Traf2- and Nck-interacting kinase (TNIK) is a serine/threonine kinase highly expressed in the brain and enriched in the postsynaptic density of glutamatergic synapses in the mammalian brain. Accumulating genetic evidence and functional data have implicated TNIK as a risk factor for psychiatric disorders. However, the endogenous substrates of TNIK in neurons are unknown. Here, we describe a novel selective small molecule inhibitor of the TNIK kinase family. Using this inhibitor, we report the identification of endogenous neuronal TNIK substrates by immunoprecipitation with a phosphomotif antibody followed by mass spectrometry. Phosphorylation consensus sequences were defined by phosphopeptide sequence analysis. Among the identified substrates were members of the delta-catenin family including p120-catenin, δ-catenin, and armadillo repeat gene deleted in velo-cardio-facial syndrome (ARVCF), each of which is linked to psychiatric or neurologic disorders. Using p120-catenin as a representative substrate, we show TNIK-induced p120-catenin phosphorylation in cells requires intact kinase activity and phosphorylation of TNIK at T181 and T187 in the activation loop. Addition of the small molecule TNIK inhibitor or knocking down TNIK by two shRNAs reduced endogenous p120-catenin phosphorylation in cells. Together, using a TNIK inhibitor and phosphomotif antibody, we identify endogenous substrates of TNIK in neurons, define consensus sequences for TNIK, and suggest signaling pathways by which TNIK influences synaptic development and function linked to psychiatric and neurologic disorders.
TNIK is an abundant kinase highly expressed in the brain and concentrated at the postsynaptic density of glutamatergic synapses (Jordan et al., 2004; Peng et al., 2004; Collins et al., 2005; Burette et al., 2015). TNIK is highly enriched in synaptosome and postsynaptic density fractions where it regulates the stability of postsynaptic density proteins (Nonaka et al., 2008; Hussain et al., 2010; Wang et al., 2011). Among its various binding partners, TNIK interacts with disrupted in schizophrenia 1, a multifunctional scaffold protein whose altered function is linked with psychiatric disorders including schizophrenia, depression, and bipolar disorder (Blackwood et al., 2001; Camargo et al., 2007; Brandon and Sawa, 2011). Transcriptional profiling, genome-wide association studies, and functional genomics network analysis further support the role of TNIK as a psychiatric risk gene (Glatt et al., 2005; Matigian et al., 2007; Potkin et al., 2009; Shi et al., 2009; Ayalew et al., 2012).
TNIK is a serine/threonine (Ser/Thr) kinase in the Ste20 family of MAP kinase kinase kinase kinases (MAP4K) (Fu et al., 1999). Although the cellular function of TNIK remains unclear, one of the most reproducible effects of TNIK overexpression in mammalian cells is kinase-dependent induction of cell rounding (Fu et al., 1999; Taira et al., 2004; Wang et al., 2011). TNIK is required for Wnt target gene induction in colorectal cancer cell lines and leukemia stem cells and thus regulates Wnt-induced proliferation (Mahmoudi et al., 2009; Shitashige et al., 2010; Schurch et al., 2012). In addition, as a downstream effector of the small GTPase Rap2A, TNIK regulates actin dynamics, cell morphology, and neuronal structure (Taira et al., 2004; Kawabe et al., 2010; Gloerich et al., 2012). In Drosophila, the single TNIK family ortholog Misshapen (Msn) negatively regulates the transforming growth factor β pathway by phosphorylating an inhibitory site in the mammalian homologs of the Drosophila protein mothers against decapentaplegic, and the C. elegans protein SMA (SMAD) transcription factors (Kaneko et al., 2011). Despite these initial clues, the functions of TNIK in the brain are not known, in large part due to the absence of known TNIK substrates in neurons.
In vitro methods for the identification of kinase substrates include utilizing peptide libraries to determine sequence specificity and protein arrays to identify putative substrates (Obata et al., 2000; Hutti et al., 2004; Mok et al., 2009). A quasi in vivo method is to incubate cell extracts with kinase of interest and [γ-32P]ATP and identify 32P-labeled proteins as putative substrates (Cohen and Knebel, 2006). A modified version of this method is to use unnatural ATP analogs and the analog-sensitive kinase bearing kinase domain mutations (Koch and Hauf, 2010). Endogenous substrates may or may not be identified by these methods because of differences with the in vivo environments in these assays. Phosphorylation site-specific antibodies have also been used successfully to identify physiologic substrates of tyrosine and Ser/Thr kinases. However, the phosphorylation consensus sequence must be known to generate suitable phosphorylation site-specific antibodies against a pool of mixed peptides matching the consensus sequence (Gronborg et al., 2002; Kane et al., 2002; Zhang et al., 2002). Therefore, substrate identification for novel Ser/Thr kinases without known consensus sequences remains challenging.
In the present study, using a novel specific small molecule inhibitor of the TNIK family along with a phosphomotif antibody found to broadly recognize TNIK substrates, we identified multiple physiologic substrates of TNIK in neurons and generalized three phosphorylation consensus sequences, pT/S-L/I/V-D/E-x-x-x-K/R, pT/S-L/I/V-x-K/R, and pT/S-L-P/Q-L/I-x-x-K/R. The identified substrates include mammalian enabled, formin-like protein 2 (FMNL2) and members of the delta-catenin family, including p120-catenin, δ-catenin, and ARVCF, supporting functions of TNIK in actin regulation, synaptic adhesion, and synaptic plasticity.
Materials and Methods
DNA Constructs and Antibodies.
EGFP-TNIK-WT, EGFP-TNIK-K54R, the empty vector HA, HA-TNIK-wild-type (WT), K54R were gifts from Kaibuchi Kariya (University of the Ryukyus). RFP and Flag-p120-catenin were purchased from Origene (Rockville, MD). HA-TNIK-T187A and T181A were generated with the Quickchange Lightning kit (Agilent Technologies, Santa Clara, CA). The scrambled and TNIK shRNAs in the pRFP-C-RS vector were purchased from Origene. The shRNA sequences are:
TNIK shRNA #1: ATACGAGACCAACCTAATGAGCGACAGGT;
TNIK shRNA #2: ATGGCACCAGAAGTTATTGCCTGTGATGA.
The HA (sc-805-G) antibody was from Santa Cruz Biotechnology (Dallas, TX). The Flag antibody (M2) was from Sigma-Aldrich (St. Louis, MO). The pS (612547), phosphoserine/threonine (pS/T; 612548), and p120-catenin (610134) antibodies were from BD Biosciences (Franklin Lakes, NJ). The TNIK antibody was generated in rabbits by an antigenic peptide located in the intermediate domain of TNIK to ensure specificity: LLRQEQAKLNEARK. The pT187 antibody was raised against the phosphorylated peptide NH2-GRRN(pT)FIGTPC-CONH2 in rabbits and purified by protein G and two steps of peptide-affinity columns with nonphosphorylated and phosphorylated peptides. The pT312 antibody was raised against the peptide CNRNSpTIENTRRHIG in rabbits and affinity purified as described previously (Kaneko et al., 2011).
Cell Line and Primary Neuronal Cultures.
HEK293T and COS7 cells were maintained in Dulbecco’s modified Eagle’s medium (Life Technologies, Carlsbad, CA) supplemented with 10% heat-inactivated fetal bovine serum (Life Technolgies). Rat hippocampal and cortical neuronal cultures were prepared from 18-day-old embryonic rats as previously described (Wang et al., 2011). Hippocampi or cortices from both male and female rat embryos were pooled for neuronal culture preparations.
Chemical Synthesis and Analysis of PF-794.
Kinase Selectivity Panels.
PF-794 (1 μM) was screened in two kinase selectivity panels with the Z-lyte assay (Life Technologies). The IC50 of PF-794 for the kinase domain of TNIK and MAP4K4 was determined by the Z-lyte assay performed at the ATP concentration that equals the ATP Km of each protein.
Computational Methods for Compound Docking.
PF-794 was docked into the TNIK X-ray crystal structure, PDB ID: 2X7F after structure preparation with the Protein Preparation Wizard in Maestro (version 9.7, Schrödinger, LLC, New York, NY) using default options. Ligand docking calculations were conducted using Glide SP (version 6.2, Schrödinger, LLC) (Friesner et al., 2004; Halgren et al., 2004). The pose was chosen based on the top-ranked poses and structure activity relationship knowledge and rendered using the PyMOL Molecular Graphic System, Version 220.127.116.11 Schrödinger, LLC (Schrödinger, LLC).
Cell Rounding Assay.
COS7 cells were seeded onto poly-l-lysine coated 12 mm coverslips (BD Biosciences) and incubated overnight to reach about 50% confluency. Cells were transiently cotransfected with RFP and EGFP or EGFP-TNIK cDNA using lipofectamine2000 (Life Technologies) per the manufacturer’s suggested protocol. For experiments using TNIK wild-type and K54R kinase dead mutants, DMSO or PF-794 (10 μM) was added to the indicated cell populations 15 hours posttransfection and incubated for 2 hours before fixation. For cell rounding inhibition experiments, cells were transfected with RFP and EGFP-TNIK. Fifteen hours posttransfection, the cells were treated with PF-794 (1 μM) and fixed at the indicated time points. For PF-794 washout experiments, cells were transfected with RFP and EGFP-TNIK. PF-794 (1 μM) was added 3 hours posttransfection and incubated for an additional 15 hours. After this incubation the original cell medium containing PF-794 was removed, and the coverslips were washed three times with prewarmed 1× phosphate-buffered saline (PBS). Fresh media was added to the well before fixation at the indicated time points. All cells were fixed using 4% paraformaldehyde/4% sucrose in 1× PBS. Coverslips were mounted onto microscope slides using prolong gold antifade mounting medium (Life Technologies) and kept in the dark at 4°C before imaging.
Images were acquired using a Zeiss (Oberkochen, Germany) widefield microscope in combination with a 20× objective. Projected cell area was determined by tracing the RFP signal of cells that were positive for the expression of EGFP, EGFP-TNIK, or EGFP-TNIK-K54R and measuring the resulting area using ImageJ analysis software. The values for projected cell area were normalized to the control COS7 cells expressing RFP and EGFP. Bar graphs represent three independent experiments with 30–40 cells being measured per condition for each trial. All values are reported as mean ± S.E.M.
For preparation of cell extracts, rat hippocampal cultures or HEK293T cells grown in 6-well plates were washed once with ice-cold PBS. Cells were then lysed in 120–150 µl of co-IP buffer (50 mM Tris-Cl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100) with freshly added protease and phosphatase inhibitor cocktails (Sigma-Aldrich). After rocking for 30 minutes at 4°C, cell lysates were collected by scraping and cleared by centrifugation at 10,000 g for 10 minutes at 4°C. Twenty microliters of protein G-Sepharose4 beads (GE Healthcare, Little Chalfont, UK) equilibrated with co-IP buffer were mixed with 1–3 µg of antibodies. After 1 hour of incubation, beads were washed one time with co-IP buffer and then mixed with cell or tissue lysates and incubated 2 hours to overnight at 4°C with rocking. The beads were washed extensively with co-IP buffer to remove unbound proteins, separated by NuPage 4–12% Bis-Tris gels (Life Technologies), and subjected to immunoblot analysis.
In Gel Digestion, TiO2 Phosphopeptide Enrichment, and Liquid Chromatography-Tandem Mass Spectrometry Analysis.
For in gel protein digestion, proteins were isolated by 1- or 2-dimensional polyacrylamide gel electrophoresis before in situ enzymatic digestion. For TiO2 enrichment, the phosphopeptides were enriched by TopTips (Glygen Corporation, Columbia, MD). For liquid chromatography-tandem mass spectrometry , an LTQ Orbitrap equipped with a Waters nanoAcquity UPLC system was used with a Waters Symmetry C18 180 µm × 20 mm trap column and a 1.7 µm, 75 µm × 250 mm nanoAcquity UPLC column (35°C) for peptide separation (Waters, Milford, MA). For database queries, all tandem mass spectrometry spectra were searched in-house using the Mascot algorithm (version 2.2.0, Matrix Science, Boston, MA) for uninterpreted tandem mass sprectrometry spectra after using the Mascot Distiller program to generate Mascot compatible files. By using the Mascot database search algorithm, a protein was considered identified when Mascot lists it as significant and more than two peptides match the same protein. The database searched is typically the NCBInr, which is chosen over genome specific databases because a match to the correct species has more significance in the larger databases and, for some incomplete genomes, a match may be found based on homology to another species.
Substrate peptides derived from p120-catenin, δ-catenin, and SMAD1 were synthesized at GenScript (Piscataway, NJ). All peptides used in the in vitro kinase assay were 20-mer peptides with the TNIK phosphorylation site as the 10th residue. The peptides used in this study are
ADP-Glo Kinase Assay.
Ten nanomoles of purified TNIK kinase domain (Carna Biosciences, Natick, MA) was incubated with 20-mer peptide of indicated concentrations in the kinase reaction buffer (50 mM Tris-Cl pH 7.5, 0.01% Triton X-100, 10 mM MgCl2, 50 µM ATP) at 25°C for 40 minutes. The ADP-Glo assay (Promega, Madison, WI) was performed as per the manufacturer's suggested protocol. Four replicates were performed for each condition, and the mean value of ADP formed in each condition was plotted as mean ± S.E.M. The substrate concentration and ADP formed was plotted in a XY graph and fitted with Michaelis-Menten equation to determine kinetic constants, Km, Vmax, and ratio of Vmax/Km.
In vitro kinase assays were performed as in the ADP-Glo assay. Ten nanomoles of purified TNIK kinase domain was incubated with 30 µM of the 20-mer peptide in the kinase reaction buffer. The ADP formed in the kinase reaction was determined by using the ADP-Glo assay. Three microliters of each kinase reaction was dotted to a glutaraldehyde-activated nitrocellulose membrane, followed by immunoblot with the pT312 antibody. The affinity of pT312 to phosphorylated peptides was determined by normalizing the dot blot signal to ADP formed in the kinase reaction. The affinity of pT312 to the phosphorylated SMAD1-TIEN, which contained the antigenic sequence of pT312, was plotted as 100, and the affinity of pT312 to other phosphopeptides was plotted as percentage of its affinity to SMAD1-TIEN. Four replicates were performed for each condition, and the mean value of ADP formed and pT312 affinity in each condition was plotted as mean ± S.E.M.
All column graphs were plotted as mean ± S.E.M. Experiments with more than two conditions were analyzed by one-way analysis of variance test, followed by post hoc Tukey test to compare the mean of each condition with the mean of every other condition, or by post hoc Dunnett test to compare the mean of each condition with a control condition. Experiments with only two conditions were analyzed by unpaired t tests.
A Tool Compound to Study TNIK Biology.
Functional studies of TNIK have been limited to overexpression and knockdown experiments because of the lack of selective pharmacological tools. In mammalian genomes, two other kinases, misshapen-like kinase 1 (MINK1) and mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4; also known as HGK), are closely related to TNIK and are potential paralogs in mammals (Fu et al., 1999; Yao et al., 1999; Hu et al., 2004) (Fig. 1A). The kinase domains of TNIK, MINK1, and MAP4K4 are highly similar with approximately 90% amino acid identity, and their ATP binding sites are identical except for a single amino acid difference at amino acid 113 (Xing et al., 2014). Because of this high ATP binding site similarity, discovery of a potent TNIK inhibitor that is selective over MINK1 and MAP4K4 has been a challenge. However, the discovery of compounds that are selective for the TNIK family over much of the kinome has been demonstrated and is hypothesized to be due to an unusual conformation adopted by the P-loop, also known as the glycine-rich loop in the kinase domains of TNIK, MINK1, and MAP4K4 (Guimaraes et al., 2011).
Using structure-based design from previously reported MAP4K4 inhibitors (Guimaraes et al., 2011), we identified PF-794 as a potent, ATP-competitive TNIK inhibitor with a cell free IC50 of 39 nM against the kinase domain (Fig. 1B). PF-794 was profiled in vitro in two kinase selectivity panels that contain representative kinases from different families that span the human kinome and only displayed greater than 60% inhibition of TNIK, MINK1, and MAP4K4 at 1 µM, indicating excellent kinome selectivity (Fig. 1C). Docking PF-794 into the X-ray crystal structure of TNIK (PBD ID: 2X7F) suggests that the aminopyridine core forms hydrogen bond (HB) interactions with the hinge residues E106 and C108 (Fig. 1D). The compound could potentially be further stabilized in the active site through interactions of the substituents with nearby residues, such as HBs with K54 and V31, a CH-π HB with G111, and a π-π stacking interaction with Y36. The 2X7F TNIK X-ray crystal structure adopts a folded P-loop in which Y36 forms a HB with Q157, which may help stabilize this conformation. As described previously (Guimaraes et al., 2011), compounds that efficiently interact with residues in the hydrophobic tunnel formed by the folded P-loop conformation, especially with the P5 residue (Y36 in TNIK) have a high selectivity against the kinome, suggesting a rational for the selectivity of PF-794 for the TNIK family. Although PF-794 is not selective for TNIK over MINK1 and MAP4K4 (Fig. 1, C and E), it is uniquely selective for the TNIK family and thus was chosen as a tool compound to study TNIK biology and signaling.
To test whether PF-794 can efficiently inhibit TNIK in cells, we developed a cell rounding assay based on the observation that TNIK induces cell rounding in a manner requiring intact kinase activity (Fu et al., 1999) (Fig. 2A). COS7 cells were cotransfected with red fluorescent protein (RFP) as a cell fill along with enhanced green fluorescent protein (EGFP), EGFP-TNIK-WT, or EGFP-TNIK-K54R, a kinase dead mutant (Fu et al., 1999). As expected, EGFP-TNIK-WT induced cell rounding as indicated by a more compact cell morphology compared with the EGFP or EGFP-TNIK-K54R transfected cells (Fig. 2, A and B). Addition of the TNIK family inhibitor PF-794 at 15 hours posttransfection (10 µM, 2 hours) inhibited TNIK-induced cell rounding and significantly increased the projected area of EGFP-TNIK-WT transfected cells, indicating that PF-794 efficiently inhibits TNIK in cells (Fig. 2, A and B). Furthermore TNIK-induced cell rounding at 15 hours posttransfection was completely reversed by 2 hours of PF-794 treatment (1 µM, 0.5 to 5 hours; Fig. 2, C and D). When PF-794 (1 µM, 15 hours) was washed out from cells expressing EGFP-TNIK-WT, the cells rounded up to a minimal projected area within 2 hours (Fig. 2, E and F). Together, these experiments show that PF-794 inhibits TNIK in cells. Moreover, TNIK-induced cell rounding and TNIK inhibition by PF-794 occur and can be reversed within 2 hours.
Detection and Isolation of TNIK Substrates in Hippocampal Neurons.
The protein phosphatase 1 and protein phosphatase 2A inhibitor okadaic acid (OA) activated TNIK in 293T cells (Kaneko et al., 2011), suggesting that TNIK may be activated by phosphorylation. Activation of many kinases is triggered by phosphorylation in the activation loop located between a highly conserved DFG (using the single-letter amino acid codes) motif and an APE motif (Endicott et al., 2012) (Fig. 3A). In the activation loop of TNIK, phosphorylated T187 has been identified by mass spectrometry (Hornbeck et al., 2012) (Fig. 3A). We hypothesized that T187 phosphorylation may be enhanced by OA and required for activation of TNIK. To test this hypothesis, we generated a pT187 phosphorylation site-specific antibody. It specifically recognized HA-TNIK and HA-TNIK-K54R, but showed no signal with the phospho-blocking mutant HA-TNIK-T187A in transfected COS7 cells (Fig. 3B). As we hypothesized, upon incubation with OA (0.5 µM, 1 hour), T187 phosphorylation in TNIK was greatly increased in rat hippocampal cultures (Fig. 3C). The observations that the kinase dead mutant HA-TNIK-K54R can be phosphorylated at T187 and that PF-794 did not inhibit T187 phosphorylation suggest that T187 is not an autophosphorylation site and is phosphorylated by a different kinase.
In Drosophila melanogaster, the single TNIK family ortholog Msn was identified by a kinome RNAi screen as an upstream kinase of MAD—the fly ortholog of the transforming growth factor β pathway effectors SMAD1, SMAD2, and SMAD3. Msn phosphorylates MAD at T312, and this phosphorylation site is conserved in the mammalian orthologs, SMAD1 and SMAD2 (Kaneko et al., 2011). A phosphoantibody pT312 was generated against a phosphopeptide containing the T312 site of MAD and was able to recognize MAD, SMAD1, and SMAD2 phosphorylation by TNIK (Kaneko et al., 2011). As OA induced TNIK activation loop phosphorylation in rat hippocampal cultures (Fig. 3, A–C), we expected that immunoblot analysis using pT312 would detect a band at ∼59 kDa corresponding to increased SMAD phosphorylation in neurons treated with OA. Surprisingly, the most prominent bands induced by OA and detected by the pT312 antibody had higher molecular masses >100 kDa (Fig. 3D). We reasoned that the phosphorylation recognition motif present around T312 in MAD and recognized by the pT312 antibody may be common to other yet unidentified TNIK substrates. Consistent with this idea, treatment of hippocampal neurons with the selective TNIK family inhibitor PF-794 decreased the intensity and abundance of pT312 immunoreactive species in OA-treated hippocampal neurons (Fig. 3D). Furthermore, whereas OA increased the abundance of protein species immunoreactive for broad spectrum phosphoserine (pS) and phosphoserine/threonine (pS/T) antibodies, as well as species immunoreactive for the pT187 TNIK activation loop phosphoantibody, the abundance of pS, pS/T, and pT187 species was not affected by PF-794 treatment (Fig. 3, D and E). Thus, pT312 recognizes multiple phosphoprotein species that are increased by OA and decreased by TNIK inhibitor PF-794, features consistent with putative TNIK substrates.
To isolate and identify potential TNIK substrates, immunoprecipitation was performed on OA-treated hippocampal neuron lysates using the pT312 antibody (Fig. 3F). Major proteins immunoprecipitated were between 107 and 195 kDa and control immunoprecipitations were performed from untreated neurons and neurons treated with both OA and PF-794 (Fig. 3F). Multiple protein species were identified in pT312 immunoprecipitates from OA-treated neuron lysates that were absent or reduced in untreated or OA+PF-794 treated neurons (Fig. 3F). Gel sections above the antibody heavy chain to ∼200 kDa in the pT312 IP lanes were isolated and subjected to in-gel trypsin digestion followed by phosphopeptide enrichment and mass spectrometry analysis (Fig. 3F). Putative TNIK substrates were defined by their presence in pT312 immunoprecipitates, increase upon OA treatment, and reduction upon incubation with the TNIK inhibitor PF-794.
Identification of TNIK Substrates and Phosphorylation Consensus Sequences.
Mass spectrometry of pT312 immunoprecipitates identified multiple phosphopeptides from hippocampal neuron lysates. We focused on phosphopeptides that were induced by OA and inhibited by PF-794 as potential TNIK substrates (Table 1). Surprisingly, bioinformatic analysis revealed three distinct consensus sequences shared by multiple peptides from different proteins—pT/S-L/I/V-D/E, pT/S-L/I/V-x-R/K, and pT/S-L-P/Q-L/I (Table 1). All phosphopeptides and phosphomimetic peptides that match these consensus sequences are listed in Table 1. Three lines of evidence support a conclusion that these seemingly different sequences are phosphorylation consensus sequences of TNIK. First, previously identified substrates of TNIK family kinases contain these sequences. For example, the Msn phosphorylation site in MAD (pTIEN) and the corresponding phosphosites in SMAD1 (pTIEN) and SMAD2 (pTVEM), match the consensus pT/S-L/I/V-D/E (Kaneko et al., 2011) (Table 1). In addition, Prickle 1 was identified as a substrate of MINK1, and the phosphosite (pTLSR) matches with the consensus pT/S-L/I/V-x-R/K (Daulat et al., 2012) (Table 1). Because TNIK, MINK1, and MAP4K4 have almost identical ATP binding sites and activation loops (Figs. 1 and 3A), they are very likely to share phosphorylation consensus sequences. Second, the pT312 antibody immunoprecipitated proteins containing each of the three consensus sequences, suggesting these phosphosites share similar spatial structures despite different primary sequences (Fig. 3F and Table 1). Third, the abundance of most of these phosphopeptides across the three different consensus sequences was increased by OA and reduced by PF-794, suggesting they are substrates of TNIK (Table 1). The abundance of the δ-catenin peptide DGWSQYHFVASSSpTIER and MAP2 peptide SDpTLQITDLLVPGSR was not reduced by PF-794 (Table 1), but we included them as potential substrates because the PF-794 treatment was only 2 hours and the half-life of individual phosphorylation sites may vary depending on kinase and phosphatase activities.
Further evidence that pT312 is a phosphomotif antibody recognizing common structures shared by the three consensus sequences came from the observation that proteins without any of the three consensus sequences and no detectable phosphoserine or phosphothreonine were also immunoprecipitated by the pT312 antibody. However, all of these proteins contained at least one phosphomimetic peptide that matches one of the above consensus sequences (Table 1; "phosphomimetic peptides"). As expected, the amount of most of the phosphomimetic peptides were neither significantly induced by OA nor inhibited by PF-794. Together these data provided strong evidence that pT312 is a phosphomotif antibody that recognizes common structures of TNIK substrates. Moreover, these findings define canonical consensus sequences for TNIK family kinases.
TNIK Phosphorylates Identified Consensus Sequences In Vitro.
To test whether the identified pT/S-L/I/V-D/E, pT/S-L/I/V-x-R/K, and pT/S-L-P/Q-L/I consensus sequences can be directly phosphorylated by TNIK, we performed in vitro kinase reactions with purified TNIK kinase domain (amino acids 1–314) followed by a luminescent ADP-Glo assay (Promega) to quantify ADP generated in kinase reactions. Twenty-mer peptides derived from SMAD1, p120-catenin, and δ-catenin were synthesized as representatives of the three consensus sequences (Table 1 and Fig. 4A). The δ-TIER peptide matches both the pT/S-L/I/V-D/E and pT/S-L/I/V-x-R/K consensus sequences, and phosphorylation of this threonine was induced by OA but not inhibited by 2 hours of PF-794 treatment (Table 1). Therefore we included it in our experiments to confirm it is a substrate site for TNIK. Kinase reactions contained 50 µM ATP. The relationship between the luminescent signal and ADP concentration was linear between 0 and 100 µM (Fig. 4B). SMAD1 is a confirmed TNIK substrate (Kaneko et al., 2011), and correspondingly the SMAD1-TIEN peptide was phosphorylated by TNIK, whereas the alanine substitution in SMAD1-AIEN prevented phosphorylation of this peptide despite the presence of other serines and threonine, demonstrating the specificity of this phosphorylation site (Fig. 4, A and C). Similar kinase reactions were performed with p120-TLTR, p120-TLPL, and δ-TIER peptides and confirmed that they can be phosphorylated by TNIK as indicated by the generation of ADP (Fig. 4C).
The 20-mer peptides were analyzed as substrates of TNIK over a wide range of concentrations. All the substrate peptides followed Michaelis-Menten kinetics in the kinase reaction (Fig. 4C and Table 2). Peptides with alanine substitution at position +1, +2, or +3 were analyzed in the ADP-Glo assay (Fig. 4, D–G, see Materials and Methods for peptide sequences). The kinetic constants for each peptide were summarized in Table 2. In all substrate peptides, replacement of the branched-chain hydrophobic residue at +1 with alanine either completely abolished or dramatically reduced phosphorylation. In p120-TLTR and δ-TIER, replacement of the arginine at +3 with alanine strongly inhibited phosphorylation. All other alanine substitutions tested showed only small effects (Fig. 4, D–G and Table 2).
Substrate specificity is frequently achieved through complementary interactions between the kinase and substrate on the basis of charge, hydrogen bonding or hydrophobic interactions (Ubersax and Ferrell, 2007). For example, cyclin-dependent kinase 2 has specificity for a basic residue at +3 that interacts with the phosphorylated threonine in the activation loop (Brown et al., 1999). Based on the finding that TNIK is phosphorylated at T187 in the activation loop (Hornbeck et al., 2012) (Fig. 3, B–D) and that a basic residue at +3 is required for phosphorylation of p120-TLTR and δ-TIER peptides, we hypothesized that a basic residue near the phosphorylation site is required for substrate interaction with activated TNIK. Alignment of all the putative substrate sites identified by mass spectrometry showed an overrepresentation of basic residues at +6 (Table 1 and Fig. 5A). Of the peptides without a basic residue at +3, 7 out of 11 have a basic residue at +6 (Fig. 5A). Indeed, replacement of the arginine at +6 with alanine in SMAD1-TIEN and p120-TLPL peptides significantly reduced the Vmax/Km ratio (Fig. 5, B and C and Table 2), which confirmed that this basic residue is required for efficient phosphorylation by TNIK.
Taken together, these results demonstrate that the consensus sequences identified by mass spectrometry are bona fide substrates for TNIK in vitro. These data also show that a branched-chain hydrophobic residue at +1 is required for all three consensus sequences to be phosphorylated by TNIK. A basic residue at +3 or +6 is also required for efficient phosphorylation by TNIK. The nonessential residues in vitro may be required for optimized phosphorylation in vivo when both the kinase and substrate concentrations are low or are positioned within a multiprotein complex. Based on the analysis of alanine substitutions, we modified the consensus sequences to pT/S-L/I/V-D/E-x-x-x-R/K, pT/S-L/I/V-x-R/K, and pT/S-L-P/Q-L/I-x-x-R/K.
The pT312 Antibody Recognized Identified Consensus Sequences In Vitro.
To confirm that the TNIK substrate phosphorylation consensus sequences identified by mass spectrometry can be recognized by the pT312 antibody, 30 µM of 20-mer peptides (Fig. 4A and see Materials and Methods for peptide sequences) were incubated with purified TNIK kinase domain in the presence of ATP in an in vitro kinase assay, followed by dot blot with the pT312 antibody (Fig. 6A). The amount of ADP generated in the kinase reactions, which represented the amount of phosphorylated peptides, is shown in Fig. 6, B–E. The relative affinity of the pT312 antibody to each phosphorylated peptide was determined by normalizing the dot blot signal to ADP formed in each kinase reaction (Fig. 6, F–I).
The pT312 antibody was raised against a 14-mer phosphorylated peptide containing T312 of MAD (see Materials and Methods). These 14 amino acid residues are completely conserved in SMAD1 (Kaneko et al., 2011). As expected, the pT312 antibody showed the highest affinity to phosphorylated SMAD1-TIEN peptide, which we plotted as a normalized value of 100 (Fig. 6, A, B, and F). The pT312 antibody showed positive dot blot signals for p120-TLTR, p120-TLPL, and δ-TIER peptides, which confirmed all three phosphorylated consensus sequences can be detected by pT312, although with lower affinities (Fig. 6, A and F–I). The pT312 antibody did not show dot blot signals for SMAD1-TAENTRR, p120-TAPLIDR, δ-TAER, or δ-TIEA, consistent with their inefficient phosphorylation by TNIK, and confirming the specificity of pT312 to phosphorylated residues (Fig. 6). It is worth noting that p120-TLALIDR and p120-TLPAIDR showed equivalent phosphorylation by TNIK but their dot blot signals were dramatically different, which ruled out the possibility that pT312 is a nonspecific pT or even pT-L/I/V antibody (Fig. 6, A, D, and H). These data provide further evidence that phosphorylated TNIK substrate sites share a common structure or epitope that can be recognized by the pT312 phosphomotif antibody.
Confirmation of p120-Catenin as an Endogenous Substrate of TNIK.
To confirm the identified substrates can be phosphorylated by TNIK in cells, we tested p120-catenin as a representative substrate. We coexpressed TNIK with p120-catenin in HEK293T cells. Phosphorylation of p120-catenin was monitored by immunoblot analysis using the pT312 phosphomotif antibody. In the absence of coexpressed HA-TNIK, phosphorylation of p120-catenin was undetectable (Fig. 7, A and B). Coexpression of HA-TNIK significantly increased pT312-detectable phosphorylation of p120-catenin. Whereas wild-type HA-TNIK increased phosphorylation of p120-catenin, the kinase dead mutant HA-TNIK-K54R failed to increase their phosphorylation. Moreover, the activation loop phosphorylation site mutants HA-TNIK-T187A and T181A only modestly increased phosphorylation of p120-catenin, which confirmed that phosphorylation of not only T187 but also T181 in the activation loop is required for full TNIK activation and phosphorylation p120-catenin (Fig. 7, A and B). Acute treatment with the selective TNIK family inhibitor PF-794 (3 µM, 2 hours) inhibited TNIK-induced phosphorylation of p120-catenin (Fig. 7, A and B). Together these data show that TNIK mediates phosphorylation of p120-catenin in cells.
To test whether p120-catenin is phosphorylated by TNIK in neurons, 15–21 days in vitro (DIV) hippocampal cultures were treated with vehicle or the selective TNIK family inhibitor PF-794 (3 µM, 2 hours). TNIK substrates were isolated from neuron lysates by pT312 immunoprecipitation before immunoblot analysis for p120-catenin (Fig. 7C). p120-Catenin was present in pT312 immunoprecipitates, and the amount of pT312-immunoprecipitated p120-catenin was significantly reduced by PF-794 (Fig. 7, C and D). These results are consistent with our mass spectrometric and in vitro data identifying p120-catenin as endogenous TNIK substrates in neurons.
To further confirm substrate specificity, two different TNIK shRNAs were transfected into HEK293T cells and produced about 50% reduction in the protein level of endogenous TNIK protein levels (Fig. 7, E and F). p120-Catenin immunoprecipitation followed by immunoblot with the pT312 antibody showed TNIK knockdown reduced endogenous p120-catenin phosphorylation (Fig. 7, E–G). The remaining phosphorylated p120-catenin was likely due to the incomplete knockdown of TNIK and could be further reduced by addition of PF-794 (Fig. 7, E–G). These data demonstrate p120-catenin is an endogenous substrate of TNIK.
Phosphorylation Consensus Sequences for TNIK Family Kinases.
In the present study, although we were looking at SMAD phosphorylation in neurons, we discovered that the pT312 phosphorylation site-specific antibody originally raised against a single phosphorylated MAD peptide was broadly reactive to and capable of immunoisolating multiple TNIK substrates (Fig. 3). Mass spectrometric analysis of pT312 immunoprecipitates identified multiple phosphopeptides from which we generalized three phosphorylation consensus sequences for the TNIK kinase family, pT/S-L/I/V-D/E-x-x-x-R/K, pT/S-L/I/V-x-R/K, and pT/S-L-P/Q-L/I-x-x-R/K (Table 1, Figs. 4 and 5).
Several lines of evidence support the assignment of pT/S-L/I/V-D/E-x-x-x-R/K, pT/S-L/I/V-x-R/K, and pT/S-L-P/Q-L/I-x-x-R/K as TNIK phosphorylation consensus sequences. First, all proteins identified here were recognized and immunoprecipitated by the pT312 antibody, which was originally raised against epitopes corresponding to the Msn phosphorylation site on MAD (Kaneko et al., 2011) (Fig. 3). Second, all proteins identified, without exception, had at least one phosphopeptide or one phosphomimetic peptide that matched one of the consensus sequences (Table 1). Third, the previously identified TNIK family substrates, SMAD1 and SMAD2, contain the consensus pT/S-L/I/V-D/E-x-x-x-R/K, whereas Prickle 1, a separate TNIK family substrate contains the consensus pT/S-L/I/V-x-R/K (Kaneko et al., 2011; Daulat et al., 2012) (Table 1). Fourth, the abundance of most phosphopeptides was increased by two to four orders of magnitude upon okadaic acid treatment and reduced significantly by the selective TNIK inhibitor PF-794. The increased abundance of substrate phosphopeptides with OA treatment was consistent with the approximately three orders of magnitude induction of kinase catalytic activity by the activation loop phosphorylation (Adams, 2001) (Fig. 3 and Table 1). Fifth, representative peptides that match each of the three consensus sequences can be phosphorylated by TNIK in vitro and recognized by the pT312 antibody (Fig. 4–6). Sixth, alanine substitutions in ADP-Glo assay demonstrated the branched-chain hydrophobic residue at +1 and basic residue at +3 or +6 are required for efficient TNIK phosphorylation (Figs. 4 and 5 and Table 2). We thus conclude that TNIK recognizes and phosphorylates the consensus sequences pT/S-L/I/V-D/E-x-x-x-R/K, pT/S-L/I/V-x-R/K, and pT/S-L-P/Q-L/I-x-x-R/K. One attractive hypothesis is that the basic residue at +3 or +6 in substrates interacts with phosphorylated T181 and/or T187 in the activation loop of TNIK family kinases, a hypothesis that can be tested in future studies by cocrystallization of TNIK with substrate peptides.
Substrate Specificity of TNIK, MINK1, and MAP4K4.
The N-terminal kinase domains of TNIK, MINK1, and MAP4K4 are highly similar (Fig. 1). Key residues in the kinase domain involved in active site interactions with ATP are 96% similar and the activation loops between DFG and APE motifs are identical, suggesting shared phosphorylation consensus sequences (Hanks and Hunter, 1995). Consistent with the sequence similarity, all three kinases phosphorylate SMAD1 when both the kinase and substrate are expressed in heterologous cells (Kaneko et al., 2011).
How might substrate specificity be achieved between these kinases? First, substrate may interact differentially with individual kinases. Prickle 1 interacts with MINK1 but not with TNIK and MAP4K4, allowing it to be a specific MINK1 substrate (Daulat et al., 2012). Second, TNIK family members maybe be differentially expressed in different cells and tissues. Third, the intermediate domains (amino acid residues from ∼300 to 1,000) of TNIK, MINK1, and MAP4K4 are less similar than the highly conserved N-terminal kinase and C-terminal CNH domains and may confer specific subcellular targeting or substrate recognition. Using TNIK shRNAs we show that endogenous p120-catenin is phosphorylated by endogenous TNIK in cells (Fig. 7). Whether individual substrates are phosphorylated by MINK1 or MAP4K4 will require future studies with shRNAs or inhibitors that are specific within TNIK family kinases.
TNIK-Induced Cell Rounding.
One of the most reproducible effects of active TNIK is cell rounding (Fu et al., 1999; Taira et al., 2004; Wang et al., 2011) (Fig. 2). A few substrates identified here could potentially mediate this effect. By interacting with the barbed ends of actin filaments and enhances the rate of F-actin polymerization, mammalian enabled regulates cell movement and shape (Gertler and Condeelis, 2011). FMNL2 is a member of the formin family, which is characterized by the ability to polymerize straight actin filaments at the barbed end (Kuhn and Geyer, 2014). FMNL2 is one of the effector proteins of small GTPases, such as Cdc42, and was shown to regulate actin dynamics and cell morphology (Kuhn and Geyer, 2014). In addition to functioning as an important regulator of classic cadherins, p120 is a main regulator of Rho GTPase and thus has a prominent effect on actin dynamics and cell morphology (Schackmann et al., 2013). Whether phosphorylation by TNIK alters the ability of these substrates to regulate actin dynamics will require further studies with phosphoblocking and phosphomimetic mutants.
Neuronal Substrates of TNIK.
Among the 16 phosphopeptides identified in pT312 immunoprecipitates by mass spectrometry, eight phosphopeptides corresponded to sequences within p120-catenin, δ-catenin, and ARVCF, three members of the delta-catenin family of armadillo-repeat proteins involved in cell adhesion and intracellular signaling (Table 1) (Pieters et al., 2012). Delta-catenin family proteins are also part of the classic cadherin adhesion complexes at synapses where they are required for synaptogenesis, spine growth, and synaptic plasticity (Kosik et al., 2005; Elia et al., 2006; Arikkath et al., 2009; Matter et al., 2009). The fact that multiple phosphorylation sites of three delta-catenins were identified suggests their phosphorylation by TNIK is of particular functional importance.
It is worth noting that almost all of the TNIK substrates identified in this study (Table 1) are genetically linked to psychiatric or neurologic disorders. Exome sequencing of sporadic autism patients identified de novo deletion of p120-catenin in an autism patient (O'Roak et al., 2012). The recent report from the Psychiatric Genomics Consortium on the largest schizophrenia genome-wide association studies to date showed that p120-catenin gene is located in one of the 108 genome-wide significant loci (Schizophrenia Working Group of the Psychiatric Genomics, 2014). δ-Catenin is one of the essential genes deleted in the neurodevelopmental cri du chat syndrome, whose symptoms include severe cognitive and language impairments, motor delays, and behavioral problems (Medina et al., 2000). ARVCF is one of the essential genes deleted in the 22q11.2 microdeletion syndrome (Sanders et al., 2005), in which the prevalence of schizophrenia and other psychotic disorders is as high as 30–40% (Karayiorgou et al., 2010; Schneider et al., 2014). A recessive mutation of metastasis suppressor 1-like (MTSS1L) caused neurodegeneration in a consanguineous family (Alazami et al., 2015). FMNL2 was one of the essential genes deleted in a severely mentally retarded patient (Lybaek et al., 2009). Frameshift mutations of inositol polyphosphate-4-phosphatase type I (INPP4A) caused neuronal loss and seizures in both a patient and the weeble mutant mouse line that arose spontaneously (Nystuen et al., 2001; Sheffer et al., 2015).
We identified delta-catenin family proteins, actin modulators, and signaling molecule as neuronal substrates of TNIK (Fig. 8). Because these substrates have been shown to be important for synaptic development and function linked to psychiatric and neurologic disorders, TNIK could modulate synaptic adhesion, synaptic actin dynamics, and synaptic signaling by phosphorylating these substrates (Fig. 8). It will be important for future studies to confirm each of the additional putative substrates and understand the consequences of TNIK phosphorylation.
A Phosphomotif Antibody to Identify Endogenous Kinase Substrates.
In this study, we found that the pT312 phosphoantibody was broadly reactive to multiple TNIK substrates in neurons (Fig. 3). Although pT312 was raised against a single phosphopeptide, this antibody was able to recognize phosphopeptides with very different primary sequences, thus enabling the identification of three substrate consensus sequences for TNIK family kinases (Table 1). The structural basis for pT312 epitope recognition is not known. Alanine substitutions in dot blot assay suggest the branched-chain hydrophobic residue at +1, the acidic residue at +2, and the branched-chain hydrophobic residue at +3 are involved in the interaction of the phosphopeptides with pT312 (Fig. 6). Differential affinities of pT312 with similarly phosphorylated p120-TLALIDR and p120-TLPAIDR demonstrated that pT312 is not just a broadly reactive pT antibody (Fig. 6, A, D, and H). In particular, it may be that different immunoglobulin species within the pT312 polyclonal antisera possess distinct phosphoepitope recognition. Alternatively, individual immunoglobulins may recognize multiple phosphoepitopes with varying affinities. In either case, the implication is that phosphoepitopes with varying primary sequences can share epitope recognition features. An understanding of the precise molecular basis of phosphorylation site recognition awaits cocrystallization of the TNIK kinase domain with substrate peptides from each of the three different consensus sequence classes (Table 1).
Prevailing methods using antibodies to identify Ser/Thr kinase substrates have generally required knowledge of phosphorylation consensus sequences (Kane et al., 2002; Zhang et al., 2002). Antibodies were generated using a pool of mixed peptides that match the consensus sequence. By using this method, novel substrates were successfully identified for Akt (Kane et al., 2002). Nonetheless, identification of initial substrates and generalization of phosphorylation consensus sequences remain one of the biggest challenges in kinase biology. Here we described a proteomics-based approach whereby endogenous TNIK substrates were identified by a phosphomotif antibody generated against a single substrate (MAD), a specific kinase inhibitor (PF-794), and a broad-acting kinase activator (OA). This relative simple method may be applicable to other kinases.
The authors thank Dr. Kaibuchi Kariya for his generous gifts of TNIK-WT and K54R constructs and Veronica Reinhart for technical assistance.
Participated in research design: Wang, Hayward, Brandon, and Ehlers
Conducted experiments: Wang, Amato, and Rubitski,
Contributed new reagents or analytic tools: Wang, Hayward, Verhoest, and Xu
Performed data analysis: Wang, Amato, Rubitsk, and Kormos
Wrote or contributed to the writing of the manuscript: Wang, Amato, Hayward, Kormos, Verhoest, Brandon, and Ehlers.
- Received October 7, 2015.
- Accepted December 1, 2015.
↵1 Current affiliation: Neuroscience iMED, AstraZeneca, Cambridge, MA
↵2 Current affiliation: TissueVision, Cambridge, MA
↵3 Current affiliation: Blueprint Medicines, Cambridge, MA
↵4 N.S.B. and M.D.E. contributed equally to this article.
- armadillo repeat gene deleted in velo-cardio-facial syndrome
- days in vitro
- enhanced green fluorescent protein
- formin-like protein 2
- hydrogen bond
- inositol polyphosphate-4-phosphatase type I
- mitogen-activated protein kinase kinase kinase kinase 4
- misshapen-like kinase 1
- metastasis suppressor 1-like
- okadaic acid
- phosphate-buffered saline
- red fluorescent protein
- mammalian homologs of SMA (C. elegans) and MAD (Drosophila)
- Traf2- and Nck-interacting kinase
- wild type
- Copyright © 2016 by The American Society for Pharmacology and Experimental Therapeutics