Drug blending as a mechanism to overcome drug resistance in cancer - PowerPoint PPT Presentation

Drug blending as a mechanism to overcome drug resistance in cancer therapy. Marc A. Marti-Renom Structural Genomics Group (ICREA, CNAG-CRG) http://marciuslab.org http://3DGenomes.org http://cnag.crg.eu

John Overtingon Bissan Al-Lazikani Francisco Martínez-Jiménez (Stratified Medical) (ICR) Francisco Martínez-Jiménez et al. (2017) Scientific Reports. Under revision.

Drug resistance is a major problem in cancer treatment Holohan, C.et al. Cancer drug resistance: an evolving paradigm. Nature Reviews. Cancer.

Mutations in drug targets is a high-frequent mechanism resistance

So… can we… Predict of the cancer-associated likelihood of a mutation? Predict the resistance-impact of the mutation?. Propose alternative treatment to the resistance?

Low-frequency mutations can drive drug resistance Schmitt, M., et al. (2015). The influence of subclonal resistance mutations on targeted cancer therapy. Nature Reviews. Clinical Oncology.

Cancer mutational landscape is complex and heterogeneous Alexandrov, L. B.et al. (2013). Signatures of mutational processes in human cancer. Nature, 500(7463), 415—21

Mutational signatures of ~30 types of cancer Alexandrov et al, Nature 500 , 415-421 (22 August 2013) doi:10:1038/nature12477 Alexandrov, L. B.et al. (2013). Signatures of mutational processes in human cancer. Nature, 500(7463), 415—21

Drug Blending concept Drug A Target A T T C C C A A T G C G T G G A G A C Gene T T T T T Cancer signature Phe Pro Met Arg Gly Asp Protein Ser Phe Cys Asp Amino acid changes Leu

Drug Blending concept Drug A Target A T T C C C A A T G C G T G G A G A C Gene T T T T T Cancer signature Phe Pro Met Arg Gly Asp Protein Amino acid changes Phe Ser Cys Asp Leu Increased sensitivity Neutral Resistance Strong resistance Drug B Drug C

Mutational probability in melanoma and colorectal cancer

ERK1/2 are promising targets for the treatment of melanoma and colon cancer Schmitt, M., et al. (2015). The influence of subclonal resistance mutations on targeted cancer therapy. Nature Reviews. Clinical Oncology.

Probability of spontaneous mutation of ERK2 VTX-11e binding-site PDB: 4QTE

Probability of spontaneous mutation of ERK2 VTX-11e binding-site melanoma colorectal cancer

ERK2 melanoma mutational landscape reveals a long-tailed distribution enriched in C>T

Colorectal cancer distribution results in 
 higher likelihood median values

what do we propose to overcome resistance? 1.Prediction of the cancer-associated likelihood. 2.Prediction of the resistance-impact. 3. Proposal of alternative non-resistant mutants

Predicting resistance using structural features and a Random Forest Classifier Increased sensitivity Neutral Resistance Strong resistance Drug B Drug C 3D-Model of mutated protein Structural Trained features of Random Forest wild type and Classifier with mutated Platinum model database Residue Mutation

Residue structural features Mol. Surf. Diff. Protein Stability Change Mol. Surf. Area MT Mol. Surf. Area WT Avg. Dist. Ligand MT Max. Dist. Ligand MT Min. Dist. Ligand MT Solvent RSA MT Solvent Acc. MT Solvent Acc. Diff Avg. Dist. Ligand WT Min. Dist. Ligand Diff. Max. Dist. Ligand WT Avg. Dist. Ligand Diff. Min. Dist. Ligand WT. RSA Diff. Solvent Acc. WT Half Sphere Exp. CA up Diff. Max. Dist. Ligand Diff. Kd nM WT SS MT Solvent RSA WT SS WT Half Sphere Exp. CA up MT Half Sphere Exp. CN MT

Platinum database http://bleoberis.bioc.cam.ac.uk/platinum/ Pires, D. E. V., Blundell, T. L., & Ascher, D. B. (2015). Nucleic Acids Research, 43(D1), D387—D391. Increased sensitivity Neutral Resistance Strong resistance Fold change Fold change Fold change Fold change > 1.2 <= 1.2 and > -1.2 <= -1.2 and >-5.0 <= -5.0 180 entries 71 entries 180 entries 180 entries

10-fold cross validation 1.0 0.9 0.8 0.7 True Positive Rate (TPR) 0.6 0.5 0.4 0.3 SRES RES 0.2 NEU 0.1 ISEN 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 False Positive Rate (FPR)

Residue structural features Mean dec. Gini Mol. Surf. Diff. Protein Stability Change Mol. Surf. Area MT Mol. Surf. Area WT Avg. Dist. Ligand MT Max. Dist. Ligand MT Min. Dist. Ligand MT Solvent RSA MT Solvent Acc. MT Solvent Acc. Diff Avg. Dist. Ligand WT Min. Dist. Ligand Diff. Max. Dist. Ligand WT Avg. Dist. Ligand Diff. Min. Dist. Ligand WT. RSA Diff. Solvent Acc. WT Half Sphere Exp. CA up Diff. Max. Dist. Ligand Diff. Kd nM WT SS MT Solvent RSA WT SS WT Half Sphere Exp. CA up MT Half Sphere Exp. CN MT 5 11 (%)

Resistance-like mutations in ERK2 for melanoma and colorectal cancer Normalized resistance score 0.0 1.0 1.0 0.8 Normalized resistance score 0.6 0.4 0.2 0.0 0.00 0.02 0.04 0.06 0.08 0.10 0.27 0.00 0.02 0.04 0.06 0.08 0.10 Predicted likelihood Predicted likelihood

Validating some predictions In-vitro identification of ERK2 VTX-11e mutants in A375 melanoma cell line P58S - Strong resistance - 0.39 Y64N - Resistance - 0.35 C65Y - Strong resistance - 0.29 G37S - Resistance - 0.42 P58L - Strong resistance - 0.40 P58T - Resistance - 0.34 Y36N - Strong resistance - 0.33 Y36H - Strong resistance - 0.32

what do we propose to overcome resistance? 1.Prediction of the cancer-associated likelihood. 2.Prediction of the resistance-impact. 3. Proposal of alternative non-resistant mutants

Predicted sensitivity map of ERK2 inhibitors to likely-and-resistant mutations

The control case, VTX11e, is predicted as sensitive to most of the screened mutations VTX11 e

del22379 seems to be unaffected by all of the screened mutations! del2237 9

The e75 compound shows a low resistant profile against most of the VTX11e resistant mutations e75

E7X series does not occupy the “resistant region” H147 D167 N154 L157 C65 Y36 D111 Y64 I56 G34 M108 P58 K54 G37 V39 I31

what do we propose to overcome resistance? 1.Prediction of the cancer-associated likelihood. 2.Prediction of the resistance-impact. 3. Proposal of alternative non-resistant mutants

Resistant mutants per tumor size (MEK1 + Selumetinib) All mutations Number of mutations with probability = 1.0 Resistance Mutations Tumor Size

All possible mutations will occur if a tumour is large enough A tumor comprising many cells can be compared to a natural population with many individuals. The amount of genetic di- versity re fm ects how it has evolved and can in fm uence its fu- ture evolution. We evaluated a single tumor by sequencing or genotyping nearly 300 regions from the tumor. When the data were analyzed by modern population genetic theory, we esti- mated more than 100 million coding region mutations in this un- exceptional tumor. The extreme genetic diversity implies evo- lution under the non-Darwinian mode. In contrast, under the prevailing view of Darwinian selection, the genetic diversity would be orders of magnitude lower. Because genetic diversity accrues rapidly, a high probability of drug resistance should be heeded, even in the treatment of microscopic tumors. Ling, S., et al. (2015). Proc Natl Acad Sci U S A, 112(47), E6496—505.

Take home messages We can use cancer signatures to predict the most likely mutations. However, we need to move towards “personalized” signatures. We can predict which of the likely mutations, are more prone to generate resistance to treatment. We can propose alternative/parallel treatments to overcome future resistance. All possible mutations will occur if a tumour is large enough


 Francisco Martínez-Jiménez Davide Baù Gireesh K. Bogu Yasmina Cuartero David Dufour Irene Farabella Silvia Galan Francesca di Giovanni Mike Goodstadt François Serra Paula Soler Yannick Spill Marco di Stefano Marie Trussart in collaboration with John Overington (Stratified Medical) & Bissan Al-Lazikani (ICR) http://marciuslab.org http://3DGenomes.org http://cnag.crg.eu