In a recent study published in Nature Communications, researchers from the United States of America (USA) investigated the potential of the transcriptome of tumor-adjacent normal lung tissue in predicting the prognosis of lung cancer.
They found that molecular profiling of the tumor-adjacent normal (TAN) lung tissue, rather than the tumor tissue, is the strongest predictor of the clinical outcomes of lung adenocarcinoma (LUAD).
Study: Inflammation in the tumor-adjacent lung as a predictor of clinical outcome in lung adenocarcinoma. Image Credit: create jobs 51/Shutterstock.com
Background
LUAD remains a leading cause of mortality worldwide despite the rapid evolution in its diagnostic and treatment methods. Given the 30% risk of disease progression even after surgical resection (the mainstay of treatment), the development of predictive models for survival in lung cancer has gained pace, predominantly using histology, gene expression, mutation, proteomics, and the microbiome.
However, the studies have been limited by a lack of clinical validation and reproducibility. Additionally, these studies majorly focus on identifying tumor signatures to stratify early-stage lung cancer. Therefore, in the present study, researchers aimed to explore the TAN lung tissue instead to predict disease progression in LUAD and survival.
About the study
The study included a cohort of 143 treatment-naïve patients with stage I LUAD from whom tumor-lung and tumor-adjacent normal lung samples were prospectively collected from the same segment, lobe, or wedge resection.
About 69% of the patients were female, averaging 70.1 years. The samples were analyzed histologically in fixed, embedded tissue. New primary tumors were identified by applying the Martini-Melamed criteria.
Simultaneously, deoxyribonucleic acid (DNA) and ribonucleic acid (RNA) from the lung tumor, TAN samples, and blood were sequenced using the PACT (short for profiling of actionable cancer targets) assay and Illumina software respectively, to identify mutations and copy number changes.
Patient samples were clustered using phenotype to ensure they came from the same patient. While various bioinformatics tools were used throughout the analysis workflow, the five-year recurrence was predicted using an in-house machine-learning-based model.
Nuclei were isolated from the samples and were subjected to single nuclei RNA sequencing (snRNA-seq).
External datasets were used to determine whether the same scoring system could be used to stratify other cancer types. Although the study did not involve randomization or blinding, patients were classified based on recurrence and progression status.
The patients were postoperatively followed up for progression status for a median of 2,284 days via computation tomography (CT) scans. The statistical analysis included Cox regression analysis, principal component analysis, and the Mann-Whitney U-test, among others.
Results and discussion
As per the results, 35% of the patients showed disease progression, with 23 patients showing the formation of a new primary tumor, 13 showing locoregional recurrence, and 14 showed systemic metastasis.
The overall survival for patients with a new primary tumor was greater than that of patients with a recurrence. Mutational and transcriptomic analysis could be successfully performed on most patients.
No significant difference was observed in progression-free survival (PFS) and recurrence-free survival (RFS) of patients categorized as per mutational status in the epidermal growth factor receptor (EGFR) gene, Kristine rat sarcoma (KRAS), or serine/threonine kinase 11 (STK11) gene.
However, a mutation in tumor protein 53 (TP53) was found to be significantly associated with LUAD recurrence. Although tumor mutational burden was found to be modest at predicting prognosis, mutations overall were found to be poor predictors of survival.
Although the transcriptomic signature of the tumor could not predict recurrence or the risk of progression, that of the TAN sample could successfully predict the recurrence of the disease and aid the stratification of patients into high- and low-risk groups.
This indicates the potential role of TAN tissue in future recurrence and its utility in predicting prognosis. However, TAN tissue could not accurately predict the formation of a new primary tumor.
The current prediction model performed equally well on data from an external dataset (from The Cancer Genome Atlas or TCGA) for predicting the prognosis of LUAD and lung squamous cell carcinoma.
When the mechanisms underlying the predictive potential of TAN samples were investigated, the researchers found that TAN tissue was associated with cancer hallmarks to an extent corroborating with previous studies.
The gene module 20 (enriched in inflammatory signaling pathways) was found to be associated with TAN as well as disease progression not only in lung cancer but also in cancers of the breast, head/neck, and kidney.
Results from the snRNA-seq analysis suggested that within the TAN samples, the highest expression of module 20 was found to be in mesothelial cells, fibroblasts, monocytes, stalk-like ECs, MAST cells, and alveolar macrophages. The module 20 scores of groups with and without progression were found to be significantly different.
Conclusion
Being the largest study in terms of cohort size and follow-up time, the study provides significant evidence regarding the potential of the TAN tissue's molecular profiling for predicting lung cancer's clinical outcomes.
It opens new avenues for developing therapeutic approaches to prevent recurrence or progression in high-risk patients.
Revolutionizing disease prediction: MILTON framework utilizes biobank datasets to identify 3,213 diseases