Structural basis of T-loop-independent recognition and activation of CDKs by the CDK-activating kinase

Abstract

Cyclin-dependent kinases (CDKs) are prototypical regulators of the cell cycle. The CDK-activating kinase (CAK) acts as a master regulator of CDK activity by catalyzing the activating phosphorylation of CDKs on a conserved threonine residue within the regulatory T-loop. However, structural data illuminating the mechanism by which the CAK recognizes and activates CDKs have remained elusive. In this study, we determined high-resolution structures of the CAK in complex with CDK2 and CDK2-cyclin A2 by cryogenic electron microscopy. Our structures reveal a T-loop-independent kinase-kinase interface with contributions from both kinase lobes. Computational analysis and structures of the CAK in complex with CDK1-cyclin B1 and CDK11 indicate that these structures represent the general architecture of CAK-CDK complexes. These results advance our mechanistic understanding of cell cycle regulation and kinase signaling cascades.

Fig. 1. Cryo-EM structures of CAK-CDK2-cyclin A2 and CAK-CDK2 complexes.

(A) Cryo-EM reconstruction of the CAK-CDK2-cyclin A2-AMP-PNP complex coloured by its constituent subunits. (B) Atomic model of the CAK-CDK2-cyclin A2-AMP-PNP complex. (C) Cryo-EM reconstruction of the CAK-CDK2 (ADP-AlF_x) complex. (D) Atomic model of the CAK-CDK2 (ADP-AlF_x) complex. Additional views of maps and models are provided in Fig. S2. (E) Close-up view of the CDK7-CDK2 interaction interface in the CAK-CDK2-cyclin A2-AMP-PNP complex (for clarity, only the two kinases are shown). Interacting residues of CDK7 and CDK2 are shown in purple and salmon, respectively. (F) Cross-section of the CDK7-CDK2 interaction interface in the CAK-CDK2-cyclin A2-AMP-PNP complex, shown as a surface representation. A full comparison of the cryo-EM structures of all CAK-CDK2-cyclin A2 and CAK-CDK2 complexes in this study and criteria for assigning interacting residues are provided in fig. S2.

Fig. 2. Analysis of the kinase-kinase interface.

(A) Superposition of the CAK-CDK2-cyclin A2-AMP-PNP complex with CDK2-Cks1 (PDB 1BUH) (26). Inset: Close-up view of the overlap between Cks1 and CDK7 (MAT1 and cyclin H are omitted for clarity). (B) Western blots against phosphorylated CDK2 and Ponceau S-stained loading controls for assays assessing the activity of the CAK towards CDK2 in the presence/absence of Cks1. Two biologically independent sets of experiments were performed, each with N = 3 technical replicates. One representative technical replicate from one set of experiments is shown here; remaining replicates are shown in fig. S4, C and D. (C) Quantification of Western blot band intensities for one set of experiments (including the data shown in panel B), presented as the mean ± standard deviation of N = 3 technical replicates. P-values obtained by two-way ANOVA and Tukey’s multiple comparisons test: 0:1 1 hr vs. 1:1 1 hr P < 0.0001; 0:1 1 hr vs. 2:1 1 hr P < 0.0001 (**** = P ≤ 0.0001). The plot for the second set of experiments is shown in fig. S4E. (D) Side view of the CAK-CDK2-cyclin A2-AMP-PNP complex. Inset: Close-up of the CDK7 β-sheet loop inserting into the CDK2 N-terminal lobe. (E) Another view of the CDK7 β-sheet loop in the CAK-CDK2-cyclin A2-AMP-PNP complex. CDK7 loop residue K44 mediates backbone contacts with CDK2 residues, while H47 and R48 insert into the CDK2 active site. (F) Western blots against phosphorylated CDK2 and Ponceau S-stained loading controls for assays assessing the activity of CAK^WT, CAK^L219R, CAK^C3A, CAK^N2A, and CAK^N3A towards CDK2. Two biologically independent sets of experiments were performed, each with N = 3 technical replicates. One representative technical replicate from one set of experiments is shown here; remaining replicates are shown in fig. S5, C and D. (G) Quantification of Western blot band intensities for one set of experiments (including the data shown in panel F), presented as the mean ± standard deviation of N = 3 technical replicates. P-values obtained by two-way ANOVA and Tukey’s multiple comparisons test: CAK^WT 2 hr vs. CAK^L219R 2 hr P < 0.0001; CAK^WT 2 hr vs. CAK^C3A 2 hr P < 0.0001; CAK^WT 2 hr vs. CAK^N2A 2 hr P > 0.9999; CAK^WT 2 hr vs. CAK^N3A 2 hr P < 0.0001 (**** = P ≤ 0.0001, ns = P > 0.05). The plot for the second set of experiments is shown in fig. S5E.

Fig. 3. A general mechanism for CAK-CDK recognition.

(A-G) AlphaFold3-predicted structures of CAKΔN (the CAK without the N-terminal 219 residues of MAT1) in complex with substrate CDKs. For clarity, only the two kinases are shown in each panel. The T-loop of the substrate is coloured red in each panel. Predicted local distance difference test (pLDDT) and predicted aligned error (PAE) plots (46) are provided in Fig. S10. (H) Cryo-EM structure of the CAK-CDK1-cyclin B1 complex. (I) Close-up of the kinase-kinase interface of the CAK-CDK1-cyclin B1 complex, superimposed with the kinase-kinase interface of the AMP-PNP-bound CAK-CDK2-cyclin A2 complex. (J) Cryo-EM structure of the CAK-CDK11 complex. (K) Close-up of the kinase-kinase interface of the CAK-CDK11 complex, superimposed with the kinase-kinase interface of the AMP-PNP-bound CAK-CDK2-cyclin A2 complex.

Fig. 4. Interaction of the CDK7 C-terminal RxL motif with cyclins.

(A) Surface representation of the apo-CAK-CDK2-cyclin A2 complex with CDK7 residues Lys340-Phe346 shown as sticks. Inset: Close-up view showing CDK7 residues Lys340-Phe346 modelled in the cyclin A2 hydrophobic patch. (B) Surface representation of the CAK-CDK1-cyclin B1 complex with CDK7 residues Lys343-Phe346 shown as sticks. Inset: Close-up view showing CDK7 residues Lys343-Phe346 modelled in the cyclin B1 hydrophobic patch. (C) Multiple sequence alignment of C-terminal sequences of CDK7 homologues from different organisms showing conservation of the RxL motif. The alignment includes the C-terminal sequence of S. cerevisiae Kin28, a homologue of CDK7 in budding yeast that possesses neither CDK-activating kinase activity nor the C-terminal extension containing the RxL motif. The complete alignment is shown in fig. S14.

Fig. 5. Increased CDK2 T-loop dynamics of CAK-CDK2 complexes.

(A) Map-model fit of the kinase-kinase interface of the CAK-CDK2-cyclin A2-AMP-PNP complex. Inset: Close-up of the CDK2 T-loop showing that it adopts a well-ordered conformation similar to that seen in isolated CDK2-cyclin complexes. (B) Atomic model of the CAK-CDK2-cyclin A2-AMP-PNP complex. Inset: Close-up of the CDK2 T-loop. Thr160, the phosphorylation target residue, sits 13.3 Å away from the γ-phosphate of the CDK7 active site nucleotide. (C) Map-model fit of the kinase-kinase interface of the CAK-CDK2 (ADP-nitrate) complex. Inset: Close-up of the CDK2 T-loop showing that it is highly flexible, as indicated by a lack of density for the segment between G147 and V164 (dotted line). The views are the same as in panel A. (D) Map-model fit of the kinase-kinase interface of the CAK-CDK2 (ADP-AlF_x) complex. Inset: Close-up of the CDK2 T-loop showing that it adopts an intermediate conformation. The views are the same as in panels A and C. (E) Schematic illustrating the proposed model of CDK recognition and activation by the CAK. MAT1 and cyclin H are not depicted for clarity.