NF-YA Overexpression in Lung Cancer: LUAD

Abstract

The trimeric transcription factor (TF) NF-Y regulates the CCAAT box, a DNA element enriched in promoters of genes overexpressed in many types of cancer. The regulatory NF-YA is present in two major isoforms, NF-YAl ("long") and NF-YAs ("short"). There is growing indication that NF-YA levels are increased in tumors. Here, we report interrogation of RNA-Seq TCGA (The Cancer Genome Atlas)-all 576 samples-and GEO (Gene Expression Ominibus) datasets of lung adenocarcinoma (LUAD). NF-YAs is overexpressed in the three subtypes, proliferative, inflammatory, and TRU (terminal respiratory unit). CCAAT is enriched in promoters of tumor differently expressed genes (DEG) and in the proliferative/inflammatory intersection, matching with KEGG (Kyoto Encyclopedia of Genes and Genomes) terms cell-cycle and signaling. Increasing levels of NF-YAs are observed from low to high CpG island methylator phenotypes (CIMP). We identified 166 genes overexpressed in LUAD cell lines with low NF-YAs/NF-YAl ratios: applying this centroid to TCGA samples faithfully predicted tumors' isoform ratio. This signature lacks CCAAT in promoters. Finally, progression-free intervals and hazard ratios concurred with the worst prognosis of patients with either a low or high NF-YAs/NF-YAl ratio. In conclusion, global overexpression of NF-YAs is documented in LUAD and is associated with aggressive tumor behavior; however, a similar prognosis is recorded in tumors with high levels of NF-YAl and overexpressed CCAAT-less genes.

Keywords: CCAAT box; LUAD; NF-YA; TCGA; alternative splicings; lung cancer; transcription factors.

Figure 1

NF-YA was overexpressed in lung adenocarcinoma (LUAD). (A) Box plots of expression levels of the three NF-Y subunits at gene level in the TCGA LUAD cohort, measured in TPMs (Transcripts per Millions). (B) Same as (A), except that the GSE40419 dataset was analyzed. (C) Same as (A), except that 91 LUAD tumors were analyzed from the previously classified non-small cell lung carcinoma (NSCLC) GSE81089 dataset. p-values were calculated using a Wilcoxon signed-rank test.

Figure 2

NF-YA short isoform was overexpressed in LUAD. (A) Box plots of expression levels of NF-YAs (“short”) and NF-YAl (“long”) in the TCGA LUAD dataset. (B) Same as (A), except that the GSE40419 dataset was analyzed. (C) Same as (A), except that LUAD tumors of the GSE81089 dataset were used. (D) Relative ratios of NF-YAs/NF-YAl in TCGA-LUAD. (E) Same as (D), except that the GSE40419 dataset was analyzed. (F) Same as (D), except that the LUAD tumors of the GSE81089 dataset were used.

Figure 3

NF-YA was overexpressed in all LUAD subtypes. (A) Box plots represent the expression of NF-YA isoforms at gene level in LUAD subgroups, measured in TPM. (B) NF-YAs/NF-YAl ratios in the subgroups. (C) Same as (A), except that expression of NF-YB at gene level is shown. (D) Box plots of expression of NF-YC isoforms at gene level in LUAD subgroups. NF-YC1—50 kD isoform. NF-YC2—37 kD isoform. p-values were calculated using a Wilcoxon signed-rank test.

Figure 4

Gene expression analysis of LUAD TCGA tumors. (A) Up- and down-regulated genes in LUAD versus normal lung tissues. NS, not significant. (B) Pscan analysis of enriched TFBS in promoters (−450/+50 bps from the TSS) of up- and down-regulated genes in LUAD. (C) Weeder analysis of enriched de novo matrices in promoters of upregulated genes. (D) Reactome pathways enriched in upregulated genes (upper panel) and down-regulated genes (lower panel) listed according to their p-value. The list was obtained using KOBAS.

Figure 5

Expression levels of NF-YA isoforms in LUAD cell lines. (A) Relative levels of expression of the two NF-YA isoforms (TPMs) of the indicated LUAD cell lines as determined by RNA-Seq experiments. (B) Western blot analysis of NF-YA protein levels in the indicated representative LUAD cancer cell lines. Vinculin was used as an internal loading control.

Figure 6

Identification and validation of a gene signature in NF-YAs/NF-YAl low ratio cell lines. (A) Overall scheme of the analysis. (B) Heatmap of expression levels of 125 genes identified as upregulated in low ratio cell lines. Color key indicates the Z-score. (C) Boxplots of NF-YAs/NF-YAl ratio distribution across TCGA LUAD samples after partitioning based on cell line signature. Analysis of the 130 centroid genes in LUAD tumors with high and low NF-YAs/NF-YAl ratios. p-values were calculated using a Wilcoxon signed-rank Z-test with Bonferroni correction.

Figure 7

LUAD subclasses have different NF-YAs/NF-YAl ratios. (A) Subgroup distribution of LUAD tumors partitioned in high (left) and low (right) NF-YAs/NF-YAl ratios, according to the cell line-based centroids. (B) Subgroup distribution of LUAD tumors partitioned on the basis of sample NF-YAs/NF-YAl ratios.

Figure 8

Clinical outcome of LUAD tumors with different NF-YAs/NF-YAl ratios. (A) Progression-free interval curves of survival probability of LUAD tumors with stratification according to quartiles of NF-YAs/NF-YAl ratios (intermediate, high, and low). (B) Hazard ratios of the three cohorts as above, according to stage (I–IV), gender, sub-type (Basal set as reference), and NF-YAs/NF-YAl ratios. p-values were calculated using a Cox proportional hazards regression analysis (See Section 2.5).