NEW YORK (GenomeWeb) – A team led by researchers at the Karolinska Institute in Sweden has analyzed the proteomic landscape of breast cancer tumors, finding similar subclasses as transcriptome-based analyses but also additional ones.
Breast tumors can be divvied into five subtypes — basal-like, normal-like, HER2, luminal A, and luminal B — based on their expression of 50 transcripts, known as the PAM50, but the researchers, led by the Karolinska Institute's Janne Lehtiö, noted that many patients are still misclassified.
They instead sought to classify breast tumors at the proteomic level and analyzed proteomic profiles for 45 breast tumors, representing each of the five PAM50 subtypes. As they reported in Nature Communications this week, they found that while clustering based on the tumors' proteomic profiles broadly agreed with the expression-based groups, they could tease out additional classifications and identify potential treatment targets.
"As breast tumors are continuously revealed to be individually unique diseases, considerations of molecular profiles will become paramount in selecting from available treatment options and developing new ones," the researchers wrote in their paper. "Though mRNA profiling has been initially dominant in this role, the 'landscape' study presented herein demonstrates the instrumental contribution analyses of the quantitative proteome will have moving forward.
For this study, the researchers selected 45 patients from the Oslo2 study cohort, whose tumors were classified into one of the five PAM50 subtype groups. They then analyzed those tumors using an LC-MS/MS-based approach to uncover 13,997 proteins derived from 12,645 genes.
Unsupervised hierarchical clustering of these proteome profiles grouped these tumors similarly to their PAM50 designations. While this clustering approach distinguished between the basal-like, normal-like, and luminal A subtypes, it found the luminal B and HER2 subtypes to overlap, a finding the researchers noted was in line with tumor transcript profiles and clinical reports of HER2-positive patients receiving conflicting mRNA-based subtype results.
Proteomic analyses could uncover further groups. For instance, a clustering analysis of high-variance proteins uncovered six consensus core tumor clusters (CoTCs). These CoTC classifications overlapped with the PAM50 subgroups of normal-like and luminal A tumors, but divided the basal-like tumors into two groups. It combined HER2 and some luminal B tumors but kept other luminal B tumors in a separate group.
These CoTC tumor groups also had distinct protein functional modules. For instance, the researchers found immune response-linked proteins were elevated in CoTC2 tumors, as compared to CoTC1 tumors, while CoTC3 and CoTC4 tumors had elevated luminal protein expression.
The researchers similarly classified these tumors based on their phosphoprotein profiles into four groups.
Additionally, they noted that the abundance of ESR1, PGR, AR, and BCL2 proteins was highly correlated, suggesting they could make tumors susceptible to treatment with a combination of hormone receptor inhibitors.
They likewise found that EGFR and MET were often coexpressed and could be a marker for basal-like and normal-like tumors, particularly in ductal carcinoma in situ.
In an integrated proteogenomic analysis, the researchers mapped peptides they identified back to the genome, including to regions that were thought to be noncoding or intronic. Nearly a third of these peptides were predicted to bind MHC class I proteins and were not present in mass spec data from normal tissues, suggesting they might be tumor specific.
Peptides originating from those regions though to be noncoding could serve as antigen targets to the tumors and be used to develop immunotherapies, the researchers said. "We postulate protein products of undescribed gene variants and noncoding regions are the consequences of cancer genome instability, and that they are strong candidates of tumor-specific targets for immunotherapies," they added.
The researchers have made their data available through a portal that is accessible at www.breastcancerlandscape.org.