Elucidation of molecular and functional heterogeneity through differential expression network analyses of discrete tumor subsets.

Intratumor heterogeneity presents a major hurdle in cancer therapy. Most current research studies consider tumors as single entities and overlook molecular diversity between heterogeneous state(s) of different cells assumed to be homogenous. The present approach was designed for fluorescence-activated cell sorting-based resolution of heterogeneity arising from cancer stem cell (CSC) hierarchies and genetic instability in ovarian tumors, followed by microarray-based expression profiling of sorted fractions. Through weighted gene correlation network analyses, we could assign enriched modules of co-regulated genes to each fraction. Such gene modules often correlate with biological functions; one such specific association was the enrichment of CD53 expression in CSCs, functional validation indicated CD53 to be a tumor-initiating cell- rather than quiescent CSC-marker. Another association defined a state of poise for stress-induced metastases in aneuploid cells. Our results thus emphasize the need for studying cell-specific functionalities relevant to regeneration, drug resistance and disease progression in discrete tumor cell fractions.

Discrete molecular classes of ovarian cancer suggestive of unique mechanisms of transformation and metastases.

PURPOSE: Tumor heterogeneity and subsistence of high-grade serous ovarian adenocarcinoma (HGSC) classes can be speculated from clinical incidences suggesting passive tumor dissemination versus active invasion and metastases. EXPERIMENTAL DESIGN: We explored this theme toward tumor classification through two approaches of gene expression pattern clustering: (i) derivation of a core set of metastases-associated genes and (ii) resolution of independent weighted correlation networks. Further identification of appropriate cell and xenograft models was carried out for resolution of class-specific biologic functions. RESULTS: Both clustering approaches achieved resolution of three distinct tumor classes, two of which validated in other datasets. Networks of enriched gene modules defined biologic functions of quiescence, cell division-differentiation-lineage commitment, immune evasion, and cross-talk with niche factors. Although deviant from normal homeostatic mechanisms, these class-specific profiles are not totally random. Preliminary validation of these suggests that Class 1 tumors survive, metastasize in an epithelial-mesenchymal transition (EMT)-independent manner, and are associated with a p53 signature, aberrant differentiation, DNA damage, and genetic instability. These features supported by association of cell-specific markers, including PAX8, PEG3, and TCF21, led to the speculation of their origin being the fimbrial fallopian tube epithelium. On the other hand, Class 2 tumors activate extracellular matrix-EMT-driven invasion programs (Slug, SPARC, FN1, THBS2 expression), IFN signaling, and immune evasion, which are prospectively suggestive of ovarian surface epithelium associated wound healing mechanisms. Further validation of these etiologies could define a new therapeutic framework for disease management.