Allosteric activation or inhibition of PI3Kγ mediated through conformational changes in the p110γ helical domain

PI3Kγ is a critical immune signaling enzyme activated downstream of diverse cell surface molecules, including Ras, PKCβ activated by the IgE receptor, and Gβγ subunits released from activated GPCRs. PI3Kγ can form two distinct complexes, with the p110γ catalytic subunit binding to either a p101 or p84 regulatory subunit, with these complexes being differentially activated by upstream stimuli. Here using a combination of cryo electron microscopy, HDX-MS, and biochemical assays we have identified novel roles of the helical domain of p110γ in regulating lipid kinase activity of distinct PI3Kγ complexes. We defined the molecular basis for how an allosteric inhibitory nanobody potently inhibits kinase activity through rigidifying the helical domain and regulatory motif of the kinase domain. The nanobody did not block either p110γ membrane recruitment or Ras/Gβγ binding, but instead decreased ATP turnover. We also identified that p110γ can be activated by dual PKCβ helical domain phosphorylation leading to partial unfolding of an N-terminal region of the helical domain. PKCβ phosphorylation is selective for p110γ-p84 compared to p110γ-p101, driven by differential dynamics of the helical domain of these different complexes. Nanobody binding prevented PKCβ mediated phosphorylation. Overall, this works shows an unexpected allosteric regulatory role of the helical domain of p110γ that is distinct between p110γ-p84 and p110γ-p101 and reveals how this can be modulated by either phosphorylation or allosteric inhibitory binding partners. This opens possibilities of future allosteric inhibitor development for therapeutic intervention.

1 0 0 kinase domain play a critical role in regulating activity, with this region referred to as the 1 0 1 regulatory motif (Rathinaswamy et al., 2021c). The p110γ isoform is unique in that it is 1 0 2 inhibited in the absence of a regulatory subunit, with this driven by an autoinhibitory 1 0 3 conformation of the regulatory motif, that is proposed to require membrane association to p110γ-p101 (Rathinaswamy et al., 2023;Shymanets et al., 2013), however, no clear 1 1 1 unique regulatory role of this difference in dynamics has been identified.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 interface of the helical and kinase domains of p110γ is a potential site for the 3 2 1 development of novel allosteric inhibitors that modulate p110γ activity.

2 2
The previously identified inhibitory nanobody (NB7) (Rathinaswamy et al., 2021b) 3 2 3 bound with high affinity and inhibited all complexes of p110γ. The nanobody interface is 3 2 4 distinct from how the nSH2 inhibits class IA PI3K activity, as its binding site is on the 3 2 5 opposite face of the helical domain (Fig. 6A). The mechanism of inhibition is also 3 2 6 distinct, as the nSH2-helical interaction plays a critical role in preventing membrane Alphafold-multimer prediction of Gβγ-p110γ (Rathinaswamy et al., 2023) shows that 3 3 2 nanobody binding does not sterically block complex formation (Fig. 6B). This is  downstream of the IgE receptor in mast cells (Walser et al., 2013), but the full details of 3 5 0 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 how this activates PI3K has been unclear. We identified an additional PKCβ 3 5 1 phosphorylation site located in the helical domain (S594/S595) (Fig. 6C). Both the S582 p110γ has shown that it is more dynamic than other class I PI3K isoforms (Burke and dramatically decreasing helical domain dynamics (Vadas et al., 2013). This putative 3 5 7 mechanism of helical domain dynamics driving PKCβ phosphorylation is consistent with 3 5 8 our observation that p101 subunits decreased p110γ phosphorylation >100-fold. PKCβ phosphorylation of p110γ leads to increased dynamics in both the helical and kinase 3 6 0 domains with increased kinase activity, although only weakly compared to full activation 3 6 1 by either membrane localised Ras or Gβγ. This increase was observed with both 3 6 2 membrane and soluble substrate, so likely is not driven by altered membrane   increase or decrease helical domain dynamics, leading to either activation or inhibition. The high-resolution structure of an allosteric inhibitor nanobody provide initial insight 3 6 9 into which pockets can specifically be targeted. Multiple ATP competitive p110γ new opportunities for targeting p110γ dysregulation in human disease.  CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Plasmids encoding Homo sapiens p110γ (human), Mus musculus p84 (mouse), Sus 3 9 8 scrofa p101 (porcine), and Gβγ were used as previously described (Rathinaswamy et between residues 320 and 321 of PKCβ). This construct was subcloned into a 4 0 6 pACEBAC Sf9 expression vector for Sf9 protein production. All constructs were cloned 4 0 7 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made terminus, with this tag included at the N-terminus of either p84 or p101 for purification of  The plasmids encoding genes for insect cell expression were transformed into genes of interest. Successful generation was identified by blue-white colony screening 4 1 7 and the bacmid was purified using a standard isopropanol-ethanol extraction method. Bacteria were grown overnight (16 hours) in 3-5 mL 2xYT (BioBasic #SD7019). Cells   (2/100 volume ratio). This amplification was allowed to proceed for 4-5 days and . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ; https://doi.org/10.1101/2023.04.12.536585 doi: bioRxiv preprint P2 viral stocks was carried out by centrifuging cell suspensions in 50 mL Falcon tubes    Frozen insect cell pellets were resuspended in lysis buffer (20 mM Tris pH 8.0, 4 5 0 100 mM NaCl, 10 mM imidazole pH 8.0, 5% glycerol (v/v), 2 mM βME), protease purify PI3Kα/β/δ, the purification protocol was performed as described above but . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Healthcare) equilibrated in gel filtration buffer. After size exclusion, the protein was 4 7 1 concentrated, aliquoted, frozen, and stored at -80ºC. For PKCβ, the protein was eluted 4 7 2 from the strep column in gel filtration buffer, and the eluate was then loaded on a 1ml 4 7 3 HisTrap™ FF column to remove his tagged LipTev. The flowthrough was collected, and  To purify phosphorylated p110γ, the purification protocol as described above was   Full length, lipidated human Gβγ (Gβ1γ2) was expressed in Sf9 insect cells and 4 9 2 purified as described previously. After 65 hours of expression, cells were harvested, and 4 9 3 the pellets were frozen as described above. Pellets were resuspended in lysis buffer (20 4 9 4 mM HEPES pH 7.7, 100 mM NaCl, 10 mM β ME, protease inhibitor (Protease Inhibitor 4 9 5 Cocktail Set III, Sigma)) and sonicated for 2 minutes (15s on, 15s off, level 4.0, Misonix 4 9 6 sonicator 3000). The lysate was spun at 500 RCF (Eppendorf Centrifuge 5810 R) to 4 9 7 remove intact cells and the supernatant was centrifuged again at 25,000 RCF for 1 hour 4 9 8 (Beckman Coulter JA-20 rotor). The pellet was resuspended in lysis buffer and sodium 4 9 9 cholate was added to a final concentration of 1% and stirred at 4°C for 1 hour. The 5 0 0 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Coulter JA-20 rotor). The supernatant was diluted 3 times with NiNTA A buffer (20 mM 5 0 2 HEPES pH 7.7, 100 mM NaCl, 10 mM Imidazole, 0.1% C12E10, 10mM βME) and Concentrator (Millipore) to < 1 mL and injected onto a Superdex™ 75 10/300 GL size pooled, concentrated, aliquoted, frozen and stored at -80 °C. Expression and purification of nanobody: Nanobody NB7-PIK3CG with a C-terminal 6X His tag was expressed from a added and mixed for 45 minutes at 4°C to induce osmotic shock. The lysate was clarified by centrifuging at 14,000 rpm for 15 minutes (Beckman Coulter JA-20 rotor).

9
For assays measuring the inhibition by nanobody, 4X kinase (final concentration: . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Glycerol, and 100mM NaCl) and incubated at 20°C for 60 minutes.

6 7
For assays comparing the difference in activation between phosphorylated and lipid. The reaction was incubated at 20°C for 60 minutes.

7 2
After the 60-minute incubation, all reactions were stopped with 4 μ L of 2X stop Antibody IRDye QC-1, covered and incubated at 20°C for 1 hr before reading the 5 7 6 fluorescence. The fluorescence intensity was measured using a SpectraMax M5 plate 5 7 7 reader at excitation 590 nm and emission 620 nm. All data was normalized against the The percent ATP turnover was interpolated using a standard curve (0.1-100 μ M ADP).

8 0
Interpolated values were then used to calculate the specific activity of the enzyme. carried out for p110γ, p110γ-p84, and p110γ-p101 (50 nM -1.9 nM), with association . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ; https://doi.org/10.1101/2023.04.12.536585 doi: bioRxiv preprint When comparing nanobody binding to phosphorylated and unphosphorylated 5 9 3 p110γ, we used a final concentration of 25 nM for both phosphorylated and non-5 9 4 phosphorylated p110γ with association occurring for 600s, followed by a 600s 5 9 5 dissociation. The K D (dissociation constant) for the different p110γ complexes was 5 9 6 calculated from the binding curves based on their global fit to a 1:1 binding model using 5 9 7 ForteBio data analysis 12.0 (Fortebio Inc.). The membrane binding dynamics of Dy647-p84-p110γ were measured in the 6 0 1 absence and presence of nanobody 7 (NB7) using TIRF microscopy. As previously  To create SLBs, a total concentration of 0.25 mM lipids was solvated in 1x PBS  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made membrane binding, Dy647-p84-p110γ was diluted into the following buffer: 20 mM  an acousto-optic tunable filter (AOTF) and laser launch built by Vortran (Sacramento, 6 3 6 CA). The power output measured through the objective for single particle imaging was 6 3 7 1-3 mW. Excitation light passing through quad multi-pass dichroic filter cube (Semrock).  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Cryo-EM Sample Preparation and Data Collection
The copyright holder for this preprint (which this version posted May 23, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 collected using a Titan Krios equipped with a Gatan K3 DED and a BioQuantum K3 6 5 5 energy filter with a slit width of 20 eV. The movies were collected at a physical pixel size 6 5 6 of 0.830 Å/pix and a total dose of 50e -/Å 2 over 50 frames. The data were processed using cryoSPARC v.3.3.2 (Punjani et al., 2017). The 6 6 0 movies were pre-processed by patch motion correction using default settings except Fourier-cropping by a factor of 2, followed by patch CTF estimation using default 6 6 2 settings. A 3D map of PI3K p110γ-p101 complex (EMD-23808) was used to create 2D features, 795,162 particles were used for multiple rounds of ab initio reconstruction and 6 6 7 heterogeneous refinement using 4 or 5 classes. 365,178 particles, which generated the 6 6 8 two best 3D reconstruction, were used to carry out Per-particle local-motion correction with 760 pixels box size later downsized to 380 pixels followed by several rounds of ab The previous structural model of full length p110γ from the complex of p110γ- using the Colabfold v1.5.2 server (Mirdita et al., 2022). The CDR loops were removed 6 8 0 from this initial model, and the remaining nanobody was fit into the map using Chimera. Phenix.real_space_refine using realspace, rigid body, and adp refinement with tight 6 8 4 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made For the dose-response phosphorylation of p110γ, p110/p84, and p110/p101,  Deconvolution of Buparlisib's mechanism of action defines specific PI3K and  CL27c improves lung function in asthma and fibrosis. Nat Commun 9:5232-16.  The regulatory subunits of PI3Kγ control distinct neutrophil responses. Sci Signal  Phosphoinositide-3-Kinase (PI3K)-γ Inhibitor (IPI-549) as an Immuno-Oncology  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 selective inhibitors bind to an active state of PI3Kγ. Nature Chemical Biology 2023. Oncogenic mutations of PIK3CA lead to increased membrane recruitment  was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 2004. High frequency of mutations of the PIK3CA gene in human cancers.   was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted May 23, 2023. ; https://doi.org/10.1101/2023.04.12.536585 doi: bioRxiv preprint CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made density of the CDR regions contoured at 3σ (blue mesh).  single peptide. Peptides colored in red are those that had a significant change in the mutants (greater . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made      . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made