Presence of a 34-gene signature is a favorable prognostic marker in squamous non-small cell lung carcinoma

The tumor immune microenvironment is a heterogeneous entity. Gene expression analysis allows us to perform comprehensive immunoprofiling and may assist in dissecting the different components of the immune infiltrate. As gene expression analysis also provides information regarding tumor cells, differences in interactions between the immune system and specific tumor characteristics can also be explored. This study aims to gain further insights in the composition of the tumor immune infiltrate and to correlate these components to histology and overall survival in non-small cell lung cancer (NSCLC). Archival tissues from 530 early stage, resected NSCLC patients with annotated tumor and patient characteristics were analyzed using the NanoString nCounter Analysis system. Unsupervised clustering of the samples was mainly driven by the overall level of inflammation, which was not correlated with survival in this patient set. Adenocarcinoma (AD) showed a significantly higher degree of immune infiltration compared to squamous cell carcinoma (SCC). A 34-gene signature, which did not correlate with the overall level of immune infiltration, was identified and showed an OS benefit in SCC. Strikingly, this benefit was not observed in AD. This difference in OS in SCC specifically was confirmed in two independent NSCLC cohorts. The highest correlation between expression of the 34-gene signature and specific immune cell populations was observed for NK cells, but although a plausible mechanism for NK cell intervention in tumor growth could be established in SCC over AD, this could not be translated back to immunohistochemistry, which showed that NK cell infiltration is scarce irrespective of histology. These findings suggest that the ability of immune cell infiltration and the interaction between tumor and immune cells may be different between AD and SCC histology and that a subgroup of SCC tumors seems more susceptible to Natural Killer cell recognition and killing, whereas this may not occur in AD tumors. A highly sensitive technique like NanoString was able to detect this subgroup based on a 34-gene signature, but further research will be needed to assist in explaining the biological rationale of such low-level expression signatures.