Background
Prostate cancer, one of the more common neoplasms in the western world, arises through the progressive development of one or more pre neoplastic lesions into adenocarcinoma, and subsequently to metastatic disease. Recent advances have identified key genetic alterations that can initiate prostate carcinogenesis, and enhance the probability of cancer progression. Foremost amongst these is the deletion or inactivation of the PTEN tumour suppressor gene, an antagonist of the phosphatidylinositol-3-kinase (PI3K/AKT) signaling pathway that promotes cell survival and proliferation. PTEN deletion in an epithelial stem cell can be an early initiating event leading to prostatic intraepithelial neoplasia (PIN), and subsequently to cancer [
1,
2]. Thus, heterogeneity in expression of PTEN in the aging prostate tissue may lead to the development of multifocal pre invasive lesions. Therapeutic and dietary approaches to target prostate cells with PTEN deletion and hyperactivated PI3K/AKT signaling may make a major contribution to reducing the incidence and progression of prostate cancer.
Isothiocyanates such as sulforaphane [SF; (-)-1-isothiocyanato-(4
R)-methylsulfinylbutane] have been shown to reduce prostate tumour growth and pulmonary metastasis in the TRAMP mouse model of prostate cancer [
3,
4], and to reduce the growth of prostate cancer xenografts in immune-deficient mice derived from the PTEN-deficient PC3 metastatic cell line [
5]. Isothiocyanates have been shown to exhibit several potential chemoprotective activities in cell and animal models [
6,
7], including the partial suppression of pAKT expression [
3,
8]. The biological activity of isothiocyanates may also provide an explanation for the inverse correlation between diets rich in cruciferous vegetables such as broccoli (the major source of SF in the diet) and the incidence and progression of prostate cancer found in both case control and prospective epidemiological studies [
9‐
12]. Moreover, in a recent human intervention study it was shown that a diet rich in broccoli resulted in changes in gene expression associated with insulin and EGF signaling in prostate tissue of men who had been diagnosed with high grade PIN (HGPIN) [
13], suggesting a potential effect of sulforaphane on PI3K/AKT signaling in humans. Thus, dietary isothiocyanates may be potential candidates to target cells with PTEN deletion or inactivation and enhanced pAKT expression in pre-cancerous prostate tissue.
In the current study, we initially show that that there is significant variation in PTEN and pAKT expression in non-neoplastic tissue of men who had previously been diagnosed with HGPIN. We then demonstrate that SF has differential effects on the viability and proliferation of human cell lines that differ in PTEN expression. We additionally report with the use of PTEN
L/L;PB-Cre4 mice [
14], that dietary intervention with SF has no effect on gene expression in mouse prostate tissue with PTEN expression, whereas in isogenic PTEN-deficient tissue SF acts to attenuate and reverse changes in PTEN deletion-mediated gene expression and induces additional changes in gene expression. We also show that there is a significant overlap in changes in gene expression induced by SF in PTEN null prostate tissue of mice with that induced in prostate tissue of men consuming a broccoli-rich diet. Finally, through the use of exon arrays, we find that SF interacts with PTEN deletion to both attenuate and promote alternative gene splicing. Our results support the finding that cells that have PTEN deletion and associated activation of PI3K/AKT signaling are hypersensitive to SF. It is possible that this leads to these cells being less competitive in growth compared to cells with wild type PTEN expression and provides an explanation of how consuming broccoli can reduce the risk of prostate cancer incidence and progression. In addition, it suggests potential therapeutic applications for sulforaphane.
Discussion
Studies on the effects of dietary components, such as isothiocyanates and fish omega-3 fatty acids, on cancer progression have focused on reporting their effects on tumour growth and development [
20‐
22]. Likewise studies on gene expression associated with carcinogenesis in animal models have frequently only centred on changes associated with the later stages of tumour growth and progression to metastasis [
14]. In this study we are concerned with changes in gene and protein expression that occur in the very early stages of prostate carcinogenesis, prior to tumour development and how they may be modulated by diet. Up to now preliminary results indicate that green tea catechins are promising in suppressing cancer progression in individuals with HG-PIN but a mechanistic explanation for this effect is not yet clear [
23]. In the present study, we sought to explore the possibility that the dietary isothiocyanate sulforaphane has contrasting effects on cells and tissues with differential PTEN expression. This may provide a partial explanation of the inverse association between diets rich in broccoli and the risk of prostate cancer [
12,
24], and also suggest potential therapeutic applications of sulforaphane and related compounds.
Initially, we found that in men that are at risk of prostate cancer following a previous diagnosis of HGPIN, there is considerable variation in PTEN and pAKT expression that is not associated with obvious histological abnormalities. Heterogeneity of expression would be consistent with the characteristic emergence of several cancer foci in the human prostate. PTEN also varied in its localisation of expression. While intense pAKT expression was associated with lack of both nuclear and cytoplasmic PTEN expression, we also observed loss of nuclear expression of PTEN without activation of pAKT. While the function of cytoplasmic PTEN is to counteract increased levels of PIP3 by dephosphorylating to PIP2, thereby preventing activation of AKT and downstream signaling, the function of nuclear PTEN is not as well defined but may be independent of its phosphatase activity [
25]. Nuclear PTEN function leads to p53-mediated G(1) growth arrest, cell death, and reduction of reactive oxygen species production [
26]. Recently, nuclear PTEN has also been linked in controlling chromosomal integrity by acting on chromatin to regulate expression of Rad51, which reduces the incidence of spontaneous DNA double-strand breaks [
27].
To determine whether PTEN and subsequent perturbation of the PI3K/AKT signaling pathway would play a significant role in sulforaphane-mediated effects on prostate carcinogenesis, we initially studied the human prostate cancer cell line PC3, which has a PTEN deletion, and found that it is more sensitive to growth inhibition by SF than the PNT1A cell line that has wild type PTEN expression. At low concentrations, SF enhanced the growth of PNT1A, similar to that previously reported for non-cancerous prostate primary epithelial cells [
7]. However, as PC3 and PNT1A cell lines would have other differences in their genetic background, we sought to investigate the biological activity of SF in the mouse prostate-specific PTEN deletion model, where progression to carcinogenesis parallels that of human disease. By comparing PTEN-deficient and WT littermates, which differ only in the expression of Cre recombinase and PTEN, we unambiguously showed that SF has much greater activity in a PTEN-deficient background, indicating selectivity towards cells that are at risk of carcinogenesis or tumour tissue itself. In addition, these effects are occurring at the early initiating stages of prostate carcinogenesis before malignant transformation has occurred.
The effects of SF on gene expression in PTEN-deficient prostate tissue were complex. After five weeks, the main effect is to ameliorate PTEN null-mediated gene expression, so that SF itself acted as a surrogate PTEN tumour suppressor, although its effects were not sufficient in this model to reverse the histopathological changes induced by the knock out of PTEN in all prostate epithelial cells. At a later stage, while SF continued to inhibit a subset of PTEN null-mediated gene expression, it also changed the expression of other genes that are associated with cell cycle arrest and apoptosis. For example, the HSF diet up-regulated cyclins A2, B1, B2, E1 and E2, down-regulated cyclin D2 and induced their associated cyclin-dependent kinases, Cdk2 and 6 (Additional file
5 - Table S5). These results are consistent with previous studies. Cell cycle regulation by SF has been reported in prostate and colon cell lines in which G1 phase cell cycle arrest was associated with protein down-regulation of cyclin D1 [
28,
29], and in non-transformed T lymphocytes where SF reduced cyclin D2 [
30]. Gene expression of cyclin D2 was also down-regulated in small intestinal polyps of ApcMin/+ mice after 3 days of SF treatment[
31]. Increased protein expression of cyclin B1 has been reported to be induced by SF in human colon and breast cells [
32,
33]. SF also resulted in a greater than 2-fold increase in expression of caspases 3 and 7 (Additional file
6 - Table S6), consistent with induction of apoptosis in a caspase-dependent manner previously observed in a variety of cell and animal models by SF [
6]. For example, SF increased caspase-3 activity in cultured PC3 human prostate cancer cells [
5,
34] and broccoli sprouts - a rich source of SF - retarded tumour growth associated with caspase-3 cleavage in TRAMP mice, an alternative model of human prostate cancer [
3].
Isothiocyanates such as SF are known activators of the NF-E2-related factor 2 (NRF2) transcription factor-signaling pathway [
6], and previous studies have reported enhanced expression of NRF2-regulated genes in small intestine and liver of wild type mice [
35,
36]. Surprisingly, we did not find any enhanced expression of NRF2-regulated genes in either wild type or PTEN null prostate tissue. This may be due to the exposure of prostate tissue in our mouse model to SF being significantly lower than that used in cell studies and other mouse studies, in order to better reflect routine dietary exposure, combined with the probable greater sensitivities of liver and small intestine compared to prostate to any given dose of SF due to topological exposure and first pass effects, respectively. However, it is noteworthy that a previous study also reported lack of enhanced expression of NRF2-regulated genes through dietary intervention with SF in intestinal polyps of APCMin/+ mice [
31], suggesting tissue specific effects. Moreover, an acute intervention with standard broccoli and a 12-month dietary intervention with a broccoli-rich diet did not induce NRF2-regulated genes in gastric mucosa [
37] and prostate tissue [
13], respectively, in human volunteers.
In this study, using the prostate-specific PTEN deletion mouse model we have provided evidence that in a PTEN null background, addition of SF to the diet can ameliorate the effects of the PTEN deletion. It is, therefore, of considerable interest to know whether dietary intervention in men who are at risk of prostate cancer, partly through PTEN deletion or activation of PI3K/AKT signaling, may have a similar effect. To test this we compared changes in gene expression induced by SF in PTEN-null mice with previously reported changes in human prostate tissue following diets enriched in either broccoli (a dietary source of SF) or peas. We unexpectedly found evidence that both diets may induce similar changes, although the overlap with the broccoli diet was of greater statistical significance. It is conceivable, then, that in addition to SF derived from broccoli, some dietary component common to both diets may be able to suppress the downstream effects of PTEN deletion. Possible factors may be lignans, such as secoisolariciresinol, and flavonoids which are present in both peas and broccoli, and which have been associated with reduction in prostate cancer risk [
38,
39]. Lignans are converted by intestinal microflora to enterodiol and enterolactone, which have been shown to suppress pAKT expression in cell models [
40] and certain flavonoids have been shown to modulate PI3K/AKT signalling [
41].
The development of exon arrays has facilitated global analysis of alternative gene splicing, and enables more precise interpretation of microarray data. Alternative splicing of primary RNA transcripts may have a variety of functions, including the expression of different protein isoforms with different functional properties and the regulation of gene expression through nonsense-mediated decay [
42]. Serine-rich (SR) proteins function as important regulators of alternative mRNA splicing; kinases and protein phosphatase 1 control the reversible phosphorylation status of these proteins that determines splice site selection and cellular location [
43,
44]. pAKT has been shown to phosphorylate the SR proteins SRp40, SF2/ASF and 9G8, which determine alternative spicing of PKC and fibronectin mRNA [
45,
46]. Furthermore, insulin has been shown to induce the alternative splicing of PKC via PI3K/AKT activation [
47]. Thus, it may be expected that hyperactivation of PI3K/AKT signaling through PTEN deletion may lead to alternative splicing of genes, which may have important biological consequences. While exon arrays have been used to compare tissue- and tumour-specific splicing in human tissues [
48,
49], we are unaware of any reported analyses that have specifically focussed on PTEN deletions or assessed the role of diet in inducing alternative splicing.
We provide evidence that the interaction between PTEN deletion and supplementation of diet with SF can result in the alternative splicing of many genes. This is likely to be a complex interaction of both PTEN deletion and SF affecting phosphatase activity, as has been shown for other isothiocyanates [
50] and overall changes in gene transcription induced by SF. Indeed, the HSF diet altered expression of two regulatory subunits of protein phosphatase 1 that may partly explain the observed effects. An example of perturbed alternative splicing is the DMBT1 gene where PTEN deletion promoted the expression of the longer isoform and SF affected this process. The human DMBT1 gene, involved in terminal differentiation of epithelial cells, is located on chromosome 10 q, a region often deleted in prostate cancer and also containing the PTEN tumour suppressor [
51]. It is also important to note that in a previous study bioinformatic analyses of the changes in gene expression in men with a broccoli-rich diet reported changes in mRNA processing, indicative of alternative splicing [
13]. However, as this previous study did not use exon arrays, it is not yet possible to quantify the effects of diet on splicing in human prostate tissue.
Materials and methods
Ethics statement
This study was conducted according to the principles expressed in the Declaration of Helsinki. The study was approved by the East Norfolk & Waveney Research Governance Committee and the Norfolk Research Ethics Committee (references 05/Q0101/9 and 09/H0311/96). All patients provided written informed consent for the collection of samples and subsequent analyses.
All animals were handled in strict accordance with good animal practice and all animal work was approved by the University of East Anglia Animal Ethics Committees and covered by the appropriate licences under the UK Home Office Animal Procedures Act, 1986 (PPL 80/1799).
Human subjects
Trans-Rectal Ultra Sound (TRUS)-guided needle biopsy tissues were obtained from men who had previously received a diagnosis of high grade prostatic intraepithelial neoplasia (HGPIN) via the Urology Clinic at the Norfolk and Norwich University Hospital as previously described [
13]. Briefly, RNA was extracted from whole biopsy tissues for gene expression analysis and adjacent biopsies were evaluated by histology. Although the latter did not contain neoplastic cells, we cannot exclude the presence of these cells in the biopsies used for gene expression analysis. Details of the volunteers (age, BMI, PSA, GSTM1 genotype) are as previously described [
13].
Cell culture and proliferation assays
The human post-pubertal prostate normal (PNT1A) cell line and the human Caucasian prostate adenocarcinoma (PC3) cell line were obtained from the European Collection of Cell Cultures (ECACC). PNT1A and PC3 cells were routinely cultured as monolayers in RPMI-1640 and HAMS media, respectively, supplemented with 10% fetal bovine serum (FBS) in a humidified atmosphere containing 5% CO2 at 37°C. Cells were grown to 70-80% confluence before incubation in complete media with various concentrations of SF. Cell viability was determined following a 24 h treatment with either SF or vehicle DMSO using the WST-1 Cell Proliferation Reagent (Roche Applied Science). Cell proliferation was assessed by ELISA following BrdU incorporation (Roche Applied Science).
Animal Husbandry and Genotyping
PTEN
L/L;C
+ mice were generated by crossing ARR2Probasin-Cre transgenic mice, PB-Cre4 [
56], to PTEN
L/L mice [
57] as described before [
14]. Only F2 generation male offspring (PTEN
L/L;C
+ ) and their littermate controls (PTEN
L/L;C
-) were used in this study. For simplicity PTEN
L/L;C
+ mice will be referred to as PTEN null and PTEN
L/L;C
- mice as wild type (WT). PB-Cre4 mice (strain B6.D2-Tg(Pbsn-cre)4Prb) were obtained from the Mouse Models of Human Cancers Consortium (MMHCC) of the National Cancer Institute, US, and PTEN
L/L mice (strain C;129S4-PTEN
tm1Hwu
/J) from the Jackson Laboratories (http://jaxmice.jax.org). Mice were housed in a room with controlled humidity, temperature and light within the Disease Modelling Unit at the University of East Anglia.
Genotyping for the Cre recombinase and PTEN genes was performed by PCR using tail tip DNA. DNA was extracted using the QIAGEN DNA blood and tissue kit (QIAGEN). For the PB-Cre4 genotyping we used two sets of primers, one specific to the transgene producing a 199 bp product (5'-ACC AGC CAG CTA TCA ACT CG-3' and 5'-TTA CAT TGG TCC AGC CAC C-3') and the other as an internal control producing a 324 bp product (5'-CTA GGC CAC AGA ATT GAA AGA TCT-3' and 5'-GTA GGT GGA AAT TCT AGC ATC ATC C-3'). Excision of exon 5 of the PTEN gene was determined by the size of the product from the PCR amplification using the primers 5'-ACT CAA GGC AGG GAT GAG C-3' and 5'- GCC CCG ATG CAA TAA ATA TG-3', where 1.2 Kb results from a wild-type PTEN allele and 1.3 kb from a mutant allele.
Animal feed consisted of basal diet (AIN-93G, control diet) or basal diet enriched with either 0.1 μMol SF/g diet (LSF diet) or 1 μMol SF/g diet (HSF diet). D, L-SF was purchased from LKT laboratories (St. Paul, MN, USA) and was incorporated into the basal diet by Harlan Teklad (Madison, WI, USA). SF levels were confirmed in the diets by LC-MS as described previously [
58]. It is noteworthy that that we chose lower SF levels compared to previous studies, ranging between 1.2-3.4 μMol SF/g [
4,
21], so that the intake would reflect more physiological dietary exposure. Homozygous male PTEN null mice and their littermate controls were randomly assigned to the experimental and control diets, which they received from weaning
ad libitum for a duration of either two weeks or five weeks. Animals were sacrificed by CO
2 asphyxiation and cervical dislocation at five and eight weeks of age. At the time of sacrifice, the total mouse weight was recorded. Mouse prostates were removed immediately and either placed in RNAlater
® and stored at -20°C for subsequent expression analysis (n = 3 per group) or fixed in 10% formalin, embedded in paraffin wax, and used for histopathological evaluations and immunohistochemichal (IHC) staining (n = 3-5 per group).
Histopathology and immunohistochemistry
Dissected tissue samples from mouse and biopsy samples from human volunteers were fixed in 10% formal-saline, embedded in paraffin wax and sections approximately 4 μm thick were cut. One slide from each tissue sample was stained with hematoxylin and eosin (H&E). For immunohistochemical assessment of both the mouse and the human tissue samples, the slides were deparaffinised in three changes of xylene and rehydrated through graded ethanols (100% to 50%). Heat-induced antigen retrieval was performed using citrate buffer pH 6.0 (Dako UK Ltd., Ely, Cambridgeshire, UK). Endogenous peroxidises were quenched using 3% hydrogen peroxide in PBS or TBS and then blocked with 1%BSA in PBST or 5% normal goat serum in TBST appropriate to manufacturer's instructions for each primary antibody. Primary antibodies were diluted in appropriate buffer according to the supplier's instructions: Anti-mouse Ki67 rat monoclonal (Dako, #M7249) diluted 1:50 in PBST; Anti-PTEN(D4.3) rabbit monoclonal diluted 1:125 in SignalStain® Antibody Diluent(Cell Signaling, New England Biolabs, Hitchin, UK); Anti-pAKT (Ser473) (D9E) rabbit monoclonal diluted 1:50 in SignalStain®; Anti-Phospho-mTOR (Ser2448)(49F9) rabbit monoclonal diluted 1:50 in TBST; (Cell Signaling antibodies: #9188; #4060 and #2976 respectively). All were incubated at 37°C for 30 mins. Detection was performed using the appropriate secondary antibody with Vector ABC Elite kits and DAB substrate kit (Vector Laboratories, Peterborough, UK).
All slides were counterstained with Mayer's haematoxylin (Surgipath Europe Ltd, Peterborough, UK), dehydrated through graded ethanols (50% to 100%), and mounted with cover slips. Slides were visualised using an Olympus BX60 (Olympus, Japan) microscope with ProgRes
® Capture Pro 2.1 software (Jenoptik, Germany). Histopathological assessment was performed according to the classification of the Bar Harbor Meeting of the Mouse Models of Human Cancer Consortium Prostate Pathology Committee [
59]. Stained sections were evaluated by two independent researchers. The immunohistochemical quickscore (Q) was determined for each by multiplying the estimated percentage of positively stained cells (P) by the intensity of staining (I). Estimates of intensity were scored as follows: 1 for weak staining; 2 for moderate staining and 3 for strong staining (Q = P × I; maximum 300).
Gene expression analysis
RNA from mouse prostate tissue stored in RNAlater
® was isolated using RNeasy Mini kit (QIAGEN) and gene expression profiling was performed using the Affymetrix GeneChip
® Mouse Exon 1.0 ST Array (Affymetrix, Santa Clara, CA) at the Nottingham Arabidopsis Stock Centre (Nottingham, UK) according to the Affymetrix protocols. This array is a whole-genome array, containing 1.2 million probesets of four Perfect Match (PM) probes each. A total of 36 arrays were hybridised (3 mice X 2 genotypes (WT and PTEN
null) X 3 diets X 2 time-points). Data were analysed using R/Bioconductor [
60] and the
aroma.affymetrix package [
18]. Linear probe level models were fit to RMA-background corrected and quantile normalised data to get gene- or exon-level summaries. For annotation we used the current custom CDF file available at the
aroma.affymetrix website containing the core probesets (17,831 transcript clusters; 224,053 probesets). One array was identified as an outlier and was removed from further analysis. Subsequent statistical data analysis to identify differentially expressed genes was performed using
limma [
61], and evaluation of alternative splicing using
FIRMA [
17] packages. Genes were identified as differentially expressed at different Benjamini and Hochberg adjusted p-values. Alternative splicing events were identified if the difference between FIRMA scores of two groups was statistically significant after Benjamini and Hochberg adjustment. For the plotting of probeset-level data for FIRMA analyses of exon array data we used the Bioconductor
GenomeGraphs package [
62]. To identify pathways that were the most over-presented in the lists of differentially expressed or alternative spliced genes, functional analyses using MAPPFinder and GenMAPP v2.1 were performed
http://www.genmapp.org/[
63]. The Database for Annotation, Visualization and Integrated Discovery v6 (DAVID;
http://david.abcc.ncifcrf.gov/) [
64] was used to identify Gene Ontology (GO) categories associated with specific gene lists (Additional file
2 - Table S2; Additional file
3 - Table S3). Microarray data generated in this study are compliant to MIAME criteria and are publicly available through ArrayExpress (Accession E-MEXP-2469).
PCR of the DMBT1 gene
Total RNA from all the mice was reverse transcribed into cDNA using random primer synthesis from the High-Capacity cDNA Reverse Transcription kit according to the manufacturer's instructions (Applied Biosystems, UK). Primer pairs for the locus flanking exon 2 were 5'- TTGTGGGGTCAAATTCTGTCT-3' and 5'-CTCCAGCATCTTCCTGGTGT-3' and for the locus flanking exon 7 were 5'-CTCAAACAAGCAGTCCCACA-3' and 5'-GTCCCTCCTGGATTCCACC-3'. PCR reactions were carried out in a total volume of 50 μl consisting of 1× green GoTaq® Flexi PCR Buffer, 0.2 mM dNTPs, 0.2 μM primer, 1 mM MgCl2, 0.025 units of GoTaq® Flexi DNA Polymerase (Promega, UK). Cycling conditions consisted of 2 min at 95°C for initial denaturation, 30 cycles of denaturation for 30 sec at 95°C, annealing for 1 min at 60°C for exon 2 and 59°C for exon 7, and extension for 1 min at 72°C, followed by 5 min final extension at 72°C. Amplified products were separated in 2% (w/v) agarose gels using 1 μg of the 50 bp DNA ladder (Invitrogen, UK) as a size marker and visualised under UV following ethidium bromide staining.
Statistical analysis of human orthologues
To estimate whether the overlap between human orthologues of mouse genes induced by SF and human genes induced by a dietary intervention was significant, we initially assumed that the total number of probes from the human Affymetrix U133 Plus 2.0 array and equivalent probes from the mouse Affymetrix Exon 1.0 ST array would be similar. We then performed a Monte Carlo simulation to assess the number of expected possible probes that both change significantly in the mouse (measured probability = 0.000467) and the human (measured probability = 0.10716). This expected number was compared to the measured number using a Binomial test in R [
65].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MHT designed and performed the animal study, analysed the array data, performed the splicing experiment. CAS performed the mouse immunohistochemistry. JFD performed the human immunohistochemistry. AM performed the cell analyses. RYB assessed the mouse and human prostate sections. RDM collected the human prostate tissue. RFM conceived the study and together with MHT prepared the manuscript. All authors read and approved the final manuscript.