Researchers collaborating among a number of French institutions have developed and validated a novel gene-expression signature that classifies colon cancers into six distinct molecular subtypes with different pathological characteristics and varying prognoses.
If the group can further validate the finding — published in PLOS Medicine this week — the classifier could be useful in predicting which patients are more likely to suffer recurrence after curative surgery for localized cancer and as a tool to inform more personalized treatment.
"The biological relevance of these subtypes is illustrated by significant differences in prognosis [and] this analysis provides possibilities for improving prognostic models and therapeutic strategies," the authors wrote.
According to the study authors, pathological staging is currently the only prognostic tool used in clinical practice to determine which patients should receive adjuvant chemotherapy. And while gene-expression studies have yielded potentially useful signatures before, these have been "poorly reproducible."
The researchers wrote that the new classifier is the first "robust transcriptome-based classification of [colon cancer] that improves current disease stratification based on clinicopathological variables and common DNA markers," like BRAF and KRAS mutations.
The team also compared the new classifier's prognostic ability to both Agendia's Coloprint, and Genomic Health's Oncotype DX Colon Cancer assays, finding that the new classifier added prognostic information that remained significant over the Oncotype DX recurrence score, while a modified version of the Coloprint signature was not prognostic at all in the research cohort.
In the study, the authors described the identification of the new subtypes based on genome-wide transcriptome analysis using microarrays, and the refinement and retrospective validation of the classifier, spanning 57 genes, in a discovery cohort of 443 patient samples and a validation cohort of 123 samples.
The group also validated the signature in a second cohort of 906 samples from seven public datasets, the authors wrote.
The researchers distinguished the six subtypes in the new classification scheme based on the predominant biological characteristics of each type— a deficient mismatch repair subtype, a KRAS mutant subtype, a cancer stem cell subtype, and three subtypes characterized by chromosomal instability with differing pathway dysregulation.
Restricting analysis to patients with stage II and III tumors who could most benefit from more reliable prognostic testing, the researchers first looked at prognosis in the discovery set and found that the subtypes differed in prognosis, based on a measure of relapse-free survival, but not significantly so — two of the groups had relatively poorer outcomes compared to the other four.
In the validation set, this prognostic difference was more pronounced, becoming statistically significant. The researchers then combined the two high-risk subtypes (C4 and C6) and four lower-risk subtypes (Subtypes C1,2, 3, and 5) to create a binary classification system. According to the authors, this led to an even stronger association with relapse-free survival in both the discovery and validation sets and in the entire cohort combined.
The binary classification remained an independent prognostic factor in multivariate analysis, and common DNA alterations were added to the group's model, the authors wrote.
The researchers also compared the prognostic ability of the new classifier against two emerging prognostic tools for colon cancer— Agendia's Coloprint, and Genomic Health's Oncotype DX Colon Cancer.
Genomic Health recently published data from the second retrospective validation of its Oncotype DX colon test, showing that the test's recurrence score was the strongest predictor of disease recurrence independent of other factors, such as stage, mismatch repair status, number of nodes examined, tumor grade, and lymphovascular invasion, in a set of 690 samples (PGx Reporter 4/17/2013)
According to the French researchers, two of the six subtypes in their new classifier were independently associated with shorter relapse-free survival, even after adjusting for biological variable like age, sex, and tumor stage, as well as the prognostic classification given by Oncotype DX.
The Oncotype DX score had prognostic value in both the group's discovery and validation cohorts, the authors wrote. However, the score was not prognostic for all the new subtypes. Both the new classifier and Oncotype DX remained independently prognostic, together with TNM staging.
The researchers also tried to measure the new classifier against a recreation of Coloprint, using available probe sets corresponding to 17 of the assay's 18 target genes. According to the authors, this 17-gene expression signature showed no prognostic value in the group's cohort.
According to the French group, their results with the new classifier have some caveats. The researchers did not include tumor grade or number of nodes examined in their multivariate analysis because this information was not available for all samples in the cohort.
The team believes it will be necessary to follow up and confirm the study results with further validation, ideally with large, prospective patient cohorts.