Enzalutamide response in a panel of prostate cancer cell lines reveals a role for glucocorticoid receptor in enzalutamide resistant disease

Abstract

Representative in vitro model systems that accurately model response to therapy and allow the identification of new targets are important for improving our treatment of prostate cancer. Here we describe molecular characterization and drug testing in a panel of 20 prostate cancer cell lines. The cell lines cluster into distinct subsets based on RNA expression, which is largely driven by functional Androgen Receptor (AR) expression. KLK3, the AR-responsive gene that encodes prostate specific antigen, shows the greatest variability in expression across the cell line panel. Other common prostate cancer associated genes such as TMPRSS2 and ERG show similar expression patterns. Copy number analysis demonstrates that many of the most commonly gained (including regions containing TERC and MYC) and lost regions (including regions containing TP53 and PTEN) that were identified in patient samples by the TCGA are mirrored in the prostate cancer cell lines. Assessment of response to the anti-androgen enzalutamide shows a distinct separation of responders and non-responders, predominantly related to status of wild-type AR. Surprisingly, several AR-null lines responded to enzalutamide. These AR-null, enzalutamide-responsive cells were characterized by high levels of expression of glucocorticoid receptor (GR) encoded by NR3C1. Treatment of these cells with the anti-GR agent mifepristone showed that they were more sensitive to this drug than enzalutamide, as were several of the enzalutamide non-responsive lines. This is consistent with several recent reports that suggest that GR expression is an alternative signaling mechanism that can bypass AR blockade. This study reinforces the utility of large cell line panels for the study of cancer and identifies several cell lines that represent ideal models to study AR-null cells that have upregulated GR to sustain growth.

Figure 1

Expression of variable and selected genes and proteins in prostate cancer cell lines. (A) Hierarchical clustering of cell lines and the thousand most variable genes reveals distinct clusters. Several of the clusters are directly related to AR function and activity. PSA (KLK3) is the most variably expressed gene in the panel, but only shows weak correlation with AR expression. Similarly, TMPRSS2 and ERG also show weak associations with AR RNA expression. An additional cluster of testes antigen genes is strongly expressed in LNCaP cells and derivatives, as well as several additional cell lines. (B) Clustering of cell lines mirrors clustering prostate tumors using genes identified by the TCGA prostate cancer. The clustering of the genes is also highly similar, although CHGA clusters with different genes in our data set. Furthermore, expression of PCOTH and GATA4 are extremely low in the cell lines, suggesting that their expression may be confined to stromal cells in vivo. (C) Western blot analysis of AR and ERG expression in the prostate cancer cell line panel. Cropped images highlighting the bands specific to AR and ERG are shown. Relative AR:GAPDH expression levels are shown below, color coded by low (blue), moderate (white), and high (red) AR:GAPDH Full, unaltered blots are available in supplementary Fig. S1.

Figure 2

Copy number analysis of prostate cancer cell lines. (A) Example of ASCAT copy number estimates (green line) for the prostate cancer cell line PC346C demonstrating regions of gain and loss. (B) IGV view of copy number changes in individual cell lines and the average copy number changes in the entire cell line panel. (C, D) GISTIC analysis of recurrent regions of copy number gain (C) and loss (D). Many of the recurrent regions of copy number alterations observed in the cell lines are the same as those seen in prostate cancer specimens as defined by the TCGA.

Figure 3

Enzalutamide response in the prostate cancer cell line panel. (A) Example dose response curves for a responsive (blue, DU145) and non-responsive (red, MDAPCa1) prostate cancer cell line. Error bars are +/− standard deviation of triplicate measurements. (B) GI50 values for each of the cell lines divides the samples into responsive (blue) and non-responsive (red) clusters. Lower bars represent lower doses of drug required to inhibit growth by 50%. For non-responsive lines, GI50 values were set to maximal dose tested. (C,D) GSEA plots for Estrogen and Androgen response elements show differences between enzalutamide responders and non-responders. The association is significant for ER associations (p < 0.05) but fails to reach significance for AR (p = 0.12).

Figure 4

Expression of other NR3 family members in prostate lines and assessment of GR as a potential target in prostate cancer cell lines. (A) Expression of NR3 nuclear receptors in the prostate cancer cell line. NR3C1 expression is largely inversely correlated with AR expression. (B) Expression of ESR1 (blue) or NR3C1 (orange) in AR-null responders (left) or non-responders (right) to enzalutamide, showing that the expression of both of these family members is significantly higher in the cell lines that responded to enzalutamide (p < 0.05). (C) Western blot analysis of glucocorticoid receptor in a subset of the prostate cell lines. Note expression was positive in all of the lines tested except for the AR positive cell line LNCaP. Full, unaltered blots are available in supplementary Fig. S1. (D) Expression of NR3 family members in the TCGA panel of prostate cell lines. Distinct clusters are evident, including tumors that lack expression of AR but show strong expression of NR3C1 (boxes). (E) Treatment of cells with dexamethasone (50 or 100 nM) results in down-regulation of GR protein, including in the presence of enzalutamide, in DU145, HH870, and PC3 cells. (F) Dexamethasone treatment significantly impacts cell growth and enzalutamide response in DU145 and HH870 cells (significant differences marked with bars to show comparisons tested and * to indicate significance). Treatment with either 50 or 100 nM dexamethasone inhibits the growth of DU145 cells (peach and yellow colored bars, p < 0.001), similar to the treatment with enzalutamide (blue bar, p < 0.001). Addition of dexamethasone to enzalutamide (light blue and dark blue bars) result in significant growth inhibition compared to enzalutamide alone (p < 0.001 in both cases). In HH870 cells, dexamethasone treatment (peach and yellow colored bars, p < 0.005) enhances cell growth compared to control. Treatment with 100 nM dexamethasone (dark blue bar) results in significant growth inhibition compared to enzalutamide alone (p < 0.05). Treatment with 50 nM dexamethasone plus enzalutamide also shows a trend towards enhancing response (p = 0.069).

Figure 5

Growth of prostate cancer cell lines treated with enzalutamide versus mifepristone. (A) Response of cell lines to enzalutamide (red) or mifepristone (blue) shows that mifepristone is more efficacious in several of the cell lines. (B) Assessment of plating efficiency of cell lines used in clonogenic assay testing mifepristone versus enzalutamide response. (C) Crystal violet stain for colony formation in DU145, PC3, and MDAPCa1 cells. 25 µM enzalutamide has minimal impact on colony formation in these cells compared to vehicle controls, whereas an equivalent dose of mifepristone almost completely inhibits the ability of these cells to form colonies. (D) Quantification of the response to mifepristone and enzalutamide in three AR null cell lines (DU145, PC3, and MDAPrCA1) shows that mifepristone inhibits colony growth more robustly than enzalutamide (at 25 µM each).