Integrated cistromic and expression analysis of amplified NKX2-1 in lung adenocarcinoma identifies LMO3 as a functional transcriptional target

Abstract

The NKX2-1 transcription factor, a regulator of normal lung development, is the most significantly amplified gene in human lung adenocarcinoma. To study the transcriptional impact of NKX2-1 amplification, we generated an expression signature associated with NKX2-1 amplification in human lung adenocarcinoma and analyzed DNA-binding sites of NKX2-1 by genome-wide chromatin immunoprecipitation. Integration of these expression and cistromic analyses identified LMO3, itself encoding a transcription regulator, as a candidate direct transcriptional target of NKX2-1. Further cistromic and overexpression analyses indicated that NKX2-1 can cooperate with the forkhead box transcription factor FOXA1 to regulate LMO3 gene expression. RNAi analysis of NKX2-1-amplified cells compared with nonamplified cells demonstrated that LMO3 mediates cell survival downstream from NKX2-1. Our findings provide new insight into the transcriptional regulatory network of NKX2-1 and suggest that LMO3 is a transcriptional signal transducer in NKX2-1-amplified lung adenocarcinomas.

Figure 1.

Genes associated with NKX2-1 amplification and overexpression in human lung adenocarcinoma. (A) Genes (12,328) are rank-ordered by differential expression score (SAM score) and plotted against the expected SAM score in 11 lung adenocarcinoma cell lines with high-level focal NKX2-1 amplification compared with 17 lung adenocarcinoma cell lines with the lowest NKX2-1 expression in a panel of 42 lung adenocarcinoma cell lines with matched copy number data from a 250K SNP array. The red circle indicates positively correlated genes with a SAM score that deviates from the expected distribution of the SAM score at a Delta slope of 0.26. The green circle indicates negatively correlated genes with score deviates from the expected distribution of the SAM score at a Delta slope of 0.26. (B) Genes (13,291) are rank-ordered by a differential expression SAM score and plotted against the expected SAM score in 47 (top 10th percentile) primary human lung adenocarcinomas with the highest expression of NKX2-1 compared with 47 (bottom 10th percentile) tumors with the lowest NKX2-1 expression from expression profiles of 470 primary lung adenocarcinomas. Genes are plotted against the expected distribution of the SAM score. The red circle indicates positively correlated genes with a SAM score that deviates from the expected distribution of the SAM score at a Delta slope of 3.9. The green circle indicates negatively correlated genes with a score that deviates from the expected distribution of the SAM score at a Delta slope of 3.9. There were more genes whose expression is positively correlated with NKX2-1 overexpression than negatively correlated. The signature tails, or overall departure from no gene effects (from 0 to 1) in the data set, for positive is 0.77 and for negative is 0.20. (C) Venn diagram depicting the overlap between positively correlated genes of the 1000 most differentially expressed genes associated with NKX2-1 amplification in lung adenocarcinoma cell lines and positively correlated genes of the 1000 most differentially expressed genes associated with the top 10th percentile highest NKX2-1 expression in primary lung adenocarcinomas (hypergeometric P-value of significant overlap <1 × 10⁻¹²). (D, left) Of the 147 overlapped genes, the expression level of the most differentially expressed 25 genes in 11 NKX2-1-amplified cell lines and 17 cell lines with the lowest NKX2-1 expression is shown in a heat map. Cell lines are ordered from left to right by copy number at the NKX2-1 locus. (Right) Of the 147 overlapped genes, the expression level of the most differentially expressed 25 genes in primary lung adenocarcinomas with the top and bottom 10th percentile NKX2-1 expression is shown in a heat map. Samples are ordered from left to right by NKX2-1 expression level. The complete list of the genes is available in Supplemental Table 1.

Figure 2.

NKX2-1 binds preferentially to promoter regions of the genes whose expression is correlated with NKX2-1 amplification. (A) The average NKX2-1 ChIP enrichment signal on the gene structure of all genes (gene bodies were normalized to 3 kb) was plotted in black. The signal is highest upstream of the TSS. The NKX2-1 ChIP signal is also plotted for genes whose expression is positively correlated with NKX2-1 amplification (red) and negatively correlated (cyan) and not correlated and low in NCI-H3122 cells (purple). NKX2-1 occupancy at TSSs is highest in genes whose expression is positively correlated. (B) NKX2-1 occupancy within 1 kb upstream of TSSs is significantly higher than expected by random chance. Of all enriched peaks, 9.2% are within 1 kb upstream of TSSs of any RefSeq genes that comprise 1.0% of the effective genome size. (C) The Venn diagram depicts overlaps of 7469 each of the most significantly enriched genomic intervals by NKX2-1 ChIP for three lung adenocarcinoma cell lines that harbor NKX2-1 amplification (NCI-H1819, NCI-H2087, and NCI-H3122).

Figure 3.

NKX2-1-occupied genes are significantly overrepresented in the NKX2-1-amplified lung adenocarcinomas. (A) ChIP signal within 0.5 kb upstream of and downstream from the center of the NKX2-1-bound peaks of 3358 common peaks that are highly significant (P < 1 × 10⁻³⁰) and 1158 peaks below the P-value of 1 × 10⁻⁶⁰ in a combined analysis for each cell line. Peaks are rank-ordered by combined P-value. The scale is shown on the right bar from 0 to 10 tag counts per 10-bp resolution. (B, left) The gene set with the genes that have the common NKX2-1 occupancy (1158 peaks with P < 10⁻⁶⁰) within 30 kb upstream of or downstream from their TSSs is significantly enriched in cell lines with high-level NKX2-1 amplification compared with the ones with the lowest NKX2-1 expression. (Right) The same gene set is also significantly enriched in primary lung adenocarcinomas with the highest NKX2-1 expression compared with the lowest.

Figure 4.

Motif analysis of NKX2-1-bound sequences identified AP-1, Forkhead, and nuclear hormone receptor-binding motifs. (A) De novo primary motif found in the most significant NKX2-1-bound sequences of the combined data shows similarity to the known binding motif of Nkx3-2 and zeste. (B) Other enriched motifs found in common NKX2-1-occupied regions in three NKX2-1-amplified cell lines (NCI-H1819, NCI-H2087, and NCI-H3122). The motifs were identified to be similar to known consensus binding motifs of AP-1 and the FOX family, respectively. (C) The known motif most significantly enriched at the fixed distance from the identified primary NKX2-1-binding motif was identified as the ERβ motif (P = 4.4 × 10⁻⁹²). Of 7538 regions that contain sequences of both the identified NKX2-1-binding motif and the ERβ consensus binding motif from NKX2-1-binding regions in any of three NKX2-1-amplified cell lines, 148 regions were found to have both motifs at a 15-bp distance. (D) Cell growth curve of an NKX2-1-amplified cell line, NCI-H3122, after suppression of FOXA1, ESR1, RARA, or LacZ control by shRNAs. Suppression of FOXA1 led to reduced cell viability, whereas that of ESR1 or RARA did not.

Figure 5.

NKX2-1 and LMO3 are the most differentially essential for cell survival in NKX2-1-amplified cell lines as compared with cell lines without amplification, and NKX2-1 is required for expression of LMO3. (A) Anti-proliferative effects were tested for five to 10 shRNAs per gene on four cell lines with NKX2-1 amplification (H1819, H2087, H3122, and HCC1833) and four cell lines without amplification or expression of NKX2-1 (H23, H1437, HCC461, and Calu-3). Individual differential essentiality scores for each shRNA (Z-score of differential effects for four cell lines vs. four cell lines) scaled on the right Y-axis are shown in red bars. The composite differential essentiality scores (weighted sum) for each gene tested scaled on the left Y-axis are shown as black bars. (B) mRNA expression of NKX2-1 and LMO3 measured by RT-qPCR in NCI-H2009 cells after introduction of two shRNA against NKX2-1 and one shRNA against GFP as control. LMO3 expression is reduced relative to the suppression of NKX2-1 expression. (*) P < 0.05; (**) P < 0.01. (C) Western blot for H2009 cell lysates with anti-NKX2-1, anti-LMO3, and anti-vinculin (loading control) antibodies after introduction of two shRNAs against NKX2-1 and one shRNA against GFP as control.

Figure 6.

NKX2-1 binds to ∼0.8 kb downstream from the TSS of the LMO3 gene and trans-activates LMO3 expression in cells with NKX2-1 amplification but not in cells without NKX2-1 amplification. (A) EMSA with in vitro translated NKX2-1 protein for two ³²P-labeled probes designed from the LMO3 gene locus. Electrophoretic mobility was shifted when incubated with NKX2-1 protein (lanes 2,6) and further shifted (supershift) by addition of anti-NKX2-1 antibody (lanes 4,8). (Lanes 3,7) Mobility shift was diminished with incubation of nonradioactive self-competitor probes. (B) Ectopic expression of NKX2-1 induced ∼10-fold expression of LMO3 in three cell lines (H3122, H2009, and H1819) that harbor NKX2-1 amplification. LMO3 expression was not induced to a similar level or reduced after ectopic NKX2-1 expression in three cell lines without NKX2-1 amplification (A549, AALE, and 293T cells). (C) ChIP enrichment (percent recovery of input) by anti-NKX2-1 antibody measured by qPCR at 0.7 kb downstream from LMO3 TSS and the SFTPB promoter for three cell lines (H1819, H2087, and H3122) with NKX2-1 amplification, A549 without NKX2-1 amplification, and A549 expressing ectopic NKX2-1. HBG1 serves as a negative control. NKX2-1 localizes to both LMO3 and SFTPB loci in NKX2-1-amplified cell lines but was not able to localize to the LMO3 locus in A549, whereas NKX2-1 was still able to occupy the SFTPB promoter region.

Figure 7.

NKX2-1 and FOXA1 colocalize at the genomic regions, including LMO3. (A) A view of the LMO3 gene locus of shifted read counts from ChIP-seq by either the anti-NKX2-1 or anti-FOXA1 antibody as well as read counts from the no-ChIP input control. The location of sequences taken for two probes designed for gel shift assay is shown as red bars. (B) The model depicting cooperative occupancy of NKX2-1 and FOXA1 on the locus downstream from the TSS of the LMO3 gene to trans-activate expression of LMO3.