Expression profiling of nuclear receptors in the NCI60 cancer cell panel reveals receptor-drug and receptor-gene interactions

Abstract

We profiled the expression of the 48 human nuclear receptors (NRs) by quantitative RT-PCR in 51 human cancer cell lines of the NCI60 collection derived from nine different tissues. NR mRNA expression accurately classified melanoma, colon, and renal cancers, whereas lung, breast, prostate, central nervous system, and leukemia cell lines exhibited heterogeneous receptor expression. Importantly, receptor mRNA levels faithfully predicted the growth-inhibitory qualities of receptor ligands in nonendocrine tumors. Correlation analysis using NR expression profiles and drug response information across the cell line panel uncovered a number of new potential receptor-drug interactions, suggesting that in these cases, individual receptor levels may predict response to chemotherapeutic interventions. Similarly, by cross-comparing receptor levels within our expression dataset and relating these profiles to existing microarray gene expression data, we defined interactions among receptors and between receptors and other genes that can now be mechanistically queried. This work supports the strategy of using NR expression profiling to classify various types of cancer, define NR-drug interactions and receptor-gene networks, predict cancer-drug sensitivity, and identify druggable targets that may be pharmacologically manipulated for potential therapeutic intervention.

Figure 1

Heat map representation of mRNA expression levels of the 48 human NRs in a panel of human cancer cell lines (NCI60) as measured by qRT-PCR. Cell lines are grouped according to tissue of origin/cancer type as labeled on rows. Leuk, Leukemia cell lines; Melan, melanomas; NSCL, non-small-cell lung cancer; Prst, prostate cancer lines. The names of the receptors are given across the columns. Dark blue rectangles represent the highest levels of expression, and white rectangles represent low/undetectable expression (cycle time >35). Standard curves were used to quantify signals. The data were normalized to 18S and processed as described in Materials and Methods. Common names and abbreviations for each NR are given in Supplemental Table 3.

Figure 2

Patterns of mRNA expression across the NCI60 panel for select receptors. A, The HNF4α profile (top) shows almost exclusive expression in colon cancer cell lines as measured by qRT-PCR. In contrast, GR (bottom) is excluded from colon cancers but is otherwise ubiquitously expressed throughout other cancer types. B, The TRβ expression profile (top) shows this receptor to be largely excluded from melanomas. In contrast, the RXRγ (bottom) profile shows exclusive expression in melanomas and reclassifies MDA-MB-435 as a melanoma cell line. The abbreviations used for tissues are as in Fig. 1.

Figure 3

Receptor-specific patterns of expression help reclassify cancer cells and target cancer growth pharmacologically. A, Hierarchical clustering analysis of NR expression. Using the Matrix 1.2 software, NR expression data across all cell lines studied were subjected to centered, PCC-driven, average linkage analysis. Hierarchical tree schema on the right show cell line clusters, whereas tree schema on the top show receptor clusters. Note color coding of cell lines by cancer of origin. Reclassified or recoded lines by NR expression discussed in the text are underlined. Please refer to the text for details. B, Inhibition of cell viability in AR-positive melanomas by an AR antagonist. Melanomas show heterogeneous expression for AR. SKMEL2, which is AR negative, fails to respond to AR antagonist nilutamide, whereas AR-positive lines SKMEL28 and UCC62 are growth inhibited by nilutamide (left). Representative results of four independent experiments are shown. Error bars represent sd from eight replicate wells. Relative receptor expression levels as measured by qRT-PCR are shown in the middle panel, and protein levels analyzed by Western blotting are shown on the right panel. Note the decrease in AR levels in the AR⁺ lines after nilutamide treatment.

Figure 4

Pairwise PCC analysis of the NRs. Centered, average linkage clustering analysis was performed on a dataset corresponding to pairwise PCC comparisons across receptors and visualized using Eisen Cluster and Tree View software programs. Positive correlations (of 1 for receptor to itself) are depicted in red with the strongest intensities corresponding to higher correlations (0.88 was highest pairwise correlation). Similarly, negative correlations are depicted in green tending toward black as they become less intense. Note that the range of positive correlations (0–0.88) is much wider than that of negative correlations (0–0.3).