NEW YORK (GenomeWeb) – Researchers from Lawrence Berkeley National Laboratory have identified a novel gene expression signature that that could help predict patient prognosis and response to therapy in several major types of cancer, including breast and lung cancers.
In a study published this week in Nature Communications, the investigators described their identification of 14 cell division-associated genes, whose expression levels predicted which patients had worse or better outcomes in a retrospective clinical dataset.
The investigators also showed that by measuring the degree of misexpression of these genes, they could predict which patients were more or less likely to respond to chemotherapeutic drugs or other interventions like radiation therapy. This suggests the approach could have significant clinical value in helping guide decisions about which patients should receive certain adjuvant treatments and which should not.
Gary Karpen, a senior scientist in the Berkeley Lab’s division of biological systems and engineering and senior author of the study, told GenomeWeb this week that the new findings were prompted by a long history of research in his lab, and by others in the field, into genes that regulate the function of centromeres and kinetochores and their impact on cell division.
Early work by Karpen and his colleagues in fruit flies linked overexpression of specific centromere-associated proteins with the chromosomal instability that is a hallmark of many cancers. In essence, depletion or overexpression of a certain protein can populate chromosomes with multiple kinetochores, causing spindles to attach to more sites than normal, breaking chromosomes apart during cell division and causing aneuploidy and other DNA damage that leads to cancer or precipitates its spread.
In its study, the Berkeley Lab team hoped to extend this research to humans by analyzing 31 genes involved in regulating the function of centromeres and kinetochores using a set of public datasets that included thousands of human clinical tumor samples from about a dozen cancer types, and associated clinical outcomes data.
Among these 31 gene candidates, the researchers were able to show that 14 were consistently overexpressed in cancer tissue compared to normal controls. They also developed a scoring method for expression of these genes, which they call the centromere and kinetochore gene expression score (CES), that could distinguish patients with better outcomes from those with poorer outcomes — either with or without specific treatments.
Interestingly, and somewhat surprisingly, high expression of genes in the signature also appeared to be predictive of response to chemotherapy and radiation therapy, Karpen said.
The group studied this in two cancer types, lung cancer and breast cancer, using clinical trial data. In the lung cancer dataset, from a trial of adjuvant chemotherapy in post-surgery stage I and II NSCLC patients, the team divided patients into thirds based on their CES.
According to the authors, being in the high-CES subgroup was associated with poor overall survival for patients without adjuvant treatment, validating the prognostic power of the CES approach.
However, adjuvant chemotherapy "effectively negated the adverse outcome associated with high CES, suggesting that the CES system also has predictive power," the group wrote.
For breast cancer, the group reported that in ER-positive cancers, CES was associated with poor survival for patients who did not receive systemic therapy, or who received tamoxifen only, without endocrine therapy. Again though, the high CES levels lost their prognostic value in those patients who received chemo, suggesting that this treatment reduced their risk.
A plausible explanation for this, Karpen said, is that higher degrees of chromosomal instability as caused by higher expression of these centromere-impacting genes may make cancer cells more vulnerable to the cytotoxicity of chemo and radiation.
The results have two main implications for adjuvant treatment. First, as a novel and independent prognostic predictor, the gene expression score method that the group developed could help physicians better distinguish which patients with early-stage cancers can forgo adjuvant treatment, and which should not.
"The cancer field has been very focused on oncogenes and tumor suppressors, but this is a completely different pathway, so it's a good thing, but also a bit surprising that without looking at any oncogenes or tumor suppressors we could develop this very robust signature," Karpen said.
There already exist gene-expression based assays that have shown prognostic potential, most notably Genomic Health's OncotypeDx in breast cancer.
According to Karpen, the hope is that he and his colleagues' CES may be complementary to other prognostic and predictive gene expression signatures that already exist. Just as targeting multiple pathways in the therapeutic field is an exciting path forward for better treatment of cancer patients, addressing multiple independent molecular pathways also makes sense as a way to capture more relevant clinical information for prognostic or diagnostic testing, he said.
Because the genes in the CES are associated with a pathway that potentially affects a wide range of cancers, it could also be more broadly applicable across cancer types than some other algorithms that already exist.
Karpen said that the group hasn't explored combining its CES with other assays yet, but that it is "very interested" in doing so in the future.
Weiguo Zhang, the study's first author and a project scientist at Berkeley Lab, stressed also that because the Berkeley Lab group's centromere signature was predictive not only of outcomes, but also actual responses to chemo and radiation, it looks like it could have added or independent utility.
"What we have here is more based on the mechanism and so it looks at which patients are more likely to relapse and then also at what treatments are more likely to work for them," he said.
According to Karpen, the results from this initial research are encouraging, but the group now needs to further validate and refine its gene expression signature and scoring methodology to have the best predictive threshold, and to demonstrate its validity in more clinical cases, including prospective clinical studies.
In addition, he said that the team is working with another research group in Nanjing, China to evaluate and develop a practical testing technology platform that could host an eventual assay based on the 14-gene signature.
As the researchers work to further validate the CES in their own lab, Karpen said that they are particularly interested in looking at and comparing different stages of cancer and pre-cancer, with the hope that the signature might be able to distinguish early-stage lesions or hyperplasias that are likely to progress to metastatic disease from those that are not.