunol 186:499–507. scription. Oncogene 25:6685–6705. 18. Chen H, et al. (2011) Activation of STAT6 by STING is critical for antiviral innate im- 37. Kim C, Cheng CY, Saldanha SA, Taylor SS (2007) PKA-I holoenzyme structure reveals munity. Cell 147:436–446. a mechanism for cAMP-dependent activation. Cell 130:1032–1043. 19. Barbie DA, et al. (2009) Systematic RNA interference reveals that oncogenic KRAS- 38. Wong W, Scott JD (2004) AKAP signalling complexes: Focal points in space and time. driven cancers require TBK1. Nature 462(7269):108–112. Nat Rev Mol Cell Biol 5:959–970.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1121552109 Ma et al. Downloaded by guest on September 30, 2021