Background: The majority of ovarian carcinomas are of high-grade serous histology, which is associated with poor prognosis and limited treatment options. Several studies have identified gene expression-based subtypes of high-grade serous ovarian carcinoma (HGSOC) as a basis for targeted therapy, yet extensive ambiguity in subtype classification impairs translation of these subtypes into clinical practice. Furthermore, although HGSOC tumors are known to be frequently polyclonal, it is unknown whether clones within the same tumor share the same subtype.
Results:We investigate whether ambiguity in subtype classification can be attributed to the polyclonal composition of HGSOC tumors, addressing the currently unresolved question whether proposed subtypes are early or late events in tumorigenesis. This hypothesis is first tested in The Cancer Genome Atlas HGSOC cases by (i) analyzing recurrent somatic copy number alterations for their association with subtypes, (ii) inferring per-alteration clonality from complementary analysis of SNP arrays and whole-exome sequencing, and (iii) testing whether subtype-associated alterations tend to predominantly occur clonally (early events) or subclonally (late events). As opposed to the genomically distinct evolution of soft-tissue sarcoma subtypes, we find that subtype association of HGSOC alterations significantly correlate with subclonality. This correlation is particularly evident for the high-purity proliferative subtype spectrum, which is also characterized by extreme genomic instability, absence of immune infiltration, and increased patient age. This is in stark contrast to the high-purity differentiated subtype spectrum, which is characterized by largely intact genome integrity, high immune infiltration, and younger patient age. Other subtypes showed intermediate levels for these characteristics. From single cell sequencing of an independent HGSOC tumor, we demonstrate that ambiguity in subtype classification extends to individual tumor epithelial cells, further supporting a developmental transition from one subtype spectrum to another.
Conclusion:We propose a novel model of HGSOC tumor development that complements the subtype perspective. In this model, individual tumors develop from an early differentiated spectrum to a late proliferative spectrum, and may exhibit characteristics of different previously defined "subtypes" at different points along a timeline characterized by increasing genomic instability and subclonal expansion. This model is more consistent with available bulk and single-cell data, and provides an explanation for ambiguity in subtype classification as the result of assigning discrete, mutually exclusive subtypes to a genomically complex process of tumor evolution.