Pervasive structural heterogeneity rewires glioblastoma chromosomes to sustain patient-specific transcriptional programs.

Glioblastoma multiforme (GBM) encompasses brain malignancies marked by phenotypic and transcriptional heterogeneity thought to render these tumors aggressive, resistant to therapy, and inevitably recurrent. However, little is known about how the spatial organization of GBM genomes underlies this heterogeneity and its effects. Here, we compile a cohort of 28 patient-derived glioblastoma stem cell-like lines (GSCs) known to reflect the properties of their tumor-of-origin; six of these were primary-relapse tumor pairs from the same patient. We generate and analyze 5 kbp-resolution chromosome conformation capture (Hi-C) data from all GSCs to systematically map thousands of standalone and complex structural variants (SVs) and the multitude of neoloops arising as a result. By combining Hi-C, histone modification, and gene expression data with chromatin folding simulations, we explain how the pervasive, uneven, and idiosyncratic occurrence of neoloops sustains tumor-specific transcriptional programs via the formation of new enhancer-promoter contacts. We also show how even moderately recurrent neoloops can relate to patient-specific vulnerabilities. Together, our data provide a resource for dissecting GBM biology and heterogeneity, as well as for informing therapeutic approaches.

TNF-α-producing macrophages determine subtype identity and prognosis via AP1 enhancer reprogramming in pancreatic cancer.

Large-scale genomic profiling of pancreatic cancer (PDAC) has revealed two distinct subtypes: 'classical' and 'basal-like'. Their variable coexistence within the stromal immune microenvironment is linked to differential prognosis; however, the extent to which these neoplastic subtypes shape the stromal immune landscape and impact clinical outcome remains unclear. By combining preclinical models, patient-derived xenografts, as well as FACS-sorted PDAC patient biopsies, we show that the basal-like neoplastic state is sustained via BRD4-mediated cJUN/AP1 expression, which induces CCL2 to recruit tumor necrosis factor (TNF)-α-secreting macrophages. TNF-α+ macrophages force classical neoplastic cells into an aggressive phenotypic state via lineage reprogramming. Integration of ATAC-, ChIP- and RNA-seq data revealed distinct JUNB/AP1 (classical) and cJUN/AP1 (basal-like)-driven regulation of PDAC subtype identity. Pharmacological inhibition of BRD4 led to suppression of the BRD4-cJUN-CCL2-TNF-α axis, restoration of classical subtype identity and a favorable prognosis. Hence, patient-tailored therapy for a cJUNhigh/TNF-αhigh subtype is paramount in overcoming highly inflamed and aggressive PDAC states.