Molecular Signatures Predict Problematic Lung Lesions Free

Historically, it has been difficult to distinguish between lung carcinoma in situ lesions that will progress to invasive squamous cell carcinoma and those that will spontaneously regress. However, a recent paper in Nature Medicine describes molecular signatures that clinicians may someday use to identify high-risk lesions requiring intervention.

“Our work shows that many changes observed in cancer are present much earlier in the disease process than previously thought,” says Sam Janes, MRCP, PhD, of University College London (UCL), UK, the study's senior author.

In the study, 85 patients with carcinoma in situ lesions underwent biopsies, after which they received an autofluorescence bronchoscopy every 4 months and a CT scan every 12 months. After determining which lesions progressed to cancer and which regressed, the researchers analyzed the biopsies for those lesions; biopsies for lesions that remained stable were not analyzed. Of 129 biopsies analyzed, whole-genome sequencing was performed on 39, methylation data gathered from 87, and gene-expression data collected from 51. (Some patients had multiple biopsies, and some samples were analyzed in multiple ways.)

The researchers found a complex set of alterations present in lesions that progressed to cancer. “There is not a clear, binary molecular switch,” explains co–lead author Adam Pennycuick, MRCP, also of UCL. “However, we do present predictive models which use sets of genes, methylation probes, or copy-number changes to predict a lesion's outcome.”

Although these predictive models could not be tested on an independent set of lung carcinoma in situ lesions, researchers did test their ability to categorize lung cancer samples from The Cancer Genome Atlas (TCGA). An assay measuring methylation at 141 probes to identify tumors matched TCGA's lung cancer versus control classifications with an AUC of 0.99. Similarly, an assay examining expression of 291 genes classified lung cancer versus controls with an AUC of 0.81, and an assay assessing copy number at 154 sites had an AUC of 0.98.

Such predictive models could transform patient care. Although only half of carcinoma in situ lesions become cancerous, many patients undergo radiation therapy or have part or all of a lung removed, notes UCL's Vitor Teixeira, PhD, co–lead author. Identifying the lesions most likely to develop into invasive tumors could help minimize unnecessary treatment.

As CT-based lung cancer screening grows more common and more lesions are detected, such predictive models would become increasingly valuable, says Geoffrey Oxnard, MD, of Dana-Farber Cancer Institute in Boston. “Methods like the proposed tests will be essential to figure out which concerning areas require intervention.”

Currently, Janes and colleagues are working with an international team to test their predictive models in a prospective trial. The team is also delving into how the molecular changes identified in the study drive cancer progression. “This could allow us to not only treat lung cancer, but potentially prevent it from ever emerging,” Janes says.

In the future, another common type of precancerous lesion in the lung—adenocarcinoma in situ—will need to be investigated, says Oxnard. Because the typical location of these lesions makes them more difficult to biopsy than squamous carcinoma in situ, they are harder to study. However, adenocarcinomas are more common in the United States than squamous cell carcinomas, making it especially important to see whether analytic methods similar to the ones used here can reveal “which adenocarcinoma in situ lesions are at high risk for progression.” –Kristin Harper