Expression profiling of a human cell line model of prostatic cancer reveals a direct involvement of interferon signaling in prostate tumor progression.

Abstract

Cancer-associated fibroblasts induce malignant behavior in genetically initiated but nontumorigenic human prostatic epithelium. The genetic basis for such transformation is still unknown. By using Affymetrix GeneChip technology, we profiled genomewide gene expression of transformed [tumorigenic benign prostatic hyperplasia (BPH1)(CAFTD)] and parental (nontumorigenic BPH1) cells. We identified differentially expressed genes, which are associated with tumorigenesis or tumor progression. One striking finding is that a significant portion of the down-regulated genes belongs to interferon (IFN)-inducible molecules. We show that IFN inhibited the tumorigenic BPH1(CAFTD) cell proliferation and colony formation in vitro and inhibited tumor growth in xenografts in vivo. Expression of the IFN-inducible molecules correlates with the growth-inhibiting effects of IFN. In addition, these genes are reported to be mapped mainly to two chromosomal regions, 10q23-26 and 17q21, which are frequently deleted in human prostate cancers. Furthermore, in silico data-mining with the GeneLogic database revealed that expression of the IFN-inducible genes was down-regulated in approximately 30% of the 49 clinically characterized samples of prostatic adenocarcinomas. Collectively, we show that there seems to be a direct link between IFN-inducible molecules and prostatic tumor progression. These findings suggest IFN-inducible molecules as potential therapeutic targets for the treatment of prostate cancer.

Full text

Page 1

Expression profiling of a human cell line model of
prostatic cancer reveals a direct involvement of
interferon signaling in prostate tumor progression
Jianyong Shou*, Robert Soriano†, Simon W. Hayward‡§, Gerald R. Cunha‡¶, P. Mickey Williams†, and Wei-Qiang Gao*?
Departments of *Molecular Oncology and†Molecular Biology, MS No. 72, Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080; and Departments of
‡Urology and¶Anatomy, University of California, San Francisco, CA 94143
Communicated by Richard H. Scheller, Genentech, Inc., Stanford, CA, December 27, 2001 (received for review October 22, 2001)
Cancer-associated fibroblasts induce malignant behavior in geneti-
cally initiated but nontumorigenic human prostatic epithelium. The
genetic basis for such transformation is still unknown. By using
Affymetrix GeneChip technology, we profiled genomewide gene
expression of transformed [tumorigenic benign prostatic hyperplasia
(BPH1)CAFTD] and parental (nontumorigenic BPH1) cells. We identified
differentially expressed genes, which are associated with tumorigen-
esis or tumor progression. One striking finding is that a significant
portion of the down-regulated genes belongs to interferon (IFN)-
inducible molecules. We show that IFN inhibited the tumorigenic
BPH1CAFTDcell proliferation and colony formation in vitro and inhib-
ited tumor growth in xenografts in vivo. Expression of the IFN-
inducible molecules correlates with the growth-inhibiting effects of
IFN.Inaddition,thesegenesarereportedtobemappedmainlytotwo
chromosomal regions, 10q23–26 and 17q21, which are frequently
deletedinhumanprostatecancers.Furthermore,insilicodata-mining
with the GeneLogic database revealed that expression of the IFN-
induciblegeneswasdown-regulatedinapproximately30%ofthe49
clinically characterized samples of prostatic adenocarcinomas. Collec-
tively, we show that there seems to be a direct link between
IFN-inducible molecules and prostatic tumor progression. These find-
ings suggest IFN-inducible molecules as potential therapeutic targets
for the treatment of prostate cancer.
P
United States. Because of histological heterogeneity and the
difficulty of conventional genetic analysis, our knowledge about
the genes specifically involved in prostate carcinogenesis is still
very limited. Gene expression profiling, using a powerful DNA
microarray technique, has recently proven to be an effective way
to globally analyze genes involved in carcinogenesis of many
types of tissues (1, 2) including the prostate (3–5).
Our previous studies have demonstrated that the nontumori-
genic human benign prostatic hyperplasia (BPH)1 cell line and
its tumorigenic sublines provide a convenient model system for
studying prostate tumorigenesis (6–8). The BPH1 line, immor-
talized by means of a large T antigen oncogene, fails to form
tumors even though it survives for several months after being
grafted into immunodeficient mice (6). However, these cells
become tumorigenic if they are recombined with carcinoma-
associated fibroblasts (CAFs) derived from human prostatic
carcinomas before grafting (7). The conversion of the nontu-
morigenic cells to tumorigenic cells elicited by prostate CAF
seems to be irreversible, because when epithelial cells isolated
from these primary tumors are grafted to new hosts in the
absence of CAF, tumors form (8). Advantages of the BPH1 cells
include their clonal derivation and relative consistency. Primary
prostate tumor tissue, in contrast, is often heterogeneous with
mixed cell types and various tumor grades.
In the present experiments, we have used RNAs extracted from
the nontumorigenic parental BPH1 line and from two tumorigenic
sublines, BPH1CAFTD-01 and -06. We examined the expression
profiles of approximately 12,000 probe sets by using Affymetrix
rostate cancer is the most frequent malignancy in aged men
and the second leading cause of male cancer death in the
(Santa Clara, CA) GeneChip U95A. We found 179 up-regulated
and 95 down-regulated probe sets in tumorigenic BPH1CAFTDcells
compared with the parental nontumorigenic BPH1 cells. It is
particularly interesting that 18 of the 95 down-regulated hits were
IFN-inducible molecules. DNA synthesis analysis revealed that
IFNs inhibited proliferation of both types of BPH cells. Soft agar
assays indicated that colony formation of BPH1CAFTDcells was
inhibited when IFN-? was included in the cultures. In addition,
xenograft experiments in nude mice showed that treatment with
IFN-?greatlyreducedtumorvolumes.Reversetranscription(RT)-
PCRanalysisrevealedthatacyclin-dependentkinaseinhibitor,p21,
issignificantlydown-regulatedinthetumorigenicBPH1CAFTDcells,
andtreatmentofthesecellswithIFN-?reverseditsexpressionback
tothebasallevelsofnontumorigenicBPH1cells.Mostimportantly,
the deceased expression of the IFN-inducible genes was confirmed
in approximately 30% of the 49 clinical samples of prostatic
adenocarcinomas.
Materials and Methods
Cell Culture and RNA Preparation. Parental BPH1 and BPH1CAFTD
cells were maintained in RPMI medium 1640 plus 5% FBS
(GIBCO?BRL) in 5% CO2-humidified tissue culture incubator.
Total RNAs were isolated by using RNeasy columns (Qiagen,
Chatsworth, CA), treated with DNase I (Roche Molecular
Biochemicals) for 40 min at room temperature, and cleaned with
RNeasy columns.
Affymetrix GeneChip Probe Array Analyses. Oligo microarray exper-
imentswereperformedandanalyzedasdescribed(9).Briefly,after
quality determination on test arrays, the samples were hybridized
for16hat45°CtoAffymetrixHumanGenomearrays(U95A).The
arrays were washed and then stained with streptavidin-
phycoerythrin (genome arrays were amplified with an anti-
streptavidin Ab). The arrays were scanned with the GeneArray
scanner (Agilent Technologies, Palo Alto, CA). Raw data were
collectedandanalyzedbyusingAffymetrix MICROARRAY SUITEand
DATA MINING TOOLS software. Experiments were done in six
replicates for two BPH1 samples. Mann–Whitney pairwise com-
parison (9) was performed to identify genes that were differentially
expressed. Each of the six parental BPH1 samples was compared
with BPH1CAFTDsamples resulting in 36 pairwise comparisons for
both samples. Genes with concordance exceeding 80.6% were
considered statistically significant (P ? 0.05). Gene lists from both
Abbreviations: BPH, benign prostatic hyperplasia; LNCaP, lymph node-derived prostate
cancer; RT, reverse transcription; CAF, carcinoma-associated fibroblasts; STAT, signal trans-
ducers and activator-1.
§Presentaddress:DepartmentofUrologicSurgery,DepartmentofCancerBiology,Vander-
bilt–IngramComprehensiveCancerCenter,VanderbiltProstateCancerCenter,Vanderbilt
University Medical Center, Nashville, TN 37232.
?To whom reprint requests should be addressed. E-mail: gao@gene.com.
The publication costs of this article were defrayed in part by page charge payment. This
article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
§1734 solely to indicate this fact.
2830–2835 ?
PNAS ?
March 5, 2002 ?
vol. 99 ?
no. 5www.pnas.org?cgi?doi?10.1073?pnas.052705299

Page 2

comparisons were then loaded into GENESPRING software (Silicon
Genetics, Redwood City, CA) to look for common Affymetrix
probe sets in both lists. This resulted in 179 probe sets increased in
common to both lists and 95 in common that decreased expression
relative to BPH1CAFTD.
GeneLogic (Gaithersburg, MD) BioExpress Database. A commercially
available database, BioExpress, was used to confirm our exper-
imental data. At the time of this work, the database included 49
human prostate tumors and 26 benign prostate hyperplasia
samples that had been analyzed on Affymetrix GeneChips (U95
series). Average difference expression values as reported by the
database were used to identify a subset of 14 prostate tumors,
which demonstrated a down-regulation of the 18 IFN-inducible
genes, when compared with prostate hyperplasia samples.
TaqMan (Perkin–Elmer?Applied Biosystems) Real-Time Quantitative
RT-PCR Analysis.TaqManRT-PCRanalysiswasperformedwiththe
5?-exonuclease assay by using fluorescent nonextendible oligonu-
cleotideprobes(TaqManPCRdetector7700)asdescribed(10,11).
Cells were grown in Falcon 96-well plates. High-quality RNA was
prepared by using Applied Biosystems PRISM 6700 Automated
Nucleic Acid Workstation and subjected to TaqMan RT-PCR.
Initial RT-PCR amplifications were also examined by agarose gel
electrophoresis to ensure that bands were visible only at the
expected molecular weights. We used specific probes and primers
for human IFN-inducible 56-kD protein, iip56 (probe, TCT CAC
AGAGTTCTCAAAGTCAGCAGCCA;forwardprimer,TGA
GCG GGC CCT GAG A; reverse primer, ATA TCT GGG TGC
CTA AGG ACC TT), human IFN-stimulated 54-kD protein, isp54
(probe, TCC ATT CTT GCC AGC CTC CAT GCT; forward
primer, CCA ATG ATA ATC TCT TCC GTG TCT G; reverse
primer, TCT GCG TCT TCA TAC TGA TCT GCT), human ifp35
(probe, TGC CCA TAT AGG AGG TCT GTA TGT TCA CCA
AC; forward primer, CAC TGG CCT GGG CTT GG; reverse
primer, ATG TGT GAC CCC TCC GCA), human cig49 (probe,
TGACTGCGCCCTGGCCCA;forwardprimerTGTCAGCAT
CTG AGC TTG AGG A; reverse primer, AGG AGC TCT CTG
GGA CTG GAG), human aprf (probe, CGG CCA GAG AGC
CAG GAG CAT C; forward primer, GGA GGC ATT CGG GAA
GTATTG;reverseprimer,CGCTACCTGGGTCAGCTTCA),
human p21 (probe, TGG CCT GGA CTG TTT TCT CTC GGC;
forward primer, AAC ACC TTC CAG CTC CTG TAA CAT A;
reverseprimer,GGGAACCAGGACACATGGG),andhuman
gapdh (probe, CTGGCATTGCCCTCAACGACCAC; forward
primer, CTCCTCCACCTTTGACGCTG; reverse primer, CAT-
ACCAGGAAATGAGCTTGACAA).
BrdUrd Immunohistochemistry, Microscopy, and Image Acquisition.
BPH1 and BPH1CAFTD-01 cells were grown either in the pres-
ence or absence of IFN-? for 24 h. BrdUrd (1:5000, Amersham
PharmaciaBiosciences)wasaddedtothemediumforthelast5h.
Cells were then fixed and processed for anti-BrdUrd immuno-
cytochemistry as described.
Tritiated-Thymidine Incorporation Assays. To measure DNA synthe-
sis, an identical number of BPH1 and BPH1CAFTD-01cells (4 ? 103
per well) were plated in 96-well plates. [3H]thymidine (1 ?Ci per
well) was added for 16 h at 24 h of culture. Cells were harvested by
using a Tomtec (Orange, CT) cell harvester as described (11, 12).
Data were collected from 8 culture wells of each experimental
group and expressed as mean ? SE. Two-tailed unpaired t test was
used for statistical analysis.
Colony Formation Assay and Xenograft Experiment. Cell cutter and
colony formation assay were performed as described (8). Briefly,
cells were plated in duplicates in 6-well plates at the density of 105,
104, 103cells per well in 0.3% agar in 1? culture medium over the
bottom layer containing 0.5% agar in culture medium. Cells were
grown for 3 weeks and fed gently with 1 ml of medium 3 times a
week. Cultures were then stained overnight with 10 ?g?ml of
tetrazolium violet in medium. Colonies larger than 200 ?m in
diameter were counted. Male athymic nude mice were injected s.c.
with5?106BPH1CAFTD-01cellsin0.2mlofculturemedium.After
grafting, mice were injected s.c. daily with 5 ? 106units?kg of body
weight (high dose) or 5 ? 105units?kg of body weight (low dose)
IFN-?. Tumor sizes were measured 3 times a week over a 27-day
period.
Results
Gene Expression Profiling in BPH1 Cells. To identify genetic ele-
ments involved in prostate tumor progression, we performed
global gene expression analysis of the BPH1-based cell models
by using Affymetrix GeneChip technology. Six independent
RNA samples were prepared respectively from parental (BPH1)
and tumorigenic (BPH1CAFTD) cell lines. To minimize the
chance of getting false positives, two independent tumorigenic
cell lines (BPH1CAFTD-01 and BPH1CAFTD-06) were used in the
study.Inadditionwefocusedonlyongenesthatshowedthesame
patterns of expression in both cell lines. Isolated RNAs were
labeled and hybridized to the Affymetrix oligo probe arrays.
Experiments and data analyses were performed by using the
U95A full-length arrays as described (9) and in Materials and
Methods.
Of a total of about 12,000 probe sets, we discovered 179 showing
significantly increased expression in both tumorigenic BPHCAFTD
cell strains. To prioritize the increased calls, the gene list generated
in this experiment was crossreferenced to the data sets generated
from a variety of different human tumor specimens that were
analyzed for gene expression by using Affymetrix GeneChip tech-
nologies. Genes that showed increased expression in tumorigenic
BPH1CAFTDcells and that were also present in primary human
prostate tumors, but absent in pooled normal human epithelial
tissues, are listed in Table 1. Among this selected population of
genes several potential contributors to the tumorigenesis were
found to be up-regulated in BPH1CAFTDcells (Table 1). For
example, matrilysin (MMP7), which has elevated expression levels
in prostate (13) and breast (14) tumors. Expression of matrilysin is
reported to be regulated by mitogenic growth factors such as
fibroblastgrowthfactor(FGF;ref.15)andepidermalgrowthfactor
(EGF; ref. 16). Matrilysin has been shown to be involved in the
cleavage of integrin (17) and implicated in cancer invasion and
metastasis (18, 19). Human cox-2 is an inducer of angiogenesis. Its
expression is up-regulated in prostate cancers (20). Inhibitors of
cox-2suppressangiogenesisandinhibitcancercellgrowth(21).Our
data also revealed a possible involvement of glypican-4 in prostate
tumorigenesis. Glypicans are one of the heparin-binding proteo-
glycans,whichaffectthebindingofheparin-bindinggrowthfactors,
such as FGF, heregulin, EGF, etc., to their receptors. Recently,
there has been rising interest in the roles of glypicans in tumori-
genesis (22, 23). Expression of glypican-1 and -4 is up-regulated in
breast cancer cells, and decreased expression of glypican-1 is found
to be associated with attenuated response of the breast cancer cells
to a variety of growth factors (24). Increased expression of these
geneswithtumorprogressionofBPH1toBPH1CAFTDcellssuggests
that the BPH1 cell models may be useful to dissect out the
molecular events underlying prostate tumorigenesis.
IFN-Inducible Genes Comprise a Major Group of Down-Regulated
Genes in Tumorigenic BPH1CAFTDCells. We also discovered a total of
122 and 209 probe sets that passed the Mann–Whitney U test to
show significantly decreased expression in BPH1CAFTD-01 and
BPH1CAFTD-06 cells, respectively, with 95 hits showing decreased
expression in both cell lines. Among this group of 95 we identified
a major set of 18, nearly 20% of the total 95 hits, which are
IFN-inducible molecules. Consistently, none of these genes ap-
Shou et al.
PNAS ?
March 5, 2002 ?
vol. 99 ?
no. 5 ?
2831
CELL BIOLOGY

Page 3

peared in the increased gene calls, indicating a dramatic inhibition
of IFN pathway in the tumorigenic BPH1CAFTDcells. These 18 hits
actually represented 11 distinct genes, which are listed in Table 2.
A quantitation of the relative decrease of expression of each of
these genes is also listed in Table 2. These genes represent down-
stream effects in a variety of biological events thought to be
regulated by IFN. Because the genes in Table 2 can be stimulated
bytypeIand?ortypeIIIFN(25),theinhibitionoftheIFNpathway
is not restricted to either type of IFN.
IFN Signaling Machinery Is Intact in BPH1CAFTDCells. To examine
whether the down-regulation of IFN-inducible genes is caused by
any defects of IFN signaling machinery, we performed immuno-
cytochemistry in cultured BPH cells with an Ab specifically recog-
nizing phosphorylated signal transducers and activator-1 (STAT1).
Previous experiments have shown that STAT1 phosphorylation is
a critical event of the IFN signaling pathway. Activation of STAT1
leads to a nuclear translocation of the phosphorylated STAT1
protein (25). As shown in Fig. 1, only weak immunostaining
reactivity was seen in the cytoplasm of the cultured BPH1 and
BPH1CAFTD-01 cells, and the nuclear staining was not detected
(Fig. 1 a and d). However, 15 min after exposure to IFN-?, intense
nuclear staining was observed (Fig. 1 c and f). Similar results were
also observed in cultures treated with IFN-? (Fig. 1 b and e). These
observationssuggestthatIFNsignalingmachineryforbothtypesof
IFNs is intact in tumorigenic BPH1CAFTDcells.
IFN Directly Inhibits Tumorigenic BPH1CAFTDCell Growth Both in Vitro
and in Vivo. To provide further supporting evidence that IFN is
directly involved in prostate tumorigenesis, we performed cell
proliferation analysis measuring [3H]thymidine and BrdUrd
incorporation. Treatment with IFN-? significantly inhibited
DNA synthesis by tumorigenic BPH1CAFTD-01 cells in a dose-
dependent manner (Fig. 2a). Similar growth-inhibiting effects
were also observed for IFN-? (data not shown). Cell counts of
mitotic (BrdUrd-positive) cells in the BPH1CAFTD-01 cell cul-
tures confirmed that inclusion of IFN-? significantly reduced the
number of proliferating cells (Fig. 2 b and c).
To determine whether IFN-? would directly influence tumor-
igenesis of BPH cells, we performed colony-formation assays in
soft agar cultures. As shown in Fig. 2d, addition of IFN-?
inhibited colony formation, whereas in control cultures many
colonies were formed.
We then extended our in vitro experiments to an in vivo setting.
Wecarriedoutxenograftexperimentsinnudemicebys.c.injection
of the tumorigenic BPH1CAFTD-01 cells. Although tumors were
formed in the mice of all three groups, including the saline control,
low dose (5 ? 105units?kg), and high dose (5 ? 106units?kg) of
IFN-?, there was a dramatic reduction in tumor growth in IFN-?
treated mice. After 27 days, 8 of 10 mice in the control group, but
only 3 of 10 in high-dose group and 4 of 10 in the low-dose group,
had tumors larger than 400 mm3. Although tumors grew aggres-
sively as a function of time in the mice receiving saline injection,
Table 1. List of genes that are increased in tumorigenic BPH cells, which are also present in primary human
prostate tumors, but not in normal epithelial tissue pools
Affymetrix IDsAccession no. Description
Fold increase
BPH1CAFTD-01 BPH1CAFTD-06
546_at
668_s_at
1069_at
1736_at
1788_s_at
33933_at
35330_at
35732_at
36491_at
36508_at
36671_at
38827_at
S76965
L22524
U04636
M62402
U48807
X63187
AJ012737
AL031427
D82345
AF030186
M27396
AF038451
Protein kinase inhibitor
Human matrilysin
Human cyclooxygenase-2 (hCox-2)
Human insulin-like growth factor binding protein 6
Human MAP kinase phosphatase (MKP-2)
HE4 extracellular proteinase inhibitor homologue
Muscle isoform filamin
Novel protein
NB thymosin beta
Glypican-4 (GPC4)
Human asparagine synthetase
Secreted cement gland protein XAG-2 homolog
1.84 ? 0.17
53.14 ? 25.21
2.47 ? 0.39
2.28 ? 0.28
3.99 ? 1.24
2.22 ? 0.35
8.87 ? 2.39
2.23 ? 0.52
4.63 ? 1.88
4.90 ? 1.26
2.16 ? 0.31
1.93 ? 0.53
3.11 ? 0.3
113.55 ? 55.23
2.98 ? 0.44
3.32 ? 0.37
2.98 ? 0.91
2.04 ? 0.28
26.55 ? 4.44
2.39 ? 0.55
13.06 ? 5.45
7.53 ? 1.55
2.19 ? 0.27
2.70 ? 0.63
Table 2. IFN-inducible genes that are down-regulated in tumorigenic BPH cells
Affymetrix IDAccession no.Description
Fold decrease
BPH1CAFTD-01BPH1CAFTD-06
909_g_at
908_at
38389_at
37014_at
1358_s_at
38584_at
464_s_at
M97935_5_at
M97935_MB_at
32859_at
33339_g_at
33338_at
289_at
39708_at
35735_at
1107_s_at
915_at
32814_at
M14660IFN-stimulated 54-kD protein6.98 ? 1.24
4.76 ? 0.89
6.91 ? 1.2
5.11 ? 0.82
4.89 ? 1.07
3.69 ? 0.85
3.65 ? 0.46
3.46 ? 1.01
2.28 ? 0.36
2.75 ? 0.42
2.44 ? 0.31
2.07 ? 0.3
1.92 ? 0.2
1.88 ? 0.17
4.53 ? 1.42
2.41 ? 0.54
2.91 ? 0.41
1.70 ? 0.24
20.03 ? 7.28
9.61 ? 1.72
21.32 ? 5.03
7.29 ? 1.16
7.31 ? 1.7
7.92 ? 1.41
4.16 ? 0.68
7.85 ? 4.57
3.22 ? 0.38
4.36 ? 0.72
3.66 ? 0.7
2.63 ? 0.33
1.92 ? 0.17
1.99 ? 0.12
45.85 ? 24.14
3.27 ? 0.76
5.69 ? 0.68
3.18 ? 0.54
X04371
M33882
U22970
AF026939
U72882
M97935
2-5A synthetase
p78 protein
IFN-inducible peptide (6–16)
CIG49
Leucine zipper protein (IFP35)
Transcription factor ISGF-3
M97936
L29277DNA-binding protein (APRF)
M55542
M13755
M24594
Guanylate binding protein isoform I (GBP-2)
IFN-induced 17-kD?15-kD protein
IFN-inducible 56-kD protein
2832 ?
www.pnas.org?cgi?doi?10.1073?pnas.052705299Shou et al.

Page 4

tumor growth was significantly inhibited in the mice receiving
IFN-? injection (P ? 0.05, Fig. 2e). No significant difference was
observed between the two groups treated with different dosages of
IFN (P ? 0.538), although tumors in the high-dose groups showed
a tendency of smaller volume (Fig. 2e).
Induction of IFN-Inducible Molecules Correlates to Growth-Inhibitory
Effects. Thus, in tumorigenic BPH1CAFTDcells, the expression of
IFN-inducible molecules was suppressed. In addition, IFNs inhib-
itedBPH1CAFTDcellgrowthinvitroandinvivo.Theseobservations
prompted us to determine the effects of IFN on the expression of
IFN-inducible molecules during growth inhibition of BPH1CAFTD
cells. We performed TaqMan RT-PCR analysis with probes and
primers specifically designed for 5 of the 11 genes listed in Table 2.
Wefoundthatallofthe5genesexaminedwereindeedup-regulated
afterIFN(100ng?ml)treatment(Fig.3a).Insharpcontrast,lymph
node-derivedprostatecancer(LNCaP)cells[showntohavedefects
in IFN signaling (26)] did not show up-regulated expression of the
5 genes tested on IFN (100 ng?ml) treatment (Fig. 3a). Further-
more,challengingLNCaPcellswithIFN-?hadnosignificanteffect
on cell proliferation as measured by [3H]thymidine incorporation
(Fig. 3b). To ensure that LNCaP cells do not respond to IFN, the
proliferation assay was also performed in defined serum-free
medium, so as to eliminate possible factors that might obscure the
effects of IFN (Fig. 3b, lower lines). Thus, we demonstrated that
thereisadirectcorrelationbetweentheabilityofIFNtoinducethe
expression of IFN-inducible molecules and its ability to suppress
prostate cancer cell growth.
IFN Treatment Elevates Expression Levels of p21 in Tumorigenic BPH
Cells. As an attempt to understand the possible mechanism by
which IFN-? regulates proliferation of BPH1CAFTDcells, we
performed quantitative RT-PCR analysis of gene expression of
cell cycle-related genes including p21 and p27, cyclin-dependent
kinase inhibitors, in the tumorigenic BPH1CAFTDcells after
IFN-? treatment. We found that although there was no change
in the expression levels of p27 (data not shown), the expression
of p21 was significantly lower in the tumorigenic BPH1CAFTD
cells, compared with the parental nontumorigenic cells (Fig. 3c).
After IFN-? treatment, however, there was an up-regulation of
p21 in tumorigenic BPH1CAFTDcells to levels comparable to the
nontumorigenic BPH1 cells (Fig. 3c).
Down-Regulation of IFN-Inducible Genes Is Validated in Clinical Sam-
ples of Prostatic Tumors. To extend our findings in more complex
clinical human prostate adenocarcinomas, the expression of the
18 IFN-inducible hits identified above was examined by querying
the GeneLogic database (see Materials and Methods). Among a
total of 49 pathologically characterized prostatic adenocarci-
noma samples, we identified a subgroup of 14 malignant tumors
(approximately 30% of the total samples) in which the expres-
sion of the IFN-inducible molecules were down-regulated as
compared with the nonmalignant human BPH samples, consis-
tent with what we discovered from our BPH1 cell model studies.
All 18 probe sets showed a trend of decreased expression in
malignant samples, and the differences in 8 of the 18 hits were
statistically significant (Welch t test, P ? 0.05, as marked by
asterisks in Fig. 4).
Discussion
By using an Affymetrix oligo microarray technique, we com-
pared gene expression profiles of tumorigenic vs. nontumori-
genic BPH1 cells. Of a total of 12,000 probe sets we identified
274 that are differentially expressed in tumorigenic compared
with nontumorigenic BPH1 cell strains. Many of the genes, such
as matrilysin, cox-2, and glypican, identified in this study have
been previously demonstrated to be associated with carcinogen-
esis. However, to our knowledge, the present work for the first
time links glypican-4 with prostate tumor progression. More-
over, the genes listed in Table 1 are also present in primary
human prostate tumor samples but absent in pooled human
Fig. 2.
Tritiated thymidine incorporation assay. (b) Labeling index of BrdUrd-positive
cells. (c) Dual labeling of BrdUrd immunocytochemistry (visualized with Texas
red-conjugatedsecondaryAb)and4?,6-diamidino-2-phenylindole(DAPI)nuclear
staining (blue) of BPH1CAFTD-01cells in control culture (c Left) and in a culture
treated with IFN-? (c Right). Note that the number of proliferating cells was
greatlyreducedinthepresenceofIFN-?(asterisksindicatingP?0.05inaandb).
(d) Colony formation assays of BPH1CAFTD-01 cells in soft agar assays. Note that
although BPH1CAFTDcells readily form colonies in control cultures, very few
colonies are formed in the cultures treated with IFN-?. (e) IFN inhibits tumor
growthafterBPH1CAFTD-01cellsaregraftedintonudemice.Notethatsignificant
reductionintumorvolumewasobservedinmicetreatedwitheitherloworhigh
doses of IFN. For ease of viewing, error bars are depicted unidirectionally in
treatment groups. [Bar ? 100 ?m (c).]
IFN-?inhibitscellproliferationandtumorigenesisofBPH1CAFTDcells.(a)
Fig. 1.
IFNs in BPH1 and BPH1CAFTDcells. Immunocytochemical labeling of antiphos-
phorylated STAT1 in cultured cells. BPH1CAFTD-01 (a–c) and BPH1 (d–f) cells
were cultured in normal medium or 15 min after treatment with IFN-? (b and
e) and IFN-? (c and f), receptively. Note that although there is only a weak
diffused cytoplasmic signal in the untreated cells, IFN induces a rapid nuclear
translocation of phosphorylated STAT1 protein. [Bar ? 50 ?m.]
Induction of STAT1 phosphorylation and nuclear translocation by
Shou et al.
PNAS ?
March 5, 2002 ?
vol. 99 ?
no. 5 ?
2833
CELL BIOLOGY

Page 5

epithelial tissues, indicating a similar trend toward expression in
malignant tissue vs. nonmalignant samples. These findings sug-
gest that the BPH1 cell model, coupled with DNA microarray
technique and in silico data mining, can be a valid way to identify
genes associated with prostate tumor progression. To uncover
novel molecules involved in prostate tumor progression, we used
Affymetrix human expressed sequence tag chips (U95B-E).
Previous studies have suggested therapeutic potential of IFNs
for the treatment of cancer (27). However, the anti-tumor
mechanisms of IFN have not been completely understood. It has
been shown that IFN can cause cell cycle arrest, down-regulation
of her2 expression, and inhibits metastatic potential in a few
metastatic prostate cancer cell lines (28, 29), but the direct link
between down-regulation of the IFN pathway and prostate
tumorigenesiswasstilllacking.Ontheotherhand,micedeficient
in the IFN-? receptor or stat1 develop spontaneous or chemi-
cally induced tumors more rapidly than wild-type mice (30).
IFN-? and lymphocytes act together to prevent various tumor
development and shape tumor immunogenicity (31). Therefore,
IFNs are mainly believed to play an indirect immunosurveillance
role that is not specific for prostate cancer. In the present study,
using genomewide gene expression-profiling approaches, we
demonstrated that the IFN signaling pathway is indeed sup-
pressed when BPH1 cells become tumorigenic. The suppression
of IFN signaling resulted in a decreased expression of a large
group of IFN-inducible molecules in tumorigenic BPH1CAFTD
cells. Our in vitro cell proliferation and colony formation assays
coupled with in vivo xenograft studies provided strong support
for the notion that IFNs can directly act on BPH1CAFTDcells to
elicit growth-inhibiting effects. Consistent with the antiprolif-
erative effects of IFN, we showed that the expression level of p21
was down-regulated in the tumorigenic BPH1CAFTDcells and
that treatment with IFN-? brought p21 expression levels back to
the basal levels of parental nontumorigenic BPH1 cells.
There are currently several models available to study prostate
carcinogenesis. For example, the TRAMP model is generated by
targeted expression of a transgene carrying the rat probasin gene
fused to the SV40 T antigen (32). This model has been used in
various prostate cancer studies, including tumor initiation, progres-
sion, metastasis, and prevention (32, 33).By using tissue recombi-
nation technology, the functional consequences of the retinoblas-
toma (Rb) tumor suppressor in development and progression of
prostate cancer and the susceptibility to hormonal carcinogenesis
have been addressed (34). Our studies using BPH1 cell lines are an
example of another model. These cells grown in culture permit a
very controllable sample. The advantage of such a sample is the
reduction of complexity found in tumors in situ, which will have a
variety of other cell types and potential necrotic sites contained in
the harvest. We readily admit that a cell culture model does have
itslimitations.OnesuchlimitationisneithertheparentalBPH1nor
its derivatives express androgen receptor (AR). The lack of AR
expression in BPH1 cells prevents them from being an ideal model
to study hormonal responses in prostate cancer. Results obtained
from any given model system are subject to further validation by
using clinical samples. In this light, the expression of the 18
IFN-induciblehitswasscreenedinclinicalsamplesbyusinginsilico
biology approaches. By mining the GeneLogic database, which is
generated by profiling gene expression in a large sets of human
clinical samples, we identified a subgroup of prostate adenocarci-
nomas in which the expression of these hits was lower than in
nonmalignant human BPHs. Because of the high heterogeneity of
humantumorsamples,wethinkthatthisportionisasignificantone,
giventhefactthatevenHer2,whichisthedrugtargetforHerceptin,
a highly effective mAb to treat breast cancer, is overexpressed only
in about 25% of the breast cancer patients (35). The complexity in
human tumor samples makes it difficult to identify genes that may
Fig.3.
in BPH1, BPH1CAFTD-01, and LNCaP cells in the absence or presence of 100 ng?ml of IFN-?. The five IFN-inducible genes include aprf, ifp35, isp54, iip56, and cig49.
Note that expression levels of all five genes are elevated after treatment of IFN-? in both BPH1 and BPH1CAFTD-01 cells. In contrast, no elevation was seen in the
LNCaP cells. (b) Tritiated thymidine incorporation assay in LNCaP cells indicates that IFN-? has no effect on the proliferation of LNCaP cells in either
serum-containing medium or serum-free medium. (c) Expression levels of p21 in cultured tumorigenic and nontumorigenic BPH1 cells. Quantitative RT-PCR
analysis of BPH1 and BPH1CAFTD-01 cells in the absence or presence of IFN-?.
CorrelationofIFN-induciblemoleculeexpressionwithgrowthinhibition.(a)QuantitativeRT-PCRanalysisofexpressionlevelsoffiveIFN-induciblegenes
Fig. 4.
probe sets. Fourteen prostate tumors were identified in the GeneLogic Bio-
Express Database that demonstrated low-level expression of the 18 IFN-
inducible probe sets. Average expression value (i.e., average of the average
differences)isdepictedforthe14prostatetumorscomparedwiththeaverage
of the 26 BPH samples. Asterisks depict those probe sets that demonstrate
statistical significance (P ? 0.05, using a Welch t test).
Comparison of tumor vs. BPH expression values for 18 IFN-inducible
2834 ?
www.pnas.org?cgi?doi?10.1073?pnas.052705299 Shou et al.

Page 6

be associated with carcinogenesis in just a subset of tumors. The
down-regulation of IFN-inducible genes in the subset of tumors
would not have been identified if only real clinical samples were
analyzed, demonstrating the value of using the cell model system to
study complex disease.
Coincidentally, cytogenetic mapping of the IFN-inducible mol-
ecules, using the tools embedded in GENESPRING, a gene expressing
analysis software used in this study, revealed two chromosomal
regions, 10q and 17q, which harbor three and two of the IFN-
inducible molecules, respectively. aprf and ifp35 are located in the
region of 17q21, whereas isp54 is mapped to 10q23–25, iip56 is
mapped to 10q25–26, and cig49 is mapped to 10q24. Furthermore,
previous genetic studies have indicated that loss of 17q21–23 or
10q23–25 is a frequent event in prostate cancer progression. These
regions are thought to harbor putative tumor suppressors and
metastasis-suppressor genes (36, 37). Although one major tumor
suppressor that has been characterized in 10q region is pten (38),
these findings together suggest that in addition to the deletion of
pten, there could be loss or down-regulation of the IFN-inducible
molecules in prostate cancers.
It is well demonstrated that ligand-dependent activation of
IFN pathway involves a cascade of protein phosphorylation,
which requires the specific type I and type II receptors, the Janus
kinases (JAKs), and STATs [for review, see Stark et al. (25)].
Phosphorylated STATs translocate to the nucleus where target
gene transcription is activated. One simple interpretation for our
finding that expression of 11 distinct IFN-inducible genes is
suppressedinthetumorigenicBPH1CAFTDcellscouldbethatthe
receptor-STAT pathway is impaired in BPH1CAFTDcells. How-
ever, our observations that STAT phosphorylation is intact and
that IFNs are able to induce gene expression in these cells are
against this model. The second explanation for the suppression
ofIFNpathwayinBPH1CAFTDcellsmightbethatparentalBPH1
cells produce more IFNs than the BPH1CAFTDcells. Although no
evidence was obtained at mRNA levels (data not shown), such
a possibility cannot be completely ruled out because there may
be other subtypes of IFNs that have not been identified yet. On
the other hand, IFN signaling has been reported to cross paths
with others. For example, ras?AP-1 signaling antagonizes the
JAK?STAT pathway of IFN signaling in the regulation of
expression of the macrophage scavenger receptor gene, via a
mechanism of competition for limited amount of CBP and p300
(39). By analogy, it is conceivable that induction of IFN-
inducible molecules in the BPH1CAFTDcells may be suppressed
by a convergence with an antagonizing signaling pathway. It is
possible that in the BPH1CAFTDcells a crossinhibition pathway
somehow suppresses, but does not eliminate, the activation of
IFN pathway. In this way, the BPH1CAFTDcells remain respon-
sive to IFN treatment. In the nontumorigenic parental BPH1
cells, however, such suppression is not activated, resulting in a
higher expression of IFN-inducible genes.
The question, then, is whether suppression of IFN signaling is
important in tumorigenesis. To address this issue, we performed
experiments to examine the correlation of growth-inhibitory
effects of IFN and the induction of the expression of IFN-
inducible molecules. To appreciate the importance of the
10q23–25 and 17q21–23 regions, five genes located in these
regions were tested (Fig. 3a). Our data showed that treatment of
the BPH1CAFTDcells with either IFN-? or -? significantly
induced the expression of all five genes. In contrast, in LNCaP
cells, which are known to have defects in IFN signaling (26),
there was no induction of these genes. Coincidentally, IFNs did
not inhibit LNCaP cell growth. Such a correlation indicates the
importance of a direct IFN effect on the suppression of tumor
progression. It is known that in IFN-mediated immunotherapies
some patients fail to respond to IFN (27). Our study provides a
possible explanation for this phenomenon, that is, the unrespon-
sive patients might have lost the IFN-inducible genes. Such a loss
would eliminate the direct effects of IFNs. In this light, our work
not only stresses the importance of pharmacogenomics but also
suggests an alternative direction for prostate cancer treatment,
that is, to target the IFN-inducible molecules, in particular, for
those patients who are resistant to IFN therapy.
We thank V. Smith, P. Polakis, and K. Hillan for giving us the access to
their Affymetrix experiment database; T. Wu, K. Jung, and Z. Zhang for
assistance in bioinformatics; M. Ostland for assistance in biostatistics;
and J. Zavala-Solorio for assistance with xenograft experiments.
1. Perou, C., Sorlie, T., Eisen, M., van de Rijn, M., Jeffrey, S., Rees, C., Pollack, J., Ross,
D., Johnsen, H., Akslen, L., et al. (2000) Nature (London) 406, 747–752.
2. Clark, E., Golub, T., Lander, E. & Ro, H. (2001) Nature (London) 406, 532–535.
3. Dhanasekaran, S., Barrette, T., Ghosh, D., Shah, R., Varambally, S., Kurachi, K.,
Pienta, K., Rubin, M. & Chinnaiyan, A. (2001) Nature (London) 412, 822–826.
4. Luo,J.,Duggan,D.,Chen,Y.,Sauvageot,J.,Ewing,C.,Bittner,M.,Trent,J.&Isaacs,
W. (2001) Cancer Res. 61, 4683–4688.
5. Welsh, J., Sapinoso, L., Su, A., Kern, S., Wang-Rodriguez, J., Moskaluk, C., Frierson,
H. J. & Hampton, G. (2001) Cancer Res. 61, 5974–5978.
6. Hayward, S. W., Dahiya, R., Cunha, G. R., Bartek, J., Deshpande, N. & Narayan, P.
(1995) In Vitro Cell. Dev. Biol. Anim. 31, 14–24.
7. Olumi, A. F., Grossfeld, G. D., Hayward, S. W., Carroll, P. R., Tlsty, T. D. & Cunha,
G. R. (1999) Cancer Res. 59, 5002–5011.
8. Hayward, S. W., Wang, Y. Z., Cao, M., Hom, Y. K., Zhang, B., Grossfeld, G. D.,
Sudilovsky, D. & Cunha, G. R. (2001) Cancer Res. 61, 8135–8142.
9. Jin, H., Yang, R., Awad, T., Wang, F., Li, W., Williams, S., Ogasawara, A., Shimada,
B., Williams, P., de Feo, G. & Paoni, N. (2001) Circulation 103, 736–742.
10. Zheng, J. & Gao, W.-Q. (2000) Nat. Neurosci. 3, 580–586.
11. Shou, J., Ross, S., Koeppen, H., de Sauvage, F. & Gao, W.-Q. (2001) Cancer Res. 61,
7291–7297.
12. Zheng, J. L., Helbig, C. & Gao, W.-Q. (1997) J. Neurosci. 17, 216–226.
13. Hashimoto, K., Kihira, Y., Matuo, Y. & Usui, T. (1998) J. Urol. 160, 1872–1876.
14. Heppner, K. J., Matrisian, L. M., Jensen, R. A. & Rodgers, W. H. (1996) Am. J.
Pathol. 149, 273–282.
15. Klein, R. D., Maliner-Jongewaard, M. S., Udayakumar, T. S., Boyd, J. L., Nagle, R. B.
& Bowden, G. T. (1999) Prostate 41, 215–223.
16. Sundareshan, P., Nagle, R. B. & Bowden, G. T. (1999) Prostate 40, 159–166.
17. von Bredow, D. C., Nagle, R. B., Bowden, G. T. & Cress, A. E. (1997) Exp. Cell Res.
236, 341–345.
18. Agnihotri, R., Crawford, H. C., Haro, H., Matrisian, L. M., Havrda, M. C. & Liaw,
L. (2001) J. Biol. Chem. 276, 28261–28267.
19. Abdel-Ghany, M., Cheng, H. C., Elble, R. C. & Pauli, B. U. (2001) J. Biol. Chem. 276,
25438–25446.
20. Kirschenbaum, A., Liu, X., Yao, S. & Levine, A. C. (2001) Urology 58, 127–131.
21. Liu, X. H., Kirschenbaum, A., Yao, S., Lee, R., Holland, J. F. & Levine, A. C. (2000)
J. Urol. 164, 820–825.
22. Kleeff, J., Ishiwata, T., Kumbasar, A., Friess, H., Buchler, M. W., Lander, A. D. &
Korc, M. (1998) J. Clin. Invest. 102, 1662–1673.
23. Filmus, J. (2001) Glycobiology 11, 19R–23R.
24. Matsuda, K., Maruyama, H., Guo, F., Kleeff, J., Itakura, J., Matsumoto, Y., Lander,
A. D. & Korc, M. (2001) Cancer Res. 61, 5562–5569.
25. Stark, G. R., Kerr, I. M., Williams, B. R., Silverman, R. H. & Schreiber, R. D. (1998)
Annu. Rev. Biochem. 67, 227–264.
26. Goldstein, D., O’Leary, M., Mitchen, J., Borden, E. C. & Wilding, G. (1991) J. Urol.
146, 1173–1177.
27. Kuratsukuri, K., Nishisaka, N., Jones, R., Wang, C. & Haas, G. (2000) Urol. Oncol.
5, 265–273.
28. Hobeika, A., Etienne, W., Cruz, P., Subramaniam, P. & Johnson, H. (1998) Int. J.
Cancer 77, 138–145.
29. Kominsky, S. L., Hobeika, A. C., Lake, F. A., Torres, B. A. & Johnson, H. M. (2000)
Cancer Res. 60, 3904–3908.
30. Kaplan, D., Shankaran, V., Dighe, A., Stockert, E., Aguet, M., Old, L. & Schreiber,
R. (1998) Proc. Natl. Acad. Sci. USA 95, 7556–7561.
31. Shankaran, V., Ikeda, H., Bruce, A. T., White, J. M., Swanson, P. E., Old, L. J. &
Schreiber, R. D. (2001) Nature (London) 410, 1107–1111.
32. Greenberg, N. M., DeMayo, F., Finegold, M. J., Medina, D., Tilley, W. D., Aspinall,
J. O., Cunha, G. R., Donjacour, A. A., Matusik, R. J. & Rosen, J. M. (1995) Proc.
Natl. Acad. Sci. USA 92, 3439–3443.
33. Gingrich, J., Barrios, R., Morton, R., Boyce, B., DeMayo, F., Finegold, M.,
Angelopoulou, R., Rosen, J. & Greenberg, N. (1996) Cancer Res. 56, 4096–4102.
34. Wang, Y., Hayward, S., Donjacour, A., Young, P., Jacks, T., Sage, J., Dahiya, R.,
Cardiff, R., Day, M. & Cunha, G. (2000) Cancer Res. 60, 6008–6017.
35. Slamon, D. & Pegram, M. (2001) Semin. Oncol. 28, 13–19.
36. Gray, I. C., Phillips, S. M., Lee, S. J., Neoptolemos, J. P., Weissenbach, J. & Spurr,
N. K. (1995) Cancer Res. 55, 4800–4803.
37. Chekmareva, M. A., Hollowell, C. M., Smith, R. C., Davis, E. M., LeBeau, M. M. &
Rinker-Schaeffer, C. W. (1997) Prostate 33, 271–280.
38. Li,L.,Ernsting,B.R.,Wishart,M.J.,Lohse,D.L.&Dixon,J.E.(1997)J.Biol.Chem.
272, 29403–29406.
39. Horvai, A., Xu, L., Korzus, E., Brard, G., Kalafus, D., Mullen, T., Rose, D.,
Rosenfeld, M. & Glass, C. (1997) Proc. Natl. Acad. Sci. USA 94, 1074–1079.
Shou et al.
PNAS ?
March 5, 2002 ?
vol. 99 ?
no. 5 ?
2835
CELL BIOLOGY