Skip to main content
Premium Trial:

Request an Annual Quote

Gene Expression Signature Under Development Predicts Lung Cancers Likely to Spread to Brain


PARIS – Researchers in Israel are working on a gene expression signature to predict which lung cancer patients are at risk for experiencing brain metastasis. The project has already suggested pathways that might be targeted by treatments to prevent the spread of cancer to the brain.

At the Worldwide Innovative Networking (WIN) Consortium's annual symposium this past weekend, Jair Bar, deputy director of the Institute of Oncology at Sheba Medical Center, discussed his team's efforts to advance a 22-gene expression signature, which has demonstrated 93 percent accuracy in predicting which non-small cell lung cancer patients are at risk of seeing their disease spread to the brain.

Bar is planning to further validate this signature in lung cancer with an eye on potentially commercializing a test, he told GenomeWeb. He and his colleagues are also looking for collaborations to study similar predictive signatures in other tumor types.

Lung cancer, breast cancer, and melanoma are primary tumors with some of the highest affinity for brain metastasis. For as many as half the patients with brain metastases, the primary tumor started in the lung. The rate of brain metastases is particularly high among those with EGFR-mutated or ALK-positive NSCLC, since as many as 60 percent will see the disease spread to their brain.

Bar cited research from a few years ago in which researchers showed that breast and lung cancer cells form gap junctions — channels that allow intercellular communication — through glial cells, called astrocytes, and use these gap junctions to signal tumors to grow and spread to the brain. In this study, researchers used inhibitors of gap junctions to reduce brain metastasis in mouse breast cancer models, providing a proof of principle that this could be a treatment strategy for treating brain metastasis. 

Using in vivo and in vitro RNAi screens, researchers have also previously identified SPOCK1 and TWIST2 as regulators of so-called brain metastasis-initiating cells. 

In trying to develop their gene signature, Bar and his colleagues are not only trying to predict brain metastasis in lung cancer patients, but their goal ultimately is to try to prevent cancer spreading to the brain in the first place. In recent years, some lung cancer patients have had reason to be optimistic given the availability of newer targeted drugs, such as Genentech's ALK inhibitor alectinib (Alecensa), which has shown to reduce progression to the central nervous system. Still, generally, brain metastasis remains a dire prognosis, with patients living on average around six months following diagnosis.

A molecular diagnostic that can predict which lung cancer patients are likely to progress to brain metastasis could be a valuable tool in the oncologist's armamentarium in terms of crafting a personalized treatment strategy.

To develop their gene expression signature, Bar's team used specimens from resected primary NSCLC tumors with follow-up data on which patients eventually developed brain metastasis and which didn't. They then conducted RNA sequencing of 91 fresh frozen tissue samples — 32 with brain metastasis and 59 controls — using Illumina HiSeq 2500 and compared the genes expressed between those tumors that did spread to the brain and those that didn't.

"It was a big ordeal to have this number of good quality specimens," Bar said at the meeting. Researchers had to extract RNA from close to 300 samples in order to have 91 that were of good enough quality.

The samples came from the Ontario Tumor Bank, Alberta Cancer Research Biorepository, and the Sheba Medical Center tumor bank. The patients who contributed these samples had not received any neoadjuvant chemotherapy or radiotherapy prior to the surgical resection of their lung cancer.

Bar's group first looked at the role of clinical factors and found that adenocarcinomas, younger age, higher stage of cancer, and the level of cancer cell differentiation were associated with brain metastases. This association between brain metastasis and cell differentiation has come up in a number of different studies Bar is working on. He noted that although this is "a really important parameter," cell differentiation isn't typically reported alongside cancer stage in pathology reports.

Using just clinical factors, his group achieved 65 percent accuracy in predicting brain metastasis with 81 percent specificity and 34 percent sensitivity. Hoping to bolster the accuracy of the signature, researchers combined clinical factors and the top 100 genes with differential expression between the samples with brain metastasis and controls. This predictor achieved an accuracy of 89 percent with 95 percent specificity and 78 percent sensitivity.

Next, Bar's team took a more focused approach and chose the 22 top differentially expressed genes, which showed even better ability to predict brain metastasis with 93 percent accuracy, 97 percent specificity, and 88 percent sensitivity. "That looks very promising," Bar said, noting that most of the 22 genes that make up the predictor aren't the obvious genes that have been correlated with brain metastasis previously.

His team wanted to make sure that this signature wasn't just capturing highly aggressive tumors that are more likely to metastasize, but specifically lung tumors that are likely to spread to the brain. Researchers had samples from patients who experienced cancer metastases to sites other than the brain, in addition to samples from patients with brain metastases, and those without any metastases, but the signature was able to segregate the patients whose cancer would spread to the brain.  

This finding suggests to Bar that this signature "may really be correlated with something that is specific to the process of forming brain metastasis."

His group lastly looked at the pathways implicated by the 22 genes in the signature and found that they were mostly involved with oxidative phosphorylation, a major process by which cells produce energy. Bar noted that while it's well known that cancer cells tend to use glycolysis for energy, some recent studies suggest that oxidative phosphorylation is also important for cancer cells.

Based on these studies and his own group's work, "it can be speculated that oxidative phosphorylation is a novel therapeutic target possibly related to the process of brain metastasis of lung cancer," he said at the meeting.

Bar hopes to publish the gene signature, but noted that this is very much a "work in progress." His group is planning to study oxidative phosphorylation in mouse models of brain metastasis and further validate the 22-gene expression signature.