Periodic Reporting for period 1 - SAMNets (Investigation of adaptive design and rewiring of Survival-Apoptosis-Mitogenic (SAM) signalling transduction network)
Période du rapport: 2017-05-01 au 2019-04-30
Driver mutations in survival, apoptotic and mitogenic signalling pathways are found in majority of human cancers. Increasing hope rests on targeted cancer therapies with kinase inhibitors, which inhibit mutated or overexpressed signalling proteins. Unfortunately, innate or acquired resistance to kinase inhibitors in cancer remains a pressing problem.
For example, although RAF inhibitors are known to be effective in treatment of metastatic melanoma with BRAFV600E mutation and wild-type RAS genes, resistance to RAF inhibitors inevitably occurs (>80% of responders) within 6-8 months. Multiple mechanisms can lead to resistance, such as oncogenic RAS mutations, upregulation of upstream receptors, RAF overexpression or amplification and BRAFV600E splice variants with the enhanced dimerization potential. This resistance cannot be overcome by existing RAF inhibitors.
A way to overcome resistance to kinase inhibitors can be the use of inhibitor combinations, but it is unclear how the best combinations can be chosen. A plethora of confounding factors, including allosteric drug–kinase interactions, phosphorylation-induced conformational changes and kinase dimerization, multiple feedback loops and different cell mutational and expression profiles hamper the intuitive reasoning on the choice of optimal drug combinations. Understanding the drugs’ mode of action and the mode of their combined actions at the network level would enable a systematic and robust design of the best combinations.
The overall objective of the project was developing data-driven computational models of signalling pathways beyond current state of the art to systematically and objectively identify best inhibitor combinations depending on mutational and expression background. The developed models revealed a new principle of kinase inhibition: inhibition of the same enzyme or closely related enzymes of the same family with two structurally different inhibitors that bind to both protomers in the asymmetric homo- and hetero dimers. The subsequent experiments on cancer cell lines corroborated model predictions.
For example, although RAF inhibitors are known to be effective in treatment of metastatic melanoma with BRAFV600E mutation and wild-type RAS genes, resistance to RAF inhibitors inevitably occurs (>80% of responders) within 6-8 months. Multiple mechanisms can lead to resistance, such as oncogenic RAS mutations, upregulation of upstream receptors, RAF overexpression or amplification and BRAFV600E splice variants with the enhanced dimerization potential. This resistance cannot be overcome by existing RAF inhibitors.
A way to overcome resistance to kinase inhibitors can be the use of inhibitor combinations, but it is unclear how the best combinations can be chosen. A plethora of confounding factors, including allosteric drug–kinase interactions, phosphorylation-induced conformational changes and kinase dimerization, multiple feedback loops and different cell mutational and expression profiles hamper the intuitive reasoning on the choice of optimal drug combinations. Understanding the drugs’ mode of action and the mode of their combined actions at the network level would enable a systematic and robust design of the best combinations.
The overall objective of the project was developing data-driven computational models of signalling pathways beyond current state of the art to systematically and objectively identify best inhibitor combinations depending on mutational and expression background. The developed models revealed a new principle of kinase inhibition: inhibition of the same enzyme or closely related enzymes of the same family with two structurally different inhibitors that bind to both protomers in the asymmetric homo- and hetero dimers. The subsequent experiments on cancer cell lines corroborated model predictions.
Objective 1. Reconstruct the network topology and connection strengths in the Survival-Apoptotic-Mitogenic (SAM) network. The network reconstruction method called Modular Response Analysis (MRA) was improved in two ways. First, it was reformulated to take into account effects of protein sequestration, which are inevitable in enzyme regulatory networks. Second, the robustness of MRA was tested with respect noise in input data and the optimal strategy of analysis of noisy data was suggested. Third, based on proteomics data obtained from collaborators, we assessed how mutant oncogenic RAS rewires intracellular signalling. The developed model of network rewiring was in agreement with available experimental data.
Objective 2. Develop site-specific dynamical models of SAM network. To achieve objective 2 we have developed a number of structure-based ODE models using rule-based modelling technique that predict whole-network responses to different types of kinase inhibitors. Kinase inhibitors were classified according to the positions of the DFG motif (DFG-IN/OUT) and a structurally conserved alpha C helix (αC-IN/OUT). The model of MAPK pathway showed that combinations of two RAF inhibitors of different types can synergistically inhibit signalling by RAF family kinases, while each inhibitor on its own is ineffective, especially in RAS-mutant tumours. The computational model of ErbB-related signalling predicted that combinations of two ErbB inhibitors of different types can synergistically inhibit phosphorylation and downstream signalling of ErbB family tyrosine kinase receptors. The computational model of JAK-related signalling suggested that combinations of two types of JAK inhibitors will show synergy in inhibition of the JAK-STAT signalling pathway in cells resistant to JAK inhibitor therapy. In all 3 cases exact types of inhibitors producing synergism were predicted by a mathematical model and depended on mutational and expression background.
Objective 3. Test the predictions of mathematical modelling of SAM network. To achieve objective 3 we performed a number of signalling, proliferation and colony formation experiments to prove the existence of drug synergy effects predicted by computational models. The experiments in RAS and RAF mutant melanoma cell lines, HER2-positive breast cancer cell lines and T-ALL cell line corroborated model predictions on drug synergy between different types of RAF, ErbB and JAK inhibitors. Thus, all key model predictions were validated experimentally.
Based on achieved results, 4 original research papers and 2 reviews were published, and 1 patent application was filed.
Objective 2. Develop site-specific dynamical models of SAM network. To achieve objective 2 we have developed a number of structure-based ODE models using rule-based modelling technique that predict whole-network responses to different types of kinase inhibitors. Kinase inhibitors were classified according to the positions of the DFG motif (DFG-IN/OUT) and a structurally conserved alpha C helix (αC-IN/OUT). The model of MAPK pathway showed that combinations of two RAF inhibitors of different types can synergistically inhibit signalling by RAF family kinases, while each inhibitor on its own is ineffective, especially in RAS-mutant tumours. The computational model of ErbB-related signalling predicted that combinations of two ErbB inhibitors of different types can synergistically inhibit phosphorylation and downstream signalling of ErbB family tyrosine kinase receptors. The computational model of JAK-related signalling suggested that combinations of two types of JAK inhibitors will show synergy in inhibition of the JAK-STAT signalling pathway in cells resistant to JAK inhibitor therapy. In all 3 cases exact types of inhibitors producing synergism were predicted by a mathematical model and depended on mutational and expression background.
Objective 3. Test the predictions of mathematical modelling of SAM network. To achieve objective 3 we performed a number of signalling, proliferation and colony formation experiments to prove the existence of drug synergy effects predicted by computational models. The experiments in RAS and RAF mutant melanoma cell lines, HER2-positive breast cancer cell lines and T-ALL cell line corroborated model predictions on drug synergy between different types of RAF, ErbB and JAK inhibitors. Thus, all key model predictions were validated experimentally.
Based on achieved results, 4 original research papers and 2 reviews were published, and 1 patent application was filed.
We have invented a novel principle of targeted drug therapy: inhibition of the same target enzyme or two closely related enzymes within the same family undergoing dimerization or oligomerization during activation with two structurally different inhibitors. Previously inhibition of one target enzyme with more than one small-molecule inhibitor was not considered as effective strategy. The synergistic combinations that we have found were never tested before, neither in vitro nor in vivo. Our approach opened the principal possibility to treat cancers in situations that are currently considered to be “undruggable”, going beyond state of the art in this field.
RAF and MEK inhibitors are known to be ineffective in treatment of RAS-mutant cancers. In general, there are no targeted inhibitors that are effective in the treatment of RAS-mutant cancers. Approximately 30% of all human cancers have RAS mutations. We showed that combination of two RAF inhibitors can effectively inhibit oncogenic signalling downstream of oncogenic RAS and suppress proliferation of RAS-mutant cells.
ErbB inhibitors are known to be only partially effective in treatment of cancers with mutations or over-expression of ErbB family receptor tyrosine kinases. We suggested combinations of two ErbB famlily inhibitors that target ErbB receptors in different conformations. These combinations were predicted to be synergistic by computational modelling. Synergy effects were corroborated experimentally. These synergy effects result in more robust inhibition of signalling pathways downstream of ErbB receptors.
JAK-STAT pathway hyper-activation is known to be driver for a number of leukaemias. Although sh-RNA experiments clearly show dependence of appropriate cell lines on JAK-STAT signalling, JAK inhibitor therapy fail to treat these disorders. Using computational modelling, we predicted that combinations of two structurally different JAK inhibitors can overcome resistance to JAK inhibitor therapy and robustly inhibit JAK-STAT signalling pathway. Our subsequent experiments corroborated model predictions.
Thus, in current project we have developed mathematical models to systematically and objectively predict synergy effects described above. Most of drug combinations that were tested consisted of already approved drugs, which implies minimal amount of additional animal experiments and necessary approvals to translate these results into clinical practice.
The developed models went beyond current state of the art as they combined rule based modelling and dynamic, mechanistic structure-based modelling with thermodynamic principles.
RAF and MEK inhibitors are known to be ineffective in treatment of RAS-mutant cancers. In general, there are no targeted inhibitors that are effective in the treatment of RAS-mutant cancers. Approximately 30% of all human cancers have RAS mutations. We showed that combination of two RAF inhibitors can effectively inhibit oncogenic signalling downstream of oncogenic RAS and suppress proliferation of RAS-mutant cells.
ErbB inhibitors are known to be only partially effective in treatment of cancers with mutations or over-expression of ErbB family receptor tyrosine kinases. We suggested combinations of two ErbB famlily inhibitors that target ErbB receptors in different conformations. These combinations were predicted to be synergistic by computational modelling. Synergy effects were corroborated experimentally. These synergy effects result in more robust inhibition of signalling pathways downstream of ErbB receptors.
JAK-STAT pathway hyper-activation is known to be driver for a number of leukaemias. Although sh-RNA experiments clearly show dependence of appropriate cell lines on JAK-STAT signalling, JAK inhibitor therapy fail to treat these disorders. Using computational modelling, we predicted that combinations of two structurally different JAK inhibitors can overcome resistance to JAK inhibitor therapy and robustly inhibit JAK-STAT signalling pathway. Our subsequent experiments corroborated model predictions.
Thus, in current project we have developed mathematical models to systematically and objectively predict synergy effects described above. Most of drug combinations that were tested consisted of already approved drugs, which implies minimal amount of additional animal experiments and necessary approvals to translate these results into clinical practice.
The developed models went beyond current state of the art as they combined rule based modelling and dynamic, mechanistic structure-based modelling with thermodynamic principles.