Loss of RUNX1 is associated with aggressive lung adenocarcinomas

Abstract

The mammalian runt-related factor 1 (RUNX1) is a master transcription factor that regulates lineage specification of hematopoietic stem cells. RUNX1 translocations result in the development of myeloid leukemias. Recently, RUNX1 has been implicated as a tumor suppressor in other cancers. We postulated RUNX1 expression may be associated with lung adenocarcinoma etiology and/or progression. We evaluated the association of RUNX1 mRNA expression with overall survival data from The Cancer Genome Atlas (TCGA), a publically available database. Compared to high expression levels, Low RUNX1 levels from lung adenocarcinomas were associated with a worse overall survival (Hazard Ratio = 2.014 (1.042-3.730 95% confidence interval), log-rank p = 0.035) compared to those that expressed high RUNX1 levels. Further immunohistochemical examination of 85 surgical specimens resected at the University of Vermont Medical Center identified that low RUNX1 protein expression was associated with larger tumors (p = 0.038). Gene expression network analysis was performed on the same subset of TCGA cases that demonstrated differential survival by RUNX1 expression. This analysis, which reveals regulatory relationships, showed that reduced RUNX1 levels were closely linked to upregulation of the transcription factor E2F1. To interrogate this relationship, RUNX1 was depleted in a lung cancer cell line that expresses high levels of RUNX1. Loss of RUNX1 resulted in enhanced proliferation, migration, and invasion. RUNX1 depletion also resulted in increased mRNA expression of E2F1 and multiple E2F1 target genes. Our data implicate loss of RUNX1 as driver of lung adenocarcinoma aggression, potentially through deregulation of the E2F1 pathway.

Keywords: RUNX; adenocarcinoma; non-small cell lung cancer; runt related transcription factor.

Figure 1

RUNX1 expression correlates with a survival benefit in LUAD patients. A, The TCGA-LUAD cohort was stratified for RUNX1 expression. A, RUNX1 (mRNA-Seq) expression distribution amongst this cohort identifies individuals in the upper (red) and lower (blue) 10th percentiles. B, Kaplan-Meyer analysis was performed comparing overall survival between High RUNX1 expressers (upper 10th percentile) and Low RUNX1 expressers (lower 10th percentile). Log-rank P value was calculated using the Mantel-Cox test. Median survival was determined to be 1,798 days for high expressers compared to 1,073 for low expressers.

Figure 2

RUNX1 immunohistochemical staining of surgically resected lung adenocarcinomas. Panel A: Image (i) demonstrates immunohistochemial of the negative control. Image (ii) demonstrates significant RUNX 1 staining (brown) throughout the adenocarcinoma specimen. Panel B: The majority of the adenocarcinomas were RUNX1 negative (59%) and were associated with significantly larger pathologic tumor sizes (P=0.0380)

Figure 3

RUNX1 levels are reduced in multiple NSCLC cell lines. A, RUNX1 transcript levels were quantified by qRT-PCR in a normal human bronchial epithelial cell line (HBE) and a panel of NSCLC cell lines. Bars represent the means and standard deviations from three replicate cell preparations. B, RUNX1 protein levels were determined by western blot. Relative gel loading was assessed by using Bio-Rad StainFree™ technology (bottom panels). The band marked “ns” represents a non-specific staining band. In addition to HBE, RUNX1 transcript levels were elevated in cell lines NCI-H292 and NCI-H1568, however only RUNX1 protein levels were observed to be elevated in NCI-H292. Therefore, our subsequent studies focused on the NCI-H292 cancer cell line and the normal epithelial-like HBE cell line. Bars represent mean ± s.d. from triplicate experiments. T-tests were performed to determine statistical significance (*P<0.05).

Figure 4

Knockdown of RUNX1 enhances proliferation of NSCLC and normal bronchial epithelial cells, and the migratory and invasive phenotype of NSCLC cell line NCI-H292. A, Lentiviral delivery of control shRNA or shRNA targeting RUNX1 shows stable knockdown of RUNX1 in NCI-H292 and HBE by western blot. Cell proliferation studies by MTS assays show that RUNX1 knockdown in NCI-H292 (B) and HBE (C) increases proliferation rate. Best-fit doubling times for shControl and shRUNX1 were determined to be 36.35 (34.29 to 38.68; 95% CI) and 31.63 (29.44 to 34.18) hours, respectively (P=0.0101). D, Migratory capacity as a consequence of RUNX1 knockdown in NCI-H292 was determined using conventional scratch assays. Scratch closure values were fit to a linear equation to quantitate migration rates, and shows RUNX1 knockdown increases rate of migration in NCI-H292. Best-fit values for closure rates were determined to be 7.004 (5.372 to 8.636; 95% CI) and 16.13 (9.996 to 22.26) pixels per hour for shControl and shRunx1, respectively (P=0.004). E, Matrigel transwell assays show increased invasion capacity when RUNX1 is depleted in NCI-H292. Crystal violet dye was extracted from stained invading cells (through a millipore membrane and quantities determined by spectrophotometry. Symbols and bars represent mean ± s.d. from a minimum of triplicate experiments. In panel E, data points represent values obtained for individual biological replicates. A Student’s t-test was performed to determine statistical significance (*P<0.05).

Figure 5

RUNX1 depletion in NCI-H292 and HBE results in alteration of multiple genes implicated in NSCLC. NCI-H292 and HBE transfected with lentivirus expressing control-, or Runx1-specific shRNA were analyzed by qRT-PCR to assess consequential changes in expression levels of genes shown to be important for several aspects of LAc pathogenesis. Data points represent values obtained for individual biological replicates. Bars represent mean ± s.d. from a minimum of triplicate experiments. T-tests were performed to determine statistical significance (*P<0.05).

Figure 6

Low RUNX1 expression in lung adenocarcinoma correlates with a phenotype driven by an E2F1-directed gene program. A, Differential gene expression analyses between Low (lower 10th percentile) and High (upper 10th percentile) RUNX1 expressers was performed using unbiased hierarchical clustering. Individual genes displaying significant (adjusted P value < 0.05) were submitted for analysis. Based on this approach, two major clusters, denoted Cluster I and Cluster II, were identified. B, Netwalk analysis reveals E2F1 as a major node of regulation for a large subset of genes that are upregulated in patients with low RUNX1 expression (cluster II). Solid red lines denote gene regulation interactions; dotted red lines denote predicted gene regulation interactions; dot-hashed red lines denote ENCODE identified interactions; blue lines denote metabolic reaction interactions. C, An independent Gene Set Enrichment Analysis that included all genes in the dataset (no filtering based on statistical significance was performed) also identified an E2F1 network as being enriched in patients with low RUNX1 expression (Normalized enrichment score −3.67, family-wise error rate P <0.0001).

Figure 7

RUNX1 depletion in NCI-H292 and HBE results in an increase in E2F1 and associated target genes. NCI-H292 and HBE transfected with lentivirus expressing control-, or Runx1-specific shRNA were analyzed by qRT-PCR to assess consequential changes in expression levels of E2F1 canonical target genes, and those identified to be negatively correlated with RUNX1 expression in LUAD patients through bioinformatic analysis. Data points represent values obtained for individual biological replicates. Bars represent mean ± s.d. from triplicate experiments. T-tests were performed to determine statistical significance (*P<0.05).