Specific inhibition of the Survivin–CRM1 interaction by peptide-modified molecular tweezers

Survivin’s dual function as apoptosis inhibitor and regulator of cell proliferation is mediated via its interaction with the export receptor CRM1. This protein–protein interaction represents an attractive target in cancer research and therapy. Here, we report a sophisticated strategy addressing Survivin’s nuclear export signal (NES), the binding site of CRM1, with advanced supramolecular tweezers for lysine and arginine. These were covalently connected to small peptides resembling the natural, self-complementary dimer interface which largely overlaps with the NES. Several biochemical methods demonstrated sequence-selective NES recognition and interference with the critical receptor interaction. These data were strongly supported by molecular dynamics simulations and multiscale computational studies. Rational design of lysine tweezers equipped with a peptidic recognition element thus allowed to address a previously unapproachable protein surface area. As an experimental proof-of-principle for specific transport signal interference, this concept should be transferable to any protein epitope with a flanking well-accessible lysine.


HRMS; m/z [M-H]
Dips in flexibility in the protein core (residues 6-90) are consistent with the known secondary structure of BIR domains 50 , where five α-helices (white rectangles α1-5), a three-stranded antiparallel β-sheet (white rectangles β1-3) and a zinc finger (residues marked by grey triangles) contribute to the stability of the tertiary structure. Survivin extends the BIR domain with a Cterminal tail featuring a two-stranded intermolecular antiparallel β-sheet (β4) and an α-helix (α6).  binding was analyzed via NMR spectroscopy. Spectra were compared to that of Survivin120 wildtype. Correct folding was examined especially in the region of 6-10 ppm resulting from amides and aromatics and below 1 ppm obtained from methyl groups (highlighted in red).
Mutant K90/103T was stable, folded correctly and hence chosen for further analysis.

Supplementary Method SI10: Additional information, computational details
For the molecular dynamics (MD) simulations, the parameters for the unmodified tweezer TW were generated with the Swissparam server. 1 These parameters have been validated in previous works 2, 3,4,5 involving TW (also known as CLR01) and several biomolecules (see more details below). For the modified tweezers TW-ELTL and TW-ELTLGEFL, we applied the parameters established for TW in combination with the peptide fragments for which we used the standard protein force field parameters of CHARMM36m. 6 The parameters for the union between the peptide motifs and the tweezer were also obtained using the Swissparam server. which do not distribute as those that can be sampled under a temperature regime and therefore a Boltzmann-like interpretation of the relative QM energy differences is not possible. We have observed in previous works 2, 3, 4 that relative QM energy differences span over a wide range that reaches tens of kcal/mol. Even so, the relative QM energy values allow assessing the relative stability, at electronic resolution, of an atomic configuration with respect to other.
As in previously reported works, 2, 3, 4 here our aim was the optimization of the geometry of tweezers complexes at the QM/MM level, e.g. using DFT to describe the whole tweezer structure (and part of the included amino acid), with an electrostatic embedding for the QM-MM treatment.
The suitability of using the above-mentioned set of parameters for generating geometries for QM/MM optimizations is evidenced by our previous work on complexes of several tweezers with amino acids and short peptides. 2 In addition, QM/MM calculations in which the initial coordinates were obtained from MD simulations using the SwissParam parameters allowed us to successfully predict (with experimental agreement) binding sites in systems as diverse as 14-3-3 proteins, p97-Nterm, the N-term of a fragment of the first exon of a Huntingtin protein, and beta amyloid peptides, among others. 2, 4, 8,9 In previous published works, we also evaluated how well this set of parameters allows sampling the tweezer forming inclusion complexes in MD simulations. 4, 9, 10, 11 The average values of structural parameters measured from simulations of TW-Lys complexes are in agreement with a crystal structure reported for a TW-Lys complex in a 14-3-3 protein (PDB ID 5OEH). 2 This agreement is also observed in the current work (see SI13). Using these parameters, we previously predicted that the formation of inclusion complexes of the tweezer with N17 (17-residue N-terminal fragment of the exon-1 domain of the Huntingtin protein) would result in the decrease of the α-helical content and amphipathic nature of N17, in agreement with REMD simulations (with Generalized Amber force field parameters for the tweezer) and experimental work. 8 Very recently, the SwissParam parameters were also used by us for simulating the formation of inclusion complexes of tweezers with model lipid bilayers and the effect of such tweezers on the membrane's structural properties, in excellent agreement with NMR measurements and biophysical experiments. 12 Nevertheless, we note that, although suitable for MD simulations and allowing a correct sampling of the geometry of tweezer complexes, such parameters should be used with caution as they may not be suitable for delivering properties such as accurate binding free energy estimations. Thus, we recently used CHARMM General Force Field (CGenFF) parameters to calculate the binding affinity in tweezer-Lys and tweezer-Arg complexes. 13 For selection of the scrambled peptide, all possible permutations (3359) of the original peptide (ELTLGEFL) were generated. The FoldX program 14, 15 was used to evaluate the relative stability of the complexes between Survivin120 and these peptides bound to the NES region. The complexes of Survivin120 with the scrambled peptides were generated using as template a representative structure of the complex Survivin120 -TW-ELTLGEFL from our simulations. The scrambled peptide LFEEGLLT followed two criteria: 1) the leucine spacing that is part of the NES consensus sequence was disrupted. 2) it does not feature reverse sequences which would simply anchor the lysine from the other side.  indicates how deep the respective lysine is threaded inside the tweezer's cavity. d(P1-N) is the distance from the phosphate group bound to C1 and the nitrogen atom of the indicated lysine.

Supplementary
The experimental values are taken from the crystal structure with PDB-ID 5OEH [https://www.rcsb.org/structure/5OEH] of a TW-lysine complex in a 14-3-3 protein.