Background
The aldo-keto reductase (AKR) enzymes comprise a functionally diverse gene family [
1]. Members of the AKR superfamily are generally monomeric (37 kD) cytosolic NAD(P)(H)-dependent oxidoreductases that convert carbonyl groups to primary or secondary alcohols, and share a common (α/β)
8-barrel structural motif (visit:
http://www.med.upenn.edu/akr) [
2]. In humans, four AKR1C isoforms have been identified; they are known as AKR1C1 [20α(3α)-hydroxysteroid dehydrogenase (HSD)] [
3], AKR1C2 (type 3 3α-HSD) [
4,
5], AKR1C3 (type 2 3α/type 5 17β-HSD) [
6,
7], and AKR1C4 (type 1 3α-HSD) [
5]. Natural substrates for these enzymes include steroids [
8,
9], prostaglandins [
10], and lipid aldehydes [
11].
Originally cloned from human prostate [
7] and placental cDNA libraries [
12], AKR1C3 shares greater than 86% sequence identity with the other three highly related human AKR1Cs [
13]. AKR1C3 catalyzes both androgen and estrogen metabolism. The relatively high 17β-HSD activity of this enzyme reduces Δ
4-androstene-3,17-dione (Δ
4-dione; a weak androgen) to yield testosterone (a potent androgen) [
14], and reduces estrone (a weak estrogen) to yield 17β-estradiol (a potent estrogen) [
13]. Using its 3α-HSD activity, AKR1C3 reduces 5α-dihydrotesterone (5α-DHT, a potent androgen) to 5α-androstane-3α,17β-diol (3α-diol, a weak androgen) [
7]. As a result, AKR1C3 may be capable of modulating the amounts of potent androgen (testosterone and 5α-DHT) and estrogen (17β-estradiol) available for the androgen receptor (AR) and estrogen receptor (ER), respectively.
Deregulated AKR1C3 expression has been associated with multiple human cancers. Suppressed AKR1C3 expression has been demonstrated in breast cancer [
15] and endometrial cancer [
16], whereas elevated expression of this enzyme has been reported in squamous cell carcinoma of the head and neck [
17]. In the prostate, low or undetectable levels of AKR1C3 are observed in normal prostate epithelium [
18], whereas levels of AKR1C3 expression are significantly elevated in localized [
19,
20], advanced [
21], and recurrent [
22] prostate cancer (PCa).
Based on the observations that expression of AKR1C3 is elevated in both localized and metastatic PCa, the enzyme might modulate significant pathological activities in cancer development or progression. Studies so far have been focused on the potential androgenic effects mediated by AKR1C3 in the prostate; and elevated AKR1C3 expression in the prostate is thought to be responsible for testosterone and 5α-DHT accumulation and AR
trans-activation in cancerous prostate [
14,
20,
21]. In this report, an androgen insensitive, AR-negative human PCa PC-3 cell line was stably transfected with AKR1C3 cDNA to establish PC3-AKR1C3 transfectants and to recapitulate elevated expression of this enzyme in PCa. Microarray, bioinformatics, and literature analyses were used to explore possible pathological consequences of elevated AKR1C3 expression in PCa cells. In addition to elevated cell growth in PC3-AKR1C3 transfectants [
23], AKR1C3 overexpression promoted angiogenic capability of PC-3 cells as evidenced by elevated levels of vascular endothelial growth factor (VEGF) expression and PC-3 cell-mediated endothelial cell (EC) tube formation on Matrigel. These results suggest that, AKR1C3-mediated steroid hormones or prostaglandin metabolism may promote aggressiveness of PCa through enhanced tumor angiogenesis.
Methods
Reagents and chemicals
PC-3 cells (CRL-1435) derived from a human bone metastatic tumor and LNCaP-FGC (LNCaP) cells (CRL-1740) were obtained from ATCC (Manassas, VA). An SV40 large T-antigen transformed human microvascular EC line, HMEC-1 [
24], was generously provided by Dr. Michael Ihnat at the University of Oklahoma Health Sciences Center. RPMI 1640 medium, F-12 nutrient mix, MCDB 131 medium, penicillin-streptomycin, fetal bovine serum (FBS), pcDNA3 vector, Trizol, and Lipofectamine 2000 were purchased from Invitrogen (Carlsbad, CA). AG1042 and LY294002 were purchased from EMD Chemicals (Gibbstown, NJ). Matrigel basement membrane matrix was purchased from BD Biosciences (Bedford, MA). Mouse anti-AKR1C3 monoclonal antibody was produced in our laboratories [
25]. Rabbit anti-insulin-like growth factor (IGF)-1 receptor (IGF-1R) β, rabbit anti-phospho-IGF-1R β (Tyr 1131), rabbit anti-Akt, and rabbit anti-phospho-Akt (Ser 473) polyclonal antibodies, as well as horseradish peroxidase (HRP)-linked goat anti-rabbit IgG were obtained from Cell Signaling Technology (Danvers, MA). Mouse anti-human β-actin antibody was obtained from Sigma-Aldrich (St. Louis, MO). HRP-conjugated goat anti-mouse IgG was obtained from KPL (Gaithersburg, MD). Bicinchoninic acid (BCA) protein assay kit and enhanced chemiluminescent (ECL) substrate reagent were obtained from Pierce (Rockford, IL). VEGF enzyme-linked immunosorbent assay (ELISA) kit was purchased from Ray Biotech (Norcross, GA).
Prostate cell culture and transfection
PC-3 cells were cultured in complete growth medium consisting of F-12 nutrient mixture, 7% FBS, and 100 units/ml penicillin-100 µg/ml streptomycin. LNCaP cells were cultured in complete growth medium consisting of RPMI-1640, 10% FBS, and penicillin-streptomycin. HMEC-1 were maintained in MCDB 131 medium supplemented with 10% FBS and penicillin-streptomycin. All cells were maintained in a humidified cell incubator at 37°C and 5% CO
2. AKR1C3 expression construct (pcDNA3-AKR1C3) was established as previously described [
23]. Transfection of PC-3 cells with either pcDNA3 empty vector or pcDNA3-AKR1C3 construct was achieved using Lipofectamine 2000. Stable transfectants were established in the presence of 600 μg/ml G418 selection. PC-3 stable transfectants were designated as PC3-mock or PC3-AKR1C3 for cells transfected with pcDNA3 or pcDNA3-AKR1C3 plasmid, respectively. Multiple independent clones were isolated and expanded in the complete growth medium containing G418. PC3-AKR1C3 clones were confirmed with elevated AKR1C3 mRNA and protein expression as compared to parental PC-3 cells and PC3-mock transfectants. LNCaP-mock and LNCaP-AKR1C3 stable transfectants were also similarly established.
RNA extraction and quality evaluation
Total RNA was isolated from 2 independent PC3-mock clones and 3 independent PC3-AKR1C3 clones for a comprehensive gene expression analysis using oligonucleotide arrays. Briefly, following PC-3 transfectants (5 × 105) seeding and adherence, total RNA was isolated from these clones using Trizol followed by RNeasy® Mini total RNA isolation kit (Qiagen, Valencia, CA) purification. Concentration of the isolated total RNA was determined with a Nanodrop scanning spectrophotometer; and quality of the total RNA was assessed using the ratio of 28:18 s rRNA by Agilent 2100 Bioanalyzer capillary gel electrophoresis system (Agilent Technologies, Santa Clara, CA).
RNA labeling, microarray hybridization, and scanning
A total of 250 ng total RNA from each clone was labeled using the Illumina Total Prep RNA Amplification kit (Ambion, Austin,. TX). Briefly, cDNA was reverse transcribed from the total RNA after priming with T7-oligo(dT); and cRNA was synthesized from the T7 promoter in the presence of biotinylated UTP. Biotinylated cRNA was hybridized to Illumina HumanWG 6 v2 BeadChips containing a total of 48,702 probes (Illumina Inc., San Diego, CA). Hybridization was detected by incubating the microarray chips with streptavidin-Cy3 (Amersham Biosciences, Piscataway, NJ) and scanned on an Illumina BeadArray Reader.
Data normalization and identification of AKR1C3 regulated genes
Raw expression values were first normalized based on a normal distribution of genes with low expression levels as previously described [
26,
27]. Briefly, normalization was performed for each array by first plotting a frequency histogram of the raw expression data for all genes. The histograms showed a right-skewed unimodal distribution curve with the mode around zero. A normal distribution curve representing random variability of the genes with low expression values was then fitted around the mode, mirroring Gaussian profile of the left part of the histogram. Two parameters [mean and standard deviation (SD)] were defined. The expression values then were normalized to the SD of the normal distribution after subtraction of the mean and then log
10-transformed. The arrays were than adjusted to each other by robust linear regression [
28,
29]. Genes with expression values less than 3.0 (or 0.477 in log
10 scale) were considered to be not expressed; this is equivalent of setting a threshold at 3 SD above noise level. Raw and normalized gene expression data are available on Gene Expression Omnibus (GEO accession GSE20956).
To identify genes that are differentially expressed between PC3-mock and PC3-AKR1C3 transfectants, an associative analysis was performed as described by Dozmorov
et al. [
30]. Briefly, in combined microarray datasets from both PC3-mock and PC3-AKR1C3 transfectants, a reference group of genes expressed above background with low variability of expression as compared with variability of all others was identified by a F-test. The variability in expression of these genes was assumed to represent the technical variability of the assay. Associative
t-test (the standard
t-test applied to the comparison of expression variability) was used to test the null hypothesis if a given gene under a given condition is associated with the reference group. Those with statistically significant variability over the reference set were assumed to represent variability due to the biological variables being manipulated.
An additional analysis was performed to minimize threshold choice effects in different statistical approaches in microarray data analysis [
31]. Class comparison between PC3-mock and PC3-AKR1C3 transfectants was performed using the BRB ArrayTools. To use the BRB Array Tools, raw expression values were log
2 transformed; and quantile normalization was applied. Differentially expressed genes between the two classes, PC3-mock and PC3-AKR1C3 transfectants, were identified by random-variance
t-test. The random-variance
t-test, implemented in the BRB ArrayTools, is an improvement of the standard
t-test since it permits sharing information among genes with within-class variation, without assuming that all genes have the same variance [
32]. A stringent significance threshold was applied to limit the number of false positive findings; genes were considered statistically significant if their
P-values were less than 0.001. Lists of differentially expressed genes from associative
t-test and class comparison analysis were compared; and genes identified by both analyses were selected for further analysis unless otherwise stated.
Functional analysis and identification of canonical pathways
For pathway analysis of the microarray data, gene lists were analyzed by Ingenuity Pathway Analysis (IPA; Ingenuity
® Systems, Redwood City, CA,
http://www.ingenuity.com), a web-based bioinformatics tool. Each gene list was tested against the full Illumina HumanWG 6 v2 dataset to identify significantly overrepresented functions and canonical pathways as compared to the same number of randomly selected genes. The most significant functions and canonical pathways with
P < 0.05 were selected; and genes overrepresented in the functions and pathways were identified. The branches of pathways containing genes that are not expressed on microarrays were not considered functional.
Analysis of 5'-flanking regions of the differentially expressed genes
To explore transcriptional regulatory mechanisms of the genes regulated in PC3-AKR1C3 versus PC3-mock transfectants, the promoter analysis and interaction network toolset (PAINT)
http://www.dbi.tju.edu/dbi/tools/paint/) [
33] was used to identify overrepresented
cis-acting elements, or transcription factor binding sites, within the identified genes against the same transcription regulatory elements (TREs) selected randomly from the whole set of genes on the microarray. Statistical significance for testing the presence of TREs was set at
P < 0.05; and additional filtering was performed by using a false discovery rate (FDR) as described in the text.
Data validation by literature analysis
Literature-based matching of genes to concepts provides an advantage over gene ontology (GO) enrichment analysis in that first, GO annotation lags literature publications, and second, more importantly it provides a much broader matching of genes to concepts, such as diseases, phenotypes, chemical compounds, and other genes frequently associated with the gene set being analyzed. To perform literature analysis, lists of differentially expressed genes between PC3-mock and PC3-AKR1C3 transfectants were analyzed by a comprehensive, computational search of published MEDLINE abstracts
via a software package called IRIDESCENT [
34‐
36]. There were 7 and 39 up- and down-regulated genes in PC3-AKR1C3 transfectants, respectively, not mentioned in the literature and were automatically discarded, since, by definition, they could not share any literature-based commonalities with other genes. This left 56 up-regulated and 108 down-regulated genes available for analysis. IRIDESCENT uses a thesaurus of names and their synonyms obtained from public databases such as OMIM, Entrez Gene, GO, CHEMID, and Disease Ontology to recognize these "objects" and their common spelling variants (e.g., "IL2" and "IL-2") within text. All available MEDLINE records (approximately 18 million at the time of this writing) were processed to identify "objects" co-occurring within the same abstracts, and within the same sentences within those abstracts. The number of co-occurrences was tallied, with a weighting based on a previous study [
35] of 0.5 for objects co-mentioned in the same abstract and 0.8 for objects co-mentioned in the same sentence, providing a network of terms connected by their relative strength of association. Once commonalities had been identified in the literature for a gene set, each was compared to an identically connected set within a random network model to calculate a ratio of how many connections were observed in the actual dataset versus how many connections would be expected by chance (Obs/Exp ratio), providing a measure of statistical enrichment for each term. The significance of enrichment was calculated using a Monte Carlo simulation and the associations reported were ≥ 3 SDs from the mean (
P ≤ 0.01).
Western blot analysis
Western blot analysis was performed to confirm IGF-1R and Akt activation in PC3-AKR1C3 transfectants identified by bioinformatics analysis. Briefly, following cell (1 × 106) seeding and adherence, PC3-mock and PC3-AKR1C3 transfectants were lysed with RIPA buffer supplemented with 5 mM EDTA, 0.1 mM phenylmethylsulphonylfluoride (PMSF), Complete Mini-Protease Inhibitor Cocktail, and PhosSTOP Phosphatase Inhibitor Cocktail (Roche Applied Science, Indianapolis, IN). Total cellular proteins were collected following centrifugation to remove cellular components and quantified using the BCA protein assay kit. Aliquots (30 µg) of the cellular proteins were separated on 10% Tris-HCl gels; and separated proteins were transferred onto PVDF membranes (Bio-Rad; Hercules, CA). The membranes were incubated with rabbit anti-human phospho-IGF-1R β (Tyr 1131) antibody (1:250), rabbit anti-human IGF-1R β antibody (1:1,000), rabbit anti-human phospho-Akt (Ser 473) antibody (1:250), or rabbit anti-human Akt antibody (1:1,500) followed by HRP-conjugated goat anti-rabbit IgG (1:2,000) incubation. Immunoreactive proteins were detected using the ECL reagent. The membranes were then stripped and incubated with a mouse anti-human β-actin monoclonal antibody (1:5,000) followed by HRP-conjugated goat anti-mouse IgG (1:10,000) incubation and ECL development.
Quantification of VEGF expression
VEGF appears to be one of the most important factors released by tumor cells as a signal for angiogenesis. VEGF expression was first compared between PC3-AKR1C3 and PC3-mock transfectants. Levels of VEGF mRNA expression were determined in these transfectants using semi-quantitative reverse transcription-polymerase chain reaction (RT-PCR) method. Total RNA isolation, first-strand cDNA synthesis, PCR amplification using pre-designed VEGF isoform-specific primer pairs and pre-determined cycling conditions followed our reported procedures [
37]. PCR amplified products were separated on 1% agarose gels and stained with ethidium bromide. Images of ethidium bromide-stained gels were acquired with Gel Doc 1000 (Bio-Rad) equipped with Quantity One
® imaging software (Bio-Rad). The involvement of IGF-1R and phosphoinositide 3-kinases (PI3K) in VEGF mRNA expression in PC3-AKR1C3 transfectants was studied by adding 20 µM AG1024 (IGF-1R selective tyrosine kinase inhibitor) and 10 µM LY294002 (non-selective PI3K inhibitor), respectively, to the cell cultures at 2 hours prior to total RNA isolation. The selected concentrations of AG1024 and LY294002 did not interfere with cell viability as determined by XTT assay (data not shown).
Total VEGF secretion by PC3-mock and PC3-AKR1C3 C1 transfectants was quantified in conditioned media using a RayBio® human VEGF ELISA kit following the manufacture's recommendations. Briefly, 1 × 106 cells were seeded onto 60 mm cell culture plates in 2 ml completed medium in duplicate. At 24 hours following cell seeding, the conditioned media were removed and centrifuged at 300 × g to clear cellular components. Aliquots (100 μl) of the conditioned media were transferred to each well of the 96-well assay plate. Known quantities of VEGF A165 were added in separate wells, run in parallel, and used to construct a standard curve. The plate was read at 450 nm on a μQuant microplate reader (Bio-Tek Instrument, Winooski, VT). Results were normalized to total protein present in the media, as determined by the BCA assay, and reported as pg of total secreted VEGF/μg total protein.
An image-based, PCa cell-mediated EC tube formation on Matrigel assay was used as an
in vitro model to assess enhanced angiogenic capability of PC3-AKR1C3 transfectants predicted by bioinformatics. Briefly, PC3-mock and PC3-AKR1C3 transfectants (8 × 10
4) were re-suspended in 0.25 ml liquid Matrigel and plated into each well of 24-well tissue culture plates in duplicate. Following overnight incubation, 8 × 10
4 HMEC-1 in 0.25 ml of the MCDB 131 complete medium were introduced on top of the solidified PC-3 transfectant-Matrigel suspension. The co-cultures on Matrigel were monitored and photographed at 6, 12, 24, 48, and 72 hours following HMEC-1 cell addition using an Olympus IX51 inverted microscope equipped with SPOT Insight CCD digital camera and SPOT Advance software (Diagnostic Instruments, Sterling Heights, MI). Three images were randomly taken from each well. Images in JPEG format were uploaded into Image Pro-Plus 6 (Media Cybernetics, Bethesda, MD), converted to green on the RGB color channel to eliminate the red color of Matrigel, and flattened. The number of tubes was counted using the dendrite/tube template in the Image Pro-Plus software; and non-tube forming cells suspended on Matrigel were mathematically excluded from counting [
37]. The co-culture HMEC-1 tube formation model was also applied to 2 clones of each from LNCaP-mock and LNCaP-AKR1C3 transfectants.
To determine the roles of IGF-1R and PI3K in prostate cell-mediated EC tube formation, parental PC-3 cells as well as PC3-mock and PC3-AKR1C3 transfectants were first incubated with 20 µM AG1024 or 10 µM LY294002 at 37°C for 2 hours. Cells were then harvested, washed with phosphate-buffered saline (PBS), and suspended in Matrigel followed by HMEC-1 addition, image acquisition, and EC tube quantification using the above described procedures.
Statistical analysis
Statistical differences between PC3-mock and PC3-AKR1C3 stable transfectants were analyzed using student t-test. Statistically significant difference was set when P < 0.05.
Discussion
Consistent with previous reports [
19‐
22], we observed low levels of AKR1C3 immunoreactivity in normal prostatic epithelium, whereas a strong immunoreactivity is observed in low grades and advanced PCa using tissue arrays. Forced expression of AKR1C3 was established in human PCa PC-3 and LNCaP cell lines to recapitulate elevated AKR1C3 expression in cancer tissues and to study potential pathological activities of this enzyme in cancerous prostate. Bioinformatics analysis and biological characterization suggest that elevated AKR1C3 expression in PCa cells not only promotes cancer cell growth [
23] but also enhances angiogenesis. In addition to the classical belief that AKR1C3 is responsible for androgen conversion and regulation of AR
trans-activation, AKR1C3 may also activate 17β-estradiol-mediated signaling pathways to promote PCa progression.
Differences in the prevalence of AKR1C3 expression in PCa were noted between tissue sections prepared from individual patients and from tissue arrays. Using radical prostatectomy samples, Nakamura
et al. reported 77.1% (54/70) [
19] and we reported 81.8% (9/11) positive immunoreactivity [
20] in adenocarcinoma. Using tissue array sections, we showed 33.3% AKR1C3 immunoreactivity in prostate adenocarcinoma (data not shown) and Stanbrough
et al. showed 5.6% (1/18) [
21]. In contrast, tissue array samples identified as androgen-independent PCa showed a higher 57.9% (11/19) immunoreactivity for AKR1C3 [
21]. Wako
et al. also reported a statistically significant positive correlation between AKR1C3 immunoreactivity and PCa stages [
22]. PCa is a heterogeneous and multi-focal disease; and the sizes, locations, and stages of PCa tissues obtained for analysis can affect interpretations. In addition, we previously described an uneven distribution of AKR1C3 in adenocarcinoma in prostatectomy specimens [
20]. Tissue array slides consist of core biopsies of 1.0 to 1.5 mm in diameter from paraffin embedded specimens; locations of the cores obtained for tissue array preparation can significantly affect immunohistochemical staining results. The differences in AKR1C3 immunoreactivity obtained from sections prepared from an individual patient and tissue arrays might result from the uneven distribution of AKR1C3 in prostate adenocarcinoma. Since AKR1C3-mediated lipophilic hormone metabolism may function through a paracrine signaling, it may not require all cancer cells to express elevated levels of AKR1C3 to promote disease progression.
In androgen metabolism, AKR1C3 acts as a 3-ketosteroid reductase converting 5α-DHT to 3α-diol [
7], and as a 17-ketosteroid reductase converting Δ
4-dione to testosterone [
14]. In addition to androgen metabolism, recombinant AKR1C3 inter-converts estrone and 17β-estradiol [
13]. Consistent with AKR1C3-mediated steroid metabolism, bioinformatics analysis confirmed that genes deregulated in PC3-AKR1C3 transfectants are related to Δ
4-dione, testosterone, 5α-DHT, and 17β-estradiol. In contrast to other microarray data showing a positive correlation between AKR1C3 and FST, FST was significantly suppressed in PC3-AKR1C3 transfectants. Ingenuity analysis of interactions between AKR1C3 and FST shows that they are connected
via testosterone and progesterone. In transfectants, AKR1C3 likely reduces Δ
4-dione to accumulate testosterone [
14], which in turn decreases FST expression [
39]. In addition, FST protein decreases FST mRNA expression [
39], but increases the secretion of progesterone [
40].
The connection from AKR1C3 overexpression to the p53 signaling suggests possible suppression of tumor suppressive effects of p53 protein and promotion of prostate cancer aggressiveness. Three down-regulated genes, SERPIN5B, PERP and PMAIP1, have been identified as exclusive members of the p53 pathway, which is further highlighted by their connections to the "p53 tumor suppressor" keyword. Expression of the
p53 gene is not detectable in PC3-mock transfectants but is detectable in AKR1C3 transfectants, albeit at a very low level. This may be due to the fact that PC-3 cells are hemizygous for chromosome 17p [
41]; and a base pair deletion of the single copy of the
p53 gene which generates a frame shift and a new in-frame stop codon may not provide the optimal hybridization between wild-type p53 probes and p53 transcripts in PC-3 cells. SERPIN5B has been identified as a suppressor of angiogenesis [
42]; and down-regulation of this gene in PC3-AKR1C3 transfectants suggests the role of AKR1C3 in promoting blood vessel formation. PERP [
43] and PMAIP1 [
44] are parts of the p53 signaling branch leading to apoptosis; and suppressed expression of these genes suggests that AKR1C3 may also promote cancer progression through apoptosis evasion. Other members of the apoptotic branch, such as FAS and Caspase 6, were also down-regulated in PC3-AKR1C3 transfectants, although their changes did not pass our strict statistical criteria. A summary of these findings are presented in Additional file
3.
The advantage of using a literature-based analysis over a GO enrichment analysis is that it permits commonalities to be identified between genes and their associated diseases, phenotypes, metabolites and chemical compounds and is complementary to GO categories. This approach greatly expands the number and nature of commonalities that can be identified and has been used in other microarray studies [
45,
46]. The downside is that associations are probabilistic, which tends to be more problematic when two terms have only been co-mentioned a few times, whereas in GO they are manually corrected and have a lower false-positive rate.
A search for biological or pathological functions of the up- and down-regulated genes in PC3-AKR1C3 transfectants with literature keywords refined and confirmed Ingenuity's findings. "Aggressiveness" is the most significant process represented by the up-regulated genes in PC3-AKR1C3 transfectants, and directly relates to other keywords: "PSA", "β-catenin" and "ER". HBEGF identified as one of EGF receptor ligands is strongly associated with large tumor size, high histoprognostic grading, and aggressiveness of breast cancer [
47]. FBLN1, a calcium-binding, acidic glycoprotein of extracellular matrix, is positively correlated with ovarian cancer progression [
48]; and elevated expression of KLK3 and TMPRSS2 have been associated with PCa aggressiveness [
49]. In the "CTNNB" (or β-catenin) keyword, both WNT5B (a representative Wnt family protein) and β-catenin are well-known pathways in carcinogenesis; and elevated expression of these molecules have been associated with AR, ER, and cancer progression [
50,
51]. In the "ER" keyword, HBEGF [
52], FBLN1 [
53], BEX1 [
54], and TMPRSS2 [
55] are regulated by 17β-estradiol through the ER. The up-regulated genes indentified in PC3-AKR1C3 transfectants surrounding these 4 keywords are interconnected and associated with more aggressive cancer phenotypes.
The majority of down-regulated genes in PC3-AKR1C3 transfectants have been identified as tumor or angiogenesis suppressors. The keyword "tumor suppressor" is the most predominant among genes down-regulated in PC3-AKR1C3 transfectants. Reduced expression of tumor suppressors is associated with development and progression of malignancies. For example, SERPINB5 inhibits cell invasion, as well as angiogenesis and promotes apoptosis [
56]. Genetic loss of the stress response
DDIT4 (or
REDD1) gene promotes anchorage-independent cell growth and elicits tumorigenesis in a mouse model [
57]. Similarly, loss of
CDH11 tumor suppressor gene shows the highest frequency in retinoblastoma tumors [
58]. Genes related to "angiogenesis" keyword are predominantly negative regulators of angiogenesis, suggesting the activation of angiogenic processes in PC3-AKR1C3 transfectants. For instance, the RUNX2 transcription factor is negatively correlated to angiogenesis [
59]. Overexpression of FST (an inhibitor of activin) in human small cell lung cancer cells suppresses metastatic colonies and microvessel density in SCID mice [
60]. "Angiogenesis" and "tumor suppressor" keywords are interconnected by the down-regulated genes identified in PC3-AKR1C3 transfectants.
Several TREs identified among up- ad down-regulated genes in PC3-AKR1C3 transfectants are also related to steroid metabolism. Steroid hormones such as progesterone up-regulates GATA-1 expression [
61]; and members of the GATA transcription factor family (GATA-2 and GATA-3) are involved in the androgen regulated PSA expression [
62]. CP2 implicated in the control of mRNA turnover and translation is involved in post-transcriptional regulation of AR expression [
63]. SREBP-1 (sterol regulatory element binding protein-1) interacts with a 17β-estradiol and insulin-sensitive SRE-1
cis-element [
64]; and up-regulation of SREBP-1 is associated with androgen-independent PCa progression [
65]. USF1 (upstream stimulatory factor 1) has been shown to act cooperatively with SREBP-1 in sterol biosynthesis [
66]. Many of the down-regulated genes shared MyoD or Elk-1 TRE. MyoD (myoblast determination protein 1) is a new component of the ERα complex [
67]. Elk-1 (an ETS-related gene whose activation is regulated by 17β-estradiol [
68]) plays a role in prostate cell proliferation [
69].
Although the involvement of AR in tumor angiogenesis of PCa remains unsettled, several reports have suggested that angiogenesis can be independent of the AR. For example, expression of interleukin (IL)-8 activates androgen-independent growth and angiogenesis in LNCaP cells [
70]. Consistent with our results identifying ER
trans-activation in AKR1C3-positive PC-3 cells, a variety of estrogenic, anti-estrogenic, and selective ER modulator (SERM) compounds has been suggested to have chemopreventive or therapeutic utilities for PCa reviewed by Ho [
71]. In addition to
trans-activation of the ER by 17β-estraldiol, 5α-androstane-3β, 17β-diol (3β-diol, an androgen metabolite formed by 3β-HSD activity of AKR1C3 [
72]), has been shown to be capable of
trans-activating ERβ and regulating PC-3 cell growth in cultures and in xenografts in nude mice [
73]. Roles of AKR1C3-mediated AR and ER
trans-activation in PCa progression require further study.
More than a dozen different proteins have been identified as angiogenic factors, meaning that they are released by tumors as signals for angiogenesis. These factors include acidic fibroblast growth factor (FGF), basic FGF, angiogenis, epidermal growth factor (EGF), granulocyte colony-stimulating factor (GM-CSF), hepatocyte growth factor (HGF), IL-8, placental growth factor (PGF), platelet-derived growth factor (PDGF), scatter factor, transforming growth factor (TGF)-α, tumor necrosis factor (TNF)-α, and VEGF. Among these angiogenic factors, VEGF is one of the most important factors, and is a main target of many anti-angiogenic agents. Although elevated VEGF expression in PC3-AKR1C3 transfectants suggests that VEGF might be responsible for promoting EC tube formation in the in vitro angiogenesis model, our results did not rule out the involvement of other angiogenic factors in promoting AKR1C3-positive cell-mediated angiogenesis. In addition, mechanisms of the secreted VEGF from PC3-AKR1C3 transfectants in activating VEGF receptors in ECs with subsequent capillary-like tube formation will need to be studied.
Although the role for estrogen in PCa progression remains controversial, our results might provide a mechanistic explanation for why a combination of 3α-diol and 17β-estradiol administration promotes prostatic epithelial hyperplasia in castrated beagle dogs mimicking human PCa [
74]. 17β-estradiol can activate IGF-1 signaling pathway in human PCa [
75]. IGF-1 [
76] and Akt [
77], well-known players in promoting PCa progression, are identified as results of elevated AKR1C3 expression in PC-3 cells. AKR1C3, therefore, can be positioned as a critical regulatory molecule in activating/inactivating steroid hormone receptors and activating growth factor signaling pathways. AKR1C3 can also be described as a molecule that provides a survival mechanism to promote angiogenesis in cancer progression. Based on this communication, inter-relationships among AKR1C3, ER, IGF-1, Akt, and subsequent VEGF expression deserve further investigation for tumor angiogenesis and progression.
Authors' contributions
MGD performed bioinformatics analysis for microarray data. JTA, QY, and JSD established stable transfectants and carried out molecular and cellular studies. JDW executed literature bioinformatics analysis. KMF analyzed the immunohistochemical data and critically analyzed the manuscript. REH, DJC, and TMP participated in study design and coordination of the study. HKL designed the study, carried out the experiments, interpreted the data, and finalize the manuscript. All authors read and approved the final manuscript.