Peptide stapling by late-stage Suzuki–Miyaura cross-coupling

The development of peptide stapling techniques to stabilise α-helical secondary structure motifs of peptides led to the design of modulators of protein–protein interactions, which had been considered undruggable for a long time. We disclose a novel approach towards peptide stapling utilising macrocyclisation by late-stage Suzuki–Miyaura cross-coupling of bromotryptophan-containing peptides of the catenin-binding domain of axin. Optimisation of the linker length in order to find a compromise between both sufficient linker rigidity and flexibility resulted in a peptide with an increased α-helicity and enhanced binding affinity to its native binding partner β-catenin. An increased proteolytic stability against proteinase K has been demonstrated.


RP-HPLC(-MS)
Analytical LC-MS was performed on an Agilent 6220 TOF-MS with a Dual ESI-source, 1200 HPLC system with autosampler, degasser, binary pump, column oven, diode array detector and a Hypersil Gold C18 column (1.9 µm, 50 × 2.1 mm) with a gradient (in 11 min from 0% B to 98% B, back to 0% B in 0.5 min, total run time 15 min) at a flow rate of 300 μL/min and column oven temperature of 40°C. HPLC solvent A consists of 94.9% water, 5% acetonitrile and 0.1% formic acid, solvent B of 5% water, 94.9% acetonitrile and 0.1% formic acid. ESI mass spectra were recorded after sample injection via 1200 HPLC system in extended dynamic range mode equipped with a Dual-ESI source, operating with a spray voltage of 2.5 kV. Nitrogen served both as the nebuliser gas and the dry gas. Nitrogen was generated by a nitrogen generator NGM 11.

ESI MS
Nano-ESI mass spectra were recorded using an Esquire 3000 ion trap mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany) equipped with a nano-ESI source. Samples were dissolved in acetonitrile and introduced by static nano-ESI using in-house pulled glass emitters. Nitrogen served both as nebuliser gas and dry gas. Nitrogen was generated by a Bruker nitrogen generator NGM 11. Helium served as cooling gas for the ion trap. The mass axis was externally calibrated with ESI-L Tuning Mix (Agilent Technologies, Santa Clara, CA, USA) as calibration standard. The spectra were recorded with the Bruker Daltonik esquireNT 5.2 esquireControl software by the accumulation and averaging of several single spectra.
DataAnalysis™ software 3.4 was used for processing the spectra.

High-resolution MS
ESI accurate mass measurements were performed using either an Agilent 6220 time-of-flight mass spectrometer (Agilent Technologies, Santa Clara, CA, USA) or a Q-IMS-TOF mass spectrometer Synapt G2Si (WatersGmbH, Manchester, UK).
Agilent 6220 TOF mass spectrometer ESI accurate mass measurements are acquired in extended dynamic range mode equipped with a Dual-ESI source, operating with a spray voltage of 2.5 kV. Nitrogen served both as the nebuliser gas and the dry gas. Nitrogen was generated by a nitrogen generator NGM 11.
Samples are introduced with a 1200 HPLC system consisting of an autosampler, degasser, binary pump, column oven and diode array detector (Agilent Technologies, Santa Clara, CA, USA) using a C18 Hypersil Gold column (length: 50 mm, diameter: 2.1 mm, particle size:

Automated column chromatography
Automated NP-column chromatography on silica gel was partly performed using a Büchi Reveleris X2 chromatography system with UV and ELSD detection.

GP1: General procedure for enzymatic bromination of tryptophan
Enzymatic bromination on a preparative scale was performed according to our previously reported procedure using halogenase-PrnF-RR-ADH combiCLEAs. [1] The biocatalyst was produced using lysed E. coli cells containing overexpressed tryptophan-7-halogenase RebH or tryptophan-6-halogenase Thal resulting from 1.5 L expression culture. After cell lysis via French-Press (three times), cell debris were spun down and flavin reductase PrnF (2.5 U mL −1 ) and alcohol dehydrogenase RR-ADH (1 U mL −1 ) were added to the supernatant. Finely   [2] Synthesis of N α -Fmoc-4-pinacolatoborono-L-phenylalanine The spectroscopic data are in accordance with literature. [3] N

Peptide synthesis
Peptide synthesis was performed on Rink Amide AM resin using Fmoc/t-Bu strategy. FITC-labelling: After N-terminal Fmoc-deprotection, the resin was swollen in DMF for 10 min and filtrated. Fmoc-β-alanine coupling was performed following the procedure listed above but using Fmoc-β-Ala-OH (5 equiv), HATU (5 equiv) and DIEA (10 equiv) in DMF. The suspension was filtrated and the procedure repeated once. The resin was washed with DMF (3 × 1 min), DCM (3 × 1 min) and DMF (3 × 1 min) followed by Fmoc-deprotection. Next, the resin was swollen in DMF for 10 min, filtrated and a solution of FITC (7 equiv) and DIEA (7 equiv) in DMF was added and the suspension shaken at room temperature for 18 h under light protection.
The suspension was filtrated, the resin washed with DMF (3 × 1 min), DCM (3 × 1 min), iPrOH (3 × 1 min) and MTBE (3 × 1 min) and dried in vacuo. FITC-labelled peptides should be stored under light protection. [7] Cleavage and purification: A degassed solution of TFA/TIS/H2O 95:2.5:2.5 including DTT (2.5% w/v) and phenol (2.5% w/v) was added to the resin and the resulting suspension incubated at room temperature for 120 min. From time to time, the suspension was mixed by S10 a gentle stream of argon. The suspension was filtrated and the procedure was repeated once.
The TFA solution obtained was added dropwise into cold MTBE and the resulting turbid suspension incubated at −20 °C for 60 min. Next, the precipitated peptide was spun down at 4 °C at 5000 rpm for 30 min, the supernatant was discarded and the residue dried by a stream of argon. The crude peptide was freeze dried and purified by preparative RP-HPLC (method A).

GP3: General procedure for on-resin Suzuki-Miyaura cross-coupling
Under Argon, a degassed mixture of DME/EtOH/H2O 9:9:2 (1 mL per 50 µmol peptide) was Synthesis"). [8] GP4: General procedure for on-resin ring-closing metathesis The resin was swollen in DCM (3 × 1 min) and DCE (3 × 1 min) and filtrated. Next, an 8.3 mM solution of Grubbs' 1st generation catalyst (0.5 equiv) in DCE was added and the suspension mixed by a gentle stream of argon for 120 min. The suspension was filtrated and the procedure repeated once. The resin was washed with DCE (3 × 1 min), DCM (3 × 1 min) and MTBE and dried in vacuo. A test cleavage was performed (see section "Peptide Synthesis") and analysed by HPLC-MS. If the conversion was incomplete, metathesis has been repeated as described above until a conversion >95% was observed. The peptide was cleaved from the resin and purified by RP-HPLC (protocols see section "Peptide Synthesis"). [7] S11 Example MALDI TOF MS of crude peptide P1b S12 Figure S1: MALDI TOF MS analysis of peptide P1b after on-resin SMC. A) Spectrum of the crude peptide after test cleavage indicating that no intermolecular dimerisation products were formed; B) Spectrum of the product fraction after purification by preparative RP-HPLC; C) Spectrum of the fraction containing the deboronated and dehalogenated species (M+H2).

SMC-stapled peptides
Scheme S2: General overview of the on-resin synthesis of peptides P1a and P1b.

S13
Scheme S3: General overview of the on-resin synthesis of peptides P2 to P5.

Peptide P1a
The synthesis was performed following the general procedures on a 50 µmol scale. Peptide   Figure S4: CD spectra of reference peptide aAxWt and stapled peptide P1b, c = 100 µM in H2O (left) and TFE/H2O 4:1 (right), respectively. As can be seen, recording CD spectra of the peptides in a mixture of TFE/H2O 4:1 induces a nice and typical curve progression of a α-helix for the wild type sequence of the CBD domain aAxWt. The helicity of stapled peptide P1b is also increased; however, compared to aAxWt the curve progression is significantly flatter indicating that the introduction of the cross-link forces the peptide in a more distorted structure.

β-Catenin expression and purification
For fluorescence polarisation assay β-Catenin was expressed and purified according to literature protocols. [9] The plasmid M57 pPET28a-TEV-full-length was transformed into

Proteinase K stability assays
The assay was performed using a modified protocol. [11] A 10 mM peptide stock solution was prepared in a mixture of Na2HPO4 buffer (100 mM, pH 7.4) and DMSO (9:1). The peptide stock solution (25 µL) was diluted with more Na2HPO4 buffer (450 µL) and thoroughly mixed.  Figure S8: Reaction monitoring via LC-MS at λ = 280 nm. A) Linear Peptide P6, samples were taken at t = 0, 10, 20, 30, 60 and 120 min. After 120 min, peptide P6 was degraded in three fragments with fragment Ac-ENPEWILDKHVQ-OH eluting at exactly the same retention time than peptide P6 (see mass spectra below at reaction times 0 and 120 min). For fragment assignments, see mass spectra below. B) SMC stapled peptide P5, samples were taken at t = 0, 10, 20, 30, 60 and 120 min. Compared to linear peptide P6, two of three cleaving sites are blocked.

LC-MS mass spectra
Peptide P6 at t = 0 min S39 Peptide P6 at t = 120 min Figure S9: ESI-MS spectra of the different fragments of peptide P6 digestion by proteinase K. As can be seen in the HPLC and mass spectra, after 120 min the linear peptide P6 is degraded by Proteinase K, which cleaves at three sites after Leu (fragment at tR = 6.5 min), Lys (fragment at tR = 5.1 min) and Gln (fragment at tR = 4.7 min): Ac-ENPEWILDKHVQRVM-NH2.

S40
Peptide 5 at t = 0 min Peptide 5 at t = 120 min Figure S10: ESI-MS spectra of the stapled peptide P5 and the only fragment resulting from digestion by Proteinase K. Compared to the linear analogue P6, two of three cleaving sites are blocked and Proteinase K is not able to hydrolyse peptide 5 after leucine and lysine anymore due to the macrocycle. Only cleavage after glutamine is still observed (fragment at tR = 5.3 min).

Density functional theory (DFT) calculations
Calculations have been performed at the density functional theory (DFT) level using Gaussian 09. [12] Standard geometry optimisation was performed using Pople's 6-31g  Table S2.

Molecular dynamics-based conformational analysis
Molecular dynamics (MD) simulations and analysis were performed within the Amber18/AmberTools 19 modelling suite. [13] All MD runs were done using the CUDA accelerated version of pmemd as implemented in Amber18. Quantum mechanical (QM) calculations for the derivation of molecular mechanics (MM) parameters were performed with the Orca QM program [14] and corresponding in-house scripts.
MM parameters for the modified amino acids (i.e., 7-subsituted tryptophan and modified lysine), where adapted from the ff14SB [15] force filed and from GAFF [16] for the linker. Standard amino acids where modeled with the ff14SB forcefield. Atomic partial charges of the new residues were obtained by following the standard RESP procedure for Amber force fileds, i.e., optimisation and derivation of electrostatic potential (ESP) at the HF/6-31G* level and following a similar procedure as in our earlier work. [17] The parameters for the dihedral angle between the indole and the linker where directly ported from our previous study. [17d] All peptides where capped at their N-and C-terminal ends with acetyl and N-methyl groups, respectively.
The same simulation protocol was applied for all peptides considered in this work. Initial structures of all the peptides were built in an arbitrary conformation and produced using a combination of tleap from the AmberTools19 and Avogadro 1.2. The structures then underwent a simulated annealing (SA) with random initial velocities for 1000 independent runs of 500 ns with a time step of 2 fs and a temperature ramp from 0K to 600K and back to 0K in the gas phase. The 1000 conformations from the SA were then energy minimized and clustered using the hierarchical agglomerative algorithm with an epsilon of 3.0 Å. The representative structures of the four most populated clusters were selected as the starting structures of the particular peptide. The four starting structures were then individually solvated in a truncated octahedron box of TIP4Pew [18] water with a buffer region of 15.0 Å using the tleap program of AmberTools19. The number of water molecules was further adjusted manually using in-house scripts to obtain the same number of atoms in all 4 boxes for each peptide. Since we analyzed three different peptides there was a total of 12 boxes. Each box was minimized and further heated up using the same strategy as in ref.
[17d]. The runs were continued in the NPT ensemble for a short conventional molecular dynamics (cMD) production simulation of 75 ns.
For each peptide, the last 50ns of each of the four independent cMDs were used to obtain parameters necessary for the corresponding accelerated molecular dynamics (aMD) simulations (i.e., the averaged total potential energy and the averaged dihedral potential energy). The structures from the last frames of each cMD were used as starting structures for the aMD runs which was performed for a period of 700 ns for every box. Parameters a1, a2, b1, and b2 of the boosting potential where taken as 3.5, 3.5, 0.175, and 0.175 kcal mol −1 , respectively (see ref. [20] for details and Table S3 for the boosting parameters used in this work). All aMD and cMD simulations were performed with a time step of 2 fs, periodic boundary S46 conditions and the obtained trajectories were saved every 20 ps. The SHAKE algorithm was applied to restrict all hydrogen containing bonds while long-range electrostatic interactions were calculated using the particle-mesh Ewald scheme as implemented in Amber18.
Temperature and pressure were controlled with Langevin dynamics (collision frequency of 2.0 ps −1 ) and isotropic position scaling, respectively.
Clustering and secondary structure analysis were performed on the last 300 ns of each aMD simulation using the cpptraj program of AmberTools19. Secondary structure analysis was additionally performed for the whole set of conformation in each structural cluster for each peptide. The average fraction of secondary structure for each amino acid were then summed up and normalize to yield the final representation as shown in Figure 3 of the main text.
Clustering was done using the hierarchical agglomerative algorithm based on root mean square deviation (RMSD) of the backbone atoms (CA, C, and N) of the peptides, completelinkage (using the maximum distance between members of two clusters) and requesting a total number of 20 clusters. Number of clusters was preferred over distance criterion due to the high flexibility of some of the peptides considered here.
RMSD of the representative structures of each cluster was calculated with respect to the active conformation of Axin's binding domain as found in the crystal structure with PDB ID 1QZ7 [19] based on backbone atoms and disregarding the first three and the last two residues of the sequence due to the rather random conformation that this part of the Axin binding domain adopts in the crystal structure.