MicroRNA expressions associated with progression of prostate cancer cells to antiandrogen therapy resistance

We used genome-wide miRNA array (1113 unique primers) profiling approach to identify specific miRNAs that are involved in development of resistance to Casodex (CDX). A clonal subline of LNCaP cells LNCaP-104S (-104S) and its androgen-independent derivative LNCaP-104R1 (-104R1) were used for monitoring differential expression of miRNAs upon treatment with CDX. LNCaP-104S cells are CDX-sensitive, whereas LNCaP-104R1 cells are not despite expressing AR at a basal level higher than LNCaP-104S cells [24]. LNCaP-104S cells require DHT for maintaining their AD status but when treated with CDX for 3 weeks in CS-FBS (charcoal-stripped FBS), CDX insensitive colonies develop that are independent of androgen (CDXR). During the first week of both CS-FBS and CDX treatments, LNCaP-104S cells grew without significant cell death, however, during the second week of CDX treatment about 60% cells died and detached. From the residual 40% cells, half of the cell population remained viable during the third week of CDX treatment. CS-FBS -treated cells did not show as much cell death during the second week of treatment and started to regrow during the third week of treatment. Cell morphology also changed as they are treated with CDX and CS-FBS (Additional file 1: Figure S1). Cells viable after third week of treatments were used for miRNA profiling experiments. Expression of AR in these cells showed gradual reduction as the treatment progresses, but the PSA expression increased, which suggests increased transcriptional activity of the residual AR in the treated cells (Figure 1).

We compared miRNA expressions in RNA extracts from untreated and CDX treated -104S cells at 0 hr, 1wk and 3wks and untreated -104R1 cells (Table 1). Hierarchical clustering of the normalized and log₂ transformed expression data (Additional file 2: Table S1) showed 5 distinct clusters of miRNAs (Additional file 3: Figure S2). Comparative analysis between samples from different conditions using two samples Welch t-test with Log₁₀ transformed data and p values of 0.05 showed significant miRNAs that are differentially expressed between conditions (Additional file 4: Table S2 and Additional file 5: Figure S3). Volcano plot (V plots) of the t-test between LNCaP-104S cells and all other samples showed 38 significant miRNAs, of which 27 miRNAs were up regulated and 11 down regulated compared to -104S (Additional file 5: Figure S3A). Comparison between untreated -104S and -104R1 cells showed 24 significant miRNAs, which includes 16 down regulated and 8 up regulated miRNAs in -104R1 (Additional file 5: Figure S3B). Differential expression of 17 significant miRNAs was observed between untreated LNCaP-104S cells and -104S cells treated with CDX, of which 13 were up regulated and 4 were down regulated CDX treated cells (Additional file 5: Figure S3C). LNCaP-104S and -104S cells treated with CSFBS also showed 9 up regulated and 5 down regulated microRNAs in CSFBS treated cells (Additional file 5: Figure S3D). Although -104R1 cells are CDX resistant there are differences in miRNA expression when -104S cells were treated with CDX (Additional file 5: Figure S3E). T-test analysis showed 24 significant miRNAs of which 18 miRNAs were up regulated and 6 down regulated in -104R1 cells. Difference in miRNA expressions was also noted between -104S cells maintained in androgen-depleted condition and AI -104R1 cells. Twenty-four significant miRNAs were identified of which 12 were up regulated and 12 down regulated in -104R1 cells (Additional file 5: Figure S3F). Comparison between androgen depletion and CDX treatment showed 5 significant miRNAs, 4 of which were up regulated and one down regulated in CDX treated cells (Additional file 5: Figure S3G).

Clustering analyses using log₂ transformed fold change (FC) values of four treatment conditions compared to -104S untreated cells showed two distinct clusters of up and down regulated miRNAs, which includes 307 down regulated and 197 up regulated miRNAs (Figure 2 and Additional file 6: Table S3). K-median clustering for the up-regulated miRNAs showed a trend of gradual increase in median FC in miRNAs in some clusters (cluster 1, 4 and 9) and a gradual decrease in some clusters (clusters 3, 7 and 8) from 1 week to 3 weeks treatments (Figure 3A and Additional file 7: Table S4). In down regulated profile there is also a trend of gradual decrease in median expression of miRNAs in some clusters (clusters 1, 2, 3 and 7) (Figure 3B and Additional file 7: Table S4). Of these lists, 100 miRNAs were chosen based on fold change and/or the z score ≥3.0 or ≤ -3.0 for validation using qPCR.

Analysis of qPCR data indicated a variable expression profile of a subset of miRNAs in LNCaP cells exposed to CDX and/or androgen withdrawal (CSFBS) for different time periods. Two sample Welch t-test with a p-value < 0.05 showed a significant change in expression of 21 miRNAs in treated samples compared to the untreated -104S cells (Figure 4A and Additional file 8: Table S5). Comparison between -104S untreated and either CDX or CSFBS treated samples identified 10 and 8 significant miRNAs respectively (Figure 4A) Comparative expression analysis between CDX 1wk and CDX 3wks also revealed significant change in expression in 3 miRNAs (Figure 4A). Venn Diagram of miRNA expression profiles following CDX treatment and androgen withdrawal indicated two common miRNAs (miR-146a and miR-759) but treatment specific differential expression was observed with 8 and 6 unique miRNAs, in CDX and CSFBS treated respectively (Figure 4B). Analysis of the up-regulated miRNAs revealed 10 miRNAs that showed 2-fold or higher expressions in all treatment conditions and time points (Figure 4C). There are also 30 miRNAs that showed 2-fold or higher expression in more than one treatment conditions. In the down regulated list, 9 miRNAs showed at least 2-fold reduction in expression in all treatment conditions and 15 miRNAs with ≥2.0-fold down regulation in more than one treatment conditions (Figure 4D).

Cluster analysis of log₂ transformed FC in expression of 100 miRNAs showed three distinct clusters displaying common expression changes in 2 time points of two eatment conditions. Two clusters contained miRNAs with increased expression and one cluster containing down regulated miRNAs (Figure 5A and Additional file 9: Table S6). K-median clustering of the up-regulated miRNAs showed a distinct trend of gradual increase (clusters 1 [32%], 2 [26%] and 4[11%]) and gradual decrease (cluster 3 14%) in FC expressions as the treatment progressed (Figure 5B and Additional file 10: Table S7). Some of the miRNAs (21) showed across the board up regulations (Table 2). Similarly, in the list of down-regulated miRNAs a gradual decrease in expression (clusters 1[18%] and 2 [21%]) could be noted as the treatment progressed (Figure 5C) and a subset of miRNAs (22) showed down-regulation in all treatment conditions (Table 3). When compared with published expression database there are supporting reports on expression patterns of specific miRNAs in different types of cancer (Tables 2 and 3). There are also opposing reports on expression of miRNAs, which includes, miR-106a [25], miR-15b [26], miR-17 [27], miR-18 [28], miR-205 [29], miR-20 [30] and miR-7 [31] for down regulated miRNAs; and let-7f-1 [32], miR-136a [33], miR-143 [34], miR-146a [35], miR-218 [36], miR-22 [37], miR-222 [38] miR-29a [39], miR-34b [40], and miR-493 [41] for up-regulated miRNAs.

MicroRNAs exhibiting up regulation or down regulation in all treatment conditions compared to untreated -104S cells (Tables 2 and 3), were used for function and disease relevance to understand the possible alterations in the ellular processes as LNCaP cells progressed towards androgen withdrawal and AR antagonist resistance. We used RT-PCR FC values of the up regulated or down regulated miRNA for the analysis using IPA software (Ingenuity Systems). Based on -log (p-values) with -values <0.05, a percentage of miRNAs were assigned to certain disorder or cellular processes (Additional file 11: Figure S4). When the functional profiles of the p-regulated miRNAs were compared, there is a decrease in the percentage of miRNAs involved in cancer, dermatological disease, and hematological disease between 3wks and 1wk CDX treated cells whereas an increase in the percentage of miRNAs involved in endocrine system disease, gastrointestinal disease hepatic system disease, reproductive system disease and metabolic disease in 3wks CDX treated S cells compared to 1wk CDX treated cells (Additional file 11: Figure S4A). In cells subjected to androgen withdrawal, there are also differences in percentage of miRNAs in different cellular processes between 1wk and 3wks treatments, which show a reduction in hematological diseases, cellular development, and reproductive system disease. An increased involvement of miRNAs in a variety of cellular processes also was noted in 3wk CSFBS treated cells, which includes cell death and survival, cell movement, endocrine system disease, gastrointestinal disease, hematological disease, hepatic system disease and metabolic disease (Additional file 11: Figure S4A).

In the list of down-regulated miRNAs, there is an increase in percentage of miRNAs involved in cell morphology, cell movement, dermatological disease, gastrointestinal disease, renal urological disease, and reproductive system disease as cells progressed from 1wk to 3wks CDX treatment. A decreased percentage of down regulated miRNAs in specific cellular processes also could be noted in these cells, which includes cellular development, endocrine system disease, hepatic system disease, tumor morphology and inflammatory response. Androgen withdrawal for 1wk to 3wks also showed changes in percentage of miRNA involvement (Additional file 11: Figure S4B). An increased percentage of down regulated miRNAs involved in cancer, cell death and survival, dermatological disease, endocrine system disease, gastrointestinal disease, metabolic disease, and tumor morphology is noted in 3wks CSFBS treated cells. 3wks androgen withdrawal also showed a decrease in miRNA percentage involved in cell growth and proliferation, cell movement, cell-cell signaling, DNA replication and repair and renal-urological disease (Additional file 11: Figure S4B).

The in silico function analysis of the target miRNAs in different treatment conditions indicated a complex interaction of a network of miRNAs and their target proteins in these cells which rendered them adaptive to the androgen withdrawal and treatments with AR antagonists. Next, we analyzed the network of interactions among target miRNAs. We used the log₂ transformed FC values of the subset of validated miRNA (Tables 2 and 3) for analysis of the functional interrelationship among miRNAs (Figure 6). The network for up-regulated miRNAs and its putative targets in 3wks CDX showed that miR-3192 directly and miR-218 indirectly through DEAD box protein DDX20 target p53. MiR-146a also directly regulates a number of Toll-like receptors (TLR1, TLR9 and TLR10), cytokine receptors and its associated proteins (IL1R1, IL12RB2, IL1RAP) and chemokine receptors (CXCR4, CCR3) (Figure 6A). In addition, miR-146a directly targets a number of growth factors and cytokines (Figure 6A). These miR-146a targets are also shown in the up-regulated network of 3wks CSFBS, in addition to BRCA, which is a direct target of miR-146a (Figure 6E). MiR-3192 targeting of p53 is also noted in 3wks CSFBS but not of miR-218 (Figure 6E). As noted in the networks, p53 also regulates a number of miRNAs, such as, miR-29b-3p, miR-22-3p, miR-22-5p and miR-221-3p (Figure 6A and E). When the networks are overlaid with diseases, a number of miRNAs in the up-regulated list of both 3wks CDX and 3wks CSFBS showed involvement in a variety of cancers including pancreatic cancer, endocrine gland tumor, prostate cancer, ovarian tumor, mammary tumor, cervical cancer and epithelial neoplasia (Figure 6B and F).

In the list of down-regulated miRNAs there is direct targeting of E2F2, E2F3, MAP3K12, FXR1 and AR interacting nuclear protein kinase HIPK3 by miR-17-5p. MiR-16p also directly targets BCL2L2, cell death suppressor BNIP2, EGFR, RAB21 and single strand purine rich DNA binding protein PURA (Figure 6C and G) in both 3wks CDX and 3wks CSFBS treated cells. Also a number of miRNAs in the down regulated list down regulate a number of common targets, such as E2F3 is targeted by miR-1244, miR-596, miR-18a-5p, miR-16-5p and miR-17-5p. EGFR is also targeted by miR-7a-5p and miR-16-5p. Overlay of diseases on networks revealed a distinct association of miR-17-5p, with various cancers including prostate cancer, breast cancer and pancreatic cancer and endocrine gland tumors. An association of miR-16-5p with prostate cancer, pancreatic cancer, pituitary cancer and endometrial cancer also could be noted (Figure 6D and H). These networks indicate a complex interplay of microRNAs and proteins during development of resistance of LNCaP cells to androgen withdrawal and treatment with CDX.

A total of 21 up regulated and 22 down regulated miRNAs in all treatment conditions were used for identification of protein targets using IPA, miRDB and TargetScan software. Venn diagram of the putative targets of up-regulated miRNAs derived from the network generated using the IPA software showed 27 proteins that are common in all treatment conditions (Figure 7A). A number of which are known tumor suppressors such as BRCA1[87], TP53 [88], RAD54L [89], IRF5 [90], DUSP2 [91], IKK1 or CHUK [92]. Other targets of up regulated miRNAs in different treatment conditions also includes proteins that inhibit tumor progression, such as DNMT3A [93], FADD [94], pTEN [95], FOXO3[96], DDX20 [97] and PA2G4 (EBP1)[98]. There are also 41 common targets of the down-regulated miRNAs in all treatment conditions (Figure 7B). Some of the targets are known oncoproteins, which includes EGFR [99], VEGFA [100], ZBTB7 (POKEMON) [101], acid phosphatase 2(ACP2) [102] and NFKB1 [103]. Additionally, proteins that are overexpressed in cancer cells are among the targets of down-regulated miRNAs in different treatment conditions, which includes Wnt3A [104], PPARα [105], UCP2 [106], CSF1 [107], and MED1 [108].

Further identification of the targets of selected miRNA (Table 2) was conducted by miRDB and TargetScan database search and proteins that are regulated by one or multiple miRNAs from our list and received higher target scores in either or both searches are presented in Tables 4 and 5. The selected proteins that are regulated by the up-regulated miRNAs include TNF receptor associated factor 6 (TRAF6) [109], interleukin receptor associated kinase1 (IRAK1), BCL6 co-repressor-like 1(BCORL1) [110], Cbl [109], neuro oncological ventral antigen1 (NOVA1) [111], coiled coil domain containing 67 (CCDC67) [112] NET1 [113], ZFAND1(Acc.# NM_024699), RGS6 [114] CDKN1B (p27Kip1) [115], IGFBP5 [116] and RNASEL [117] (Table 4). The proteins that are potentially regulated by the down-regulated miRNAs include, ABHD3 [118], FGD4 [119], CCNJ [120], CHAMP1 [121], Myb [122], PIK3CD [85], VEGFA [123], SPOPL [124], RAB9B [125], EGFR [126], E2F1 [127] and DOK4 [128]. Western blot analysis confirmed the predicted expression profiles of some of the targets. Expression of Cbl, p27Kip1, TRAF6, IRAK1, and ZFAND1 were significantly and progressively down regulated in cells treated with CDX or subjected to androgen deprivation (Figure 8). Similarly, expression of FGD4, VEGFA, EGFR, DOK4 and ABHD3 were up regulated to moderate to high levels in cells treated with CDX or CSFBS (Figure 9).

The androgen responsive LNCaP-104S and androgen independent LNCaP-104R1 cells were generous gifts from Dr. Shutsung Liao, University of Chicago. These LNCaP sublines were isolated and characterized as previously described [146]. LNCaP-104S cells were maintained in DMEM (Gibco) containing 10% FBS (Atlanta Biologicals), 1nM DHT (Sigma-Aldrich), 1% Antibiotic/Antimycotic (Invitrogen). LNCaP-104R1 cells were maintained in DMEM containing 10% charcoal-stripped FBS (CSFBS), 1% Antibiotic/Antimycotic. Isolation of androgen independent cells was conducted by passaging LNCaP-104S cells in DMEM/10% CSFBS, supplemented with or without 5 μM Bicalutamide (Fluka) (Casodex, CDX) for 3 weeks (detail treatment method is in the Additional file 12 supplemental method section). Cells were harvested for protein or RNA extraction at 7 and 21 days post treatment.

Total RNA was extracted from untreated and treated cells using the Cell-to-Cts kit (System Biosciences) according to the manufacturer’s instructions. Total RNA was converted to cDNA utilizing the QuantiMir RT Kit (System Biosciences) according to the manufacturer’s instructions. Briefly, small RNAs were first tagged with polyA-tails; oligo-dT primers were annealed next, and converted to cDNA by reverse transcription.

Expression of mature miRNAs in untreated and treated LNCaP cells was determined by quantitative real-time PCR (qRT-PCR) using the miRNome microRNA Profiling Kit (System Biosciences) and cDNAs according to the manufacturer’s instruction. The kit provides specific primers for 1,113 mature miRNAs and 3 internal control snRNAs. MiRNA IDs listed in the text are based on Sanger miRBase identifiers. Primers were designed to maintain uniform amplification efficiencies. qRT-PCR was conducted using the Applied Biosystems 7900HT thermal cycler and data analyzed using and SDS2.3 software. DNA concentrations were reported through SYBR Green fluorescence and normalized to that of the passive reference dye, ROX.

Ct values generated by the SDS2.3 software were normalized according to the average Ct values of the three internal controls provided with the miRNome Profiler kit using qbasePLUS software (Biogazelle). In order to ensure the integrity of the ∆Ct values, we utilized the Genorm software (Biogazelle) to identify 7 additional miRNAs displaying stable expressions between samples. The Ct values of all miRNAs were then normalized to these 10 controls. The relative expression values were then generated using qbasePLUS software and used in additional analysis (detail explanation is in the Additional file 12 supplemental method section). The Ct values were used to derive ΔΔCt values using the miRNome analysis software (SBI). Candidate miRNAs with higher fold change values in each treatment conditions were determined by z score calculation as described in the Additional file 12 supplemental method section.

Clustering was performed using log₂ transformed average ΔΔCt values and the Cluster 3.0 software (Michiel de Hoon, Univ of Tokyo, based on Eisen Lab Cluster software). Hierarchical clustering of the miRNAs was based on the average linkage of the Pearson’s correlation values. The relative expression values were Log₂ transformed and used for evaluation of differences among groups of treated and untreated cells by performing a Welch t-test using MultExperiment Viewer (MEV) software. Results of the t-test were displayed in Volcano plots using -log₁₀ P values of the log₂ transformed values. The K-median clustering of the normalized values were performed using MEV software.

Next, identification of target mRNAs, creation of miRNA-protein networks, and identification of altered cellular processes were conducted using the IPA software (Ingenuity Systems). The ∆∆Ct values were used for a core analysis by selecting all tissue and cell types and using a stringent filter and generating direct relationships only. From the core analysis, mRNA targets were identified for each sample and these mRNAs were used to generate the Venn diagrams using online tool (Venny: http://​bioinfogp.​cnb.​csic.​es/​tools/​venny/​index.​html). The miRNA-protein network generated by the core analysis was overlayed with different cancer types and members of the network that displayed alterations in those cancers were identified using the Interactive Pathway Analysis (IPA) software (Ingenuity Systems). All connections made in the networks are based on previously published results.

Cells harvested at different time points were lysed and total protein extracts were used for western blotting using antibodies specific for AR (US Biological, Salem, MA), PSA (Santa Cruz Biotechnology, Dallas, TX), Cbl (C-15) (Santa Cruz Biotechnology, Dallas, TX), TRAF6 (Millipore, Temecula, CA), p27Kip1 (C-19) (Santa Cruz, Biotechnology), IRAK1 (F-4) (Santa Cruz, Biotechnology), ZFAND1 (A-14) (Santa Cruz Biotechnology), FGD4 (Epitomics, Burlingame, CA), ABHD3 (Biorbyt, Cambrige, UK), DOK4 (C-16) (Santa Cruz Biotechnology), EGFR (1005) (Santa Cruz Biotechnology), VEGFA (A-20) (Santa Cruz Biotechnology), α-tubulin (Cell Signaling, Danvers, MA), GAPDH (Sigma-Aldrich, St. Louis, MO). Total extracts (30–50 μg) were directly mixed with Lammeli sample buffer and separated on SDS-PAGE. Immunoblotting was performed using appropriate primary and horseradish peroxidase conjugated respective secondary antibodies. Positive signals were detected using a chemiluminiscence ECL kit (Pierce, Rockford, IL).