Molecular basis of Tousled-Like Kinase 2 activation

Abstract

Tousled-like kinases (TLKs) are required for genome stability and normal development in numerous organisms and have been implicated in breast cancer and intellectual disability. In humans, the similar TLK1 and TLK2 interact with each other and TLK activity enhances ASF1 histone binding and is inhibited by the DNA damage response, although the molecular mechanisms of TLK regulation remain unclear. Here we describe the crystal structure of the TLK2 kinase domain. We show that the coiled-coil domains mediate dimerization and are essential for activation through ordered autophosphorylation that promotes higher order oligomers that locally increase TLK2 activity. We show that TLK2 mutations involved in intellectual disability impair kinase activity, and the docking of several small-molecule inhibitors of TLK activity suggest that the crystal structure will be useful for guiding the rationale design of new inhibition strategies. Together our results provide insights into the structure and molecular regulation of the TLKs.

Fig. 1

Architecture and characterization of TLK2. a TLK2 domain architecture and constructs used in this study. For a complete list see Supplementary Figure 2. b Size exclusion chromatography (SEC) coupled with Multi-Angle Laser Light Scattering (MALLS) profiles were used to assess the oligomerization state of the phosphorylated and unphosphorylated constructs with or without the C-tail for the kinase domain and c ΔN-TLK2 constructs. The effect of the catalytically inactive mutant ΔN-TLK-KD (D613A mutation) is also included. d The indicated TLK2 mutants were overexpressed in AD293 cells by transient transfection and pulled down from cell lysates using Streptavidin resin. Protein–protein interactions were analyzed by western blotting for TLK2, TLK1, ASF1 and LC8 (right panel). Input levels are shown in the left panel including the Ponceau red-stained membrane showing similar total protein levels. The KD in this experiment bears the D592V mutation instead of D613A. e Kinase assays were performed from Streptavidin pull-downs performed as in d. Kinase complexes were incubated with either MBP or ASF1a in the presence of ³²P-ATP and autophosphorylation (TLK2) and substrate phosphorylation assessed in dried SDS-PAGE gels exposed to a phosphoimager. f Immunofluorescence analysis of exogenous TLK2 in transiently transfected AD293 cells. Localization of TLK2 and mutant forms was determined by staining with anti-FLAG antibodies and co-staining of nuclei with DAPI. Cytoplasmic localization of the ∆N-TLK2 mutant confirms the presence of the NLS in the N-terminus of the protein. The scale bar represents 10 μm in all the pictures. Uncropped gels and blots are shown in Supplementary Figures 10–15

Fig. 2

Catalytic activity depends on TLK2 autophosphorylation. a Autoradiograms of TLK2 autophosphorylation activity for the kinase domain and the ΔN-TLK2 constructs after a 30-min or 14-h exposure, respectively (see Supplementary Figure 4a for SDS-PAGE and the initial velocity plots). b Quantification of the autophosphorylation assay showing that the kinase domain is not efficient as the ΔN-TLK2 constructs. The unphosphorylated ΔN-TLK2 displays the larger activity in autophosphorylation. Data points indicate the relative autophosphorylation normalized to the autophosphorylation activity of ΔN-TLK2 (mean ± s.d., n = 3 biological replicates). See Supplementary Figure 4 for the details of the autophosphorylation rates of KdomL-p and ΔN-TLK2-p. c The autophosphorylation of ΔN-TLK2 is a unimolecular reaction. The increasing ratios of the ΔN-TLK2-KD inactive mutant diminished the activity of the kinase. The data points indicate the relative autophosphorylation of the ΔN-TLK2/ΔN-TLK2-KD reactions normalized to the activity of point 1 (mean ± s.d., n = 3 biological replicates). d, e Phosphorylation of TLK2 substrates. The activity is represented in a histogram for MBP and ASF1a phosphorylation by various TLK2 constructs. The data points indicate the relative substrate phosphorylation for the TLK2 constructs normalized to the TLK2 construct with the highest activity (mean ± s.d., n = 3 biological replicates). See Supplementary Figure 4 for autoradiograms and SDS-PAGE for MBP and ASF1a phosphorylation

Fig. 3

TLK2 undergoes both cis- and trans-phosphorylation in the dimer. a Hierarchical clustering analysis of the phosphorylation sites of the ΔN-TLK2 constructs expressed in E. coli displayed in a heat map. Log-transformed and row-normalized intensities of phosphosites are shown in triplicates for the heterodimer kinase-dead subunit (HeteroKD), the heterodimer active subunit (Hetero), the active homodimer (ΔN-TLK2) and the kinase-dead homodimer (ΔN-TLK2-KD). b Volcano plot validation showing the regulation and significance of phosphosites between the ΔN-TLK2 homodimer and the kinase-dead ΔN-TLK2-KD homodimer. Phosphosites are labelled in black (significant), light orange (highly significant) and red (most highly significant). c All detected phosphorylation sites mapped on the ΔN-TLK2 domain scheme. The drawing does not imply a parallel arrangement of the dimer. Sites cannot be assigned to individual molecules of the dimer. Boxed phosphosites have been detected in both HEK293 and E. coli-expressed ΔN-TLK2, while unboxed sites were only found in the E. coli-expressed ΔN-TLK2. Font colour represents significance as shown in Fig. 3b for the E. coli-expressed ΔN-TLK2 (black, light orange and red). The unboxed green sites represent phosphosites observed exclusively in the ΔN-TLK2 expressed in HEK293 cells. d Schematic representation depicting the unique phosphorylation sites observed in the heterodimer ΔN-TLK2 active subunit and in the ΔN-TLK2-KD kinase-dead subunit. The phosphorylation sites in T208, T213, S218, S226, S289, T300, S307, T357, T380, S761 and T762 were observed in both subunits of the heterodimer (see Supplementary Data 1 and Fig. 6a–d for volcano plots displaying distributions of peptides when comparing the different constructs)

Fig. 4

Crystal structure of TLK2 kinase domain. a Ribbon diagram of the TLK2 kinase domain structure in complex with ATPγS. The protein fold shows the classical N- and C-lobes and the important regulatory features associated with the kinase domain. The singular activation loop of TLK2 can be fully traced in the structure (see Table 1 for data collection and refinement). b To correlate structural and sequence conservation, we generated a sausage plot, showing a variable tube representation of the Cα trace after a PDB query for performed with ENDSCRIPT. For this drawing, homologous protein structures (1291 PDB files) were superposed onto the TLK2 kinase domain structure with ProFit. The size of the tube is proportional to the mean r.m.s. deviation per residue between Cα pairs. The white to red colour ramping is used to visualize sequence conservation. c A detailed view of the structure showing the ATPγS, catalytic and activation loops, P+1 loop and the αF helix. The key regulatory residues such as the DFG loop, the HYD, E647 and K661 are depicted in sticks. An omit map for the ATPγS molecule at 1.25 σ level is included in the figure

Fig. 5

Phosphorylation sites in the kinase domain. a Nine out of the 15 phosphorylation sites in the kinase domain can be modelled on the structure. The sites are highlighted in stick and sphere representation and lie in the N-lobe, activation loop and C-lobe. The six phosphosites on the C-tail cannot be observed because the crystallized construct lack that segment. b In vitro kinase assay of the indicated phosphorylation mutants of TLK2 isolated from AD293 cells by Streptavidin pull-downs showing relative autophosphorylation and MBP phosphorylation (see Supplementary Figure 9a for the autoradiograms and Supplementary Figure 9b IP pull-downs). The KD in this experiment bears the D592V mutation instead of the D613A. c Structure of the kinase domain highlighting the position of the intellectual disorder and cancer mutants. The amino acids where mutations have been reported are shown in stick–sphere representation. d In vitro kinase assay of the different ID and cancer mutants using MBP and ASF1a as substrates. e Quantification and comparison with the wild-type protein is represented in a histogram. The data points indicate the relative kinase activity of the TLK2-mutants normalized to the activity of the wild-type protein (mean ± s.d., n = 3 biological replicates). See Supplementary Figure 9c–f for MBP and ASF1a phosphorylation and autophosphorylation activities by the mutants

Fig. 6

TLK2 inhibition and activation model. a In vitro kinase assay of TLK2 isolated from AD293 cells by Streptavidin pull-downs in the presence or absence of the substrate MBP and the indicated kinase inhibitors. All inhibitors were used at 10 µM and the Coomassie stained gel shows equal loading of TLK2 and MBP. b Detailed view of the ATP-binding pocket in the TLK2 kinase structure showing the bound ATPγS molecule. c Zoom of the ATP-binding pocket containing the modelled molecules shown to inhibit TLK2 kinase activity. The best score modelling results displayed no steric clashes. d Model of TLK2 activation