Background
Since the 1940s advanced prostate cancer has been treated with surgical or chemical castration in order to reduce systemic androgen levels [
1]. The cumulative experience is that such androgen deprivation therapy (ADT) leads to efficient regression of invasive prostate cancer and to reduced levels of the serological marker prostate-specific antigen (PSA). Unfortunately, ADT seems not to increase long-term overall survival of prostate cancer [
2], and castration-resistant prostate cancer (CRPC) in patients on ADT is typically diagnosed by rising serum PSA levels. Patients with CRPC have a poor prognosis [
3], and patients with metastases have shown median overall survival of ≤19 months [
4]. Androgens, in particular dihydrotestosterone, are activating ligands of the androgen receptor (AR) transcription factor. Novel highly potent drugs that block either androgen production or its stimulation of the AR have shown effect in CRPC and are associated with an extended median survival of several months [
1,
5]. Nonetheless, CRPC remains incurable and progresses in spite of any current therapy. The AR has been shown to be critical to proliferation and survival of the bulk population of prostate cancer cells both in early prostate cancer and in CRPC, but different mechanisms are at play. In physiological prostate homeostasis the prostate epithelium is dependent upon a paracrine mechanism according to which androgen stimulates the stromal AR to induce expression of diffusible growth factors such as FGF7, FGF10, IGF1 and EGF which are essential for prostate basal epithelial cell proliferation [
6]. Epithelial basal cell expression of the AR with androgen available leads to proliferation arrest and luminal terminal cell differentiation. During progression of prostate cancer the AR switches from an epithelial anti-proliferative transcription factor to an oncogene. This may occur in a stepwise fashion by still incompletely understood molecular mechanisms. Several possibly independent steps in CRPC cell generation encompass the loss of ligand-bound AR-dependent inhibition of proliferation, the oncogenic addiction to AR signaling and the replacement of paracrine AR signaling by autocrine growth factor signaling [
7‐
9].
The molecular mechanisms that underlie
AR transcriptional induction in normal prostate epithelial homeostasis and to which extent these mechanisms are retained in putative prostate cancer stem cells (CSCs) are not understood. One hypothesis that could explain that prostate cancer invariably escapes from ADT and androgen targeted therapy (ATT) would be the existence of a subpopulation of prostate CSCs that are AR negative and therefore insensitive to androgen deprivation. Evidence has been found to support the paradoxical possibility that ADT and ATT could lead to expansion of the pool of prostate CSCs [
3] hypothetically due to loss of negative feedback by more differentiated cancer cells. Additional consequences of ADT and ATT could be to induce reprogramming plasticity of CSCs such as epithelial to mesenchymal transition (EMT) or neuroendocrine transdifferentiation [
1,
5].
The understanding of essential molecular mechanisms of putative prostate CSCs is hampered by the low number of these cells in patient materials. If those cells are AR negative and AR non-responsive and give rise to AR positive and AR-dependent cells it is possible that some features of normal prostate cells are retained, although with loss of abilities to terminal differentiation and apoptosis induction. Better understanding of normal differentiation is likely to offer new insights into tumor initiation and may help explain the functional significance of common genetic alterations seen in prostate cancer [
10]. Utilizing a previously published model of stepwise prostate carcinogenesis [
11‐
15] and prostate cancer cell lines we therefore undertook a further examination of conditions for the restriction of AR and classical AR target gene expression in different cellular contexts.
Methods
Reagents, antibodies, cell culture and cell lines
Primary Prostate Epithelial Cells (PrECs; American Type Culture Collection (ATCC); Cat# ATCC-PCS-440-010) and prostate cancer cell lines LNCaP (ATCC-CRL-1740), VCaP (ATCC-CRL-2876) and 22Rv1 cells (ATCC-CRL-2505) were bought from LGC Standards GmbH (Wesel, Germany). The prostate cell lines EP156T, EPT1, EPT2 and PrECs were grown in MCDB153 medium (Biological Ind. Ltd., Israel) with 1 % for EP156T and PrECs, and 5 % fetal calf serum (FCS) for EPT1 and EPT2 cells, and supplemented with growth factors and antibiotics as described elsewhere [
13,
15]. EPT3 cells were grown in Ham’s F12 medium (Lonza, Basel, Switzerland, Cat# 3 MB147) with 5 % FCS. Cells with exogenous AR were grown in equivalent medium but without androgens and with charcoal stripped FCS. LNCaP and 22Rv1 cells were grown in RPMI-1640 (Lonza, Cat# BW12-702 F) with 10 % FCS. VCaP were grown in DMEM (Lonza, Cat# BE12-604 F) with 10 % FCS. For experiments investigating the effect of high calcium, cells were grown in standard MCDB-153 medium supplemented with 1 % FCS, 1 % FCS and 600 μM Ca(NO
3)
2, 10 % FCS or grown in RPMI-1640 with 10 % FCS. To study epigenetic restriction cells were grown in standard medium with 10 μM 5-Aza-2′-deoxycytidine (5-Aza-dC) (Sigma Aldrich, St. Louis, MO, USA, Cat# A3656) for five days with addition of 250 nM trichostatin A (TSA) (Sigma Aldrich, Cat# T1952) the last two days. Medium was changed each day. DNA microsatellite validation of progeny identity of EP156T, EPT1, EPT2, EPT3-PT1 and EPT3-M1 cells has been published previously [
15]. Matrigel-overlay cultures were performed with modifications based on Debnath J et al. [
16] with a bed of growth factor reduced (GFR) Matrigel (Cat# 356231, BD Biosciences) and 2 % GFR Matrigel in the medium, medium was changed every 3–4 days. Cells were grown in a humidified atmosphere containing 5 % CO
2 at 37 °C. Primary antibodies; AR (Cat# ab133273, ab9474), actin (Cat# ab8226), GAPDH (Cat# ab181602) and PSA (Cat# ab53774) were purchased from Abcam (Cambridge, UK).
Vectors, transfection and transduction
The pLenti6.3/V5-DEST-AR expression clone was generated by LR recombination reaction between the entry clone pDONR-AR (Genecopoeia™, Rockville, MD, United States, Cat# GC-E2325), and the destination vector pLenti6.3/V5-DEST. Correct insertion of the AR gene was verified by sequencing with CMV forward primer and V5(C-term) reverse primer, according to the manufacturer’s protocol (Invitrogen, Life Technologies, Carlsbad, CA, United States, Cat# V533-06).
The pLenti6.3/V5-DEST-AR and ViraPower™ Packaging Mix were co-transfected in the 293FT producer cell line, according to the manufacturer’s protocol (Invitrogen, Cat# K370-20). EP156T and EPT3-PT1 cells were seeded in six-well plates and infected with the viral supernatant. After 48 h incubation the supernatant was removed and cells were maintained in androgen-free MCDB medium with 2 μg/ml blasticidine for the selection of stably transduced EP156T-AR and EPT3-PT1-AR cells. Negative control cells were made for each cell type using the pLenti6.3/V5-GW/lacZ control vector (Invitrogen, Cat# K370-20).
Indirect immunofluorescence assay (IF) and Western blotting (Wb)
For IF, cells were grown on 12 mm glass coverslips (Assistent, Sondheim v. d. Rhön Germany, Cat. # 1014/12/1001) in 24 well plates, then washed with PBS, fixed (4 % fresh formaldehyde in PBS for 20 min. at room temperature), permeabilized (0.5 % Triton X-100 for 10 min.), blocked (100 mM glycin for 10 min) and stored (in PBS at 4 °C) with PBS washes between each step. Following blocking with 0.5 % BSA/PBS for 15 min. primary antibodies were added at room temperature for 1 hour at indicated dilutions in 0.5 % BSA/PBS. The FITC-labelled secondary anti-rabbit or mouse IgG (Southern Biotech, Cat# 4050–02, 1030–02) was added for 30 minutes at room temperature in 0.5 % BSA/PBS. Coverslips were mounted in Prolong Gold with DAPI (Molecular Probes, Life Technologies, Cat# P-36931) on glass slides and analyzed using Leica DM IRBE fluorescence microscopy.
For Wb analysis cells were lysed in RIPA-buffer with 1:100 Protease Inhibitor Cocktail Set I (Calbiochem, Cat# 535142). Protein concentrations were measured using the Pierce BCA Protein Assay Kit (ThermoFisher Scientific, Waltham, MA, Cat# 23225), and 5 μg protein lysates were separated by SDS electrophoresis in NuPAGE® 10 % Bis-Tris Gels (LifeTechnologies, Carlsbad, CA, United States, Cat# NP0303BOX) followed by blotting to PVDF membranes (GE Healthcare Life Sciences, Cat# RPN1416F) using Pierce 1-Step Transfer Buffer (ThermoFisher, USA, Cat# 84731) and Pierce G2 Fast Blotter (ThermoFisher). Membranes were blocked for one hour in PBS 0.1 % Tween and 5 % Skim milk powder (Sigma Aldrich, St. Louis, MO, USA, Cat# 70166). Primary antibodies were incubated for 1 hour in blocking buffer at RT, and HRP-labelled secondary antibodies (GE Healthcare, Little Chalfont, UK, Cat# NA931V, NA934V), were incubated as the primary antibodies 1/10000. Pierce ECL Western Blotting Substrate (ThermoFisher, Cat# 23106) or SuperSignal West Femto Maximum Sensitity Substrate (ThermoFisher, Cat# 34096) was used for detection with Chemidoc XRS using Quantity One 4.6.5 (Bio-Rad). Molecular weight marker used was MagicMark XP (Life Technologies, Cat# LC5602).
PSA quantification assay
Cell culture supernatants were centrifuged in an Eppendorf centrifuge at 14 000 x g for 2 minutes at room temperature, and 0.5 ml of the supernatants were analyzed using the Elecsys total PSA immunoassay (#04641655 190) in a Cobas analyzer (Roche, Basel, Switzerland) according to the kit manual and according to the accredited routines of the Laboratory of Clinical Biochemistry (LKB) Haukeland University Hospital. The lower detection limit is 0.003 ng/ml total PSA. Values above 100 ng/ml are considered above the measuring range.
RNA purification, TaqMan real-time RT-qPCR and Agilent microarrays
Total RNA was extracted using the miRNeasy kit from Qiagen (Qiagen, Venlo, Netherlands, Cat# 217004). The total RNA was DNase treated, ss-cDNA was synthesized and the RT-qPCR was run and analyzed as previously described [
17], using pre-designed Taqman probes (Life Technologies) with the following Assay ID numbers:
ACTB (Hs99999903_m1),
AR (Hs00171172_m1),
KLK3 (Hs02576345_m1),
NKX3-1 (Hs00171834_m1),
TMPRSS2 (Hs00237175_m1). The Agilent Human Whole Genome (4x44 k) Oligo Microarray with Sure Print Technology (Agilent Technologies, Palo Alto, CA, US, Design # G4112-60520 G4845-60510), was used to analyze samples in the present study. Total RNA purification, cDNA labeling, hybridization and normalization have been described previously [
17,
18]. Following normalization, significance analysis of microarray (SAM) of the J-Express program package (
http://www.molmine.com) [
19] was used for identification of differentially expressed genes. Only genes that changed at least 2.0 fold with FDR below 10 % were considered as differentially expressed genes in cell lines. ArrayExpress ID for the EP156T and EPT1 cells is (ID: E-TABM-949), EPT2 and EPT3 cells is (ID: E-MTAB-1521) [
15] and for the EP156T, EP156T-LacZ, EP156T-AR, LNCaP, VCaP and 22Rv1 cell lines (ID: E-MTAB-3715).
RNA sequencing (RNA-seq)
Total RNAs were included for RNA-seq if RIN (RNA Integrity Number) was above 9 and total RNA was at least 500 ng according to the Agilent 2100 Bioanalyzer™. Illumina HiSeq™ 2000 (Illumina) RNA-Seq was performed according to manufacturer’s instructions and according to StarSeq™ (Mainz, Germany) protocols. Prior to cDNA synthesis rRNA depletion of total RNA was done. The Qubit™/Bioanalyzer™ instruments were used for concentration and quality control and fragmentation and sizing was achieved using the CovarisTMS2 (Brighton, UK) kits and instrumentation according to instructions. cDNAs were tagged with barcoded adapters for multiplexing. Paired-end sequencing with read length 150 base pairs and 100 million reads per sample were chosen for raw sequence data acquisition. Raw data were formatted in BAM files and mapped to the December 2013 build of the UCSC Human genome browser. The following module versions were used in the TopHat and Cufflinks analyses for alignment and to estimate expression levels: TopHat2 v2.0.7, Bowtie 0.12.9, Cufflinks 2.1.1, Isaac Variant Caller 2.0.5, Picard tools 1.72. RNA-seq data is available at Gene Expression Omnibus (ID: GSE71797).
Statistical analysis
Results from real-time RT-qPCR were analyzed using the RQ Manager v1.2 software and DataAssist v3.01 (both Applied Biosystems, Foster City, CA, USA). Error bars show 95 % confidence intervals. 95 % confidence intervals were analyzed for secreted PSA values using Microsoft Excel 2011 (Redmond, WA, USA).
Discussion
AR negative (AR
−
) prostate epithelial stem cells divide asymmetrically to self-renew and to differentiate into either non-proliferating AR
−
neuroendocrine cells or TP63
+
/AR
−
transient amplifying (TA) cells in the normal adult prostate. The basally located AR
−
TA cells undergo a limited number of amplifying rounds of proliferation before maturing into TP63
+
/PSCA
+
intermediate cells [
7,
29‐
31]. When AR expression is induced by incompletely understood mechanisms and with sufficient androgen available, intermediate cells terminally differentiate into luminal-secretory cells. An important aspect of the terminal differentiation is that androgen-bound AR represses MYC to inhibit proliferation and activates a large number of luminal secretory target genes [
7,
9].
Many groups have investigated
in vitro the molecular events associated with replication and differentiation of prostate basal cells to luminal secretory cells and found restricted AR and AR target gene expressions [
26,
32‐
38]. One study has reported that co-treatment of prostate basal cells with clorgyline, 1,25-dihydroxyvitamin D
3, all-trans retinoic acid and TGF-β1, induced expression of AR and loss of the basal marker KRT14 [
39]. Lamb et al. found that confluent monolayers of primary prostate basal cells treated with dihydrotestosterone and FGF7 for 2–3 weeks differentiated into a top layer of luminal cells with expression of AR and classical AR target genes together with additional markers of terminally differentiated luminal secretory cells [
21]. They found, however, that once cells reached passage 5, the efficiency of luminal differentiation was dramatically reduced. On only one occasion were they able to induce luminal differentiation in a patient-derived immortalized basal cell line. This is consistent with our results with the hTERT immortalized EP156T cells and the failure to demonstrate AR and classical AR target gene expression using either the conditions described by Lamb et al. or additional conditions including long-term confluent cultures, 3D Matrigel cultures, co-cultures with mesenchymal type cells and different combinations of biologically active compounds.
A few studies have reported morphological features of prostate basal cell differentiation using different treatments, such as the cell monolayer becoming stratified or forming gland like buds [
40], but with either lack of AR expression or persistent expression of basal cell markers [
35,
40‐
42]. Of interest, the original publication on the establishment of the EP156T cell line found that it formed glandular like structures in Matrigel and with indirect immunofluorescence detection of AR and KLK3 [
20]. The EP156T cells, received at passage 37 in our laboratory and with a carefully kept passage history, still form glandular like structures in Matrigel, but using both highly sensitive real-time RT-qPCR assays and PSA detection assays we have been unable to detect AR and KLK3 production by these cultures. Additionally, treatment with the epigenetic modifying drugs 5-Aza-dC and TSA was not able to induce AR transcription, indicating that restriction of AR expression is not predominantly epigenetic but rather may be due to lack of cofactors.
In order to examine further the nature of the restriction of AR and AR target gene expression in EP156T cells, EP156T-AR cells stably expressing exogenous AR were selected. These cells were passaged in androgen depleted medium due to the potential of exogenous AR to induce terminal differentiation and growth arrest [
7]. When androgen was added to EP156T-AR cells, both monolayer cultures and Matrigel cultures produced
KLK3 mRNA and protein. Several previous
in vitro differentiation studies of prostate basal epithelial cells have noted a late restriction where
AR and
KLK3 mRNAs can be detected without the corresponding proteins [
37,
43‐
45]. The PSA assay verified that EP156T-AR cells secreted PSA to the supernatant in an androgen-dependent way. There was therefore no evidence of restricted translation in this model.
According to genome-wide microarray analyses, the addition of androgen to EP156T-AR cells induced the classical AR target genes both in monolayer and Matrigel cultures. This was in contrast to androgen treated EP156T cells or in EP156T-AR cells with androgen depleted medium. The endogenous AR mRNA remained repressed, however, both in monolayer cultures and in Matrigel. The possibility that androgen-bound AR could have a positive feedback effect on endogenous AR transcription was therefore not supported by the present studies. It remains a high priority future task to identify the precise molecular mechanisms of endogenous AR transcription activation in prostate basal epithelial cells. Possibilities include lack of essential cofactors, epigenetic repression or selection of mutants.
Downregulation of basal cell integrins α6β4 and α3β1 is considered a critical event in luminal differentiation [
21], but it is unclear whether AR represses integrin mRNA transcription or whether loss of integrin expression must precede AR expression [
30]. Interestingly, androgen addition to EP156T-AR cells was followed by downregulation of
ITGA3,
ITGA4 and
ITGA6. LAMC2 was downregulated in an androgen-dependent way in EP156T-AR cells. LAMC2 is the laminin that binds integrins α6β4 and α3β1 in basal prostate cells. Several additional integrins and laminins changed their expression in androgen treated EP156T-AR cells, indicating that the AR is involved in co-ordinated changes of integrins and laminins in differentiating prostate basal cells.
The lineage hierarchy of prostate epithelial differentiation remains inadequately defined [
46]. The origin and relationship between the benign prostate cells that initiate cancer and the cancer stem-like cells that propagate tumors are still vigorously investigated [
10,
47,
48]. Recent reports suggest that luminal epithelial stem cells can act as the cell of origin of prostate cancer in the form of a castration-resistant Nkx3-1-expressing cell (CARN) [
49]. Additionally both mouse and human epithelial luminal cells can establish prostate organoids
in vitro [
50,
51].
AR is central to growth and survival of both benign and malignant prostate epithelial cells, but the mechanisms seem to be very different in normal prostate homeostasis and cancer growth. In normal prostate epithelial cells the requirement for androgen is mediated through AR in the prostate stromal cells. Stromal androgen-bound AR induces secreted growth factors, so-called andromedins such as IGF-1, EGF, FGF7 and FGF10, to promote growth and survival of the epithelium [
6]. During prostate carcinogenesis AR expression in the stroma decreases concurrently with increased AR expression in the tumor cells as prostate cancer progresses [
52], and stromal cells surrounding metastatic prostate cells are AR negative, suggesting that cancer cells themselves start to supply the necessary andromedins, releasing themselves from the requirement of AR-positive stromal cells and androgens. Prostate carcinogenesis and progression therefore seem to involve acquisition of autocrine growth signals in addition to a switch of the AR from being a cell intrinsic inhibitor of proliferation to becoming a stimulator of proliferation [
6,
8,
52].
The AR is critical for proliferation and survival of the bulk population of prostate cancer cells both at early stages and during CRPC as reflected by the effect of AR-inhibiting therapy [
53‐
55]. But prostate cancer always escapes from these treatments in support of the hypothesis that a small sub-population of AR
−
and androgen-independent prostate CSCs is the source of more differentiated AR
+
bulk population of prostate cancer cells [
56]. The hypothesis that ADT may lead to a “rebound” increase in the number of AR
−
cells with basal cell and CSC features, is reviewed elsewhere (see [
3] and references therein [
57‐
61]). Additional studies have reported androgen-independent early human prostate adenocarcinoma cells and prostate CSCs with low AR [
48,
62‐
67].
The existence of AR
−
prostate CSC with basal cell features could help to explain the recurrence of transdifferentiated neuroendocrine cancers following highly potent ATT of CRPC [
68,
69]. It is not clear, however, if either loss of negative feedback by differentiated prostate cancer cells on CSC proliferation or if therapeutic inhibition of the AR could contribute to increase the pool of prostate CSCs [
3,
70,
71] or contribute to induction of EMT, epithelial mesenchymal plasticity and increased aggressiveness and reprogramming potential in prostate CSCs [
5,
69,
72‐
74]. In this regard, it is noteworthy that mesenchymal type EPT3-AR cells, in contrast to epithelial type EP156T-AR cells, were androgen non-responsive. They were unable to produce detectable PSA in the culture supernatants even with higher levels of exogenous AR protein than in EP156T-AR and LNCaP cells for up to 2 weeks in androgen containing growth medium. The restricted PSA expression in the mesenchymal context suggests that if ADT increases the pool of mesenchymal type prostate cancer cells, then this might go undetected during PSA monitoring of disease progression.
Abbreviations
ADT, androgen deprivation therapy; ATT, highly active androgen targeted therapy; ChIP-chip, chromatin immunoprecipitation with DNA microarray chip; CRPC, castration - resistant prostate cancer; CSC, cancer stem cell; E, epithelial; EMT, epithelial to mesenchymal transition; FDR, false discovery rate; FPKM, fragments per kilobase of exon per million reads mapped; IF, indirect immunofluorescence; M, mesenchymal; PrECs, human primary prostate transit amplifying cells with basal cell features; RNA-seq, RNA-sequencing; RT-qPCR, reverse transcription quantitative polymerase chain reaction
Acknowledgements
Beth Johannessen is acknowledged for cell culture and RNA purification work and Hua My Hoang for excellent RNA purification, target labeling and gene expression microarray hybridizations. Anders Molven and Solrun Steine performed for DNA microsatellite typing of cells used in the present study. We thank Kari Rostad for help with RT-qPCR and Terje Ertkjern and Mia Helen Hansen Hjelle, Laboratory of Clinical Biochemistry, Haukeland University Hospital for PSA assaying.