Modulation of the substrate specificity of the kinase PDK1 by distinct conformations of the full-length protein

Abstract

The activation of at least 23 different mammalian kinases requires the phosphorylation of their hydrophobic motifs by the kinase PDK1. A linker connects the phosphoinositide-binding PH domain to the catalytic domain, which contains a docking site for substrates called the PIF pocket. Here, we used a chemical biology approach to show that PDK1 existed in equilibrium between at least three distinct conformations with differing substrate specificities. The inositol polyphosphate derivative HYG8 bound to the PH domain and disrupted PDK1 dimerization by stabilizing a monomeric conformation in which the PH domain associated with the catalytic domain and the PIF pocket was accessible. In the absence of lipids, HYG8 potently inhibited the phosphorylation of Akt (also termed PKB) but did not affect the intrinsic activity of PDK1 or the phosphorylation of SGK, which requires docking to the PIF pocket. In contrast, the small-molecule valsartan bound to the PIF pocket and stabilized a second distinct monomeric conformation. Our study reveals dynamic conformations of full-length PDK1 in which the location of the linker and the PH domain relative to the catalytic domain determines the selective phosphorylation of PDK1 substrates. The study further suggests new approaches for the design of drugs to selectively modulate signaling downstream of PDK1.

Fig. 1. PDK1 phosphorylates substrates by different mechanisms.

The phosphorylation of substrates by PDK is regulated by the ability of substrates to interact with PDK1. (Top) Substrates such as S6K or SGK are phosphorylated and activated in vitro in a manner that does not require PIP₃ or the PH domain of PDK1. Phosphorylation of a hydrophobic motif (HM) on the substrate triggers the docking interaction with the PIF pocket and a neighbouring phosphate-binding site on the small lobe of the catalytic domain of PDK1 (24, 97, 98). The inset shows the polypeptide PIFtide (derived from the HM of the protein kinase C-related protein kinase PRK2) in complex with the PIF pocket of PDK1 (PDB 4RRV) (31). Substrates of PDK1 phosphorylated at the activation loop achieve their active conformation, in which the C-terminal HM binds intramolecularly to the equivalent PIF pocket site on the substrate. (Middle) The phosphorylation of Akt downstream of PI3K requires the co-localization of PDK1 and Akt with PIP₃, which interacts with the N-terminal PH domain on Akt and the C-terminal PH domain in PDK1 at the cell membrane. (Bottom) Akt can also be phosphorylated by PDK1 in the absence of PIP₃ (38) and therefore a different mechanism for their interaction should be possible. Although a role of the PH domain of PDK1 in the PIP₃-dependent phosphorylation of Akt is established (middle), the involvement of the PH domain of PDK1 in the phosphorylation of other substrates is unknown.

Fig. 2. HYG8 inhibits the PDK1-mediated activation of Akt but not that of SGK.

(A) Chemical structures of biotin-PIP₃ and the inositol phosphates and derivatives used. (B) Effect of HYG8 (1 μM) on the PDK1-mediated activation of the Akt ΔPH [S473D] construct lacking the PH domain, full-length Akt [S473D], and full-length SGK [S422D]. The activity of PDK1_1-556 was measured indirectly by its ability to activate GST-Akt [S473D] and GST-SGK [S422D]. The activity of Akt and SGK were measured using [γ³²P]ATP and KK-Crosstide as substrates. The HM phosphorylation sites in GST-Akt [S473D] and GST-SGK [S422D] are mutated to Asp. Before the assay, the two substrates were dephosphorylated in vitro to ensure that the activity of the Akt and SGK constructs depended only on the phosphorylation of their activation loops by PDK1. N=2 independent experiments. (C) In vitro substrate phosphorylation by PDK1_1-556 was evaluated in the presence of increasing HYG8 concentrations using phospho-specific antibodies. Lane 1 shows basal phosphorylation; lane 2 shows phosphorylation in the presence of excess PDK1; lanes 3-8 show phosphorylation by PDK1 with increasing HYG8 concentrations. Quantification of fluorescence from Western blots is shown. N=4 independent experiments. One representative Western blot is shown.

Fig. 3. Inositol phosphates and synthetic derivatives differentially displace PIP₃ from PDK1_1-556 and from the PH domain of PDK1 and thermally stabilize PDK1_1-556 to differing extents.

(A,B) Schematic representations of the AlphaScreen interaction assays. Interaction of biotin-PIP₃ with GST-PDK1_1-556 (A) and with GST-PDK1_360-556, which includes only the linker and PH domain of PDK1 (B). (C) AlphaScreen interaction between biotin-PIP₃ and GST-PDK1_360-556 (4 nM) or GST-PDK1_1-556 (4 nM). N=3 independent experiments. (D-G) Effect of HYG14 and AMR1474 on the interaction of PIP₃ (20 nM) with GST-PDK1_1-556 (D) or GST-PDK1_360-556 (E). Effect of HYG8 and IP₅ on the interaction of PIP₃ (20 nM)with GST-PDK1_1-556 (F) or GST-PDK1_360-556 (G). N=2 independent experiments. GST-PDK1 constructs were used at 3 nM. (H-J) Differential scanning fluorimetry was used to assess the effect of HYG8 on the thermal stability (ΔTm) of PDK1_1-556 (H) and PDK1_50-359 (I) and of the indicated inositol phosphates and derivatives at 1 μM (white) or 20 μM (black) on PDK1_1-556 (J). N=3 independent experiments.

Fig. 4. Inositol phosphates and synthetic derivatives inhibit the formation of PDK1 dimers.

(A) Schematic representation of constructs used: full-length PDK1 (PDK1_1-556), the catalytic domain of PDK1 (PDK1_50-359) and the C-terminal construct comprising the linker region and the PH domain (PDK1360-556). (B) AlphaScreen interaction assay to measure the formation of PDK1 dimers. N=2 independent experiments. (C,D) Effect of inositol phosphates and derivatives on the interaction between HisPDK1_1-556 and GST-PDK1_360-556 (C) and on the interaction between His-PDK1_1-556 and GST-PDK1_1-556 (D). N=3 independent experiments. (E) Effect of PIFtide on the dimer interaction between His-PDK1_1-556 and GST-PDK1_360-556 and the dimerization displacement by HYG8 (500 nM). N=3 independent experiments. (F). Effect of PS210 and PS48 on the dimer interaction between His-PDK1_1-556 and GST-PDK1_360-556. N=3 independent experiments. (G) Interaction of PDK1 with PIFtide and effect of HYG8, HYG14, IP₅, or AMR1474 on the PDK1-PIFtide interaction. N=3 independent experiments.(H) The effect of the presence or absence of HYG8 (300 nM) on the PDK1-PIFtide interaction. N=2 independent experiments. (I) Differential scanning fluorimetry was used to assess the effect of PIFtide and HYG8 individually or together on the thermal stability of PDK1_1-556. N=3 independent experiments for each group. **** p<0,0001 One-way ANOVA followed by Tuckey post hoc test with Bonferroni correction. (J,K) Ab initio low resolution structure determination from experimental SEC-SAXS data using DAMMIF for apo His-PDK1_1-556 (J) and His-PDK1_1-556 + HYG8 (K). Out of 10 to 20 runs, the 3 most representative DAMMIF models according to SUPCOMB are depicted.

Fig. 5. Valsartan inhibits PDK1 dimerization by interacting with the PIF pocket of PDK1.

The AlphaScreen PDK1 dimerization assay (His-PDK1_1-556 and GST-PDK1_360-556) was used to screen a compound library (Prestwick Chemical Library®) for small molecules that affect the interaction. (A) Chemical structures of the hit compound identified, valsartan, and the inactive related compound, des(oxopentyl) valsartan. (B and C) The effect of valsartan and des(oxopentyl) valsartan on the interaction between His-PDK1_1-556 and GST-PDK1_1-556 (B) and on the interaction between His-PDK1_1-556 and GST-PDK1_360-556 (C) was assessed. N=2 independent experiments. (D) Thermal stability of PDK1_1-556 and PDK1_50-359 in the presence of valsartan and des(oxopentyl) valsartan. N=3 independent experiments. (E) The effect of DMSO, valsartan (200μM) and PIFtide (2μM) on the specific activity of the catalytic domain of PDK1 (PDK11-359). N=2 independent experiments. (F) The effect of valsartan was tested on the activity of the catalytic domain of PDK1 (PDK11-359; catalytic domain, CD), on WT full-length PDK1 (PDK1_1-556; FL) and full-length PDK1 mutated at a central residue in the PIF pocket, L155E (PDK1 FL [L155E]). N=3 independent experiments. (G) Effect of valsartan and des(oxopentyl) valsartan on the interaction between GST-PDK1_1-556 and biotin-PIFtide as assessed with AlphaScreen technology. N=3 independent experiments. (H) Effect of valsartan on the thermal stabilization of PDK1_1-556 by HYG8. N=3 independent experiments. (I) Crystal structure of PDK1 in complex with valsartan (PDB: 8DQT). (J) Magnified view of the PIF pocket of PDK1 depicting the PIFtide binding mode (top) (PDB: 4RRV (31)) and of the same orientation of the pocket in complex with the three molecules of valsartan. The 2mFo-DFc electron density map of the valsartan molecules is shown at the σ=1.0 level (bottom).

Fig. 6. Hydrogen/deuterium (H/D) exchange identifies the site of interaction of the linker-PH domain with the catalytic domain that is stabilized by HYG8.

H/D exchange was assessed for the catalytic domain of PDK1 (His-PDK1_50-359), full-length PDK1 (His-PDK1_1-556) and the full-length protein in the presence of HYG8 for 0.5, 1, 5, 10, and 100 min. The H/D exchange of selected polypeptides corresponding to the catalytic domain from PDK1_50-359 (CD, red), PDK1_1-556 (blue) and PDK1_1-556 + HYG8 (green) are shown. The structure of the catalytic domain of PDK1 (PDB: 3HRC) is used to indicate the location of polypeptides. N=3 independent experiments.

Fig. 7. Molecular model of full-length PDK1 in complex with HYG8.

(A) The modeled PDK1 structure (77-549) in the conformation bound to HYG8. The catalytic domain (dark grey) is modeled from PDB 4RRV and the PH domain (blue) is modeled from PDB 1W1G. The linker region (white cartoon) is a snapshot from an ensemble of conformations. Other colored regions are further explored with H/D exchange in fig. S11. (B) The interaction of the PH domain helix 434-443 (blue) with the catalytic domain. The conformation of the helix comprising (434-443) is from docking simulations using ClusPro. (C) HYG8 (orange, sticks) bound to the PH domain (docking). (D) The linker–catalytic domain interface was modeled based on the high affinity binding site of the pseudosubstrate inhibitor peptide PKI (5-24) from PKA (PDB 1CDK) (cyan) complexed to the catalytic domain of PKA (green). The PDK1 linker region (372-381) including Tyr³⁷³ and Tyr³⁷⁶ was modeled as a helix (yellow) occupying the equivalent hydrophobic pocket at the side of the catalytic domain that is occupied by PKI in PKA. (E) The modeled structure of PDK1 (residues 77-549) is shown fitted into the most representative ab initio low resolution structure DAMMIF (Fig. 4K) determined from experimental SAXS data for His-PDK1_1-556 + HYG8 (light grey). The DAMMIF corresponding to the low resolution structure of the catalytic domain PDK1_50-359 is shown in a grey mesh. The top part of the ab initio structure, not occupied by the model, is considered to be occupied by be the N-terminal 76 residues of PDK1 not present in the model and additional N-terminal residues from the His- and myc- tags present in the protein used for SAXS.

Fig. 8

Different conformations of full-length PDK1 determine substrate specificity. Small molecules HYG8 and valsartan stabilized different monomeric conformations of full-length PDK1, defined by the relative position of the linker-PH domain. The different conformations of full-length PDK1 had distinct biochemical characteristics. The conformations of PDK1 stabilized by HYG8 protected the “back” of the kinase catalytic domain, had unmodified in vitro activity towards some substrates, such as SGK, but were impaired in the phosphorylation of Akt (also termed PKB) and another particular peptide substrate.