Identification of the first structurally validated covalent ligands of the small GTPase RAB27A

Rab27A is a small GTPase, which mediates transport and docking of secretory vesicles at the plasma membrane via protein–protein interactions (PPIs) with effector proteins. Rab27A promotes the growth and invasion of multiple cancer types such as breast, lung and pancreatic, by enhancing secretion of chemokines, metalloproteases and exosomes. The significant role of Rab27A in multiple cancer types and the minor role in adults suggest that Rab27A may be a suitable target to disrupt cancer metastasis. Similar to many GTPases, the flat topology of the Rab27A-effector PPI interface and the high affinity for GTP make it a challenging target for inhibition by small molecules. Reported co-crystal structures show that several effectors of Rab27A interact with the Rab27A SF4 pocket (‘WF-binding pocket’) via a conserved tryptophan–phenylalanine (WF) dipeptide motif. To obtain structural insight into the ligandability of this pocket, a novel construct was designed fusing Rab27A to part of an effector protein (fRab27A), allowing crystallisation of Rab27A in high throughput. The paradigm of KRas covalent inhibitor development highlights the challenge presented by GTPase proteins as targets. However, taking advantage of two cysteine residues, C123 and C188, that flank the WF pocket and are unique to Rab27A and Rab27B among the >60 Rab family proteins, we used the quantitative Irreversible Tethering (qIT) assay to identify the first covalent ligands for native Rab27A. The binding modes of two hits were elucidated by co-crystallisation with fRab27A, exemplifying a platform for identifying suitable lead fragments for future development of competitive inhibitors of the Rab27A-effector interaction interface, corroborating the use of covalent libraries to tackle challenging targets.

19  Figure S1. Evaluating Rab27A (PDB: 3BC1, chain A) hotspots using FTMap server. 1 This server identifies binding pockets within a protein surface by evaluating binding energy of molecules with different physicochemical properties. A) and B) demonstrate that organic molecule clusters predominantly occupy the nucleotide binding site and the WF pocket.

Protein expression and purification
All Rab27A constructs contain the sequence for human Rab27A (UniProt entry P51159, residues 1-192, mutations: Q78L and C123S or C188S or both as specified), which was cloned into a pET15b vector (Invitrogen) including a N-terminal His-tag followed by a Tobacco Etch Virus (TEV) recognition site (ENLYFQ¦G). Fusion constructs also contain the C-terminus of Slp2a SHD1 (SFLTEEEQEAIMKVLQRDAALKRAEEER (residues 5-32)) linked to the Nterminus of Rab27A via a flexible poly glycine-serine linker (GSGSGSG). For protein expression, plasmids were transformed to E. coli BL21 cells and spread on LB agar plates containing 100 mg/L Ampicillin for selection. Single colonies were picked for amplification and incubated overnight into LB media containing 100 mg/L Ampicillin at 37°C, shaking. Big scale cultures were inoculated using these overnight cultures at 1% v/v, and grown at 37 °C until absorbance at 600 nm reached 0.7. Protein expression was induced using 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 37 °C for 3 hours. Subsequently cells were pelleted at 4k rpm for 10 min, then re-suspended in lysis buffer containing 500 mM NaCl, 10 mM imidazole, 5 mM MgCl2 and 50 mM Tris at pH 8.0 Cells were lysed with a cell disruptor at 25K psi and centrifuged at 15k rpm for 45 min. The supernatant was loaded on Ni 2+ -NTA resin equilibrated with lysis buffer, washed extensively and eluted using buffer containing 500 mM NaCl, 300 mM imidazole, 5 mM MgCl2 and 50 mM Tris at pH 8.0 The protein was dialyzed for 6 h using 100 mM NaCl, 5 mM MgCl2 and 50 mM Tris, pH 8.0 Afterwards TEV protease (obtained in-house as previously described 5  The peaks corresponding to Rab27A constructs were analysed by SDS-PAGE, pooled, concentrated and flash frozen using liquid nitrogen. All Rab27A constructs containing exposed cysteines at C123 or C188 were purified in buffer containing an additional 0.1% β-mercaptoethanol (βME) during Ni 2+ -NTA steps.

Protein labelling and purification
To a 15 mL falcon tube were added 600 μL of desired construct (100 μM stock), 100 μL of 50% w/v TCEP-agarose beads (ThermoFisher), and 1.8 mL of 100 mM NaCl, 5 mM MgCl2, and 20 mM HEPES pH 8.0. 50 μL of ligand (50 mM stock) were diluted with 2.45 mL of 100 mM NaCl, 5 mM MgCl2, and 20 mM HEPES pH 8.0, followed by centrifugation (2400 rpm, 5 min). The supernatant was added to the protein mixture and incubated at 4 °C. The reaction was monitored as described in the QIT protocol (v. infra). When the labelling reached 90%, the labelled protein solution was concentrated to 0.5 mL by using a Vivaspin 20 filter (5000 MWCO).
The protein was diluted with 4.5 mL of 100 mM NaCl, 5 mM MgCl2, and 20 mM HEPES pH 8.0 and concentrated again to 0.5 mL (5x), to remove excess compound, then purified by superdex S-75 gel filtration at a flow rate of 1 mL/min in 150 mM NaCl, 5 mM MgCl2, and 20 mM Tris at pH 8 for crystallography.

Crystal diffraction, data collection and data processing
Data collections were carried out at i02 beamline in Diamond Light Source (Oxford, UK) at 100K of temperature, wavelength 0.9795 Å, and using a Pilatus detector. Data were collected at 0.2°-0.5° oscillations per image and 200° total oscillation per crystal. Data was integrated, scaled and reduced using DIALS. 6 Initial phases were calculated using the molecular replacement program Phaser. 7 The coordinates of Rab27A from chain A of the Rab27A-Slp2a complex (PDB:3BC1) without the nucleotide and the magnesium ion were used as the search model. Subsequently, the initial model generated by phaser was refined through an iterative cycle using COOT 9 and REFMAC5. 10 Final model structures were validated using the Molprobity server 11 at http://molprobity.biochem.duke.edu. All structure images were prepared using Pymol (DeLano Scientific LLC, http://pymol.sourceforge.net/). X-ray data collection, processing and refinement statistics are given in Table S1.

Molecular dynamics simulations
The Rab27A structure (PDB: 3BC1) was simulated with bound GTP and Mg 2+ . The structure was parametrised using the latest CHARMM36 force field 4 , solvated with Tip3p waters 12 and neutralised with Na + and Clions at a concentration of 150 mM. Temperature was coupled for 100 ps at 300 K with the V-rescale method, 13 with positional restraints on the protein heavy atoms. Pressure was then coupled at 1 bar for another 100 ps with position restraints, using the Berendsen algorithm. 14 The Particle mesh Ewald method 15 was used for electrostatic interactions, and LINCS 16 to define the constraints. The integration timestep was 2 fs. The final production simulation was extended for 250 ns. The simulation and data analysis were carried out using the GROMACS simulation package. 17 qIT assay for screening and hit validation 126 electrophilic acrylamides (see supplementary excel file) were screened in the qIT assay adapted from Craven et al 18 (1k rpm, 1 min) and incubated for 60 min at room temperature and then fluorescence intensity (excitation/emission: 384/470 nm) was measured on an EnVision™ plate reader. Data analysis: All analyses were conducted using Prism 9.0 software (Graphpad). Each fluorescence readout was normalized to the average of the DMSO controls. The normalized fluorescence was plotted against time. A one phase exponential decay was fitted to each plot (constraints: Y(0) > 0.8; 0 < plateau < 0.3; k > 0). Data from at least three independent assay replicates were used to generate the graphs in Fig. S5.

General Information
All chemicals were purchased from Sigma-Aldrich, Apollo Scientific, Acros Organics, Alfa Aesar and used without further purification unless otherwise indicated. All reactions were monitored by thin layer chromatography (TLC) using UV for visualisation unless otherwise stated. Compounds were purified using either an automated system using pre-packed silica cartridges with UV detection or by manual columns using an appropriate solvent mixture as detailed. 1

5-(4-methoxyphenyl)-3,4-dihydro-2H-pyrrole (2)
Ketone 1 (0.59 g, 2.0 mmol) was stirred in neat TFA (1.5 mL) at rt for 3.5 h. After the reaction was complete by TLC, the mixture was cooled to 0 °C and 50% w/v NaOH solution was added to the mixture until pH 13-14. The aqueous layer was extracted with CH2Cl2 (3x 25 mL), then the combined organic layers were dried over MgSO4, filtered, and concentrated in vacuo to afford the title compound 2 as a white crystalline solid which was used without further purification (0.34 g, 96%).

2-(4-methoxyphenyl)pyrrolidine (3)
To a stirring solution of imide 2 (0.30 g, 1.7 mmol) in MeOH/H2O 4:1 (2.0 mL) was added NaBH4 (78 mg, 2.0 mmol) and the reaction was stirred for 20 h at rt. Additional NaBH4 (20 mg, 0.8 mmol) was added and the reaction was stirred until completion as monitored by TLC. The reaction mixture was acidified with 1 M HCl to pH 1-3 and stirred for an additional 30 min, then 1 M NaOH was added until pH 13-15. The aqueous layer was extracted with CH2Cl2 (3x 10 mL). The combined organic layers were dried over MgSO4, filtered, and concentrated in vacuo to give the title compound 3 as a yellow oil (81 mg, 95%), which was used without further purification.