Tracking the mechanism of covalent molecular glue stabilization using native mass spectrometry

Molecular glues are powerful tools for the control of protein–protein interactions. Yet, the mechanisms underlying multi-component protein complex formation remain poorly understood. Native mass spectrometry (MS) detects multiple protein species simultaneously, providing an entry to elucidate these mechanisms. Here, for the first time, covalent molecular glue stabilization was kinetically investigated by combining native MS with biophysical and structural techniques. This approach elucidated the stoichiometry of a multi-component protein–ligand complex, the assembly order, and the contributions of covalent versus non-covalent binding events that govern molecular glue activity. Aldehyde-based molecular glue activity is initially regulated by cooperative non-covalent binding, followed by slow covalent ligation, further enhancing stabilization. This study provides a framework to investigate the mechanisms of covalent small molecule ligation and informs (covalent) molecular glue development.


Fig. S3
Extended td-FA data for the binary, FC-A stabilized, and MG1 stabilized 14-3-3/Pin1 complexes. Raw td-FA data is shown for five timepoints (A-O), and the K D profile is shown for the total experiment timespan (P-R). Open datapoints represent DMSO control (binary system); filled datapoints represent stabilized (ternary system). Open datapoints represent DMSO control (binary system); filled datapoints represent stabilized (ternary system).

Fig. S11
Native mass spectrometry data showing MG1 does not bind Pin1. 14-3-3 (1 eq.), Pin1 (5 eq.), and MG1 (5 eq.) were incubated for 20 hours at pH 8.0 prior to the measurement. A lower m/z range was set to monitor binding. No binding was observed between MG1 and Pin1 in the low m/z region.

Fig. S14
Comparison between native MS and td-FA data. 14-3-3 (1 eq.), Pin1 (5 eq.), and MG1 (5 eq.) were incubated at pH 8.0 and native mass spectra were acquired over time (A). The abundances of the indicated species were summed. Comparison of the K D values measured by the td-FA experiments and abundances measured in native MS(/MS) at timepoints 2 and 4h. Both experiments were conducted at pH 8.0.

Protein complex PDB code Preparation
Binary protein complexes were prepared by mixing 12 mg/mL 14-3-3σΔC (truncated after T231 to reduce flexibility) in a 1:2 ratio with Pin1 17mer peptide (LVKHSQSRRPSpSWRQEK, N terminal acetylated, Genscript) 1 in complexation buffer (20 mM HEPES pH 7.5, 2 mM MgCl 2 , 2 mM β-mercaptoethanol, filtered (0.2 µm)), followed by overnight incubation at 4 °C. The binary protein complexes were crystallized in a hanging drop set up (4 °C), whereby the complexation solution (250 nL) was mixed with precipitation buffer (250 nL; 95 mM HEPES pH 7.1, 28% PEG400, 190 mM CaCl 2 , 5% glycerol, filtered (0.2 µm)). The crystals grew within two weeks, and were subsequently soaked for 7 days with FC-A or MG3 (final concentration 10 mM, final DMSO ≤ 1%) to form the 14-3-3σΔC/Pin1/FC-A and 14-3-3σΔC/Pin1/MG3 crystals, respectively. Crystals were directly flash-frozen in liquid nitrogen after fishing and data acquisition took place at ID23-1. Settings were 1440 images, 0.25°/ image, 15% transmission and 10 ms exposure time. autoPROC (version 1.1.7) was used to index and integrate the diffraction data. 6 The data was further processed using the CCP4i2 suite (version 8.0.002). 7 Scaling was performed using AIMLESS. 8,9 MolRep 10,11 was used for phasing, using the binary 14-3-3σΔC/Pin1 structure (PDB code: 7AOG) as a template. Using the SMILES code of each compound, a three dimensional structure of the compound was generated using AceDRG 12 , which was thereafter build in the 14-3-3σΔC/Pin1 structure utilizing the F o -F c and 2F o -F c electron density maps in COOT (version 0.9.8.1). 13 Alternating cycles of model improvement and refinement were performed using COOT, REFMAC (version 5) 14,15 , and phenix.refine (phenix software suite, version 1.20.1) 16,17 . Figures were generated with PyMOL (version 2.5.2). 2F o -F c electron density maps were contoured at 1σ. . Wells containing FITC labeled Pin1 peptide in FA buffer were used to calculate the G-factor. The data was plotted using Origin 2020 software. Using the same software, sigmoidal functions were fitted to the data using the following formula: anisotropy = start + (end -start)•(x n ) / (k n + x n ); start = bottom asymptote; end = top asymptote; x = titrant concentration; k = K D ; n = Hill coefficient. The datapoint with the highest titrant concentration was always excluded from the fitting procedure, as this datapoint suffers from a buffer mismatch.
All results are based on three independent experiments from which the average and standard deviation for each fitted K D value (Tables S4-9) were calculated using Microsoft Excel. For the stabilized (ternary) 14-3-3/Pin1 complexes, the fitted K D values represent apparent K D values (K D app ).

General materials
All reactions were prepared using analytical grade (AR) grade solvents. All reagents were purchased from TCI, or Sigma-Aldrich and were used without further purification. Solvents were removed in vacuo using a Buchi rotary evaporator connected to a diaphragm pump. All other used solvents were of analytical grade and supplied by Biosolve. Reaction glassware was dried at 130 °C for more than 24 hours prior to use. TLC was carried out on aluminum-backed silica (Merck silica gel 60 F254) plates supplied by Merck. Visualization of the plates was achieved using an ultraviolet lamp (λ max = 254 nm). Preparative HPLC was performed using a Gemini S4 110A 150 x 21.20 mm column using miliQ water with 0.1% formic acid (FA) and acetonitrile (ACN) with 0.1% FA. Analytical (LR) HPLC-MS analysis was performed on a system using comprising a C4 Jupiter SuC4300A 150 x 2.0 mm column using miliQ water with 0.1% FA and acetonitrile with 0.1% FA, using a gradient of 5% to 100% ACN over 10 minutes, connected to a Thermo Fisher LTQ XL Linear Ion Trap Mass Spectrometer. The purity of the samples was assessed using PDA (254 nm) and MS (positive mode, m/z 100 -1000). Unless otherwise stated all final compounds were ≥95% pure as judged by HPLC. High resolution mass spectra (HRMS) were recorded using a Waters ACQUITY UPLC I-Class LC system coupled to a Xevo G2 Quadrupole Time of Flight (Q-TOF) mass spectrometer equipped with a Phenomex kinetex® 2.6 μm EVO C18 100 x 2.1 mm column. Proton ( 1 H) and carbon ( 13 C) NMR spectral data were collected on a 400 MHz Bruker Cryomagnet. Chemical shifts (δ) are quoted in parts per million (ppm) and referenced to the residual solvent peak. Coupling constants (J) are quoted in Hertz (Hz) and splitting patterns reported in an abbreviated manner: app.