Small Molecule Modulator of the mTORC2 Pathway Discovered from a DEL Library Designed to Bind to Pleckstrin Homology Domains

Abstract

Pleckstrin homology (PH) domains are structural motifs critical for cellular processes, such as signal transduction and cytoskeletal organization. Due to their involvement in various diseases, PH domains are promising therapeutic targets, yet their highly charged and hydrophobic binding sites are not ideal for traditional small drugs. In this study, we designed a DNA-encoded library (DEL) mimicking phospholipids to identify novel modulators targeting PH domains with uncharted chemical properties. Screening against several PH domains led to the discovery of 2DII, a small molecule that selectively binds to mSin1^PH. This compound can modulate mTORC2 activity by impairing mTORC2’s membrane interactions, resulting in reduced AKT1 phosphorylation. A micromapping via Dexter energy transfer based on 2DII bearing an iridium catalyst (2DII-Ir), along with a biotin-diazirine small molecule was used for target identification by proteomics, which confirmed mSin1 as the primary intracellular target of 2DII, demonstrating its potential for selective mTORC2 pathway modulation. These findings introduce a novel strategy for targeting PH domains and provide a foundation for the development of therapeutic interventions that modulate PH-domain-dependent signaling pathways.

Introduction

Pleckstrin homology (PH) domains are critical structural motifs found in a wide variety of proteins involved in intracellular signaling and cytoskeletal organization. Initially identified in the platelet protein pleckstrin,¹ these prevalent, approximately 120 amino acid domains represent the 11th most abundant distinct 3D protein fold. PH domains typically bind phosphatidylinositol lipids with high specificity, thereby playing a pivotal role in the localization of proteins to cellular membranes. This recruitment is essential for the function of many signaling pathways, resulting in the high effective concentration necessary for pathway activation, including those regulating cell growth, differentiation, and survival.

Despite their similar structure, PH domains exhibit significant diversity in their binding affinities and specificities for different phosphoinositide, as well as in their roles within cellular processes.² This diversity endows PH domains with versatility for a context-dependent response. The discovery of key mutations in various proteins’ PH domains highlights their implications in human diseases.³ For instance, a somatic mutation in the AKT1 PH domain (AKT1^PH) enhances binding affinity to phosphoinositide, thus overactivating the downstream signaling cascade and resulting in oncogenesis.⁴ On the contrary, mutations in Bruton’s tyrosine kinase PH Domain (BTK^PH) that reduce binding to PIP₃ have profound implications for its function in B-cell development and are closely associated with X-linked agammaglobulinemia (XLA), a primary immunodeficiency disorder.³ Beyond simply acting as a plasma membrane localization domain, protein–protein interactions with PH domains can also regulate the function of the interacting protein in a PIP-dependent fashion. An example is the Dbl^PH where PIP binding augments its guanine nucleotide exchange activity and has implications on Cdc42 activity.⁵

The accumulating evidence implicating PH domains in human diseases⁶⁻¹⁰ suggests that targeting PH domains with small molecules to modulate the protein–lipid interaction could have therapeutic benefits.¹¹ Indeed, compounds targeting the PH domain of AKT1¹²⁻²³ as well as the PH domains of other disease-related proteins, such as dynamin I, cytohesin, CNK1, and BRAG2²⁴⁻²⁷ have been reported. Notwithstanding these precedents, the number of reported inhibitors is largely inferior to other therapeutically relevant protein families such as kinases or bromodomains.²⁸ PH domains generally bind phosphoinositide in a shallow and highly charged surface pocket with an adjacent lipophilic patch,²⁹ which may benefit from drug properties that are poorly represented in historically small molecule screening decks. PH domains typically consist of two perpendicular antiparallel β-barrel sheets, each formed by seven β-strands, followed by an α-helix (Figure 1a, structure of AKT1^PH and mSin1^PH as representative examples). The core of this secondary structure consists of a series of highly conserved hydrophobic residues. A positively charged pocket with an arrangement of lysines and arginines interacts with anionic PIP headgroups, whereas the diversity of the loops that connect the β-strands broadens the range of functions of PH domains³⁰⁻³⁵ with hydrophobic residues, in particular, playing a crucial role in the degree of penetration into the membrane which affects lipid interaction.³⁶ For instance, in AKT1^PH, the hydrophobic residues Y18 and I19 in the β1−β2 loop insert into the membrane, engaging with the acyl tail of PIP₃ and a cholesterol molecule, which boosts the stability and affinity of the PH domain for the membrane.³⁷ Additionally, the presence of long-chain bases (LCBs) has been found to enhance the affinity of some PH domains for specific PIPs.³⁸ In the case of ACAP1, the β1−β2 loop penetrates the lipid bilayer, significantly contributing to protein–lipid binding.³⁹ Similarly, the PH domain of PLEKHA7 exhibits a greater affinity for PIP lipids in membranes compared to soluble inositol phosphates, aided by hydrophobic residues L181 and M179 in the β1−β2 loop, which deeply insert into the membrane core.⁴⁰ Electron paramagnetic resonance-guided MD simulations of GPR1 have shown that the PH domain’s orientation at the membrane surface is stabilized by hydrophobic interactions between side chains and the membrane core.⁴¹ These findings highlight the essential role of both polar and hydrophobic interactions in the membrane binding of the PH domains.

Figure 1.

Discovery of a Selective PH Domain Binder. (a) Structure of AKT1^PH (PDB: 7MYX) and mSin1^PH(PDB: 3VOQ), Graphical representation of AKT1^PH and mSin1^PH docked with PI(3,4,5)P₃ in PyMOL. Blue patches: phosphoinositide binding site, AKT1^PH: R23, K14, R25,R86; mSin1^PH: K464, R393, H392, Q427, K428. Magenta patches: lipid binding site, AKT1^PH: Y18, I19. mSin1^PH: I425, F423. PI(3,4,5)P₃ structure is represented in black. (b) PH domain directed screening using a PNA encoded fragment-based libraries mimicking phospholipid discovering a novel mSin1^PH binder affecting the mTORC2 pathway.

To target this binding site of PH domains, we envisioned library pairing fragments that could act as a surrogate of the headgroup, mimicking the negative charges of the phosphoinositol with cell-permeable groups, and a second set of fragments targeting the hydrocarbon tail (Figure 1b). The combination of these fragments should deliver novel PH ligands in a molecular diversity space that is underrepresented in commercially available or reported libraries (vide infra). To discover such inhibitors, we used a DNA-encoded library (DEL)⁴²⁻⁴⁶ where the fragments are linked to individual encoding tags (PNA),⁴⁷⁻⁴⁹ thus facilitating the combinatorial fragment pairing by hybridization.⁵⁰⁻⁵³ We focused our efforts on mSin1^PH, a regulatory and substrate recognition subunit of mTORC2 which occludes the active site pocket of mTOR and this inhibition is relieved upon membrane binding (Figure 1b).^54,55 We hypothesized that small molecule binders of mSin1^PH could enhance mTORC2 kinase activity (by relieving steric hindrance of the active site) or inhibit kinase activity (by preventing binding to an endogenous ligand i.e. PIP₃)–either property could be interesting for fundamental studies and potentially as clinical leads.

Results

DEL Targeting Pleckstrin Homology (PH) Domain Identifies mSin1 PH Domain Binders

The library is composed of two sets of fragments, designed to interact with the hydrophobic patch (480 fragments) and the binding site for the headgroup (500 fragments), respectively (Figure 2a). The first set of fragments includes different classes of hydrophobic building blocks such as saturated and unsaturated fatty acids, alkyl-derivatized heterocycles, fused rings, triterpenoids that are appended at the N-terminus of the encoding PNA (N-Ter-Lipo⁴⁸⁰, see Supplementary Scheme 1 for full structures). The second set of fragments includes polar and negatively charged moieties such as sulfonamides, sulfonic acid, phosphonates, and sulfamides. Inspired by the report of heterocyclic sulfonamide mimicking the phosphoinositide headgroup, adopting a binding pose that closely overlays with PIP₃ in its interaction with AKT1^PH,56 this fragment set is biased for this structural motif. These fragments were coupled to the C-terminus of the encoding PNA (C-Ter-Sulfo⁵⁰⁰, Supporting Information Scheme 2). Hybridization of the two sets of fragments onto a library of DNA templates presented in a microarray format results in combinatorial pairing with the adjacent presentation of the fragment for affinity screening, resulting in 240,000 unique combinations.

Figure 2.

Design and chemical space exploration of a phospholipid mimic fragment based PNA encoded library. (a) Influences and design of the combinatorial PNA encoded fragment-based library; N-Ter-LIPO⁴⁸⁰ (in red) encoded library hydrophobic moieties, and C-Ter-Sulfo⁵⁰⁰ (in blue) polar and negatively charged moieties. (b) DEL Exploring physicochemical properties by Molecular weight, clogP and tPSA analysis; Chemical diversity of reported DEL libraries (black), 49,675 data points plotted generated from reporting DEL chemical space distribution analysis. Chemical diversity of PH domain-targeting library (magenta), 65,532 unique combinations out of 240,000 plotted representing the overall physicochemical properties scope of the library. 3D graph plotted with OriginPro 8.5, Speed mode: on, for graphical representation coherence (1000 data points).

The diversity space covered by this PH-targeting library was compared to previously reported DNA-encoded libraries’ chemical space analysis⁵⁷⁻⁵⁹ by examining three molecular properties: molecular weight (MW), topological polar surface area (TPSA), and calculated logP (cLogP). As shown in Figure 2b, the product of this library (using a covalent linker between the fragments) lies in a diversity space that is poorly populated by previous libraries, with overall larger MW and tPSA values and a wider clogP distribution.

The fragment-based library was screened against mSin1^PH, AKT1^PH which is involved also in the mTORC2 pathway, and three other PH domains were chosen for their significance in major cellular functions: βARK^PH, IRS^PH, and DyII^PH. In addition, P97^ND1L, which lacks a PH domain, was used as a negative control. The library was arrayed by hybridization onto a custom DNA microarray and the different proteins, each expressed with a His-tag, were subsequently introduced followed by washes to remove excess protein and protein associated with poor binders. Incubation with a Cy5-labeled anti-His mAb was used to quantify the amount of protein retained by an individual fragment combination (see Supplementary Figure 1 for a graphical representation of the workflow). The results are presented as a heat map with the fragments of the N-Ter-Lipo⁴⁸⁰ as rows, fragments of the C-Ter-Sulfo⁵⁰⁰ as columns (Figure 3A for mSin1^PH compared to AKT1^PH and Supplementary Figure 2 for βARK^PH, IRS^PH, and DyII^PH). All tested PH domains showed distinct patterns of binding, while the negative control (P97^ND1L) did not show significant binding, suggesting that the inclusion of a lipidic fragment did not result in promiscuous binding or physisorption of proteins. The fitness of a given fragment can be assessed from the median of intensities across all of its combinations (Figure 3b). This fitness ranking was used to generate a focused library pairing the selected fragments with various covalent linkers (25 lipidic fragments × 5 sulfonamide fragments × 5 linkers, see Supplementary Figure 3 and Supplementary Scheme 3 for structures of the fragments and details of synthesis). The results of the screen of this focused library against mSin1^PH and AKT1^PH are shown in Figure 3c. The comparison of binding fitness between both PH domains shows that while there are parallels in binding fitness, some combinations show a preference for mSin1^PH. For instance, 4RV is the best binder for both mSin1^PH and AKT1^PH while 2DII and 4DII show much stronger binding for mSin1^PH than AKT1^PH. The results strongly suggest that both fragments contribute to the observed binding. For instance, lipidic fragment D, an unsaturated C16 acyl chain, is insufficient for binding mSin1^PH if not in combination with sulfonamide fragment II. Likewise, the covalent linker plays an important role, suggesting there is an optimal distance and geometry between the two fragments, and linkers 2 and 4 are the best. Five molecules were selected for resynthesis without the tag and further biochemical validation: 4RV, 4HIII, and 4YIII as high-affinity binders but without selectivity and 4DII and 2DII selective compounds for mSin1^PH (Figure 3d). The focused library was also screened against βARK^PH, IRS^PH, DyII^PH, and P97^ND1L. Analysis of the heat map further corroborates the unique interaction of 2DII with mSin1^PH (Supplementary Figure 4) and the lack of binding for a protein lacking a PH domain.

Figure 3.

mSin1^PH DNA microarray screening of the phospholipid mimic PNA encoded fragment-based library and Focus PNA encoded library. (a) mSin1^PH (300 nM) and AKT1^PH (300 nM) heat map screening with combinatorial PNA/DNA encoded fragment-based N-Ter-LIPO⁴⁸⁰/C-Ter-Sulfo⁵⁰⁰ libraries display (240k members) in PBS-DDM (0.7 mM DDM), fluorescence intensities (Cy5 channel, from Anti-polyHis Cy5 mAb), The heat maps have been normalized over the corresponding background for comparison purposes. (b) Median fluorescence intensities normalized over the background to discriminate individual fragments affinity, N-Ter-LIPO⁴⁸⁰ fragments (top) C-Ter-Sulfo⁵⁰⁰ fragments (bottom) from mSin1PH (orange) and AKT1^PH (black) screening. Highlight of the fragment selected for focused library synthesis: N-Ter-LIPO⁴⁸⁰ (A-Z)/ C-Ter-Sulfo⁵⁰⁰ (I–V). (c) mSin1^PH (300 nM) and AKT1^PH (300 nM) heat map screening with Focused library (625 members) in PBS-DDM (0.7 mM DDM), fluorescence intensities (Cy5 channel, from from Anti-polyHis Cy5 mAb), The heat maps have been normalized over the corresponding background for comparison purposes. d, Hit-compounds emerged from the focused library screening.

Hit-Compound 2DII Alters mTORC2 Pathway Activity

The activity of the five off-tag compounds was assessed in the human breast adenocarcinoma cell line MCF7. Since mSin1 is a core subunit of mTORC2, we tested the influence of the five hit-compounds on mTORC2 signaling, by following the phosphorylation of AKT1 at S473, a bona fide substrate of mTORC2.^17,18 Treatment with the compounds for 45 min exhibited a notable reduction of AKT1^S473 phosphorylation only with 2DII, and Torin, an mTOR inhibitor used as a positive control.⁶⁰ Compared to the vehicle control (DMSO) 2DII showed a 75% reduction in the signaling pathway at 40 μM with reduced potency at lower concentrations giving cellular EC₅₀ of 8.50 ± 1.45 μM (Figure 4a, Supplementary Figure 5, 6, and 7).

Figure 4.

2DII Activity. (a) AKT1 phosphorylation inhibition assay; follow of the AKT1S473 phosphorylation upon incubation of 2DII (40 to 0625 μM) compared to DMSO and Torin (mTOR kinase inhibitor) by WB (pS473 AKT1 Rb-Antibody to detect AKT1S473 phosphorylation and AKT1 pan m-Antibody to detect AKT1). (b) In vitro affinity pulldown of mSin1^PH, AKT1^PH, DyII^PH, βARK^PH, and IRS^PH (300 nM) with 2DII-Biotin compared to Biotin, washing: PBS-DDM (0.7 mM) WB analysis (m-Anti His antibody/Rb-Anti m HRP antibody to detect polyHis tag PH domain). (c) K_d measurements by 2DII-FITC (75 nM) fluorescence polarization with mSin1^PH, DyIIPH, AKT1^PH, IRS^PH, and βARK^PH. Measured mP normalized to 2DII-FITC free rotation.

To further validate the selective interaction of 2DII with mSin1^PH, the molecule was synthesized with biotin installed at the oligonucleotide tag position, a position that does not interfere with target engagement. This allowed us to perform in vitro affinity pull-down assays with different PH domains. For the in vitro selectivity study, we have selected distinct PH domains involved in major cell function regulation as representative members of the large family of PH domains: AKT1^PH, βARK^PH, IRS^PH, and DyII^PH. 2DII-Biotin was able to pull-down mSin1^PH but also AKT1^PH. By contrast, no detectable pull-down with DyII^PH was observed. Unfortunately, the intrinsic affinity of βARK^PH and IRS^PH for streptavidin beads did not allow us to conclude on the retention of these two PH domains by 2DII-Biotin affinity pull-down (Figure 4b).

The biotin moiety was replaced by a FITC group and the K_d of 2DII was measured by fluorescence polarization (FP) revealing a strong change of anisotropy exclusively for mSin1^PH, with a calculated K_d of 1.22 ± 0.35 μM (Figure 4c). A smaller anisotropy response was observed for AKT1^PH with a calculated K_d five-fold higher than that with mSin1^PH. No significant change in anisotropy was observed in response to IRS^PH and DyII^PH. Aligning with the affinity pull-down results, the “sticky” behavior of βARK^PH led to unspecific interactions affording a linear change in anisotropy with no inflection point.

Endogenous mSin1 Identified as a Significant Target of 2DII in Live Cells by Proteomic Analysis

Encouraged by the intracellular activity of 2DII and its in vitro selectivity, we aimed to evaluate its selectivity in a cellular context by chemical proteomics using photoaffinity pull-down.

First, we pursued a photo-cross-linking approach using a diazirine cross-linker.⁶¹ We prepared four different analogues of 2DII with the diazirine moiety installed in various positions, aiming to identify a modification that would not interfere with binding but would be close in space to the target to ensure target trapping rather than reaction with the lipid bilayer or hydrolysis (Supplementary Figure 8a). An additional alkyne handle was incorporated instead of the PNA tag to facilitate labeling with azide-modified fluorophores for visualization by SDS-PAGE or biotin for pull-down enrichment. The 2DII-Diazirine probes were individually incubated with mSin1^PH and cross-linked by UV irradiation. Subsequently, click chemistry was used to attach N₃–Cy3 or N₃–Biotin. Probes 3 and 4 showed the strongest labeling by SDS-PAGE (Supplementary Figure 8b,c). In vitro, a cellular extract spiked with the recombinant mSin1^PH was incubated individually with probes 3 and 4 and irradiated, but unfortunately, very poor labeling of mSin1^PH was observed (Supplementary Figure 8d). It should be noted that prior attempts to characterize PI(3,4,5)P₃ binders using functionalized active probes of this phospholipid failed to map key interactors such as mSin1 and AKT1 by affinity-based proteomics.^62,63 We hypothesized that in a cellular context, cross-linking may be challenging due to the presence of lipids. Based on this, we turned to microenvironment mapping via Dexter energy transfer, as recently developed by MacMillan et al.⁶⁴⁻⁶⁷ The catalytic nature of the technology may be able to overcome unproductive trapping of the carbenes with the membrane by multiplying the cross-linking sites along the target protein and not being restrained in the diazirine PAL stoichiometry. We opted for the second-generation Ir-photocatalyst ([Ir(dFCF₃ppy)₂(dMebpyAcOH)])⁶⁵ with enhanced cell permeability and demonstrated use in live cells.⁶⁷ A new analogue of 2DII was synthesized bearing this second generation-iridium catalyst (2DII-Ir, iridium catalyst: [Ir(dFCF₃ppy)₂(dMebpyAcOH)]), along with the biotin-diazirine small molecule (Figure 5b).

Figure 5.

Intracellular Target Identification by microenvironment mapping via Dexter energy transfer. (a) Scheme representing principle of microenvironment mapping via Dexter energy transfer. (b) Molecule toolbox synthesized for microenvironment mapping (2DII-Ir and Biotin-Diazirine). (c) In vitro labeling using Dexter energy transfer with 2DII-Ir on mSin1^PH, WB (Streptavidin AlexaFluorTM 680 to detect labeling and m-Anti His antibody/Rb-Anti m HRP antibody to detect mSin1^PH). (d) Volcano Plot representation of the label-free proteomic MS data obtained from in cellulo microenvironment mapping, left: comparison of 2DII-Ir ± 2DII (off-compete), right: comparison of 2DII vsIr-NHBoc (free-Ir). Threshold: significance (−Log10(p-value)) > 1.4, Fold change >1.4. Revealed mSin1 as a top enriched protein.

To test the labeling efficiency of the Dexter energy transfer probe, we pursued in vitro protein–probe engagement (Figure 5c). The recombinant PH domain of mSin1 was successfully labeled via Dexter transfer energy. Preincubation with the parent compound led to a slight decrease in biotinylation efficiency while controls omitting light irradiation or the catalyst led to negligible signal. The lack of complete inhibition of labeling with the preincubation with 2DII is consistent with a dynamic exchange between 2DII and 2DII-Ir in the binding site of mSin1 during the time of the experiment.

Following the successful in vitro photoaffinity labeling, we pursued target identification of 2DII in live cells. MCF7 cells were treated with 2DII-Ir followed by the addition of biotin-diazirine. Upon biotin-diazirine incubation, cells were subjected to 455 nm irradiation then washed thoroughly to eliminate any nonconverted Biotin-Diazirine and lysed. Biotinylated proteins were enriched by affinity pull-down using streptavidin-coated magnetic beads, isolated proteins were immediately subjected to digestion and the peptides thus obtained were analyzed by label-free proteomics.

The intracellular labeling directed by 2DII-Ir was compared against two distinct control experiments; a competition where the parent compound 2DII was preincubated before 2DII-Ir addition (off-compete) and a labeling directed only by the iridium catalyst without conjugation to 2DII (Ir-NHBoc) (free-Ir).

Out of more than 250 different PH domains in human cells, 138 were identified through proteomic analysis of an MCF7 cell lysate (Supplementary Table 1). When comparing the intracellular labeling mediated by 2DII-Ir against the two control experiments; 34 of these PH domains exhibited signs of enrichment showing a positive fold change (FC) and significance value (−Log₁₀(p-value)). However, in both chemoproteomic comparative analyses (2DII-Ir ± 2DII and 2DII vs Ir-NHBoc), only mSin1 stood out as a significant and prominently enriched PH domain-containing protein with a notable FC > 1.4 and a statistical significance (−Log₁₀(p-value)) > 1.4 (Figure 5d and Supplementary Figure 9). The enrichments of mSin1^PH support the notion that compound 2DII is a selective interactor rather than a general pleckstrin homology recognition motif.

2DII Exhibits Its Activity through Impairment of mTORC2-Membrane Contact Sites

Collectively, the data point to the fact that 2DII selectively binds mSin1; however, this does not explain how this binding translates into downstream pathway inhibition. Inspired by the proposed mechanisms of action of Perifosine on the PH domain of Akt,⁶⁸ and knowing that mTORC2 partially resides on the inner leaflet of the plasma membrane (PM)⁶⁹ and that mTORC2 is sensitive to PM tension,^70,71 we hypothesized three possible, nonmutually exclusive scenarios: 2DII binds to mTORC2 that displaces the complex from the membrane and thus proximity to AKT1; 2DII is partially intercalating in the plasma membrane and disrupts its lipid packing influencing membrane tension; binding prevents an endogenous ligand from releasing mSin1 from the mTOR kinase pocket and thus diminishes the kinase activity of mTORC2. We started to challenge these hypotheses by employing flotation assays using artificially made lipid vesicles as membrane mimetics.⁷² When vesicles and proteins are mixed and centrifuged together, lipids end up in the top fraction of the tube (assured through the inclusion of fluorescent lipid tracers), whereas proteins follow the centrifugal gradient to the bottom. However, if the protein of interest interacts strongly with corresponding lipids it will float with the vesicles and be found in the top fraction (Figure 6a). Comparison of the distribution of protein in the assay performed with DOPC vesicles or DOPC vesicles loaded with PI(3,4,5)P₃ or PI(3,4)P₂ clearly shows a specific interaction of mSin1^PH with PIPs, with PI(3,4,5)P₃ yielding the most protein in the top fraction (Figure 6b). The addition of 2DII to DOPC vesicles did not change the top:bottom protein distribution, whereas its addition to vesicles loaded with PI(3,4,5)P₃ resulted in a dramatic change (4-fold reduction) in floated protein. Collectively, these results demonstrate that 2DII abrogates the mSin1^PH-PI(3,4,5)P₃ interaction. However, the DOPC vesicles with a high loading of PI(3,4,5)P₃ may not recapitulate the more heterogeneous and complex environment of a cellular membrane.

Figure 6.

Deciphering the mechanism of action of 2DII. (a) Principle of the liposome flotation assay (b) results of the liposome flotation assay; WB mSin1^PH interaction with DOPC (dioleoylphosphatidylcholine) vesicles and DOPC vesicles loaded with PIPs in presence or absence of 2DII; representative Western Blot of protein content in the top (T) and bottom (B) fractions (left) and quantification of the top/bottom signal ratio (right) (c) localization of 2DII in live cell following it is fluorescent analogue 2DII-Cy3 (d) displacement of the fluorescent analogue (2DII-Cy3) by the parent compound (2DII). (e) Pretreatment with the parent compound (2DII) block the incorporation of the fluorescent analogue (2DII-Cy3). (f) Effect of 2DII to 1.2 mSin1-GFP isoform localization: full media: complete DMEM (24 h growth); starvation: DMEM w/o FCS (overnight starvation 18 h); +FCS: DMEM w/o FCS → FCS (15 min); +2DII (40 μM): DMEM w/o FCS → 2DII (45 min); +2DII (40 μM) + FCS: DMEM w/o FCS → 2DII (45 min) → FCS (15 min).

To further investigate this point, we designed and synthesized fluorescent analogue 2DII-Cy3 and followed its localization in live cells. 2DII-Cy3 localized preferentially at the PM (Figure 6c and Supplementary Movie 1). To ensure that the introduction of a fluorophore does not interfere with the natural 2DII distribution, cells were treated with 2DII post 2DII-Cy3 incubation, which induced the displacement of 2DII-Cy3. Conversely, pretreatment with the parent compound does not allow the full incorporation of 2DII-Cy3 at the PM. These experiments suggest that 2DII accumulates in the native plasma membrane⁷³ (Figure 6d,e, Supplementary Movies 2, and 3).

Finally, to track mSin1 localization in live cells, we expressed full-length mSin1 isoform 1.2 fused with a GFP protein in MCF7 cells. In full media, mSin1-GFP localized homogeneously on the plasma membrane (Figure 6f). Upon starvation, mSin1-GFP was partially displaced from the PM forming additional puncta inside the cell. Subsequent addition of growth factors restored the PM-localization of the mSin1-GFP signal. Although the localization of mTORC2 is controversial, these observations are consistent with previous reports that active mTORC2 is prominently localized to the PM.⁶⁹ Intriguingly, when the cells in the serum starvation state were treated with the compound 2DII, or 2DII plus growth factors, mSin1 partially relocalized back to the PM, albeit with a less uniform/more granular distribution. We note that MCF7 cells are notoriously difficult to transfect, and these observations are based on a limited number of cells per condition (see Figure S10 for additional examples). Although preliminary, together with the in vitro flotation assays, these results suggest that 2DII abrogates the correct recruitment of mSin1 (and ostensibly mTORC2) to the PM by both outcompeting the endogenous interaction with PI(3,4,5)P₃ and by acting as a decoy ligand at the membrane.

Discussion

Pleckstrin homology (PH) domains are known to be crucial for anchoring proteins to cellular membranes, specifically binding phosphoinositides such as PIP₂ and PIP₃, as well as in modulating protein active site accessibility and protein–protein interactions. These multifaceted functions render PH domains pivotal in cellular processes, enzymatic activities, and signal transduction pathways. Despite the prominent biological function of PH domains, small molecule modulators are sparse. We hypothesized that the classical constraints used in small molecule library design such as ‘rules of 5’ might yield unfavorable biases to the identification of chemical matter to compete for PIPs binding. Based on the unique physicochemical properties of PIPs, we designed a DNA-encoded library combining fragments that recapitulate these properties with one set of fragments having a high lyophilic character while the other set of fragments contains chemical functionalities that can mimic the phosphoinositol headgroup. A broad screen against five representative members of the PH domain (mSin1^PH, AKT1^PH, βARK^PH, IRS^PH, and DyII^PH) was used to inform the design of a focused library which allowed the identification of a selective binder of mSin1^PH, compound 2DII. mSin1^PH was selected based on the role of this protein in the mTORC2 signaling pathway. Cellular assays demonstrated that 2DII partially inhibited the mTORC2 pathway, reducing the level of downstream activation of AKT1 in MCF7 cells, while biochemical assays showed that 2DII selectively bound mSin1^PH and competed for PIP₃ binding. Chemical proteomic analyses using microenvironment mapping identified mSin1 as a likely target of 2DII. Further investigation into the mode of action indicated that 2DII predominantly localizes to the membrane and affects membrane localization of mSin1.2, and, by extension, mTORC2. These results collectively suggest that the compound could act as a decoy for mTORC2, triggering its aberrant recruitment at the plasma membrane compromising its proximity to AKT1 and thus attenuating phosphorylation. At present, we cannot rule out that 2DII acts through additional means, for example, by obstructing the release of the PH domain of mSin1 from the active site of the mTOR kinase.

Despite its modest potency, 2DII represents a novel strategy to selectively modulate the mTORC2 pathway. We hope that these results will encourage further work in selective PH ligands. Such efforts could contribute to the emergence of a novel mode of action for therapeutic intervention. While 2DII is clearly outside of the recommended range of chemical descriptors (MW = 1066.75 Da, cLogP = 7.94, tPSA = 253.66), it is active in cell-based assays, highlighting the importance of exploring uncharted chemical space. The identified compound 2DII contains a C16 acyl chain. Medicinal chemistry efforts around sphingosine-1 have shown that a flexible aliphatic chain can often be replaced by more rigid cycles resulting in better target affinity and selectivity.⁷⁴ While our library explored 480 lipidic fragments, a deeper exploration will likely yield better affinity binders. The use of DNA-encoded libraries greatly facilitates downstream validation work beyond the affinity screen since biotin, fluorophore, or microenvironment-mapping-catalyst conjugates can be prepared using the same conjugation position as the encoding tag.

Method

PNA/DNA Screening

The PNA library was hybridized on an Agilent DNA microarray slide (designs 086165-244000 spots or 074019-15000 spots) on which the DNA sequences complementary to the 14mer PNA tags encoding our library members are arrayed. The PNA-tagged library was diluted at 5 μM in the hybridization buffer (1 × −1.2 M lithium chloride; 0.3 M Li-MES; 0.012 M EDTA; 3% Li-DS; 5% Triton X100), complemented with 10 μg/mL salmon sperm DNA. The hybridization was carried out for 18 h at 60 °C. The hybridized slide was washed with 2× SSC, 0.1% SDS buffer for 5 min, 0.2× SSC, 0.1% SDS for 5 more min, briefly rinsed with mQ water, and dried at RT. The recombinant protein was diluted to 300 nM in binding buffer (PBS-DDM (0.7 mM DDM)). The protein was incubated for one hour at RT on the hybridized slides, and then the nonbound protein was washed by soaking the slide in 300 mL of the same buffer. The presence of the protein on the microarray features was detected by immunostaining the slide. Mouse anti His antibody DyLigth ^Tm649 conjugated (Rockland Immunogenics, Cat# 200 343 382) diluted 1/5000 in PBS-T, 0.5% BSA. The slide was incubated with the diluted antibody for 30 min at RT, PBS-T washed, dried, and scanned on a Genepix 4100A Personal Scanner. The scanned array was quantified using GenepixPro7 Software (molecular devices), and the median fluorescence corresponding to each feature was used to create the represented heat maps on Microsoft Excel.

Pulldown

Streptavidin-coated magnetic beads Dynal (Dynabeads Thermo Fischer ref: 11206D) were functionalized with 2DII-Biotin. Twenty μL magnetic beads were washed 2 times and blocked 1 h with PSB-5%BSA. Then the magnetic beads were resuspended on 30 μL of 10 μM biotinylated compounds diluted in PBS. The beads were functionalized for 30 min with gentle agitation and washed 3 times with PBS. PH domains were diluted in PBS-DDM (0.7 mM DDM) to the desired concentration (300 nM), and the functionalized magnetic beads were resuspended on the protein solution. The beads were then incubated with gentle rotation for 1 h and washed three times with binding buffer (PBS-DDM (0.7 mM DDM)), and the retained protein was heat eluted on SDS buffer, 95 °C for 5 min. The samples were directly loaded and separated on SDS-PAGE. The gel was transferred to a nitrocellulose membrane (Whatman) and nitrocellulose membranes were blocked and incubated with PBS-Tween 0.05 and 5% BSA. Antibody: primary antibody: mouse anti polyHis antibody DyLigth ^Tm649 conjugated (Rockland Immunogenics, Cat# 200 343 382) (1:1000). Secondary antibody: Rb-Anti mouse HRP antibody (Invitrogen catalog no. 31457) (1:5000). Chemiluminescence was recorded on a Fusion FX7.

Fluorescence Polarization

A series of two-fold dilutions into protein dilution buffer (PBS) starting at 12.5 μM protein concentration were made by sequentially mixing 99 μL of buffer with 99 μL of the previous protein solution in a new microcentrifuge tube. One μL of 2DII-FITC stock solution (7.5 μM) was added into each dilution of the protein dilution tubes, and also to 99 μL of dilution buffer without protein, for a final concentration of 75 nM fluorescent probe. The contents of each tube were gently vortexed for 10 min. The different solutions were transferred into a Nunc 96-well polypropylene plate (Thermo Scientific: Cat#267342) and also a solution of protein dilution buffer (as an assay blank for background fluorescence). The plate was loaded into the plate reader (TECAN spark plate reader) to measure polarization changes. All measurements were performed in triplicate. The measured mP was normalized to 2DII-FITC in the buffer. K_d values were determined from the one-site-specific binding curve fitting with Prism 9 (GraphPad Software).

Direct Labeling

A 100 μL solution of 1 μM mSin1^PH and 20 μM 2DII diazirine probe was prepared in a buffer (PBS-DDM (0.7 mM DDM)). The solution was shaken on ice (or at 4 °C) for an hour. The mixture was irradiated for 5 min on ice by UV light 1000 W·Hg lamp with no filter). Stock solutions were prepared as follows: N₃-Cy3/N₃-biotin linker 1.25 mM in DMSO, TCEP 14 mg·mL^–1 in water, TBTA 1.6 mM in tBuOH/DMSO 8/2, and CuSO₄ 50 mM in water. For 100 μL of protein, a solution of 2 μL of TCEP, 4 μL of N3-Cy3/N3Biotin, 2 μL of CuSO₄, and 6 μL of TBTA was prepared. The mixture was vortexed for 30 s and added to the protein solution. The reaction mixture was shaken at 25 °C for 45 min to a maximum of 1 h and stopped by acetone protein precipitation. For 100 μL of protein solution, 400 μL of cold acetone was added and kept 1 h at −20 °C. The heterogeneous solution was centrifuged for 10 min at 20,000g at 4 °C. After centrifugation, a pellet formed at the bottom. The supernatant was discarded without the pellet. The pellet was air-dried and redissolved in SDS buffer, and the samples were directly loaded and separated on SDS-PAGE. The gel was transferred to a nitrocellulose membrane (Whatman) and nitrocellulose membranes were blocked and incubated with PBS-Tween 0.05 and 5% BSA. Antibody/Streptavidin: Primary antibody: Rb anti mSin1 (Sigma-Aldrich cat. no. 07 2276) (1:1000)/ Secondary antibody: Goat anti-rabbit Cy5(Invitrogen Cat# A10523) (1:5000), Streptavidin Cy3 (Rockland Immunogenics Cat# S000 04) (1:100). For Cy3 fluorescence from the probe, it was directly read on SDS-PAGE. Fluorescence was recorded on Fusion FX7.

Protein Expression

Genes encoding PH domains of mSin1^PH, DyII^PH, AKT1^PH, IRS^PH, and βARK^PH, were codon-optimized for expression in E. coli and cloned into a pET-GST-TEV-6xHIS or in a pET-GST-TEV-GFP vector using Gibson assembly. Plasmids were sequenced for verification and transformed in E. coli (BL21*) for protein expression. Cells were grown in a 2×YT medium supplemented with 50 mg/L Kanamycin at 37 °C until the measured OD₆₀₀ reached a value of 0.6. Protein expression was then induced by adding Isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.1 mM, and cells were left to grow overnight at 18 °C. Cells were harvested by centrifugation at 4000 rpm for 45 min at 4 °C. Pellet was washed with PBS and frozen at −80 °C.

Protein Purification

PH domains were purified by affinity chromatography using a batch purification approach with glutathione Sepharose 4B beads (GE Healthcare). Cells were suspended in lysis buffer (50 mM Tris–HCl pH 8.0, 450 mM NaCl, 2 mM β-mercaptoethanol, 0.02 mg mL^–1 DNase, 0.01 mg mL^–1 Lysozyme supplemented with complete EDTA-free protease inhibitor cocktail (Roche)) and lysed using the Avestin Emulsiflex C3. The lysate was centrifugated at 16,000 rpm for 45 min at 4 °C, and the soluble fraction was collected and incubated with Glutathione 4B beads. After 2 h at 4 °C on a wheel, the beads were washed with lysis buffer and eluted with elution buffer (50 mM Tris–HCl pH 8.0, 450 mM NaCl, β-mercaptoethanol, 50 mM l-glutathione). The eluted fraction was desalted against lysis buffer to remove glutathione. The Tobacco Etch Virus protease was added at a ratio of 1/40 (1 mL TEV protease bound to a GST-tag/40 mL protein solution), and the reaction was performed at 6–8 °C for 10 h. The cleaved protein was separated from the GST-tag, protease-GST-tag, and the contaminants by reapplication to the Glutathione 4B beads. After concentration (Amicon Ultra Centrifugal Filters, <3 kDa), at 4000 rpm for 15 min at 4 °C), the remaining impurities were removed by size exclusion chromatography (ÄKTA Purifier) on Superdex 75 (GE Healthcare; separation range: 3000–70,000 Da). Proteins were concentrated once more when required, and the concentration was quantified with Bradford Protein Assays (Bio-Rad Protein Assay Dye Reagent) using a BSA standard. The obtained protein solution was aliquoted and stored at −80 °C by snap freezing in liquid nitrogen.

Flotation Assay

Lipids used in this study were purchased from Avanti Polar Inc. Calculated amounts of chosen lipids (DOPC 18:1 (Δ9-cis) 850375P, PI(3,4,5)P3 18:1 (Δ9-cis) 850156P, PI(3,4)P2 18:1 (Δ9-cis) 850153P, PI(4,5)P2 18:1 (Δ9-cis) 850155P, and Rhodamine PE 18:1 (Δ9-cis) 810150C) were mixed and dried by solvent evaporation under argon flow and vacuum. Unilamellar vesicles were formed upon hydration of dried lipids in liposome buffer (50 mM Tris–HCl pH 7.25, 110 mM NaCl) (4 mg mL^–1 of total lipid concentration), sonication for 15 min, and seven freeze/thaw cycles. Formed liposomes were incubated with the purified mSin1^PH domain (final concentration 4 μM) for 2 h at 4 °C and mixed with the corresponding amount of 62% sucrose solution (final concentration 41% sucrose); 35% and 8.5% sucrose solutions were stacked on the bottom layer, forming a sucrose gradient. Samples were centrifuged at 55,000 rpm for 1 h at 4 °C, and the localization of liposomes was monitored under UV light. Protein distribution inside the sucrose gradient was determined by Western Blot using a primary antipolyhistidine antibody (monoclonal anti-poly-Histidine antibody produced in mouse, Sigma-Aldrich, H1029) and corresponding secondary antibody (IRDye 680RD Donkey anti-Mouse (H+L), Li-Cor Biosciences, 926-68072). In the competition assay, the protein was preincubated with the compound (final concentration of 20 μM) for 15 min, and the procedure was repeated as explained above.

Immunoblotting

Western blot was conducted according to the procedure described in Rispal et al. SDS-gels (4–20% Mini-PROTEAN TGX Precast Protein Gels, 15-well, 15 μL, Bio-Rad, 4561096; 4–20% Mini-PROTEAN TGX Precast Protein Gels, 10-well, 50 μL, Bio-Rad, 4561094) were transferred using a fast dry transfer iBlot system (Invitrogen, IB21001). Nitrocellulose membranes were blocked and incubated with PBS-Tween 0.05 and 5% BSA. The primary antibody (Monoclonal Anti-polyHistidine antibody produced in mouse, Sigma-Aldrich, H1029) was incubated with the membranes (dilution 1:1000) at RT for 2 h. The incubation with the secondary antibody (IRDye 680RD Donkey anti-Mouse (H+L), Li-Cor Biosciences, 926-68072; IRDye 800CW Donkey anti-Mouse (H+L), Li-Cor Biosciences, 926-32212) was performed for 30 min at RT (dilution 1:10,000). Three washes of 5 min with PBS-Tween 0.05% followed the primary and the secondary incubations. The membranes were imaged, and the signal was quantified using the LI-COR Odyssey Classic Imaging System.

mTOR Activity Assay upon Compound Treatment

MCF7 cells (MCF7 ATCC HTB-22) were seeded in 12-well plates (100,000 cells per well in DMEM containing 10% FBS, Life Technologies). After 24 h, cells were subjected to overnight starvation with DMEM not containing FBS. Cells were then treated with compounds for 45 min without media exchange. 250 nM Torin1 was used as a positive control and incubated with the cells for 45 min. Following the compound treatment, FBS was added to the wells (final concentration 10%). After 15 min, cells were rinsed with ice-cold PBS, lysed using 80 μL of lysis buffer (10 mM Na-PPi, 10 mM Na-β-glycerophosphate, 40 mM HEPES, 4 mM EDTA, 1% Triton X-100; pH 7.4) supplemented with Halt protease/phosphatase inhibitor (Thermo Fisher Scientific, 78441) and frozen at −20 °C until further analysis. The lysates were subjected to centrifugation (14,000 rpm, 20 min, 4 °C), and supernatants were collected. Subsequently, total protein concentrations were measured using a Pierce BCA Protein Assay (Thermo Fisher Scientific, 23227). The loading buffer was added to the supernatants (250 mM Tris–HCl, 4% SDS, 30% glycerol, 100 mM DTT), and the samples were denatured at 95 °C for 7 min. Next, SDS PAGE was used to resolve the proteins. Samples were run in gradient polyacrylamide gels (4–20% Mini-PROTEAN TGX Precast Protein Gels, 10-well, 50 μL, Bio-Rad, 4561094) at 60–110 V. The gels were immunoblotted for total Akt and p473 Akt1 (Akt pan (40D4) Mouse mAb (dilution 1:1000), cell signaling, 2920; Phospho-Akt (Ser473) (D9E) XP Rabbit mAb (1:1000), Cell Signaling, 4060, respectively).

Intracellular Labeling (Microenvironment Mapping)

MCF7 cells (MCF7 ATCC HTB-22) were seeded in ϕ 60 mm plates (800,000 cells/well) in 5 mL of DMEM containing 10% FBS (Life Technologies). After 24 h, cells were subjected to overnight starvation with FluoroBrite DMEM without FBS. Cells were then treated with 2DII-Ir (20 μM) or Ir-NHBoc (10 μM) respectively. The final DMSO concentration was always kept below 0.5%. The plates were incubated at 37 °C for 3 h, and the media were removed and replaced. Diazirine-Biotin was added (200 μM) and plates were incubated at 37 °C for an additional 20 min. The plates were subsequently irradiated (without the lid) at 455 nM (LED 1W) for 30 min at 37 °C. After, cells were rinsed with ice-cold PBS, lysed using 200 μL of lysis buffer (10 mM Na-PPi, 10 mM Na-β-glycerophosphate, 40 mM HEPES, 4 mM EDTA, 1% Triton X-100; pH 7.4) supplemented with Halt protease/phosphatase inhibitor (Thermo Fisher Scientific, 78441) and frozen at −20 °C until further analysis. The lysates were subjected to centrifugation (14,000 rpm, 20 min), and supernatants were collected. Subsequently, total protein concentrations were measured using a Pierce BCA Protein Assay (Thermo Fisher Scientific, 23227). In the competition experiment, the cells were incubated with the compound 2DII (50 μM) for 45 min prior to the addition of 2DII-Ir (10 μM).

To begin the protein precipitation process, 4 volumes of cold acetone were added to the lysate sample (100 μg), vortexed thoroughly, and the sample was in the freezer for 1 h. After incubation, the sample was centrifuged to pellet the proteins, carefully removing the supernatant, and resolubilized in PBS, ensuring it was completely dissolved. For the pulldown, 200 μL of beads (magnetic streptavidin beads (NEB) were washed three times with PBS, and then the sample was incubated with the prepared beads overnight at 4 °C with gentle agitation. Following the overnight incubation, on-bead digestion was performed. The beads were washed twice with PBS, followed by two washes with 100 mM ammonium bicarbonate in water. A solution of 6 M urea in 25 mM ammonium bicarbonate was prepared and 10 mM DTT was added. The beads were incubated with this urea/DTT solution for 30 min at 55 °C. Then, iodoacetamide was added to a final concentration of 27 mM in 25 mM ammonium bicarbonate and incubated at RT in the dark for 30 min. After incubation, the supernatant was removed and the beads were washed twice with PBS and twice with 25 mM ammonium bicarbonate. Next, the beads were covered with Trypsin in 25 mM ammonium bicarbonate at a protease-to-protein ratio of 1:50 and incubated overnight at 37 °C. After this overnight incubation, trypsin was again added at a ratio of 1:100 and incubated for an additional 1 h at 37 °C. The supernatant containing the digested peptides was collected, the beads were washed once with 25 mM ammonium bicarbonate, and this wash was combined with the collected supernatant. Finally, 1% formic acid was added to the combined supernatant to stop the digestion. The samples were stored at −80 °C until proteomics was conducted. Samples were desalted prior to running (10.1101/2021.08.02.454797).

Recombinant Protein Labeling

To a solution of mSin1^PH domain (2.5 μM) was added 2DII-Ir (20 μM) and incubated for 1 h at 4 °C. To the sample was added diazirine-biotin (200 μM), gently mixed, and irradiated (455 nm, LED 1 W) on a 96-well plate for 1 h at 4 °C. In the competition experiment, the sample was incubated with the compound 2DII (50 μM) for 45 min prior to 2DII-Ir addition. The first control was missing 2DII and the second control sample was not irradiated. Samples were collected, diluted with the sample buffer, and subjected to immunoblotting.

2DII-Cy3 Localization Assay

MCF7 cells (MCF7 ATCC HTB-22) were seeded in ϕ 35 mm plates (150,000 cells/well) in 2 mL of DMEM containing 10% FBS (Life Technologies). After 24 h, cells were subjected to overnight starvation with FluoroBrite DMEM without FBS. Cells were then treated with 2DII-Cy3 (5 μM) for 20 min followed by the addition of 2DII (25 μM) or vehicle control (DMSO) for 20 min. Distribution of the fluorescent signal was captured in a time-lapse of 1 image per minute on a STELLARIS 8 FALCON (FAst Lifetime CONtrast). The experiment was repeated in reverse order under the same settings; cells were treated with 2DII (25 μM) or vehicle control (DMSO) for 20 min, followed by the addition of 2DII-Cy3 (5 μM) for 20 min.

Full-Length 1.2 mSin1-GFP Localization Assay

The constructs used in the study were purchased from Addgene (72907 N1-mSin1.1-GFP, 72908 N1-mSin1.2-GFP, 72909 N1-mSin1.5-GFP) and generated using standard molecular biology protocols. MCF7 cells (MCF7 ATCC HTB-22) were seeded on glass coverslips in a 12-well plate (100,000 cells/well) in 2 mL of DMEM containing 10% FBS (Life Technologies) and grown overnight. Cells were then transfected with the corresponding DNA using Lipofectamine 2000 Transfection Reagent (11668019 Invitrogen) according to the manufacturer’s guidelines. After 24 h cells were subjected to overnight starvation with DMEM without FBS. Subsequently, 2DII (20 and 40 μM, respectively) or DMSO was added to wells for 45 min followed by 15 min treatment with FBS (final concentration 10%). Cells were then washed with PBS and fixed with 4% PFA in PBS for 20 min at RT. After washing, cells were stained with DAPI Nucleic Acid Stain (D1306 Invitrogen, 1:1000) in PBS for 15 min. Coverslips were mounted (ProLong Glass Antifade Mountant, P36980 Invitrogen) and imaged on a STELLARIS 8 FALCON (FAst Lifetime CONtrast).

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acschembio.4c00597. The raw data associated with experiments can be found at https://zenodo.org/records/14052295.

• Synthetic procedures and physical characterization of compounds used in the studies with explicated structures; recombinant PH domain sequences and characterization (SDS-PAGE) (PDF)

Author Contributions

^§ These authors contributed equally: A.G. and J.G.

The authors declare no competing financial interest.