Original Paper |
|
|||
|
Acta Biochim Biophys |
||||
|
doi:10.1111/j.1745-7270.2006.00205x |
Characterization of Genes
Associated with Different Phenotypes of Human Bladder Cancer Cells
Yu-Cong YANG, Xu LI, and Wei
CHEN*
Center
for Experimental Medicine, the First Affiliated Hospital, School of Medicine, Xi’an
Jiaotong University, Xi’an 710061, China
Received: March 22, 2006
Accepted: June 20, 2006
This work was supported by a grant
from the National Natural Science Foundation of China (No. 30370660)
*Corresponding
Author: Tel, 86-29-85323528; E-mail, [email protected]
Abstract To identify genes associated with
morphological phenotypes of human bladder transitional cell carcinoma, we used
suppression subtractive hybridization (SSH) to create a subtractive cDNA
library of two established cell lines, BLZ-211 and BLS-211, derived from a
patient with transitional cell carcinoma of the bladder, then to screen for
differentially expressed genes. Real-time reverse transcription-polymerase
chain reaction was used to further confirm the selected differentially
expressed genes. Forward and reverse subtractive cDNA libraries yielded 168 and
305 putative clones, and among them more than 90% contained the inserts. After
differential screening, 36 different transcripts were obtained from 64 cDNA
clones of a forward and reverse subtraction library. Among them, 17 were
identified as known genes by homology, for example, Vimentin, Keratin7,
DDH and UCH-L1. The remaining 19 were unknown expressed genes,
and were collected as new expressed sequence tags by the GenBank dbEST database
with the accession numbers DR008207, DR010178, DR159652–DR159660,
DY230447–DY230448, and DY505708–DY505713. Their function will be studied
further. Thus, SSH appears to be a useful technique for identifying
differentially expressed genes between cell lines or clones. Our results, as
revealed by SSH, also suggest that differences in gene expression of
cytoskeletal proteins might contribute to the different morphologies in BLZ-211
and BLS-211 cells.
Key words bladder cancer; phenotypes; suppression subtractive
hybridization
Bladder cancer is one of the most commonly diagnosed malignancies in
the urinary system. Eighty percent of
these tumors are characterized by transitional cell carcinomas (TCC) with a
high recurrence rate (50%–80%) [1]. Intensive research has been conducted to explore the
mechanisms involved in their pathological characteristics. For example, tumor
heterogeneity is an important factor that might affect clinical features of the
disease [2]. Phenotypic differences associated with genetic heterogeneity of
the cancer cells might lead to differences in manifested pathology. Thus,
analyzing the genetic heterogeneity and gene expression could provide insight
for a biological basis of bladder cancer.
We have previously established two different cell lines, named
BLZ-211 and BLS-211, originally derived from the surgical sample of a
55-year-old male patient with grade II TCC of the bladder through routine
pathological diagnosis [3]. We have previously demonstrated that these two
cell lines were clonally established and appeared to be useful models with
respect to the questions of genetic heterogeneity in bladder cancer [4–6]. In order to
exclude the effect of culture in vitro, we detected all cell cultures
biological characteristics within 60 passages. These two cell lines share some
similar characteristics. For example, the DNA index is triploidy in both lines
and the values, determined by flow cytometry, are 1.602 and 1.591 for BLZ-211
and BLS-211 cell lines, respectively. They contain at least five shared
structural anomalies as revealed by detailed karyotyping [4] and both are
sensitive to cisplatin, but not to vincristine, in drug sensitivity testing
[5].
However, these two cell lines show remarkable differences in cell
morphology. BLZ-211 cells are polyhedral and plump, with large vesicular nuclei
and nucleoli; and the cells grow in a tightly connected manner with virtually
no space between them. By comparison, BLS-211 cells are spindle-shaped with
elongated and partly twisted characteristics, and do not have a polarized
epithelial phenotype or tight connection between them. BLS-211 cells are also
more sensitive in response to trypsin digestion than BLZ-211 cells. In addition,
cell adhesion assay revealed that the amount of BLZ-211 cells adhering to
fibronectin is higher than that of BLS-211 cells [6] and in vitro scrape
wound assay showed that the speed of BLS-211 migration is faster than that of
BLZ-211. Cloning efficiency in soft agar is 12% and 6% for BLS-211 and BLZ-211
cells, respectively. As these two cell lines were originally cloned from the same
tumor sample and had similar chromosomal alterations, we speculate that the difference
in gene expression might contribute to these morphological differences.
Identification and characterization of the differences in gene expression
between them could provide important clues for the understanding of the
pathological basis and the discovery of new molecular targets associated with
the pathogenesis of bladder cancer.
A number of different
techniques have been used to characterize the differences in gene expression.
These include subtractive hybridization [7], differential display [8],
representational difference analysis (RDA) [9], cDNA array hybridization [10]
and serial analysis of gene expression [11]. The technique of cDNA microarray
has been commonly used to allow the generation of differential expression
profiles, but it cannot find novel sequences that are not yet available on the
microchips. One recently described method, suppression subtractive
hybridization (SSH) [12], is similar to RDA but with some extra advantages. For
example, the SSH procedures allow the identification of differentially
expressed genes, particularly for those genes whose levels of expression are
“significantly” different between two target populations without
amplification of commonly expressed genes. SSH can isolate unknown cDNAs for
the study of gene expression.
In the present work, we used
SSH to explore the differences in gene expression between BLZ-211 and BLS-211
cell lines, which might determine the genetic components associated with their
morphological variations.
Materials and Methods
Cell storage and culture
The BLZ-211 and BLS-211 cell lines were originally established in
1993 from a 55-year-old male patient with a grade II TCC of the bladder. The
cells were kept in liquid nitrogen. For cell culture, they were maintained in
RPMI 1640 supplemented with 2 mM L-glutamine, 10% fetal calf serum, 400
U/ml penicillin, and 200 mg/ml streptomycin, at 37 ºC in a 5% CO2
atmosphere.
Scrape wound
migration/proliferation assay
The potential for migration and proliferation of BLS-211 and BLZ-211
cells were assessed using an in vitro scrape wound assay. The two kinds
of cells were plated in a 50 ml culture bottle (5105 cells).
At confluence, monolayers were scraped using a sterile stick to make a straight
groove in the cell monolayer. Cell culture media were aspirated and replaced by
8 ml of fresh RPMI 1640. At 0, 24 and 48 h after wounding, cells were viewed
and photographed.
RNA isolation and cDNA library
generation for SSH
Cells were grown to 80%–90% confluence. Total RNA was isolated using
Trizol reagent (Invitrogen, Carlsbad, USA) following the protocols provided by
the supplier, and then dissolved in nuclease-free water. mRNA was extracted
using a Poly(A) pure kit (Qiagen, VIC, Australia). Using 2 mg of mRNA, cDNA
synthesis was carried out with a PCR-select cDNA subtraction kit (Clontech,
Palo Alto, USA) according to the recommendation from the supplier. SSH of the
cDNA was then carried out using the advantage cDNA polymerase included in the
same kit. By reversing the tester and the driver, two subtractive libraries
were generated in order to isolate upregulated transcripts in both BLZ-211
and BLS-211 cells. The SSH-cDNA repertoire for each experiment was cloned into
the pGEM-T easy vector (Promega, Madison, USA) and transformed into Escherichia
coli JM109, then stored at –80 ºC.
Differential screening using
subtracted probes
Colonies from each library were selected and the inserts amplified
by PCR using adaptor primers (5‘-TCGAGCGGCCGCCCGGGCAGGT-3‘ and 5‘-AGCGTGGTCGCGGCCGAGGT-3‘)
and Taq polymerase. The PCR products were analyzed, sized by gel
electrophoresis, then arrayed in duplicate by manually spotting onto nylon
membranes and fixed at 80 ºC in an oven. The arrays were screened by
hybridization using 32P-labeled unsubtracted and subtracted cDNA
probes which were generated from BLZ-211 and BLS-211 SSH-cDNA by random priming
(Clontech). Prior to labeling, the adaptor sequences were removed from the
SSH-cDNA by digestion with RsaI and SmaI, followed by
purification using a QIAquick kit (Clontech). Membranes were hybridized with 32P-labeled probes at 42 ºC and were detected by autoradiography.
Colonies that showed differences 2-fold greater in signal intensity as assessed
by densitometry were taken for further analysis.
DNA sequencing
Sequencing of all cDNA clones in the two SSH libraries was carried
out on ABI 377 DNA Sequencer at Shanghai Huanuo Biotechnology Company (Shanghai,
China) using M13 forward primer. Nucleic acid sequence homology searches were
performed by comparison to the National Center for Biotechnology Information’s
RefSeq, GenBank, and expressed sequence tags databases (dbEST) using BLAST (http://www.ncbi.nlm.nih.gov/blast).
Reverse transcription (RT)-PCR
analysis
Total RNA (1 mg) from the cells was reversely transcribed to cDNA by SuperScript
II reverse transcriptase (Invitrogen) in a 20 ml reaction volume. PCR was
performed with 0.5 ml cDNA template, 1´reaction
buffer, 0.2 mM dNTP mix, 25 mM each primer and 0.3 u Taq DNA polymerase
(TaKaRa, Dalian, China), and subjected to 30 cycles of amplification at 94 ºC
for 30 s, 54 ºC for 30 s, and 72 ºC for 40 s. The G3PDH was used as an
internal control in each PCR amplification. The PCR primers for each individual
gene were synthesized by TaKaRa and Shanghai Huanuo Biotechnology Company. The
PCR products were separated by electrophoresis in 1.5% agarose gel stained
with ethidium bromide and photographed under ultraviolet light.
Real-time RT-PCR analysis
Using real-time RT-PCR, four genes were chosen for further analysis.
Primers of target genes were designed and synthesized by TaKaRa, and are listed
in Table 1. A 7300 real-time PCR system was used (Applied Biosystems,
Foster City, USA). PCR reactions were performed in a total volume of 25 ml containing
12.5 ml 2´SYBR premix ExTag (TaKaRa), 0.5 ml 50´Rox Reference Dye (TaKaRa), 1.0 ml each primer (200 ng) and
1.0 ml cDNA of BLS-211 or BLZ-211 cells in each sample. In order to
guarantee the comparability of the calculated gene mRNA expression in all
analyzed samples, the housekeeping gene G3PDH was amplified. The
amplification program included a denaturation step (10 s at 95 ºC) and an
amplification step (5 s at 95 ºC, 15 s at 59 ºC, 31 s at 72 ºC, 40 cycles),
followed by a melting curve program.
Results
Morphologic characteristics
under light microscopy
As shown in Fig. 1, BLZ-211 cells appeared to have a
polarized epithelial phenotype and grow in a tightly connected manner [Fig.
1(A)], whereas BLS-211 cells were spindle-shaped and loosely distributed, as
previously described [Fig. 1(B)].
Scrape wound
migration/proliferation assay
Fig. 2 shows representative images of
the wound scrape after 0, 24 and 48 h for each group of cells. After 24 h, BLS-211
cells had completely healed the wound, but BLZ-211 cells had only healed
approximately 60% of it. After 48 h, BLZ-211 had healed 90% of the wound.
BLS-211 cells have the ability to migrate and proliferate more quickly than
BLZ-211 cells.
cDNA library generation
We applied SSH to mRNAs isolated from BLZ-211 and BLS-211 cell
lines. The subtractions were conducted in libraries for both directions
(BLZ-211 minus BLS-211) and (BLS-211 minus BLZ-211). Two SSH cDNA libraries
were constructed. They were considered to be of good quality in determining the
efficiency of the subtraction. We used 2 ml secondary PCR products
and 1 ml T-easy vector for ligation, and 50 ml high efficiency competent
cells (1´108 cfu/mg DNA) were used
for transformations. The subtractive forward and reverse cDNA libraries had 168
and 305 white colonies, respectively. White colonies should contain inserts,
whereas blue colonies should not. We found that over 90% of the white colonies
contained inserts.
Differential screening
cDNA clones from each library were arrayed in a duplicate manner by
spotting onto nylon membranes and screened with reverse hybridization using
subtracted and unsubtracted cDNA probes. The screening of the subtracted clones
permitted ready identification of the genes that were differentially expressed
and substantially increased the efficiency with which colonies of interest
could be identified (Fig. 3). Of the clones analyzed, 64 (37 in BLZ-211
and 27 in BLS-211 cells) were detected as overexpressed in one cell line
relative to the other, based on the criteria of a greater than 2-fold
difference in hybridization signal intensity.
Identification of
differentially expressed genes
Thirty-six different transcripts were found among the 64 clones
isolated. Tables 2 and 3 show the upregulated genes in BLZ-211
and BLS-211 cells, respectively. Among them, 17 were identified as known genes
by homology, and 19 were genes of unknown function, as determined by searching
the GenBank database. These 19 were unknown transcripts, and were collected as
new expressed sequence tags by the GenBank dbEST database with the accession
numbers DR008207, DR010178, DR159652–159660, DY230447–230448, and DY505708–DY505713.
RT-PCR analysis
We used RT-PCR to further examine the isolated genes (Fig. 4).
With RT-PCR, we found that Vimentin (PCR product 119 bp) was
overexpressed in BLS-211 cells and Keratin7 (PCR product 111 bp) and Fibronectin
(PCR product 143 bp) were overexpressed in BLZ-211 cells.
Real-time RT-PCR analysis
mRNA expression profiles of the four candidate genes were analyzed.
Following amplification, CT (threshold cycle), DCT (normalization of CT for target gene relative to G3PDH
CT) and DDCT (comparing differences in the DCT values
of BLS-211 to BLZ-211) values were calculated and are shown in Table 4.
Relative quantifications of the expression levels of the four candidate genes between
BLS-211 and BLZ-211 were analyzed using real-time RT-PCR. We confirmed that
expression levels of Vimentin and TAL6 increased in BLS-211, Fibronectin
and Keratin7 decreased in BLS-211 (Table 4).
Discussion
SSH is a powerful approach for the identification of genes that are
differentially expressed in one cell population compared with another. Although
cDNA microarray is increasingly applied for massive parallel analysis of gene
expression, SSH is still widely used as it enables the recovery of abundant, as
well as low copy-number, mRNA transcripts. In addition, SSH not only permits
efficient identification of phenotype-associated genes with known function, but
also allows the unbiased isolation of novel sequences that are not yet
available on the microchips. But the most tedious step of this method is
validation of the clones.
In this work, we arrayed and screened initial SSH clones using
reverse hybridization that could minimize the number of non-relevant clones. We
then further verified some genes by real-time RT-PCR. In order to exclude the
potential genomic DNA contamination in RT-PCR, we used DNase to isolated RNA,
and a negative control (contains RNA and all other reagents except for the
reverse transcriptase) during the cDNA synthesis portion was included. Using
these tools, we compared two TCC cell lines characterized by different
morphology. We found that 17 known genes and 19 unknown genes were
differentially expressed. Two cytoskeletal
genes, Keratin7 and Vimentin, were identified. Keratin7
was overexpressed in BLZ-211 cells, whereas Vimentin was overexpressed
in BLS-211 cells. Both encode intermediate filament proteins. Normal adult
urothelium is known to express keratin subtypes with the characteristic of
“simple” epithelia (K7, K8, K18, K19 and K20) and
“stratified” epithelia (K13). K7 is retained by all bladder TCCs, and
is regarded as a bladder epithelial tumor marker [13]. Vimentin is regarded as
a mesenchymal marker and is not expressed in normal adult urothelium. But in
our study, Vimentin was upregulated in the BLS-211 cell line. The
cytoskeletal gene expressions are consistent with phenotypic differences in
cell shape, which suggests that they might contribute to the spindle shape for
BLS-211 cells and the epithelial shape for BLZ-211 cells. We also detected Vimentin
and Keratin7 expression in two cell lines of human cervical carcinomas,
CS1213 and RJC. These two cell lines displayed anchor-dependent or
anchor-independent characteristics. The anchor-independent cell line (CS1213)
displayed elliptic or spindle cell shape and overexpressed the Vimentin
gene, but the anchor-dependent cell line (RJC) showed epithelial phenotype and
highly expressed the Keratin7 gene, which confirmed that cytoskeletal
genes can affect cell shape and growth.
We found that, in in vitro scrape wound assays,
healing from high-strength induced wound in
BLS-211 cells took place at 24 h, but the process took more than 48 h in
BLZ-211 cells. This suggests that the migration of BLS-211 cells is faster than
that of BLZ-211 cells. In addition, cloning efficiency in soft agar is higher
in BLS-211 (12%) than that in BLZ-211 (6%). This might suggest that BLS-211
cells are relatively more invasive. Furthermore, BLS-211 cells are established
from bladder TCC and also overexpress epithelial marker TAL6, a
tumor-associated antigen L6, or transmembrane 4 superfamily member 1. TAL6 has
been reported to be associated with cell motility and metastatic potential of
solid tumor [14,15]. This would suggest that BLS-211 is a type of
epithelial-mesenchymal transition (EMT) cells. In vitro, EMT results in
increased cell motility, loss of epithelial morphology and acquisition of
mesenchymal characteristics such as expression of intermediate filament
vimentin [16]. Therefore, the BLS-211 cell line can be a model for studying EMT
in bladder TCC.
Mitochondrial genes that overexpressed in BLZ-211 cells included NADH
dehydrogenase, ATP synthase 6 and cytochrome c oxidase. These
genes are associated with cell energy metabolism. This reflects that BLZ-211
cells are metabolically highly active. These findings are in accordance with
previous electron microscopy results showing that BLZ-211 cells have larger and
increased numbers of mitochondria [3].
Certain genes encoding enzymes were overexpressed in BLS-211 cells,
such as UCHL1 (ubiquitin thiolesterase), TXNRD1 (thioredoxin
reductase 1), and DDH (dihydrodiol dehydrogenase 1). They might also be
involved in phenotypic differences in BLS-211 morphology. UCHL-1 protein is a
thiol protease that hydrolyzes a peptide bond at the C-terminal of ubiquitin
and is involved with the processing of ubiquitin precursors and ubiquinated
proteins. Studies have suggested that ubiquitin plays an important role in EMT.
The ubiquitination of E-cadherin is essential for the shuttling of E-cadherin
to the lysosome and its degradation. E-cadherin is a type I transmembrane
protein of the adherens junctions and is degraded in EMT [17]. Thioredoxin
reductase 1 (TrxR1) is a cytosolic enzyme that plays a central role in
controlling cellular redox homeostasis. TrxR1 can transduce regulatory redox
signals through NADPH-dependent reduction of thioredoxin (Trx), which is able
to reduce a broad spectrum of target enzymes and regulate the activity of
several transcription factors (e.g., p53 and NF-kB). The TrxR1/Trx system is
involved in every step of cancer biology, ranging from transformation and
progression to invasion, metastasis and resistance to therapy [18]. Therefore,
they might contribute to the morphological phenotype of BLS-211. Dihydrodiol
dehydrogenase is a member of the aldo-keto reductase superfamily, which changes
the aldehyde or ketone moiety to a corresponding alcohol by using NADH or NADPH
as a cofactor [19]. Overexpression of DDH has been found in some
cancers, for example, breast, lung, esophageal, cervical and prostate cancers
[20–24],
and has been associated with disease progression. Our finding that the
overexpression of these genes might be related to the morphological features of
BLS-211 appears to be the first time such a relationship has been described in
bladder cancers, and needs to be confirmed further.
In summary, these results suggest that phenotypic differences in
cell shape could be related to genes regulating the cytoskeleton and enzymes.
Differentially expressed genes found in this work provide an important
direction for further exploration of human bladder TCC progression.
References
1 Kurth KH, Bouffioux C, Sylvester R, van der
Meijden AP, Oosterlinck W, Brausi M. Treatment of superficial bladder tumors:
Achievements and needs. The EORTC Genitourinary Group. Eur Urol 2000,
37(Suppl): 1–9
2 Gildea JJ, Golden WL, Harding MA, Theodorescu
D. Genetic and phenotypic changes associated with the acquisition of
tumorigenicity in human bladder cancer. Gene Chromosome Cancer 2000, 27: 252–263
3 Li X, Chen W, Nan XY, Yan CF, Yang YC, Cheng
XL. The biological characteristics of a newly established human bladder
transitional cell carcinoma cell line BLZ-211. Chin J Urol 1997, 18: 328–331
4 Yan CF, Li X, Chen W. Karyotypic
characteristics of three new established human bladder transitional cell
carcinoma lines. Journal of Xi’an Medical University 1997, 18: 427–429
5 Lin R, Li X, Chen W. Chemosensitivity testing
of three novel bladder transitional cell carcinoma lines derived from an
untreated patient and BIU-87 cell line. Journal of Xi’an Medical University
2000, 21: 39–41
6 Chen W, Li X, Cheng XL, Lin R. Study on
invasion related factors of three human urinary bladder cancer cell lines.
Journal of Xi’an Jiaotong University (Medical Sciences) 2004, 25: 89–91
7 Davis MM, Cohen DI, Nielsen EA, Steinmetz M,
Paul WE, Hood L. Cell-type-specific cDNA probes and the murine I region: The
localization and orientation of Ad alpha. Proc Natl Acad Sci USA 1984, 81: 2194–2198
8 Liu YG, Xiong YZ, Deng CY. Isolation,
sequence analysis and expression profile of a novel swine gene differentially
expressed in the Longissimus dorsi muscle tissues from Landrace´Large white cross-combination. Acta Biochim
Biophys Sin 2005, 37: 186–191
9 Lisitsyn N, Lisitsyn N, Wigler M. Cloning the
differences between two complex genomes. Science 1993, 259: 946–951
10 Schena M, Shalon D, Davis RW, Brown PO.
Quantitative monitoring of gene expression patterns with a complementary DNA
microarray. Science 1995, 270: 467–470
11 Velculescu VE, Zhang L, Vogelstein B, Kinzler
KW. Serial analysis of gene expression. Science 1995, 270: 484–487
12 Diatchenko L, Lau YF, Campbell AP, Chenchik A,
Moqadam F, Huang B, Lukyanov S et al. Suppression subtractive
hybridization: A method for generating differentially regulated or
tissue-specific cDNA probes and libraries. Proc Natl Acad Sci USA 1996, 93:
6025–6030
13 Langner C, Wegscheider BJ, Rehak P, Ratschek
M, Zigeuner R. Prognostic value of keratin subtyping in transitional cell
carcinoma of the upper urinary tract. Virchows Arch 2004, 445: 442–448
14 Hashida H, Takabayashi A, Tokuhara T, Hattori
N, Taki T, Hasegawa H, Satoh S et al. Clinical significance of
transmembrane 4 superfamily in colon cancer. Br J Cancer 2003, 89: 158–167
15 Kao YR, Shih JY, Wen WC, Ko YP, Chen BM, Chan
YL, Chu YW et al. Tumor-associated antigen L6 and the invasion of human
lung cancer cells. Clin Cancer Res 2003, 9: 2807–2816
16 Arias AM. Epithelial mesenchymal interactions
in cancer and development. Cell 2001, 105: 425–431
17 Palacios F, Tushir JS, Fujita Y,
D’Souza-Schorey C. Lysosomal targeting of E-cadherin: A unique mechanism for
the down-regulation of cell-cell adhesion during epithelial to mesenchymal
transitions. Mol Cell Biol 2005, 25: 389–402
18 Gorreta F, Runfola TP, Vanmeter AJ, Barzaghi
D, Chandhoke V, Del Giacco L. Identification of thioredoxin reductase
1-regulated genes using small interference RNA and cDNA microarray. Cancer Biol
Ther 2005, 4: 1079–1088
19 Vogel K, Bentley P, Platt KL, Oesch F. Rat
liver cytoplasmic dihydrodiol dehydrogenase. Purification to apparent
homogeneity and properties. J Biol Chem 1980, 255: 9621–9625
20 Ji Q, Aoyama C, Nien YD, Liu PI, Chen PK,
Chang L, Stanczyk FZ et al. Selective loss of AKR1C1 and AKR1C2 in
breast cancer and their potential effect on progesterone signaling. Cancer Res
2004, 64: 7610–7617
21 Hsu NY, Ho HC, Chow KC, Lin TY, Shih CS, Wang
LS, Tsai CM. Overexpression of dihydrodiol dehydrogenase as a prognostic marker
of non-small cell lung cancer. Cancer Res 2001, 61: 2727–2731
22 Wang LS, Chow KC, Wu YC, Lin TY, Li WY.
Inverse expression of dihydrodiol dehydrogenase and glutathione-S-transferase
in patients with esophageal squamous cell carcinoma. Int J Cancer 2004, 111:
246–251
23 Ueda M, Hung YC, Chen JT, Chiou SH, Huang HH,
Lin TY, Terai Y et al. Infection of human papillomavirus and
overexpression of dihydrodiol dehydrogenase in uterine cervical cancer. Gynecol
Oncol 2006, 102: 173–181
24 Stanbrough M, Bubley GJ, Ross K, Golub TR,
Rubin MA, Penning TM, Febbo PG et al. Increased expression of genes
converting adrenal androgens to testosterone in androgen-independent prostate
cancer. Cancer Res 2006, 66: 2815–2825