Signaling pathways control cell-fate decisions that ultimately determine the behavior of cancer cells.Therefore, the dynamics of pathway activity may contain prognostically relevant information different from that contained in the static nature of other types of biomarkers. To investigate this hypothesis, we characterized the network that regulated stress signaling by the Jun N-terminal kinase (JNK) pathway in neuroblastoma cells. We generated an experimentally calibrated and validated computational model of this network and used the model to extract prognostic information from neuroblastoma patient-specific simulations of JNK activation. Switch-like JNK activation mediates cell death by apoptosis. An inability to initiate switch-like JNK activation in the simulations was significantly associated with poor overall survival for patients with neuroblastoma with or without MYCN amplification, indicating that patientspecific simulations of JNK activation could stratify patients. Furthermore, our analysis demonstrated that extracting information about a signaling pathway to develop a prognostically useful model requires understanding not only components and disease-associated changes in the abundance or activity of the components, but how those changes affect pathway dynamics.

SummaryAmino acid hydroxylation is a post-translational modification that regulates intra- and inter-molecular protein-protein interactions. The modifications are regulated by a family of 2-oxoglutarate- (2OG) dependent enzymes and, although the biochemistry is well understood, until now only a few substrates have been described for these enzymes. Using quantitative interaction proteomics, we screened for substrates of the proline hydroxylase PHD3 and the asparagine hydroxylase FIH, which regulate the HIF-mediated hypoxic response. We were able to identify hundreds of potential substrates. Enrichment analysis revealed that the potential substrates of both hydroxylases cluster in the same pathways but frequently modify different nodes of signaling networks. We confirm that two proteins identified in our screen, MAPK6 (Erk3) and RIPK4, are indeed hydroxylated in a FIH- or PHD3-dependent mechanism. We further determined that FIH-dependent hydroxylation regulates RIPK4-dependent Wnt signaling, and that PHD3-dependent hydroxylation of MAPK6 protects the protein from proteasomal degradation.

Cilia are highly conserved microtubule-based structures that perform a variety of sensory and motility functions during development and adult homeostasis. In humans, defects specifically affecting motile cilia lead to chronic airway infections, infertility and laterality defects in the genetically heterogeneous disorder Primary Ciliary Dyskinesia (PCD). Using the comparatively simple Drosophila system, in which mechanosensory neurons possess modified motile cilia, we employed a recently elucidated cilia transcriptional RFX-FOX code to identify novel PCD candidate genes. Here, we report characterization of CG31320/HEATR2, which plays a conserved critical role in forming the axonemal dynein arms required for ciliary motility in both flies and humans. Inner and outer arm dyneins are absent from axonemes of CG31320 mutant flies and from PCD individuals with a novel splice-acceptor HEATR2 mutation. Functional conservation of closely arranged RFX-FOX binding sites upstream of HEATR2 orthologues may drive higher cytoplasmic expression of HEATR2 during early motile ciliogenesis. Immunoprecipitation reveals HEATR2 interacts with DNAI2, but not HSP70 or HSP90, distinguishing it from the client/chaperone functions described for other cytoplasmic proteins required for dynein arm assembly such as DNAAF1-4. These data implicate CG31320/HEATR2 in a growing intracellular pre-assembly and transport network that is necessary to deliver functional dynein machinery to the ciliary compartment for integration into the motile axoneme.

BackgroundThe oncogenic receptor tyrosine kinase (RTK) ERBB2 is known to dimerize with other EGFR family members, particularly ERBB3, through which it potently activates PI3K signalling. Antibody-mediated inhibition of this ERBB2/ERBB3/PI3K axis has been a cornerstone of treatment for ERBB2-amplified breast cancer patients for two decades. However, the lack of response and the rapid onset of relapse in many patients now question the assumption that the ERBB2/ERBB3 heterodimer is the sole relevant effector target of these therapies.MethodsThrough a systematic protein-protein interaction screen, we have identified and validated alternative RTKs that interact with ERBB2. Using quantitative readouts of signalling pathway activation and cell proliferation, we have examined their influence upon the mechanism of trastuzumab- and pertuzumab-mediated inhibition of cell growth in ERBB2-amplified breast cancer cell lines and a patient-derived xenograft model.ResultsWe now demonstrate that inactivation of ERBB3/PI3K by these therapeutic antibodies is insufficient to inhibit the growth of ERBB2-amplified breast cancer cells. Instead, we show extensive promiscuity between ERBB2 and an array of RTKs from outside of the EGFR family. Paradoxically, pertuzumab also acts as an artificial ligand to promote ERBB2 activation and ERK signalling, through allosteric activation by a subset of these non-canonical RTKs. However, this unexpected activation mechanism also increases the sensitivity of the receptor network to the ERBB2 kinase inhibitor lapatinib, which in combination with pertuzumab, displays a synergistic effect in single-agent resistant cell lines and PDX models.ConclusionsThe interaction of ERBB2 with a number of non-canonical RTKs activates a compensatory signalling response following treatment with pertuzumab, although a counter-intuitive combination of ERBB2 antibody therapy and a kinase inhibitor can overcome this innate therapeutic resistance.Electronic supplementary materialThe online version of this article (10.1186/s13058-019-1127-y) contains supplementary material, which is available to authorized users.

Biological systems are known to be both robust and evolvable to internal and external perturbations, but what causes these apparently contradictory properties? We used Boolean network modeling and attractor landscape analysis to investigate the evolvability and robustness of the human signaling network. Our results show that the human signaling network can be divided into an evolvable core where perturbations change the attractor landscape in state space, and a robust neighbor where perturbations have no effect on the attractor landscape. Using chemical inhibition and overexpression of nodes, we validated that perturbations affect the evolvable core more strongly than the robust neighbor. We also found that the evolvable core has a distinct network structure, which is enriched in feedback loops, and features a higher degree of scale-freeness and longer path lengths connecting the nodes. In addition, the genes with high evolvability scores are associated with evolvability-related properties such as rapid evolvability, low species broadness, and immunity whereas the genes with high robustness scores are associated with robustness-related properties such as slow evolvability, high species broadness, and oncogenes. Intriguingly, US Food and Drug Administration-approved drug targets have high evolvability scores whereas experimental drug targets have high robustness scores.

Osteosarcoma (OS) is an aggressive sarcoma, where novel treatment approaches are required. Genomic studies suggest that a subset of OS, including OS tumour cell lines (TCLs), exhibit genomic loss of heterozygosity (LOH) patterns reminiscent of BRCA1 or BRCA2 mutant tumours. This raises the possibility that PARP inhibitors (PARPi), used to treat BRCA1/2 mutant cancers, could be used to target OS. Using high-throughput drug sensitivity screening we generated chemosensitivity profiles for 79 small molecule inhibitors, including three clinical PARPi. Drug screening was performed in 88 tumour cell lines, including 18 OS TCLs. This identified known sensitivity effects in OS TCLs, such as sensitivity to FGFR inhibitors. When compared to BRCA1/2 mutant TCLs, OS TCLs, with the exception of LM7, were PARPi resistant, including those with previously determined BRCAness LoH profiles. Post-screen validation experiments confirmed PARPi sensitivity in LM7 cells as well as a defect in the ability to form nuclear RAD51 foci in response to DNA damage. LM7 provides one OS model for the study of PARPi sensitivity through a potential defect in RAD51-mediated DNA repair. The drug sensitivity dataset we generated in 88 TCLs could also serve as a resource for the study of drug sensitivity effects in OS.

Migrating cells need to coordinate distinct leading and trailing edge dynamics but the underlying mechanisms are unclear. Here, we combine experiments and mathematical modeling to elaborate the minimal autonomous biochemical machinery necessary and sufficient for this dynamic coordination and cell movement. RhoA activates Rac1 via DIA and inhibits Rac1 via ROCK, while Rac1 inhibits RhoA through PAK. Our data suggest that in motile, polarized cells, RhoA-ROCK interactions prevail at the rear whereas RhoA-DIA interactions dominate at the front where Rac1/Rho oscillations drive protrusions and retractions. At the rear, high RhoA and low Rac1 activities are maintained until a wave of oscillatory GTPase activities from the cell front reaches the rear, inducing transient GTPase oscillations and RhoA activity spikes. After the rear retracts, the initial GTPase pattern resumes. Our findings show how periodic, propagating GTPase waves coordinate distinct GTPase patterns at the leading and trailing edge dynamics in moving cells.

Signalling pathways exert finely tuned control over cell fate decisions that ultimately determine the behaviour of cancer cells. It could therefore be expected that the dynamics of key pathway activation may contain prognostically relevant information over and above that which is contained in the static nature of traditional biomarkers. To investigate this hypothesis we have characterised the network architecture regulating JNK stress signalling in neuroblastoma cells and applied an experimentally calibrated and validated computational model of this network to extract prognostic information from neuroblastoma patient-specific simulations of JNK activation. Survival analysis based upon the dynamics of these simulations revealed that an inability to initiate switch-like JNK signalling in silico was significantly associated with poor overall survival for both MYCN amplified and non-amplified neuroblastoma patients. Furthermore, our analysis demonstrated that in order to extract prognostic information from a signalling pathway, deciphering the extant network structure is a vital consideration in model development. We also show that network based analysis can lead to the discovery of new therapeutic targets. Citation Format: Walter Kolch, Dirk Fey, Melinda Halasz, Nora Rauch, Amaya Garcia Munoz, Ruth Pilkington, Boris N. Kholodenko, David R. Croucher, Sean P. Kennedy, Jordan F. Hastings, Frank Westermann, Daniel Dreidax, Matthias Fischer, David Duffy, Aleksandar Krstic, Thomas Schwarzl. Personalized cancer diagnostics and therapeutics based on the computational modeling of signal transduction networks. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr CN05-03.