Abstract B20: Prediction of lung cancer survival by genes in the PAM50 breast cancer panel.

Accumulating evidence shows that lung tumors are responsive to the steroid hormones β-estradiol and progesterone, and variables related to hormone receptor status have been reported to affect lung cancer survival. These findings suggest that there may be similarities in how hormonal pathways control tumor progression between lung cancer and breast cancer. We examined the relationship between disease-free survival (DFS) and expression of genes included in three published breast cancer survival signatures in a cohort of 104 early-stage (IA and IB) non-small cell lung cancers. Cases were selected from the University of Pittsburgh Lung Cancer SPORE Tissue Bank based on the following criteria: tumor tissue originated from a completely resected primary T1N0 or T2N0 adenocarinoma or squamous cell lung cancer; fresh-frozen lung cancer tissue was available; no neo-adjuvant therapy given; clinical variables (smoking history, age at diagnosis, sex, size of tumor, pleural invasion, site of recurrence) and outcome were known. RNA isolated from banked frozen tumor tissue was used to assess mRNA expression with the Illumina Human HT-12 v4 BeadChip; all mRNA analyzed had passed quality checks. Data were also subjected to background subtraction and quantile normalization. The Illumina BeadChip contains 16 out of 16 published Oncotype DX genes, 61 out of 66 MammaPrint genes, and 49 of 50 PAM50 genes. Supervised Principal Component Analysis was conducted to evaluate the ability to predict disease-free survival (DFS) for each gene panel as a group and also by using the top genes from the gene panels that showed differential expression in the lung cancer cohort. All results were subjected to 10-fold cross-validation. Genes from the Oncotype DX panel showed no differential expression in the cohort and had no ability to separate cases based on DFS. For the MammaPrint genes, three probes (ALDH4A1, FLT1, RECQL5) showed significant differences in expression in the cohort (p < 0.05), and the gene panel showed a relationship to DFS (HR 1.72, p = 0.07). Examination of the top differentially expressed genes did not improve prediction for the MammaPrint genes. For the PAM50 gene panel, three probes (CXXC5, FGFR4, FOXC1) showed significant differential expression (p < 0.05) and 23 probes showed differential expression at p < 0.1. The PAM50 gene panel as a group was able to separate lung cancer cases based on DFS (HR 1.9, p = 0.03). The PAM50 genes with differential expression at p < 0.1 also separated the cohort into four prognosis groups (p = 0.008 for trend). The worst prognosis group had a HR of 4.76 (p = 0.002). These genes were then analyzed in a subset of 44 cases in the Stage I cohort for whom ERβ protein expression status was known from immunohistochemistry (IHC) analysis. Comparing ERβ high (Allred score of >4, N= 20) and ERβ low (Allred score of 4 or less, N = 24) cases, the top PAM50 genes were more informative in ERβ high cases (HR 11.7, p = 0.0007) compared to ERβ low cases (HR 3.38, p = 0.045). For ERβ high cases, Kaplan-Meier survival curves showed that 50% of the high-risk group relapsed by 22 months, while 50% relapse was not reached in the low-risk group at 60 months. Similar analyses in the MammaPrint and Oncotype DX gene panels did not improve DFS prediction. We next validated the top PAM50 genes in a different cohort of 64 NSCLC cases, which included all stages and histologies, and had available ERβ IHC data, expression data from the Illumina Human HT-12 v4 BeadChip, and survival information. The top PAM50 genes were also able to predict DFS in this cohort: HR 2.19, p = 0.034 in all cases and HR 3.24, p = 0.042 in ERβ high cases. Pathway analysis indicated that the informative PAM50 genes describe a network that contains the estrogen and progesterone receptors, HER2, and cyclin E. The top canonical pathways identified were HER2/HER3 signaling and estrogen signaling. These results suggest that genes involved in interactions between hormonal and HER signaling may be predictive of lung cancer survival, especially in early-stage lung cancer, and provide further evidence for the importance of hormonal pathways in the biology of lung cancer. Supported by NCI SPORE in Lung Cancer P50 040990 and by the University of Pittsburgh Medical Center Genomics Initiative.

Citation Format: Jill M. Siegfried, Yan Lin, Sanja Dacic, Brenda Diergaarde, Laura P. Stabile, Hui-Min Lin, Marjorie Romkes. Prediction of lung cancer survival by genes in the PAM50 breast cancer panel. [abstract]. In: Proceedings of the AACR-IASLC Joint Conference on Molecular Origins of Lung Cancer; 2014 Jan 6-9; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2014;20(2Suppl):Abstract nr B20.