Oncogenic stimuli trigger the DNA damage response (DDR) and induction of the alternative reading frame (ARF) tumor suppressor, both of which can activate the p53 pathway and provide intrinsic barriers to tumor progression. However, the respective timeframes and signal thresholds for ARF induction and DDR activation during tumorigenesis remain elusive. Here, these issues were addressed by analyses of mouse models of urinary bladder, colon, pancreatic and skin premalignant and malignant lesions. Consistently, ARF expression occurred at a later stage of tumor progression than activation of the DDR or p16INK4A , a tumor-suppressor gene overlapping with ARF. Analogous results were obtained in several human clinical settings, including early and progressive lesions of the urinary bladder, head and neck, skin and pancreas. Mechanistic analyses of epithelial and fibroblast cell models exposed to various oncogenes showed that the delayed upregulation of ARF reflected a requirement for a higher, transcriptionally based threshold of oncogenic stress, elicited by at least two oncogenic 'hits', compared with lower activation threshold for DDR. We propose that relative to DDR activation, ARF provides a complementary and delayed barrier to tumor development, responding to more robust stimuli of escalating oncogenic overload.

ZBTB7A (Pokemon) is a member of the POK family of transcriptional repressors. Its main function is the suppression of the p14ARF tumour suppressor gene. Although ZBTB7A expression has been found to be increased in various types of lymphoma, there are no reports dealing with its expression in solid tumours. Given that p14(ARF) inhibits MDM2, the main negative regulator of p53, we hypothesized that overexpression of ZBTB7A could lead indirectly to p53 inactivation. To this end, we examined the status of ZBTB7A and its relationship with tumour kinetics (proliferation and apoptosis) and nodal members of the p53 network in a panel of 83 non-small cell lung carcinomas (NSCLCs). We observed, in the majority of the samples, prominent expression of ZBTB7A in the cancerous areas compared to negligible presence in the adjacent normal tissue elements. Gene amplification (two- to five-fold) was found in 27.7% of the cases, denoting its significance as a mechanism driving ZBTB7A overproduction in NSCLCs. In the remaining non-amplified group of carcinomas, analysis of the mRNA and protein expression patterns suggested that deregulation at the transcriptional and post-translational level accounts for ZBTB7A overexpression. Proliferation was associated with ZBTB7A expression (p = 0.033) but not apoptosis. The association with proliferation was reflected in the positive correlation between ZBTB7A expression and tumour size (p = 0.018). The overexpression of ZBTB7A in both p53 mutant and p53 wild-type cases, implies either a synergistic effect or that ZBTB7A exerts its oncogenic properties independently of the p14(ARF)-MDM2-p53 axis. The concomitant expression of ZBTB7A with p14(ARF) (p = 0.039), instead of the anticipated inverse relation, supports the latter notion. In conclusion, regardless of the pathway followed, the distinct expression of ZBTB7A in cancerous areas and the association with proliferation and tumour size pinpoints a role for this novel cell cycle regulator in the pathogenesis of lung cancer.

Centrosome abnormalities are observed in human cancers and have been associated with aneuploidy, a driving force in tumour progression. However, the exact pathways that tend to cause centrosome abnormalities have not been fully elucidated in human tumours. Using a series of 68 non-small-cell lung carcinomas and an array of in vitro experiments, the relationship between centrosome abnormalities, aneuploidy, and the status of key G1 to S-phase transition cell-cycle molecules, involved in the regulation of centrosome duplication, was investigated. Centrosome amplification and structural abnormalities were common (53%), were strongly related to aneuploidy, and, surprisingly, were even seen in adjacent hyperplastic regions, suggesting the possibility that these are early lesions in lung carcinogenesis. Cyclin E and E2F1 overexpression, but not p53 mutation, was observed to correlate with centrosome abnormalities in vivo (p = 0.029 and p = 0.015, respectively). This was further strengthened by the observation that cyclin E was specifically present in the nucleus and/or cytoplasm of the cells that contained centrosome aberrations. The cytoplasmic cyclin E signal may be attributed, in part, to the presence of truncated low-molecular-weight isoforms of cyclin E. In order to isolate the effect of cyclin E on the appearance of centrosome abnormalities, a U2OS tetracycline-repressible cyclin E cell line that has a normal centrosome profile by default was used. With this system, it was confirmed in vitro that persistent cyclin E overexpression is sufficient to cause the appearance of centrosome abnormalities.

A major challenge in cancer treatment is predicting the clinical response to anti-cancer drugs for each individual patient. For complex diseases such as cancer, characterized by high inter-patient variance, the implementation of precision medicine approaches is dependent upon understanding the pathological processes at the molecular level. While the "omics" era provides unique opportunities to dissect the molecular features of diseases, the ability to utilize it in targeted therapeutic efforts is hindered by both the massive size and diverse nature of the "omics" data.Recent advances with Deep Learning Neural Networks (DLNNs), suggests that DLNN could be trained on large data sets to efficiently predict therapeutic responses in cancer treatment. We present the application of Association Rule Mining combined with DLNNs for the analysis of high-throughput molecular profiles of 1001 cancer cell lines, in order to extract cancer-specific signatures in the form of easily interpretable rules and use these rules as input to predict pharmacological responses to a large number of anti-cancer drugs. The proposed algorithm outperformed Random Forests (RF) and Bayesian Multitask Multiple Kernel Learning (BMMKL) classification which currently represent the state-of-the-art in drug-response prediction. Moreover, the in silico pipeline presented, introduces a novel strategy for identifying potential therapeutic targets, as well as possible drug combinations with high therapeutic potential. For the first time, we demonstrate that DLNNs trained on a large pharmacogenomics data-set can effectively predict the therapeutic response of specific drugs in different cancer types. These findings serve as a proof of concept for the application of DLNNs to predict therapeutic responsiveness, a milestone in precision medicine.

This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 2.0 France Licence.