NEW YORK (GenomeWeb) – Building on earlier finding showing population and gender differences in the expression of microRNA isoforms in healthy individuals, researchers from Thomas Jefferson University have reported new data showing that such differences occur in breast cancer and potentially other diseases.
The findings, which also suggest that these isoforms can be used to differentiate cancer subtypes, represent "a call to arms in the sense that there is a lot of complexity that … we did not realize," Isidore Rigoutsos, director of TJU's Computational Medicine Center and senior author of the report, said. "The point of the paper is to show that understanding this complexity in a disease context is warranted, if not necessary."
It has been thought that each arm of the hairpin-shaped molecules that give rise to miRNAs could only produce one mature miRNA — a so-called archetype miRNA. Recent studies, however, have shown that miRNA isoforms with slightly different 5' or 3' termini can also be generated.
In a study published last year, Rigoutsos' lab reported that the expression of these so-called isomiRs could differ in the same cell type, depending on the race and gender of an individual. "That begged the question, 'What does this mean for disease?'" he told GenomeWeb this week.
To answer this, Rigoutsos and his colleagues focused on breast cancer, using datasets on 22 normal and 294 disease samples from the National Cancer Institute's Cancer Genome Atlas (TCGA) repository. The researchers specifically looked at subtypes of breast cancer based on hormonal profile — whether they were positive or negative for estrogen receptor, progesterone receptor, and human epidermal growth factor receptor. Quantitation of these markers is routinely used to classify tumors and guide treatment regimens.
They also factored in the race of the patients, comparing datasets from subjects identified as either white or black women. While informative, this approach was not without its limitations, Rigoutsos cautioned.
"TCGA uses self-declared race … and the concept of race is not one that is easy to define in the absence of genetics," he explained. "So we knew that within [the race] category, there were subcategories, but clearly we don't have enough data to speak of those subcategories. By necessity, we stayed further up in the [race] hierarchy, so to speak."
The investigators performed RNA sequencing on the samples, considering 5,353 human miRNA precursors, including ones contained in the latest version of miRBase and others they had independently identified in earlier experiments.
When they catalogued all possible isomiRs that were detectable by RNA sequencing, they found that in virtually all cases of known miRNAs, there was "a clear deviation from the one-arm-one-miRNA dogma," they wrote in their paper, which appeared in Nucleic Acids Research this month. In the case of miR-21-5p, for example, as many as 43 distinct isomiRs are produced.
In light of this complexity, Rigoutsos and his group then investigated whether looking at the full isomiR profile of a sample could distinguish normal from cancer tissue better than archetype miRNAs alone.
Using two common breast cancer subtypes, the scientists demonstrated that in fact the full isomiR profile was a more powerful descriptor than archetype miRNAs. However, the isomiRs that were responsible for differentiating normal from disease in white women were not the same as those in black women. For instance, several isomiRs of miR-183-5p were upregulated in white but not black women who were all negative for all three hormone receptors.
Notably, when Rigoutsos and his team overexpressed three of these miR-183-5p isomiRs in a breast cancer cell line, microarray analysis revealed that each isomiR had a distinct impact on the cellular transcriptome, suggesting a complex interplay between isomiRs that can drive disease.
The researchers also looked at whether isomiR profiles could be used to differentiate subtypes of cancer. They performed an integrated analysis of the isomiRs and mRNAs in luminal A and luminal B tumors with the same hormonal profile and found that the isomiR profile could "easily discriminate" the two subtypes.
Additionally, the team examined the possible factors influencing isomiR expression levels, discovering that neither the hairpin precursor nor the arm of origin determined isomiR abundance. "Indeed, isomiRs from different hairpins and different arms formed clusters based on other characteristics, namely the position of the 5' endpoint relative to the reference coordinate [found in miRBase] and the relative position of the 3' endpoint," they wrote in their paper.
Rigoutsos noted that the findings add greater complexity to the miRNA picture than may have previously been expected, but said he believes there are advantages to be found in this.
"People worry about complexity, but this complexity has always been there," he said. "My view is that knowing about the complexity is giving us power. … It allows us to be more flexible in many ways."
In the end, "we're saying that we need to be more detailed in the question we ask in a given disease setting."
Rigoutsos said his group is currently working to extend their findings to other diseases, but declined to comment on these efforts.