Jonathan Dauvé

Background/Objectives: Ovarian cancer is a complex disease with poor outcomes that affects women worldwide. The lack of successful therapeutic options for this malignancy has led to the need to identify novel biomarkers for patient stratification. Here, we aim to develop the outcome predictors based on the gene expression data as they may serve to identify categories of patients who are more likely to respond to certain therapies. Methods: We used The Cancer Genome Atlas (TCGA) ovarian cancer transcriptomic data from 372 patients and approximately 16,600 genes to train and evaluate the deep learning survival models. In addition, we collected an in-house validation dataset of 12 patients to assess the performance of the trained survival models for their direct use in clinical practice. Despite deceptive generalization capabilities, we demonstrated how our model can be interpreted to uncover biological processes associated with survival. We calculated the contributions of the input genes to the output of the best trained model and derived the corresponding molecular pathways. Results: These pathways allowed us to stratify the TCGA patients into high-risk and low-risk groups (p-value 0.025). We validated the stratification ability of the identified pathways on the in-house dataset consisting of 12 patients (p-value 0.229) and on the external clinical and molecular dataset consisting of 274 patients (p-value 0.006). Conclusions: The deep learning-based models for survival prediction with RNA-seq data could be used to detect and interpret the gene-sets associated with survival in ovarian cancer patients and open a new avenue for future research.

Lumican, a small leucine-rich proteoglycan (SLRP) of the extracellular matrix (ECM), displays anti-tumor properties through its direct interaction with MMP-14. Lumican-derived peptides, such as lumcorin (17 amino acids) or L9M (10 amino acids), are able to inhibit the proteolytic activity of MMP-14 and melanoma progression. This work aimed to visualize the interactions of lumican-derived peptides and MMP-14. Molecular modeling was used to characterize the interactions between lumican-derived peptides, such as lumcorin, L9M, and cyclic L9M (L9Mc, 12 amino acids), and MMP-14. The interaction of L9Mc with MMP-14 was preferential with the MT-Loop domain while lumcorin interacted more with the catalytic site. Key residues in the MMP-14 amino acid sequence were highlighted for the interaction between the inhibitory SLRP-derived peptides and MMP-14. In order to validate the in silico data, MMP-14 activity and migration assays were performed using murine B16F1 and human HT-144 melanoma cells. In contrast to the HT-144 melanoma cell line, L9Mc significantly inhibited the migration of B16F1 cells and the activity of MMP-14 but with less efficacy than lumican and lumcorin. L9Mc significantly inhibited the proliferation of B16F1 but not of HT-144 cells in vitro and primary melanoma tumor growth in vivo. Thus, the site of interaction between the domains of MMP-14 and lumcorin or L9Mc were different, which might explain the differences in the inhibitory effect of MMP-14 activity. Altogether, the biological assays validated the prediction of the in silico study. Possible and feasible improvements include molecular dynamics results.

Targeted therapies and, more precisely, EGFR tyrosine kinase inhibitors (TKIs) have been a major improvement in the therapeutic management of EGFR-mutated non-small-cell lung cancers (NSCLCs). Earlier administration of these TKIs throughout tumor progression is imperative to improve patient outcomes. Consequently, studies have focused on refining the characterization of biomarkers, especially concerning the resistance mutation p.Thr790Met of EGFR. Herein, we developed peptide nucleic acid (PNA)-mediated PCR clamping followed by pyrosequencing, favoring enrichment of the mutated fraction. A preamplification step was first added to increase the amplifiable DNA fraction. Throughout the application of our method on DNA extracted from FFPE samples of 46 patients with NSCLC who had relapsed under first-generation EGFR TKI, we evaluated a sensitivity of 93.3% and a specificity of 100%. All 19 patients who were positive for the p.Thr790Met mutation with NGS were also found to be positive with our protocol. The only discordant case was a sample with no mutation detected with NGS, but which was positive with PNA. This protocol allows for the detection of the p.Thr790Met mutation with a sensitivity of 0.5% which will permit earlier detection and an improvement of therapeutic management.

Background: Fluoropyrimidines (FPs) carry around 20% risk of G3-5 toxicity and 0.2-1% risk of death, due to dihydropyrimidine dehydrogenase (DPD) deficiency. Several screening approaches exist for predicting toxicity, however there is ongoing debate over which method is best. This study compares 4 screening approaches. Method: 472 patients treated for colorectal, head-and-neck, breast, or pancreatic cancers, who had not been tested pre-treatment for FP toxicity risk, were screened using: DPYD genotyping (G); phenotyping via plasma Uracil (U); phenotyping via plasma-dihydrouracil/uracil ratio (UH₂/U); and a Multi-Parametric Method (MPM) using genotype, phenotype, and epigenetic data. Performance was compared, particularly the inability to detect at-risk patients (false negatives). Results: False negative rates for detecting G5 toxicity risk were 51.2%, 19.5%, 9.8% and 2.4%, for G, U, UH₂/U and MPM, respectively. False negative rates for detecting G4-5 toxicity risk were 59.8%, 36.1%, 21.3% and 4.7%, respectively. MPM demonstrated significantly (p < 0.001) better prediction performance. Conclusion: MPM is the most effective method for limiting G4-5 toxicity. Its systematic implementation is cost-effective and significantly improves the risk-benefit ratio of FP-treatment. The use of MPM, rather than G or U testing, would avoid nearly 8,000 FP-related deaths per year globally (500 in France), and spare hundreds of thousands from G4 toxicity.

In physiological conditions, self-DNA released by dying cells is not detected by intracellular DNA sensors. In chronic inflammatory disorders, unabated inflammation has been associated with a break in innate immune tolerance to self-DNA. However, extracellular DNA has to complex with DNA-binding molecules to gain access to intracellular DNA sensors. IL-26 is a member of the IL-10 cytokine family, overexpressed in numerous chronic inflammatory diseases, in which biological activity remains unclear. We demonstrate in this study that IL-26 binds to genomic DNA, mitochondrial DNA, and neutrophil extracellular traps, and shuttles them in the cytosol of human myeloid cells. As a consequence, IL-26 allows extracellular DNA to trigger proinflammatory cytokine secretion by monocytes, in a STING- and inflammasome-dependent manner. Supporting these biological properties, IL-10–based modeling predicts two DNA-binding domains, two amphipathic helices, and an in-plane membrane anchor in IL-26, which are structural features of cationic amphipathic cell-penetrating peptides. In line with these properties, patients with active autoantibody-associated vasculitis, a chronic relapsing autoimmune inflammatory disease associated with extensive cell death, exhibit high levels of both circulating IL-26 and IL-26–DNA complexes. Moreover, in patients with crescentic glomerulonephritis, IL-26 is expressed by renal arterial smooth muscle cells and deposits in necrotizing lesions. Accordingly, human primary smooth cells secrete IL-26 in response to proinflammatory cytokines. In conclusion, IL-26 is a unique cationic protein more similar to a soluble pattern recognition receptor than to conventional cytokines. IL-26 expressed in inflammatory lesions confers proinflammatory properties to DNA released by dying cells, setting up a positive amplification loop between extensive cell death and unabated inflammation.