Ng, S. W. K. | Search bioRxiv

bioRxiv2025

The subclonal footprint of pervasive early dissemination in pancreatic cancer

Early is too late describes a central problem in pancreatic cancer referring to its relentless drive to disseminate and highlights the need to understand how this disease spreads so rapidly. In this study, we profiled 1,013 samples from 277 donors, including tissue whole genome and RNA sequencing combined with plasma whole genome sequencing at [~]25x. Strikingly, local primary tumours, including some reaching 7cm, were found to shed little to no circulating tumour DNA (ctDNA). Instead, metastatic burden in the liver but not extrahepatic sites, was a main physiologic determinant of ctDNA levels. Whole-genome duplication (WGD), high cell cycle activity, and non-glandular differentiation emerged as tumour-intrinsic features related to increased ctDNA shedding. By contrast, decreased shedding was related to extrinsic features including a reactive microenvironment and unexpectedly, humoral immunity. Analysis of tumour clonal architecture showed that in patients with low ctDNA levels, the signal disproportionately originated from subclones and this signal persisted even when the primary tumour was removed. Disseminated subclones were a significant source of ctDNA in early-stage patients. Longitudinal analysis of patients revealed that subclones seeded metastases and shed ctDNA in the blood years before detection. This first report of paired tissue and plasma whole genomes in pancreatic cancer is a unique resource and has broad implications for disease surveillance, treatment monitoring, and early detection in this disease.

cancer biology↗

bioRxiv2024

Distinct gene regulatory networks govern hematopoietic and leukemia stem cells

The underlying gene regulatory networks (GRN) that govern leukemia stem cells (LSC) in acute myeloid leukemia (AML) and hematopoietic stem cells (HSC) are not well understood. Here, we identified GRNs by integrating gene expression (GE) and chromatin accessibility data derived from functionally defined cell populations enriched for HSC and LSC. We analyzed n=32 LSC+ and n=32 LSC-cell fractions from n=22 AML patients, along with n=7 stem and n=10 progenitor enriched cell populations sorted from human umbilical cord blood (hUCB), producing a database of n{approx}17,000 transcription factor (TF) regulatory interactions for hUCB-HSPC and AML. We developed an iterative algorithm that associates the degree of chromatin openness with TF binding preferences, and the GE of candidate TF and target genes within 100kb upstream of transcription start sites. A putative regulatory structure was found to be enriched in HSC-enriched cell populations, comprising TF-target gene interactions between ETS1, EGR1, RUNX2, and ZNF683 oriented in a self-reinforcing configuration. A regulatory loop comprising FOXK1 and MEIS1, rather than the 4-factor HSC subnetwork, was detected in the LSC-specific GRN. The core HSC and LSC TF networks were extended using protein-protein interaction (PPI) data to determine connectivity with interacting genes whose expression strongly associated with LSC/HSC frequency estimates, producing a database of n=103,516 PPI target pathways. The effect of perturbing genes along the identified pathways on functional HSC and LSC frequency was predicted based on statistical regression analyses. To validate GRN predictions, we used pharmacologic and CRISPR targeting, in addition to re-examining published functional data associated with several network nodes that were predicted to impact stemness. Notably, we found that inhibition of CDK6 in AML samples markedly reduced LSC numbers as assessed in de novo serial xenotransplantation studies (fold change {approx} 10), as predicted by the LSC GRN model. Additionally, in-house CRISPR-based knockdown of ETS1 resulted in a significant decrease in HSC quiescence-associated microRNA-126 expression, and increased HSC frequency. Taken together, our models provide a comprehensive view of the underlying regulatory structures governing functional human HSC and LSC. This approach has translational potential as it can be used as a high-throughput in-silico screening tool for the systematic identification of gene targets for LSC elimination and HSC expansion.

bioinformatics↗