Colin Am Semple

Pleural mesothelioma (PM) is a tumour type showing high structural complexity but the full spectrum of variants present and their impacts are poorly understood. We comprehensively describe the mutational landscape of mesothelioma, exploiting a large (N=443) cohort profiling the tumour genome and transcriptome in parallel. We find that the majority of candidate driver variants are mediated by copy number alterations (CNAs) and discover high frequencies of complex rearrangements across most tumours, dominated by chromothripsis (74% of tumours) and chromoplexy (59%). Chromothripsis contributes to significant CNA hotspots to delete known PM tumour suppressor genes such as CDKN2A, NF2 and BAP1 , thought to be among the earliest events in PM development; and the presence of chromothripsis is associated with significantly worse patient survival. We also discover multi-faceted crosstalk between the mitochondrial and nuclear genomes during PM evolution, involving alterations to both genomes, and mitochondrial expression changes associated with subtypes and chromothripsis.

More than twenty years ago, the FANTOM consortium uncovered that mammalian genomes are pervasively transcribed, revealing multitudes of RNAs with unknown functions. A subset of these transcripts has since then been linked to transcriptional control and to chromatin organization via their ability to interact with DNA, suggesting that chromatin-associated RNAs could be key players in genome regulation. Although recent technological advances now enable the mapping of genome-wide RNA-DNA contacts, a lack of analyses integrating these methods with other genomic features and across multiple cellular contexts hinders our comprehensive understanding of the principles underlying RNA-DNA interactions and of their biological importance. As part of the FANTOM6 project, we thus generated RNA-DNA interaction maps in 16 different human cell types, then combined these contacts with multiple layers of other genomic data to investigate how patterns of interaction between RNA and DNA relate to chromatin organization and function. We show that the RNA-DNA interactome is highly dynamic yet reproducibly organized in cell-type specific networks, constituted of a great diversity of interactions that vary in function of their distance, the nature of their sources and the chromatin state of their targets. In particular, we detected numerous regulatory elements that exhibit marked changes in activity when differentially bound by transcripts, implying that thousands of RNA-DNA interactions can play a mechanistic role in gene expression. This regulatory function correlates with RNA-protein interactions and significantly associates with cell type-relevant and disorder-related traits. In addition to providing essential resources for future research in RNA-mediated chromatin regulation, cellular biology and human diseases, our study thus establishes the RNA-DNA interactome as a new genome regulatory layer that defines and maintains cellular identity and behavior.

Background & Aims: Haematoxylin and eosin (H&E) staining remains the diagnostic gold standard for solid cancers, including hepatocellular carcinoma, and is increasingly complemented by genomic profiling for precision medicine. Inferring genomic alterations directly from H&E images could streamline testing, but heterogeneity and biases in human training data limit interpretation of genotype-phenotype associations. Here, we aimed to relate histologic to genomic pathology to provide biological explainability for mutation prediction models and assess the impact of germline variation on model performance. Methods: We analysed 597 murine liver tumours with matched whole-genome sequencing and histopathology (163,835 image tiles; 22.9 million nuclei). Our controlled in vivo design accounted for germline variation, biological sex, and causal mutagen (N-diethylnitrosamine), removing confounding factors present in human cohorts. We trained and evaluated deep learning and supervised machine learning models to predict germline variation and cancer driver alterations from H&E. Results: Modelling accurately predicted germline and somatic alterations from histology, at both locus-specific and genome-wide scales. Quantitative image analysis revealed an unexpected association between Egfr driver mutations and hepatic steatosis, linking genotype to an interpretable morphological phenotype. While model performance declined when applied to tumours from unrepresented genetic backgrounds, this limitation was biologically informative, revealing strain-dependent differences in tumour evolution, notably the prevalence of whole-genome duplication. Conclusions: Machine learning integration of histological and genomic pathology enables accurate, interpretable inference of genetic alterations from H&E, potentially reducing reliance on costly ancillary molecular assays. Our predictions are supported by human-interpretable biological features, addressing concerns around “black-box” technologies. However, caution is required when applying such methods to samples with a genetic background that, even if closely related, is beyond the genetic horizon of training data.

Relationship between subtyping methodologies. A, Comparison of transcriptomic subgrouping approaches: composition of Tothill subtypes across each of the TCGA subtypes; labeled P value represents comparison of Tothill subtype frequency across all TCGA subtypes by the Chi-squared test. B, Distribution of homologous recombination repair (HRR)-centric subtypes across each of the TCGA transcriptomic subtypes; labeled P value represents comparison of BRCA1/2m frequency across all groups by Chi-squared test; Bonferroni-adjusted P = 0.009. C, Distribution of HRR-centric subtypes across each of the Tothill transcriptomic subtypes; labeled P value represents comparison of BRCA1/2m frequency across all groups by the Chi-squared test; Bonferroni-adjusted P = 0.003. BRCA2m, BRCA2 mutant; BRCA1m, BRCA1 mutant; EMSY-overxp; overexpression of EMSY; nBRCA-HRRm, non-BRCA1/2 HRR gene mutation; CCNE1g, gain of CCNE1; HRRwt, non-CCNE1g HRR wild-type.

PTEN and RB loss in HGSOC. A, Frequency of loss of PTEN protein expression across HRR-centric subtypes. B, Frequency of loss of RB protein expression across HRR-centric subtypes. C, Impact of RB loss on survival in patients based on HRR status. Multivariable hazard ratio (mHR) for HRR-ab: RB loss vs. HRR-ab RB-intact = 0.50; 95% CI, 0.30–0.84; mHR for HRRwt/CCNE1g: RB loss vs. HRRwt/CCNE1g: RB-intact = 0.71; 95% CI, 0.48–1.06. BRCA2m, BRCA2 mutant; BRCA1m, BRCA1 mutant; EMSY-overxp; overexpression of EMSY; nBRCA-HRRm, non-BRCA1/2 HRR gene mutation; CCNE1g, gain of CCNE1; HRRwt, non-CCNE1g HRR wild-type. HRR-ab, HRR-aberrant: BRCA1m, BRCA2m, EMSY-overxp, or nBRCA-HRRm.

Clinicopathologic features of patients with HGSOC.

Molecular landscape of HGSOC. 1Mutation in non-BRCA1/2 HRR genes: 3 BRIP1, 2 CHEK2, 1 RAD51C, 1 PALB2, 1 concurrent BAP1 and NBN. CCNE1 CN gain, ≥4 copies by TaqMan CN assay. EMSY CN amplification, ≥6 copies by TaqMan CN assay.

HRR pathway (HRR)–centric subtyping of high-grade serous carcinoma. A, HRR-centric classification taxonomy. B, OS profile of HRR-centric subtypes. C, Chemosensitivity of HRR-centric subtypes at first-line treatment (left) and treatment for disease relapse (right) as determined by CA125 tumor marker (top) and radiology (bottom). BRCA2m, BRCA2 mutant; BRCA1m, BRCA1 mutant; EMSY-overxp; overexpression of EMSY; CCNE1g, gain of CCNE1; HRRwt, non-CCNE1g HRR wild-type.

Tumor-infiltrating immune cells across high-grade serous carcinoma subtypes. A, CD3+ infiltration across HRR-centric subtypes; labeled P value represents comparison of BRCA2m and CCNE1g groups using the Mann–Whitney U test. B, CD3+ infiltration across TCGA transcriptomic subtypes; labeled P value represents comparison of IMR and PRO groups using the Mann–Whitney U test. C, CD3+ infiltration across Tothill transcriptomic subtypes; labeled P value represents comparison of C2 and C5 groups using the Mann–Whitney U test. BRCA2m, BRCA2 mutant; BRCA1m, BRCA1 mutant; EMSY-overxp; overexpression of EMSY; CCNE1g, gain of CCNE1; HRRwt, non-CCNE1g HRR wild-type.

Mitochondrial DNA (mtDNA) mutations are prevalent across cancer genomes, and growing evidence implicates their multifaceted role in energy metabolism with tumorigenesis. Ovarian cancer, in particular, demonstrates high mtDNA copy numbers and increased incidences of truncating and missense mtDNA mutations, with heteroplasmy levels predictive of prognosis. This review provides a comprehensive description of published mtDNA sequencing data in ovarian cancer, the majority being high-grade serous samples, encompassing both coding and non-coding regions. MtDNA mutations within non-coding regions, such as the D-loop control region, can affect mtDNA replication and transcription, hence affecting overall mtDNA copy numbers, while mtDNA mutations within coding regions can directly impact respiratory complex function and downstream metabolic pathways. MtDNA mutations may serve as clinically valuable diagnostic biomarkers for ovarian cancer and predictors for chemoresistance. We also explore ongoing efforts to deepen our understanding of mitochondrial oncogenetics through the creation of novel cancer models enabled by mitochondrial gene editing techniques. Developing robust human ovarian cancer cell models will be critical to elucidate mechanistic and phenotypic consequences of mtDNA mutations, assess drug response and resistance and identify new therapeutic targets to advance precision oncology in this emerging field.