Altered pathways and colorectal cancer prognosis

Slattery et al. 7] describe in BMC Medicine a pathway-based approach to analyze whole genome expression
changes in colon cancer tissues when compared to normal adjacent tissue samples. The
main conclusion of the study is that having more deregulated pathways is associated
with good prognosis 7]. This research is innovative in the analytical approach. The study initially compared
RNA-seq expression data from 175 colon tumors with their paired adjacent normal mucosa.
Differentially expressed genes found were then assigned to pathways, and the relevant
explanatory variable analyzed was the number of deregulated genes within pathways.
Using a simple method, each patient was assigned a deregulation score based on the
number of altered genes for each pathway. Interestingly, they found that having a
high score was associated with better survival in 16 pathways, after adjusting for
age, stage, sex, and tumor molecular phenotype (MSI, TP53, KRAS, and CIMP). The most
significant pathways involved functions related to cell signaling and growth.

The pathway approach used is an interesting strategy, since deregulation of different
genes could converge in the same pathway. This allows tumors exhibiting dissimilar
gene expression patterns to achieve a similar phenotype. In consequence, collapsing
genes into pathways could be a useful tool not only to summarize a large list of genes
into more comprehensive functional entities, but also to classify a priori molecularly
different tumors into subtypes. From a translational point of view this categorization
has a special relevance, since tumors exhibiting different gene expression patterns
could behave in an analogous way regarding treatment response or patient outcome.
Indeed, similar pathway-oriented approaches have been described to be more useful
than those based on expression of individual genes 8] and reported to be informative about prognosis in breast cancer 9].

The authors of this study reported cell cycle as the most important pathway related
to CRC prognosis along with metabolic pathways and others classically related to colon
carcinogenesis like the Wnt pathway. Other studies on molecular subtypes using gene
expression signatures have described “epithelial” tumors as highly proliferative and
with better prognosis than tumors with a “mesenchymal” phenotype 10]-12]. It is intriguing that the epithelial-to-mesenchymal transition (EMT) pathway did
not appear as a relevant pathway in their analysis, which should be related to poor
prognosis. However, the authors identified TGFB1 as an upstream regulator of genes
showing differential expression using a network analysis. TGFB1 is a tumor suppressor
gene regarding tumor initiation, but also induces EMT and acts as a promoter of metastasis
13],14]. This dual role of TGFB1 makes the interpretation of its function difficult, but
EMT induced by TGFB1 expression usually has been related to poor prognosis 10],13]-15]. Also, regarding the functional interpretation of the data of Slattery’s study, a
limitation is that gene expression in normal tissue adjacent to the tumor has been
shown to be altered, when compared to the expression in healthy subjects 16]. Probably there is a cross-talk between tumor and normal adjacent tissue in addition
to the tumor micro-environment cross-talk, and many of the gene expression signals
captured by RNA-seq derive from stroma and not from the tumor cell 17]. This does not deny the potential prognostic value of the differences, but it could
modify the interpretation of some of the pathways’ effects and the strategy to design
targeted therapies.

Another interesting finding revealed in Slattery’s study is that pathway deregulation
is not only related to prognosis but is also inversely related to stage. Tumors from
patients diagnosed in stage I had more differentially expressed genes than those from
patients with advanced cancer, though the differences were not statistically significant.
This finding, if confirmed in other studies, opens new questions related to CRC progression.
Why do less advanced tumors have more differentially expressed genes? Are metastatic
tumors more specialized than early stage ones and need less altered genes to survive
once they have acquired metastatic potential? Large gene expression changes in cell
cycle observed in stage I and II tumors could suggest that these tumors are able to
proliferate and grow in size but are unable to develop more sophisticated functions
necessary to invade and disseminate. As the authors propose, perhaps increased gene
expression changes could be destabilizing for the tumor and, in consequence, responsible
for better prognosis.

Despite the interesting findings reported, a major limitation of the study is that
the authors have not been able to provide a validation of their results in an independent
dataset. There are available public datasets on colon cancer expression that simultaneously
have analyzed tumor and paired mucosa that can be used to validate differential expression
genes 18], but this dataset is restricted to stage II tumors and has not published data on
prognosis yet. Other interesting datasets like The Cancer Genome Atlas (TCGA) consortium
study 19] have analyzed very few normal tissues, and this hinders its utility to validate results
that focus on differences between tumor and normal tissue such as the specific alterations
observed in Slattery’s study.