Oncogenic mutations of PIK3CA lead to increased membrane recruitment driven by reorientation of the ABD, p85 and C-terminus

Abstract

PIK3CA encoding the phosphoinositide 3-kinase (PI3K) p110α catalytic subunit is frequently mutated in cancer, with mutations occurring widely throughout the primary sequence. The full set of mechanisms underlying how PI3Ks are activated by all oncogenic mutations on membranes are unclear. Using a synergy of biochemical assays and hydrogen deuterium exchange mass spectrometry (HDX-MS), we reveal unique regulatory mechanisms underlying PI3K activation. Engagement of p110α on membranes leads to disengagement of the ABD of p110α from the catalytic core, and the C2 domain from the iSH2 domain of the p85 regulatory subunit. PI3K activation also requires reorientation of the p110α C-terminus, with mutations that alter the inhibited conformation of the C-terminus increasing membrane binding. Mutations at the C-terminus (M1043I/L, H1047R, G1049R, and N1068KLKR) activate p110α through distinct mechanisms, with this having important implications for mutant selective inhibitor development. This work reveals unique mechanisms underlying how PI3K is activated by oncogenic mutations, and explains how double mutants can synergistically increase PI3K activity.

Fig. 1. Structure of p110α/p85α and location of PIK3CA oncogenic mutations.

A Domain schematic of both p110α/p85α, with the catalytic core, activation loop (orange), regulatory motif (green) and C-terminus (magenta) of p110α annotated. Same colour scheme has been used to map these features on the structures below. B Oncogenic mutations in p110α mapped on the structure of WT p110α/p85α (PDB: 4OVU). The frequency of oncogenic mutations from the COSMIC database as described in panel C is coloured according to the legend with any mutation with a frequency >50 indicated as a sphere. Other features are coloured according to the domain schematic in panel A. The p85 subunit is shown as a transparent blue surface. Cartoons of the two views of PI3K highlighting these features are shown to the right of each structural model. These same cartoon views are used to map all further HDX-MS data. C Frequency of mutations across the primary sequence of PIK3CA from the Catalog of Somatic Mutations in Cancer (COSMIC) database (data from January 2022). D C-terminus in H1047R adopts a unique confirmation compared to WT (PDB: 3HHM). Features are coloured the same as in panel B. E Above: the sequence of the C-terminus is shown, with mutants coloured blue, and the membrane binding WIF motif in bold. The sequence of a frameshift mutant (N1068KLKR) is also shown. Below: Orientation of the C-terminus in the WT (PDB: 4OVU) and H1047R (PDB: 3HHM) crystal structures. The relative positioning of additional oncogenic mutants (M1043I/L, N1044K, G1049R/S) are indicated. The reorientation of the C-terminus (coloured in magenta) that occurs upon H1047R mutation is indicated by the red arrow.

Fig. 2. Conformational changes in p110α core compared to full-length p110α- p85α, and comparison to changes upon pY/Ras membrane recruitment.

A Peptides in p110α that showed significant differences in HDX (>0.4 Da and 5% difference, with a two-tailed t-test p < 0.01) between the catalytic core and the full-length complex are mapped on the structure of p110α-p85α complex (PDB: 4OVU [https://www.rcsb.org/structure/4ovu].) according to the legend. The regions of the ABD (pink) and p85 (green) that are missing in the p110α core are shown as a surface. Cartoon models representing the differences between states are shown next to the structures. A more extensive set of peptides are shown in Supplementary Fig.4. B Peptides in p110α that showed significant differences in both p110α and p85α between free (with pY) and membrane-bound (pY, membrane Ras) (data adapted from Siempelkamp et al.) are mapped onto the structure of p110α-p85α (PDB: 4OVU) according to the legend. Data are presented as the mean, with error bars representing SD (n = 3). C The sum of the number of deuteron difference for all peptides analysed over the entire deuterium exchange time course for p110α core compared to full-length p110α- p85α. Peptides coloured in red are those that had a significant change (>0.4 Da and 5% difference at any timepoint, with a two-tailed t-test p < 0.01). Each point represents a single peptide, and error bars are shown as the sum of S.D. across all time points (n = 3 for each time point). D Selected p110α peptide in the ABD-RBD linker that showed increases in exchange in p110α core compared to full-length p110α-p85α (left), and upon membrane binding of full-length PI3K (right, data adapted from Siempelkamp et al.). (Mean is shown, with error bars representing S.D., n = 3)., with smaller than the size of the point. A more extensive set of peptides comparing the full-length p110α-p85α with p110α core are shown in Supplementary Fig. 4, with the full list of all peptides and their deuterium incorporation in the source data file.

Fig. 3. Enhanced membrane binding of p110α core compared to full-length p110α/p85α, and mapping of the p110α membrane binding interface.

A Cartoon schematic describing the protein-lipid FRET assay, where tryptophan’s in the protein are excited at 280 nm, with emission at 350 nm, which upon membrane binding can excite the dansyl moiety in dansyl phosphatidylserine lipids, leading to emission at 520 nm. B Protein-lipid FRET measurements of membrane recruitment comparing p110α core and full-length p110α/p85α complex as well as full-length p110α/p85α complex and p85a apo under basal and pY activated conditions on PE/PS/PIP₂ liposomes containing 5% brain PIP₂, 65% egg yolk PE, 25% brain PS and 10% Dansyl-PS (error bars are S.D., n = 3). Experiments were carried out with 1 µM pY, 0.5 µM PI3K and 16.65 µg lipid vesicles. The values were normalised to WT apo. Two-tailed t-test p-values represented by the symbols as follows: **<0.001; *<0.05; n.s.>0.05. C Peptides in p110α core that showed significant differences in HDX (>0.4 Da and 5% difference, with a two-tailed t-test p < 0.01) upon binding to 5% PIP₂/PS/PE vesicles were mapped onto the catalytic core of p110α (PDB: 3HHM) according to the legend. D The sum of the number of deuteron difference for all peptides analysed over the entire deuterium exchange time course for p110α core upon binding membranes. Peptides coloured in red are those that had a significant change (>0.4 Da and 5% difference at any timepoint, with a two-tailed t-test p < 0.01). Each point represents a single peptide, and error bars are shown as the sum of S.D. across all time points (n = 3 for each time point). A more complete set of peptides comparing the full-length p110α-p85α with p110α core are shown in Supplementary Fig.4, with the full list of all peptides and their deuterium incorporation in the source data file. E Selected p110α peptides in the kinase domain that showed decreases in exchange in the p110α core upon binding membranes. (Mean is shown, with error bars representing S.D., n = 3). Source data for this figure are provided in the Source Data file.

Fig. 4. Biochemical analysis of C-terminal PIK3CA mutations and their effect on membrane binding and ATPase assays.

A Measurement of ATP turnover performed with different p110α constructs in solution. Experiments were performed with 200 to 0.186 nM PI3K, with 100 µM ATP in the absence of lipid vesicles. B Specific activity values normalised to WT for the ATPase assay performed with different p110α constructs in solution (technical replicates, data is presented as mean values, error bars are S.D., n = 10 (H1047R) n = 11 (N1068fs) or n = 12 (WT, M1043L, G1049R.). Experiments were performed with 19.5 to 0.59 nM PI3K, and 100 µM ATP in the absence of lipid vesicles. Two-tailed t-test p-values represented by the symbols are as follows: **<0.001; *<0.05; n.s. > 0.05. C Protein-Lipid FRET assay performed with different p110α and p85α constructs under basal and pY activated states on PM mimic liposomes containing 5% PIP₂, 10% Dansyl PS, 15 % PS, 40% PE, 10% cholesterol, 15% PC and 5% SM. Experiments were carried out at saturating concentrations of PI3K (0.5–1 µM) and 16.65 µg/ml of lipid vesicles (mean is shown, with error bars representing S.D., n = 3). Two-tailed t-test p-values represented by the symbols are as follows: **<0.001; *<0.05; n.s. > 0.05. Source data for this figure are provided in the Source Data file.

Fig. 5. Structural difference between various c-terminal mutants of p110α compared to WT p110α/p85α, and mapping of the p110α membrane binding interface.

A–C HDX comparing p110α/p85α WT vs H1047R (A), G1049R (B) and ΔCter (1–1048) (C). Significant differences in deuterium exchange are mapped on to the structure of p110α/p85α H1047R according to the legend (PDB: 3HHM] (A + B) and 4OVU (C)). The graph of the sum of number of deuteron difference in deuterium incorporation for p110α in each experiment is shown below, with each point representing a single peptide. Peptides coloured in red are those that had a significant change in the mutants (>0.4 Da and 5% difference at any timepoint, with a two-tailed t-test p < 0.01). Error bars are shown as the sum of S.D. across all time points. (n = 3 for each time point). D, E HDX comparing p110α/p85α WT vs M1043L (D) and N1068fs (E). The graph of the #D difference in deuterium incorporation for p110α in each experiment is shown below, with each point representing a single peptide. Error bars are shown as the sum of S.D. across all time points (n = 3 for each time point). For all panels, the HDX-MS data for p85 subunits is shown in Supplementary Fig. 5, along with representative peptides showing significant changes. The full HDX-MS data is available in the source data.

Fig. 6. Molecular mechanism of activation of WT PIK3CA, and how oncogenic mutations mimic this process.

Summary of molecular mechanisms of PI3K inhibition by the ABD and regulatory subunit, how activation occurs for wild-type p110α-p85 (A), and how oncogenic mutations can alter this process (B–D). A Proposed mechanism of activation of wild-type p110α-p85α. Activation is initiated by nSH2 disengagement through binding pYXXM motifs (pY), followed by ABD-p85 disengagement, followed by membrane binding (which can be promoted through binding to membrane localised Ras). B Activation of p110α-p85α by oncogenic mutants that promote the nSH2 disengagement step of PI3K activation (nSH2-helical hot spot-mutants, E542K, E545K). See Supplementary Fig. 7 for complete list of mutants. C Activation of p110α-p85α by oncogenic mutants that promote the ABD-p85 disengagement step of PI3K activation (ABD, ABD-RBD linker, C2-iSH2 mutants). See Supplementary Fig. 7 for complete list of mutants. D Different molecular mechanisms driving activation of C-terminal mutations in p110α. The regulatory motif is coloured green, with the C-terminus coloured red, and the activation loop in black (inactive) or red (orange). The C-terminus when in its closed conformation has the membrane binding WIF motif oriented away from the membrane surface. Membrane binding requires the reorientation of this tail, with the membrane itself likely involved in this conformational change. Mutations that disrupt this interface (H1047R, G1049R) are in an open conformation, leading to greatly increased membrane binding. In the N1068fs mutant there is no change in conformation in solution, but the added KLKR motif dramatically increases membrane recruitment.