Differential gene expression in the peripheral zone compared to the transition zone of the human prostate gland

Abstract

Gene expression profiles may lend insight into whether prostate adenocarcinoma (CaP) predominantly occurs in the peripheral zone (PZ) compared to the transition zone (TZ). From human prostates, tissue sets consisting of PZ and TZ were isolated to investigate whether there is a differential level of gene expression between these two regions of this gland. Gene expression profiling using Affymetrix Human Genome U133 plus 2.0 arrays coupled with quantitative real-time reverse transcriptase-PCR was employed. Genes associated with neurogenesis, signal transduction, embryo implantation and cell adhesion were found to be expressed at a higher level in the PZ. Those overexpressed in the TZ were associated with neurogenesis development, signal transduction, cell motility and development. Whether such differential gene expression profiles may identify molecular mechanisms responsible for susceptibility to CaP remains to be ascertained.

Introduction

The human prostate is conventionally divided into regions or tightly fused zones known as the peripheral zone (PZ), the central zone (CZ) and the transition zone (TZ).¹ The multifocal entity of prostate adenocarcinoma (CaP) arises mostly in the PZ whereas the non-malignant overgrowth of benign prostatic hypertrophy occurs exclusively in the TZ; the CZ seems to be relatively immune to both pathologies.¹ CaP is the most common male cancer in Europe, North America and some parts of Africa,² and epidemiological studies of populations that migrated from low- to high-risk regions point to dietary and/or lifestyle factors playing an important role in its aetiology.³

CaP progression is a multistage process from latent carcinoma(s) of low-histological grade to high-grade metastatic disease.⁴ Whether this pathology is predominantly PZ-associated remains to be ascertained but divergent hormone responsiveness of primary putative PZ- and TZ-derived stem cells⁵ or expression of hormone-metabolising enzymes (for example, CYP1B1)⁶ have been suggested factors in this morphological susceptibility. The CZ may have a different embryogenic origin (that is, the mesonepheric duct; the urogenital sinus is believed to give rise to the PZ and TZ¹) and consistent with this, its protein profile differs from that of the other zones.⁷ Paradoxically, PZ and TZ glandular epithelia were found to express similar protein profiles⁷ and have a similar proliferative index and incidence of apoptosis.⁸ Following analysis by Fourier-transform infrared microspectroscopy, TZ epithelium was unexpectedly found to be more likely to exhibit a susceptibility-to-adenocarcinoma spectral signature than PZ.⁹

If epithelial cells lining the branched tubuloacinar glands of the PZ and TZ are equally susceptible to initiating events, other effects (for example, growth promotion by hormone-associated genes or the epigenetic silencing of protective genes) might give rise to CaP progression. To investigate this, we examined the differential gene expression profile of CaP-free tissue sets consisting of PZ and TZ. Employing oligonucleotide microarrays and candidate gene verification using quantitative real-time reverse transcriptase (RT)-polymerase chain reaction (PCR), our aim was to identify genes whose zone-specific preferential expression might be associated with susceptibility or resistance to CaP.

Materials and methods

Study participants

Informed consent to obtain tissue for research was obtained (LREC no. 2003.6.v; Preston, Chorley and South Ribble Ethical Committee). Study participants were selected from patients undergoing radical retropubic prostatectomy (RRP) for CaP on the basis of a low prostate-specific antigen (<20 μg/l serum) and low volume of disease (⩽2 core biopsies positive for CaP per 8 taken). A further CaP-free tissue was obtained after a radical cystoprostatectomy for removal of muscle-invasive bladder carcinoma.

Tissue retrieval and storage

After surgical resection, prostate tissue was transported to the pathology laboratory (<3 min) and dissected using aseptic techniques. Provisionally, tissue assumed to be cancer-free was selected (from the lobe from which biopsy cores were negative for cancer); formalin-fixed sections stained with hematoxylin and eosin were checked retrospectively to assess whether CaP was likely to be absent. Using a forceps and scalpel, the prostate tissue was sliced from the upper part of the gland starting just above the area of the verumontanum. The tissues (2 cm in length and 0.3 cm in width) isolated from the most peripheral and postero-lateral aspect of the gland were designated as PZ while the tissues (1.5 cm in length and 0.3 cm in width) isolated from the area immediately lateral to the urethra (peri-urethral) were designated as TZ. Tissue sets (PZ and TZ per individual patient) were immediately placed in RNAlater (Qiagen Ltd, Crawley, West Sussex, UK) and stored at −85°C. Remaining prostate was formalin fixed for histopathology.

Tissue sets consisting of PZ and TZ were obtained from 12 patients (53–69 years); 11 of these underwent RRP while patient 12 had a cystoprostatectomy (Table 1). On digital rectal examination, carried out by a single assessor, eight participants were characterized as benign (non-cancerous-feeling glands—stage T₁c) and the remaining patients had a malignant-feeling gland in one lobe (stage T₂). A single pathologist assigned the Gleason grade; in one tissue (PZ of patient no. 4), a node of malignancy was found following retrospective histopathology.

Table 1 Details of study participants and tissue samples examined

RNA extraction

Towards transcript profiling, total RNA was extracted using TRIzol reagent (Invitrogen, Paisley, UK) according to the manufacturer's instructions and purified by ethanol precipitation to remove any residual reagents. RNA concentration was quantified using the Nanodrop ND-1000 spectrophotometer, and quality was assessed by observing the integrity of 28S and 18S bands using an Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). For the purposes of this study, the samples were grouped by prostate zone. Samples (n=6) from three tissue sets (PZ and TZ per individual patient) were stored at −80°C until use and contributed separately (that is, one sample per chip) towards three values per prostate region (Figures 1a and b).

Figure 1

Preferentially expressed genes in the (a) PZ compared to the TZ and (b) TZ compared to the PZ of human prostate. A conditional tree-view of differentially expressed genes is shown in which the vertical axis corresponds to the candidate genes and the horizontal axis to the tissue samples. The colour-bar scale gives an indication of the level of expression with red representing preferentially expressed transcripts and green to blue representing lower expression. RNA samples from individual tissues contributed separately towards three values grouped by prostate zone (that is, derived from three samples applied to three different chips). Significance was set at a level of P<0.01.

Towards quantitative real-time RT-PCR, total RNA extraction was performed using the Qiagen RNeasy Kit in combination with the Qiagen RNase-free DNase kit (Qiagen Ltd).

Based on experience, these two extraction methods were conducted independently in different laboratories. However, the finding that both sets of analyses were confirmatory (see below) lends strong support to the robustness of the datasets generated.

Transcript profiling and data analysis

A One Cycle Eukaryotic Target Labeling Assay (Affymetrix, Santa Clara, CA, USA) was used for amplification and biotinylation of tissue-extracted RNA. Sample preparation and hybridization were carried out following the manufacturer's instructions. Briefly, 5 μg of total RNA was reverse transcribed using T7 promoter sequence-tagged random hexamers and the resulting double-stranded cDNA was used as a template to generate biotin-labelled antisense cRNA by in vitro transcription. After clean-up, 20 μg of full-length cRNA was fragmented by metal-induced hydrolysis. The quality of the fragmented cRNA was verified on an Agilent 2100 Bioanalyzer (Agilent Technologies). Samples were hybridized for 16 h at 45°C on the Affymetrix Human Genome U133 plus 2.0 high-density microarray containing 54 675 probes (that is, 24 325 Unigene clusters). Following post-hybridization washes, the arrays were scanned using the Affymetrix GeneChip Scanner 3000 7G and expression data were analysed using GeneSpring v7.2 software (Silicon Genetics, Redwood City, CA, USA). Raw data were normalized per chip (50th percentile) and per gene (median). A 2-fold cut-off value was set with genes filtered on flags to remove those absent in all samples and to only conserve the present and marginal-flagged genes. Statistical filtering was then applied to the remaining list; a variance analysis (one-way ANOVA test) of the expression values was performed with a level of significance set at P<0.01. From the resulting list of statistically significant genes, differentially expressed (that is, between PZ and TZ per individual patient) candidate genes were identified and individually checked for consistent differential expression. Finally, ontological analysis using the Gene Ontology option available on GeneSpring v7.2 software as well as the Human Gene Nomenclature Database (www.genecards.org) was carried out to investigate the function of the selected genes and their relevance to CaP progression.

Quantitative real-time RT-PCR

RNA (0.4 μg) was reverse transcribed in a final volume of 20 μl containing Taqman reverse transcription reagents (Applied Biosystems, Warrington, Cheshire, UK): 1 × Taqman RT buffer; MgCl₂ (5.5 mM); oligo d(T)₁₆ (2.5 μ M); dNTP mix (dGTP, dCTP, dATP and dTTP; each at a concentration of 500 μ M); RNase inhibitor (0.4 U/μl); RT (MultiScribe; Applied Biosystems) (1.25 U/μl) and RNase-free water. Reaction mixtures were then incubated at 25°C (10 min), 48°C (30 min) and 95°C (5 min).

cDNA samples were stored at −20°C before use. Primers (Table 2) for candidate genes, and endogenous control β-ACTIN was chosen using Primer Express software 2.0 (Applied Biosystems) and designed so that one primer spanned an exon boundary. Specificity was confirmed using the NCBI BLAST search tool. Quantitative real-time PCR was performed using an ABI Prism 7000 Sequence Detection System (Applied Biosystems). Reaction mixtures contained 1 × SYBR Green PCR master mix (Applied Biosystems); forward and reverse primers (Invitrogen) at a concentration of 300 nM; for the candidate gene amplification 20 ng cDNA template or for β-ACTIN amplification 5 ng cDNA template; made to a total volume of 25 μl with sterile H₂O. Thermal cycling parameters included activation at 95°C (10 min) followed by 40 cycles each of denaturation at 95°C (15 s) and annealing/extending at 60°C (1 min). Each reaction was performed in triplicate (using the same master mix) and ‘no-template’ controls were included in each experiment. Dissociation curves were run to eliminate non-specific amplification, including primer-dimers.

Table 2 Primers used for quantitative real-time RT-PCR analyses of candidate genes

Results

Transcript profiling in the PZ and TZ

A genome-wide approach was employed to investigate whether a differential gene expression signature associated with the PZ versus TZ of normal human prostate could be identified. For transcript profiling, three CaP-free tissues were selected (patient nos. 10, 11 and 12; see Table 1); following total RNA extraction, these samples contributed separately towards gene expression microarray analysis. By comparing tissue sets consisting of PZ and TZ, a list of genes showing differential expression between both zones was established; 45 probes representing 43 genes were identified as differentially regulated in the PZ compared to the TZ, including 33 under- and 12 overexpressed (Figures 1a and B). The majority of the differentially expressed genes were preferentially expressed in the TZ compared to PZ. Each gene was assigned to a biological process group (Table 3) based on the Gene Ontology database.

Table 3 Biological classification of genes differentially expressed between the PZ and the TZ of human prostate

Genes preferentially expressed in the PZ compared to the TZ included bone morphogenetic protein (BMP) antagonist gremlin (GREM1, 7.6- to 11.6-fold difference), secreted frizzled-related protein 4 (SFRP4, 5.5- to 6.1-fold difference), close homologue of L1 (CHL1, 2.8- to 3.4-fold difference) and zinc finger protein 185 (ZNF185, 2.2-fold difference) (Figure 1a). Those identified to be preferentially expressed in the TZ included neural epidermal growth factor-like like 2 (NELL2, 10-fold difference), bone morphogenetic protein-5 (BMP5, 10-fold difference) and S100A4 (5-fold difference) (Figure 1b). These fold differences ranged from 2.09 to 11.65 for preferentially expressed PZ genes and 2.01 to 11.52 for preferentially expressed TZ genes (see Supplementary Material).

Verification of differential candidate gene expression in the human prostate

Quantitative real-time RT-PCR was employed to verify the differential gene expression of four candidate genes (GREM1, SFRP4, CHL1 and ZNF185) preferentially expressed in the PZ compared to TZ and three (NELL2, BMP5 and S100A4) in the TZ (Table 4). Averaged threshold cycle (C_T) values of amplified cDNA were in the 27–35 range for GREM1, 24–30 for SFRP4, 28–35 for CHL1, 31–38 for ZNF185, 25–33 for NELL2, 24–36 for BMP5 and 21–26 for S100A4 (data not shown). Except for a small proportion of samples, inter-individual variations in candidate gene expression were not marked. Preferential expression in PZ compared to TZ of GREM1 (2- to 62-fold, except for two tissue sets), SFRP4 (2- to 11-fold, except for a separate two tissue sets), CHL1 (2- to 9-fold, except for yet another tissue set) and ZNF185 (2- to 36-fold, except for five tissue sets) was observed (Table 4). In comparison, lower expression in the PZ compared to the TZ of NELL2 (2- to 33-fold, except for one tissue set), BMP5 (2- to 100-fold) and S100A4 (2- to 10-fold) was verified (Table 4).

Table 4 Relative quantitative expression in prostate zones by real-time RT-PCR

Discussion

Although morphologically similar,⁸ it remains unknown whether there is a specific underlying mechanism that may confer susceptibility to CaP progression in the PZ or, conversely, resistance in the TZ in normal human prostate.¹⁰ Prostatic glandular tissue possesses the metabolic machinery to activate pro-carcinogens or hormones to DNA-reactive electrophilic species,^{6, 11} and such agents induce measurable levels of damage in isolated primary prostate epithelial cells.¹² For this study, stringent attempts were made to isolate and examine CaP-free tissues but given the age range of the participants (53–69 years) and the fact that most were undergoing RRP for organ-confined disease, there was always a distinct possibility that disease-progressive changes were present even though they may not be observable by histology. However, an advantage of this approach would be that gene expression profiles responsible for disease progression may be present that would be absent in tissues isolated from a younger cohort.¹³

Genes preferentially expressed in the PZ were associated with neurogenesis development (GREM1 initiates epithelial–mesenchymal signalling interactions during organogenesis), signal transduction (SFRP4 is a negative regulator of Wnt signalling), embryo implantation and cell adhesion (CHL1 promotes cell migration, axonal growth and synaptic remodelling) (Table 3). GREM1-mediated expression is essential to initiate epithelial–mesenchymal signalling interactions during embryonic organogenesis.¹⁴ BMP is a potential regulator of ureter development and antagonism by GREM1 expression may confer survival advantages. SFRP proteins bind directly to Wnt ligands and methylation silencing of SFRP4 may play important roles in aberrant Wnt activation.¹⁵ CHL1, located on human chromosome 3p26.1, is expressed in neurons and glia of both the central and peripheral nervous system, and promotes neurite outgrowth and neuronal survival; its role may be in integrin-dependent cell migration.¹⁶ Methylation-specific PCR has shown ZNF185 inactivation by cytosine-phosphate-guanine dinucleotide methylation suggesting that epigenetic alterations resulting in transcriptional silencing may be a useful biomarker of CaP progression.¹⁷ Recent studies suggest that ZNF185 may function as a tumour-suppressor protein by associating with the actin cytoskeleton through its N-terminal region.¹⁸ Our findings suggest that ZNF185 overexpression may be an important block to CaP progression in the PZ and its epigenetic inactivation may remove this protection.

Those genes preferentially expressed in the TZ were associated with neurogenesis, signal transduction (NELL2 may promote survival through modulation on mitogen-activated protein kinases (MAPKs)), cell motility (the S100 calcium-binding protein of S100A4) and development (BMPs of BMP-5 are secreted signalling proteins that induce ecotopic bone) (Table 3). NELL2 appears to promote neuronal survival by modulating MAPKs and consequently is highly expressed in hippocampus and cerebral cortex; its gene product is a thrombospondin-1-like protein exhibiting six epidermal growth factor-like domains,¹⁹ most probably acting as a multi-domain adhesive in the extracellular matrix and participating in cell–cell communication.²⁰ In the central nervous system, NELL2 is expressed in nascent, post-mitotic neurons as they start to differentiate whereas, in the peripheral nervous system, it is also expressed in subsets of progenitor cells.²⁰ BMP5 may be expressed early in skeletal development and local expression may prefigure specific anatomical structures in the vertebrate skeleton.²¹ It may also play a role in the control of apoptosis during limb development with Smad proteins and MAPK p38 acting as intracellular effectors.²² S100A4 overexpression has been associated with cancer (including CaP) progression, invasion and metastasis; its gene product is a member of the S100 calcium-binding protein family whose upregulation is inversely correlated with E-cadherin expression.²³

A check of exon results consisting of Affymetrix array data for the LNCaP, DU-145, PC-3 and 22RV1 prostate cancer cell lines and one primary sample (unpublished data, Medical Oncology Centre, Institute of Cancer, University of London) was conducted. From this dataset a number of the genes (CCDC8, HOXD13, SPON1, PTGS1, HSD11B1, ZFHX1B, SYNPO, UNC5B, S100A4) observed to be overexpressed in normal TZ tissues, which were hitherto identified to be downregulated in prostate cancer cells and CaP tissues, were noted.^{24, 25, 26, 27} No genes preferentially expressed in normal PZ tissues were noted on the exon array (unpublished data). However, SFRP4 was actually downregulated in the exon array but the literature reports that this is overexpressed in CaP.²⁸

Marked differences in intra-individual gene expression patterns in the PZ compared to the TZ were observed (Table 4). Of note was the observation that many of these differentially expressed genes have previously been associated with cell growth and division, differentiation and cell migration. The expression patterns of these genes may result in an environment favouring zone-specific susceptibility or resistance to CaP progression. Alternatively, alterations of such gene expression profiles (for example, through transcriptional silencing) may be the pivotal pre- or post-initiation events.