Molecular subtypes of high-grade serous ovarian cancer: the holy grail?
Molecular subtypes of high-grade serous ovarian cancer: the holy grail?
复制标题
高级别浆液性卵巢癌的分子亚型：圣杯？
DOI：
10.1093/jnci/dju297
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Birrer,Michael
中科院分区：
文献类型：
--
作者：
Waldron,Levi;Riester,Markus;Birrer,Michael
With the development of robust genomic platforms and extensive genetic profiling of tumors, the identification of previously unrecognized cancer subtypes has become a reality. Molecular subtypes can reflect important biology, developmental origins, and most importantly have clinical utility. While genomic subtyping efforts have rapidly produced results for some cancers, the identification of molecular subtypes has been difficult for high-grade serous ovarian cancer (HSOC). What makes subtyping so difficult? The Cancer Genome Atlas (TCGA)(1) showed that in HSOC, hundreds of genes are affected by recurrent focal copy number and promoter methylation events, as well as by a small number of recurrent somatic short variant mutations. These extensive genetic abnormalities are likely due to a profound abnormality in DNA repair, resulting in genomic chaos and, in addition to the recurrent driver events, a large numbers of passenger events. It is predictable that HSOC would be genetically plastic with rapid evolution during the disease course, with extensive heterogeneity at the time of initial diagnosis. This would make the identification of specific tumor subtypes particularly challenging. In the face of such complexity, patients can be grouped based on combinations of genomic or epigenetic events in an almost arbitrary number of ways. So how can high-throughput transcriptomic data help in identifying clinically relevant subtypes? First, it can identify groups of patients whose disparate genomic events have similar expression footprints. Second, it can help to prioritize alterations with strong expression phenotype over ones with only small effect on gene expression. Unsupervised clustering of transcriptome data organizes tumors into discrete groups based on these two criteria, and it is important to acknowledge that this process will almost always succeed in identifying clusters discovery or even random data (2). In HSOC, the first study to report subtypes was the Australian Ovarian Cancer Study (AOCS)(3). This unsupervised microarraybased effort analyzed a cohort of tumors of mixed histology, mixed tumor grade, mixed sampling locations, and variable amounts of stroma. The Cancer Genome Atlas (TCGA) later reported largely overlapping subtypes and titled these “immunoreactive,”“differentiated,”“proliferative,” and “mesenchymal” but was unable to show any difference in clinical outcome between the subtypes. Independent of these efforts, other contemporaneous large array–based studies of high-grade advanced stage serous ovarian cancers generated prognostic signatures, but could not describe any molecular subtypes (4, 5). Thus, the robustness of these subtypes has remained controversial.In this issue of the Journal, Konecny and colleagues (6) present more evidence for the existence and survival association of four HSOC molecular subtypes as proposed by TCGA, in a new microarray dataset of 174 patients with clinical follow-up at the Mayo clinic. Consistent with a recent meta-analysis by Verhaak et al.(7), the authors convincingly demonstrated that patients classified as “immunoreactive” have on average best prognosis, whereas the “mesenchymal” subtype is associated with poor outcome, with an adjusted hazard ratio (HR) comparing these two groups of 1.84 (95% confidence interval [CI]= 1.15 to 2.94, P=. 01). A difference in patient survival reinforces that subtypes exist, because differences in overall survival are expected to originate from key biological distinctions. The authors also developed a novel subtyping system for HSOC based on their Mayo cohort expression data. Robust clustering utilizing the 1850 genes with highest variability in …
