Golgi Protein GOLM1 Is a
Tissue and Urine Biomarker
of Prostate Cancer1,2
Bharathi Laxman*, Rohit Mehra*,†, Qi Cao*,
Saravana M. Dhanasekaran*, Scott A. Tomlins*,
Jill Granger*, Adaikkalam Vellaichamy*,
Arun Sreekumar*, Jianjun Yu‡, Wenjuan Gu‡,
Ronglai Shen‡, Debashis Ghosh‡,
Lorinda M. Wright§, Raleigh D. Kladney¶,
Rainer Kuefer#, Mark A. Rubin**, Claus J. Fimmel§,
and Arul M. Chinnaiyan*,†,††
*Michigan Center for Translational Pathology, Department
of Pathology, University of Michigan Medical School,
Ann Arbor, MI 48109, USA;†Comprehensive Cancer
Center, University of Michigan Medical School,
Ann Arbor, MI 48109, USA;‡Department of Biostatistics,
University of Michigan, Ann Arbor, MI 48109, USA;
§Edward Hines VA Medical Center, Hines, IL and Division
of Gastroenterology, Hepatology and Nutrition, Loyola
University, Stritch School of Medicine, Maywood, IL 60153,
USA;¶Division of Urologic Surgery, Washington University
School of Medicine, St. Louis, MO 63110, USA;#Department
of Urology, University of Ulm, Ulm, Germany; **Department
of Pathology and Laboratory Medicine, Cornell University,
New York, NY 10021, USA;††Department of Urology,
University of Michigan Medical School, Ann Arbor,
MI 48109, USA
Prostate cancer is the most common type of tumor found in American men and is the second leading cause of
cancer death in males. To identify biomarkers that distinguish prostate cancer from normal, we compared multiple
gene expression profiling studies. Through meta-analysis of expression array data from multiple prostate cancer
studies, we identified GOLM1 (Golgi membrane protein 1, Golm 1) as consistently up-regulated in clinically local-
ized prostate cancer. This observation was confirmed by reverse transcription–polymerase chain reaction (RT-PCR)
and validated at the protein level by immunoblot assay and immunohistochemistry. Prostate epithelial cells were
identified as the cellular source of GOLM1 expression using laser capture microdissection. Immunohistochemical
staining localized the GOLM1 signal to the subapical cytoplasmic region, typical of a Golgi distribution. Surprisingly,
Abbreviations: Golm1, Golgi membrane protein 1; PSA, prostate-specific antigen; PIN, prostatic intraepithelial neoplasia; BPH, benign prostatic hyperplasia; LCM, laser cap-
Address all correspondence to: Arul M. Chinnaiyan, MD, PhD, Department of Pathology, University of Michigan Medical School, 1400 E. Medical Center Dr. 5316 CCGC,
Ann Arbor, MI 48109-0602. E-mail: email@example.com; and Sooryanarayana Varambally, PhD, Department of Pathology, University of Michigan Medical School, 1400 E.
Medical Center Dr. 5430 CCGC, Ann Arbor, MI 48109. E-mail: firstname.lastname@example.org
1This research was supported in part by the National Institutes of Health Prostate SPORE P50 CA69568 (to A.M.C.); Department of Defense Grants PC040517 (to R.M.),
PC051081 (to A.M.C. and S.V.). C.J.F. was supported by a VA Merit Review.
2This article refers to supplementary materials, which are designated by Tables W1 and W2 and are available online at www.neoplasia.com.
Received 4 August 2008; Revised 2 September 2008; Accepted 2 September 2008
Copyright © 2008 Neoplasia Press, Inc. All rights reserved 1522-8002/08/$25.00
Volume 10 Number 11November 2008pp. 1285–1294
GOLM1 immunoreactivity was detected in the supernatants of prostate cell lines and in the urine of patients with
prostate cancer. The mechanism by which intact GOLM1 might be released from cells has not yet been elucidated.
GOLM1 transcript levels were measured in urine sediments using quantitative PCR on a cohort of patients present-
ing for biopsy or radical prostatectomy. We found that urinary GOLM1 mRNA levels were a significant predictor of
prostate cancer. Further, GOLM1 outperformed serum prostate-specific antigen (PSA) in detecting prostate can-
cer. The area under the receiver-operating characteristic curve was 0.622 for GOLM1 (P = .0009) versus 0.495
for serum PSA (P = .902). Our data indicating the up-regulation of GOLM1 expression and its appearance in pa-
tients’ urine suggest GOLM1 as a potential novel biomarker for clinically localized prostate cancer.
Neoplasia (2008) 10, 1285–1294
Prostate cancer is the most commonly diagnosed malignancy and is
a leading cause of cancer-related death in the Western male popu-
lation [1,2]. Early diagnosis is critical for the effective treatment of
malignant tumors. High-throughput technologies such as DNA and
protein microarray have enabled the identification of genes and their
corresponding proteins that are differentially regulated in malignant
conditions [3,4]. These high throughput studies have offered re-
searchers a better understanding of the disease and the molecular
circuitries that are dysregulated in cancer. Our previous DNA mi-
croarray studies using prostate cancer tissue RNA have identified
multiple genes, including alpha methylacyl coenzyme A racemase
(AMACR), Enhancer of Zeste Homolog 2 (EZH2), tumor protein
TPD52, and ERG, as dysregulated in localized and metastatic pros-
tate cancer [5–8]. Many of these findings provided new insights
into the biology of prostate carcinogenesis and have identified the
next generation of candidate biomarkers and potential therapeutic
targets [9–12]. In the case of AMACR, a peroxisomal fatty acid–
metabolizing enzyme, the up-regulated protein was detectable in
patients’ urine, which indicated its potential use for noninvasive di-
agnostic studies . Furthermore, our studies indicated an immune
response to AMACR in prostate cancer patients and autoantibodies
directed against AMACR that could be detected in the patient’s se-
GOLM1 (Golm 1, NM_016548) is a resident cis-Golgi membrane
protein of unknown function. The first evidence of its up-regulation
was shown in the hepatocytes of patients with acute and chronic
forms of hepatitis and hepatocellular cancer . GOLM1 has a sin-
gle N-terminal transmembrane domain and an extensive C-terminal,
coiled-coil domain that faces the luminal surface of the Golgi appa-
ratus. N-terminal cleavage by a furin proprotein convertase resulted
in the release of the C-terminal ectomain and its appearance in serum
. The cleaved form of GOLM1 was detectable in the serum of
patients with hepatocellular cancer, a finding that may have diagnos-
tic value . Initial gene expression array studies in our laboratory,
and by others [3,18] suggested increased expression levels of GOLM1
mRNA in prostate cancer tissues. We subsequently detected GOLM1
transcripts in patient urine samples. By multiplexing with other urine
markers, we demonstrated that GOLM1 mRNA levels can serve as
significant predictors of prostate cancer . Our study confirms
the epithelial cell–specific expression of GOLM1 in prostate cancer
tissues. Furthermore, we show that the GOLM1 protein is released
from prostate cell lines in vitro and is detectable in the urine of pa-
tients with established prostate cancer. The secretion from cell lines
could be inhibited by treating cells with the protein transport inhib-
itor brefeldin A [20,21]. Our observation suggests that GOLM1 has
potential as a noninvasive biomarker of localized prostate cancer.
Materials and Methods
Quantitative Real-time Polymerase Chain Reaction
To validate GOLM1 overexpression observed in multiple gene ex-
pression profiling studies, we performed quantitative real-time poly-
merase chain reaction (qPCR) for GOLM1 expression using SYBR
green [6,22]. Briefly, cDNA was made with the total RNA isolated
from 11 benign prostatic hyperplasia (BPH), 27 localized prostate
cancers, and 8 metastatic prostate cancer samples. The quantification
of cDNA in each sample was performed by interpolating a Ctvalue
from a standard curve of Ctvalues obtained from serially diluted,
commercially prepared cDNA pooled normal prostate samples
(Clontech, Palo Alto, CA). This calculated quantity of GOLM1
was then normalized against the quantity of the housekeeping gene
glyceraldehyde-3 phosphate dehydrogenase (GAPDH) expressed in
that sample. “No–reverse transcription” controls were included to ex-
clude amplification of the genomic DNA. The primer sequences
used for GOLM1 were: 5′-CAGCGTGAAAAGCGGAATC-3′ and
5′-TCGGCCCTGTTGTGAAATA-3′. GAPDH primers were pre-
pared as described by Vandesompele et al. .
Urine samples were obtained from 333 patients with informed
consent after a digital rectal examination before either needle biopsy
(n = 269) or radical prostatectomy (n = 64) at the University of
Michigan Health System with institutional review board approval
as described earlier (Table W1) . Isolation of RNA from urine
and Transplex Whole Transcriptome Amplification (WTA) were as
described . Quantitative polymerase chain reaction was used to
detect GOLM1 and the control transcripts PSA and GAPDH from
WTA amplified cDNA essentially as described . The primer
sequences for GOLM1 , GAPDH , and PSA  were
previously described. Quantitative polymerase chain reaction was
performed on WTA cDNA from urine collected from 129 biopsy-
negative patients and 204 patients with prostate cancer (140 biopsy-
positive patients and 64 prostatectomy patients). Samples that had
PSA Ctvalues > 28 were excluded to ensure sufficient prostate cell
collection, leading to 124 biopsy-negative and 195 samples from pa-
tients with prostate cancer in the analysis. This resulted in a final data
GOLM1 Is a Tissue Biomarker of Prostate Cancer Varambally et al. Neoplasia Vol. 10, No. 11, 2008
set of samples from 195 patients with prostate cancer (133 positive
needle biopsy and 62 radical prostatectomy) and 124 biopsy-negative
patients. Threshold levels were set during the exponential phase
of the qPCR reaction using Sequence Detection Software version
1.2.2 (Applied Biosystems, Foster City, CA), with the same baseline
and threshold set for each plate, to generate threshold cycle (Ct) va-
lues for all genes for each sample. We adjusted GOLM1 was against
its mean urine PSA and GAPDH values (2(CtPSA + CtGAPDH)/ 2 −
CtGOLM1) × 1000).
The receiver-operating characteristic (ROC) curve and a Box plot
for GOLM1 were generated using Statistical Package for the Social
Sciences 11.5 (SPSS, Inc., Chicago, IL). The area under the ROC
curve (AUC-ROC) was then calculated for GOLM1 and serum
Laser Capture Microdissection
Normal and cancerous prostate epithelial cells were collected using
laser capture microdissection (LCM) as described previously .
Briefly, for LCM, the SL Microtest device with μCUT software
was applied (MMI GmbH, Heidelberg, Germany). A total area of
6 mm2was cut for all of the samples and was collected with the
aid of the adhesive surface lid from a specially manufactured tube
(MMI GmbH). Approximately 10,000 cells collected from each
sample were lysed with 30 μl of 5× SDS-reducing sample buffer
and clarified by centrifugation at 12,000g for 10 minutes. The sam-
ples were then loaded onto a gel for immunoblot analysis.
Normal and prostate cancer tissues were homogenized in NP-40
lysis buffer containing 50 mM Tris-HCl, pH 7.4, 1% Nonidet P-40
(Sigma, St. Louis, MO) and complete protease inhibitor cocktail
(Roche, Indianapolis, IN). Fifteen micrograms of protein extract was
mixedwith SDS sample buffer and run on a 10% SDS–polyacrylamide
onto polyvinylidene fluoride membranes (Amersham Pharmacia Bio-
tech, Piscataway, NJ), which were incubated for 1 hour in blocking
buffer [Tris-buffered saline with0.1% Tween (TBS-T) and 5% nonfat
dry milk]. GOLM1 rabbit polyclonal antibody (raised against amino
acids 41-400) was applied at a 1:50,000 dilution in blocking buffer
overnight at 4°C. After washing with TBS-T buffer, the membrane
for 1 hour at room temperature. The signals were visualized with the
ECL detection system (GE Healthcare) and autoradiography. For
GAPDH and actin Western blots, the GOLM1 antibody probed
membranes were stripped with western reprobe buffer (Geno-tech,
St. Louis, MO), blocked in TBS-T containing 5% nonfat dry milk,
and incubated with either rabbit anti-GAPDH antibodies (1:50,000
dilution; Abcam, Cambridge, MA) or rabbit anti–actin antibodies
(1:5000 dilution; Sigma) for 2 hours. For laser capture microdissected
prostate epithelial cell immunoblot, the cells were lysed in sample
buffer and analyzed for GOLM1 protein expression by immunoblot
analysis as described above.
At the time of initial diagnosis, urine samples (n = 52) from biopsy-
proven, clinically localized prostate cancer patients (mean ± SD age,
58.1 ± 0.60 years) were collected with informed consent. Clinical
and pathology data from patients with biopsy-proven clinically local-
ized prostate cancer used in this study is provided in Table W2.
No treatment was administered to the subjects in the interval be-
tween biopsy and urine sample collection. As controls, urine samples
from 50 male subjects (mean ± SD, 57.9 ± 0.1 years) with no known
history of prostate cancer were collected from the University of
Michigan Clinical Pathology Laboratories. Within 2 days of collec-
tion, the urine samples were concentrated at 4°C using a Biomax
Ultrafree concentrator (5000 MW cutoff; Millipore Corporation,
Bedford, MA). The protein content of each sample was estimated
with Bradford reagent following the manufacturer’s instructions
(Bio-Rad Laboratories, Hercules, CA). The concentrated urine sam-
ples were frozen at −20°C until use. Twenty micrograms of total pro-
tein from each urine sample was electrophoresed and analyzed for
GOLM1 protein expression by immunoblot analysis as described
above. The intensity of the 74-kDa GOLM1 band in each sample
was scored visually by a researcher who was blinded to the diagnosis
of samples and the pattern of loading in the gels. A GOLM1-positive
control sample (LNCaP total cell extract) was used for immunoblot
experiments. While scoring, the band in the sample lanes with high-
est reactivity was assigned a score of 4 (highest level of reactivity),
whereas absence of a band was scored 0 (non reactivity). Intermediate
band intensities were assigned as follows: weak (score = 1), interme-
diate (score = 2), and high (score = 3). The mean values for GOLM1
reactivity were presented as population mean values with 95% con-
fidence intervals. Student’s t test (2-sided) was used to test for statis-
tically significant differences in GOLM1 reactivity between patients
with prostate cancer and control subjects. No adjustment for multi-
ple testing was made during the analysis. P values less than or equal
to 0.05 were considered statistically significant. Receiver-operating
characteristic curves were used to assess the sensitivity and specificity
of urine-associated GOLM1 to detect prostate cancer.
Totestthesecretion oftheGOLM1protein, aflag-tagged GOLM1
construct was overexpressed in HEK293 Phoenix cells. Cell culture
supernatants were collected 12 and 24 hours after transfection. For
the analysis, 5 μl of culture supernatant from GOLM1-transfected,
vector-transfected, and control protein-transfected cells were run on
SDS-PAGE. Immunoblots were probed with either flag or GOLM1
antibodies. Likewise, the presence of GOLM1 was also assessed in
the culture supernatants of LNCaP (prostate cell line) and DU145
(prostate cancer cell line). Prostate cancer cell lines were cultured
in RPMI medium containing 10% FBS, and the medium was
changed to 5% FBS 12 hours before the collection of the superna-
tant to reduce the bovine serum protein content. Culture superna-
tants were collected at 24 hours and analyzed by SDS-PAGE and
immunoblot analysis with GOLM1-specific antibody. To investi-
gate the specificity of the secretion, the GOLM1 antibody was pre-
treated with 5 μg of recombinant GOLM1 protein . Cell-free
medium served as the negative control.
In Vitro Overexpression of GOLM1
A mammalian expression construct of GOLM1 was generated by
subcloning the PCR product into the pACRSVpLpA(−)loxP-SSP
vector (UMICH vector core). Primers used were GOLM1-KpnI-6-
CTTGGGAAACGGGCGTCGCAGC-5′ and GOLM1-XbaI-
Phoenix cells were transfected with the flag-GOLM1 construct as
Neoplasia Vol. 10, No. 11, 2008 GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al.
well as vector control using FuGENE6 (Roche) transfection reagent.
Culture supernatants were collected 12 and 24 hours after transfec-
tion for immunoblot analysis.
For the development of the tissue microarray (TMA), prostate
cancer and normal prostate tissues were embedded in paraffin. Study
pathologists (M.A.R. and R.M.) reviewed the slides of all cases and
designated areas of interest. These slides were used as a template for
TMA construction. All TMAs were assembled using the manual tis-
sue arrayer (Beecher Instruments, Silver Spring, MD). At least three
tissue cores were sampled from each donor block. Histologic diagno-
sis of the tissue cores was verified by standard hematoxylin and eosin
staining of the initial TMA slide. This radical prostatectomy series is
part of the University of Michigan Prostate Cancer Specialized Pro-
gram of Research Excellence Tissue Core, with informed consent of
the patients and prior institutional review board approval. Standard
biotin-avidin complex immunohistochemistry was performed using
a polyclonal anti-GOLM1 antibody. Digital images were acquired
Figure 1. GOLM1 expression in prostate cancer. (A) GOLM1 transcript levels were collected from DNA microarray analysis of 101 pros-
tate samples. Cy3/Cy5 ratios indicate the expression of GOLM1 in prostate tissue RNA compared to RNA from normal prostate pool.
BPH indicates benign prostatic hyperlasia; MET, metastatic prostate cancer; NAP, normal adjacent prostate; PCA, prostate cancer. (B)
GOLM1 expression in prostate cancer profiling studies from different publicly available cancer microarray data sets in Oncomine
[3,18,26,40] and Yang et al. (unpublished data). Expression array analysis of multiple prostate cancer microarray data sets were collected
and analyzed, and statistical significance was calculated. (For details visit: www.oncomine.org). Class 1 represents the GOLM1 expres-
sion in normal tissues and class 2 represents GOLM1 expression in cancer. (C) Quantitative polymerase chain reaction confirmation of
increased expression of GOLM1 transcripts in prostate cancer. Quantitative polymerase chain reaction was done using RNA from benign,
prostate cancer, and metastatic prostate cancer tissue samples. Quantitative polymerase chain reaction of glyceraldehyde-3-phosphate
dehydrogenase served as the internal control.
GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al.Neoplasia Vol. 10, No. 11, 2008
using the BLISS Imaging System (Bacus Laboratory, Lombard, IL)
and evaluated using a previously validated Web-based tool (TMA
Profiler, University of Michigan, Ann Arbor, MI) . Staining was
scored as negative (score = 1), weak (score = 2), moderate (score =
3), and strong (score = 4) based on the intensity of staining of tumor
cells using a similar system that has been validated previously .
Immunofluorescence Staining and Confocal Microscopy
Prostate tissue sections were soaked in xylene for 1 hour to remove
paraffin. Antigen retrieval was achieved by heating the slides in citrate
buffer (pH 6.0) for 15 minutes in a pressure cooker. The slides were
then blocked in PBS-T containing 5% normal donkey serum for
1 hour. Slides were incubated overnight at 4°C with a mixture of
mouse anti–E-cadherin (BD Biosciences, San Diego, CA) and rabbit
GOLM1 antibodies at 1:200 and 1:10,000 dilutions, respectively,
washed, and followed by secondary antibodies (anti–mouse Alexa
555 and anti–rabbit Alexa 488 at 1:1000 dilutions) for 1 hour. Slides
were mounted using Vectashield mounting medium containing DAPI
(Vector Laboratories, Burlingame, CA) after washing with PBS-Tand
PBS. Confocal images were taken with a Zeiss LSM510 META (Carl
Zeiss, Göttingen, Germany) imaging system using an ultraviolet, ar-
gon, and helium neon 1 light source. The triple color images were
exported as TIFF images.
Gene expression profiling studies from our group using prostate
tissue RNA demonstrated that GOLM1, a type II Golgi membrane
protein–coding gene, was up-regulated and consistently overexpressed
in prostate cancer (Figure 1A). Oncomine  (www.oncomine.org)
analysis of various prostate cancer microarray data sets revealed up-
regulation of GOLM1 in prostate cancer samples compared to normal
(Figure 1B). Quantitative real-time PCR revealed consistent up-
regulation of GOLM1 transcripts in prostate cancer samples confirm-
ing the gene expression profiling study findings (Figure 1C). The
up-regulation of GOLM1 expression was corroborated at the protein
level by immunoblot analysis with lysates from prostate tissues
(Figure 2A). There is a marked up-regulation of GOLM1 protein
in prostate cancer tissues compared to benign adjacent specimens,
whereas metastatic tissue displayed a moderate up-regulation. Further
verification and data on cell type–specific expression was obtained by
performing immunoblot analysis on laser capture microdissected
samples. Epithelial cells from BPH, prostatic intraepithelial neoplasia
(PIN), localized prostate cancer (PCA), and hormone refractory met-
astatic prostate cancer (MET) obtained by LCM were analyzed by
Western blot analysis with anti-GOLM1 and subsequently with
anti–β-actin (loading control) antibodies (Figure 2B). Laser capture
microdissected samples showed up-regulation of GOLM1 protein in
PCA and MET samples, demonstrating that the up-regulation ob-
served in whole tissue lysates (Figure 2A) was caused by an increase
in GOLM1 expression within the cancer epithelia. The apparent de-
crease in GOLM1 expression in MET compared to PCA was intrigu-
ing and was previously noted by Chandran et al. .
Immunofluorescence staining and confocal imaging of prostate tis-
sue section stained with GOLM1-specific antibody not surprisingly
revealed a Golgi staining pattern. There was a clear increase in
GOLM1 staining in prostate cancer epithelia compared to normal
epithelial cells (Figure 3A, pink arrows). High-density TMA analyses
indicated moderate-to-strong GOLM1 protein expression in clini-
cally localized prostate cancer samples with predominant cytoplasmic
localization (Figure 3B). Expression levels of GOLM1 protein in
malignant epithelia were greater compared to those of benign tissue.
Scoring of the TMAs indicated that the maximum staining (highest
score of 4) was observed in prostate cancer, whereas PIN demonstrated
moderate increases when compared to normal prostate epithelium
We next sought to detect GOLM1 protein in prostate cell culture
supernatant to probe whether it is secreted by prostate cancer cells. A
previous study that detected the presence of GOLM1 in serum of
hepatocellular cancer patients  supported this notion. Hence,
Figure 2. Immunoblot analyses of the GOLM1 protein in prostate tissue lysates. (A) Immunoblot analysis of GOLM1 protein expression
using benign and prostate cancer tissue lysates. Total prostate tissue lysates (10-μg total protein) were analyzed using anti–GOLM1
rabbit polyclonal antibody. Ten micrograms of total tissue lysate from benign prostate tissue, prostate cancer, and metastatic prostate
cancer tissues were analyzed by immunoblot. The blot was reprobed with GAPDH antibody to confirm equal loading. (B) The prostate
epithelial cells were isolated by LCM technique and were analyzed by immunoblot analysis. The cells from benign adjacent tissue, PIN,
prostate cancer, and metastatic cancer were used for this analysis. The blots were reprobed with anti–β-actin antibody.
Neoplasia Vol. 10, No. 11, 2008 GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al.
Figure 3. Immunofluorescence and immunohistochemical analyses of GOLM1 expression. (A) GOLM1 staining and confocal imaging of
prostate cancer epithelial cells. Representative GOLM1 staining is shown in green. E-cadherin (red) was used for costaining, and nuclei
were stained with DAPI (blue). Pink arrows indicate the overexpressed GOLM1 protein stained in green. The white arrow indicates a
lower level of GOLM1 expression in normal epithelium. (B) The TMA analyses of prostate tissues using anti–GOLM1 antibody. Intense
staining of GOLM1 is seen in cancerous prostate epithelial cells. (C) Correlations of GOLM1 staining intensity in benign, PIN, and pros-
GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al.Neoplasia Vol. 10, No. 11, 2008
the presence of a specific band in immunoblots of culture superna-
tant from HEK293 Phoenix cells that overexpressed Flag-tagged
GOLM1, with either FLAG antibody or GOLM1-specific antibody
(Figure 4A), suggested the presence of a secretory form of GOLM1.
This signal was abrogated when GOLM1 antibody was preincubated
with purified recombinant protein, showing the specificity of this sig-
nal (Figure 4B). In addition, immunoblots also revealed secretion of
endogenous GOLM1 by DU145 and LNCaP prostate cell lines,
which was inhibited by treatment with 0.5 μM brefeldin A (Fig-
ure 4C), a fungal lactone and inhibitor of protein trafficking. This
result suggests secretion of Golm1 through the anterograde secretory
pathway [20,21]. Brefeldin also blocked the secretion of prostate-
specific antigen (PSA) by LNCaP cells (data not shown).
To determine the subcellular localization of GOLM1 in prostate
cells, we performed fluorescence immunostaining with GOLM1
antibody using LNCaP cell line. The immunostaining indicated spe-
cific Golgi staining as anticipated (Figure 4D, panels i and ii), and
coimmunostaining with Golgi marker Golgin 97 confirmed this ob-
servation (Figure 4D, panel iii). However, cells treated with brefeldin
A showed diffuse staining (Figure 4D, panels iv and v) and colocali-
zation with the endoplasmic reticulum marker protein disulfide
isomerase (Figure 4D, panel vi). This observation suggested interfer-
ence with anterograde transport and increased retrograde transport
[20,21] and redistribution of GOLM1 from the Golgi apparatus
to the endoplasmic reticulum, which explains the reduced GOLM1
in the culture supernatant after brefeldin A treatment.
GOLM1, we further investigated the possibility of detecting GOLM1
in patient serum and urine samples. Immunoblot analysis of urine
samples from prostate cancer patients showed a specific band
corresponding to GOLM1 comparable in molecular weight to
LNCaP lysate. A representative immunoblot is depicted in Fig-
ure 5A. Of note, inhibition of GOLM1 reactivity in the urine
was observed after preincubation of GOLM1 antibody with the
Figure 4. Detection of GOLM1 protein in cell culture medium. (A) Culture supernatants were collected from HEK293 Phoenix cells over-
expressing Flag-GOLM1 construct and immunoblotted with either anti–Flag mouse monoclonal antibody or GOLM1 rabbit polyclonal
antibody to detect the presence of GOLM1. Total protein was stained with Ponceau S showing equal loading of culture supernatant.
Vector-transfected cells were used as controls. Un-T is culture supernatant from untransfected cells. (B) Immunoblot using anti–GOLM1
antibody that was preincubated with recombinant GOLM1 protein. For immunoblot, culture supernatants from LNCaP and DU145 cell
lines were used. (C) Culture supernatants from Du145 and LNCaP prostate cancer cell lines were collected and analyzed by immunoblot.
Addition of secretory blocker brefeldin A reduced the GOLM1 protein in the culture supernatant. Representative immunoblots from
multiple experiments are shown here. (D) Immunostaining of GOLM1. Panels i and ii show untreated LNCaP cells stained with GOLM1
(green) and E-cadherin (red) antibodies. Nuclei were stained with DAPI (blue). Panel iii shows colocalization of GOLM1 with a Golgi
marker protein, Golgin 97 (red). When treated with brefeldin A, GOLM1 shows diffused staining as shown in panels iv and v. Panel
vi shows staining of protein disulfide isomerase (red), an endoplasmic reticulum enzyme distribution in brefeldin A–treated cells.
Neoplasia Vol. 10, No. 11, 2008 GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al.
recombinant GOLM1 protein, similar to the results obtained for
culture supernatants (Figure 4B; data not shown). The mean score
for urine-associated GOLM1 reactivity in prostate cancer patients
(mean = 2.77) was significantly greater (P < .0001) than in the
control subjects (mean = 0.96). Significantly greater percentage of
prostate cancer urine samples (75%) had GOLM1 reactivity score
in the range of 2 to 4 in contrast to controls (28%; Figure 5B). An
ROC curve was generated for GOLM1 reactivity, and an optimum
cutoff point was selected at the region where the slope of the curve
had the highest value (Figure 5C). At this cutoff point, GOLM1
had the best discriminatory power in distinguishing between urine
from prostate cancer patients and control populations (AUC =
GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al. Neoplasia Vol. 10, No. 11, 2008
0.785, 95% confidence interval = 0.693-0.876, P < .0001) repre-
senting a sensitivity and specificity of 75% and 72%, respectively
(Figure 5C). Overall, 39 of 52 urine samples from patients with
clinically localized prostate cancer and 14 of 50 control samples
were considered positive for GOLM1 reactivity. Immunoblot anal-
ysis of prostate cancer patient sera did not indicate the presence of
GOLM1 protein (data not shown). Furthermore, GOLMI mRNA
transcripts showed significant association for discriminating patients
with prostate cancer from the negative needle biopsy cohort, as re-
ported in our earlier study  (Figure 5D). We assayed a total of
333 samples, a much larger sample size than previously reported
 and tested for the ability to detect prostate cancer based on
the ROC curves, GOLM1 (AUC = 0.622, P = .0009) outperformed
serum PSA (AUC = 0.495, P = .902; Figure 5E), suggesting the use
of urine-based GOLM1 mRNA measurements for the noninvasive
detection of prostate cancer. The sensitivity, specificity, and positive
and negative predictive values for the detection of GOLM1 in urine
were 0.594, 0.709, 0.732, and 0.490, respectively.
Gene expression analysis facilitated the identification of gene sig-
natures that are dysregulated in a given cancer. Multiple microarray
studies identified GOLM1 (GP73 or GOLPH2), a resident Golgi
protein, to be transcriptionaly up-regulated in prostate cancer. Here,
we demonstrate an up-regulation of GOLM1 in prostate cancer
epithelial cells. Further, a secretory form of this protein was iden-
tified in culture supernatants of prostate cell lines. On the basis
of the detection of the N-terminal FLAG tag in luminal GOLM1,
it seems that the release of GOLM1 into the lumen does not in-
volve N-terminal cleavage at the proprotein convertase cleavage site
(R52VRR55) . This suggests that the release mechanism used
by prostate cancer cells may differ from hepatocytes. The secretory
mechanism of this apparent “full-length GOLM1” is intriguing and
remains to be determined. The urinary GOLM1 protein may also
be derived from sloughed prostate cancer cells, although such cells
are typically only encountered in advanced stages of the disease and
afterprostate massage .This, however,doesnotexplain therelease
lines could be inhibited by brefeldin A, a protein transport inhibitor
[20,21], which suggests that this process could involve secretion. Al-
ternatively, GOLM1 could be released within exosomes  or per-
haps microvesicles . Regardless of the mechanism, the value of
and digital rectal examination is promising and needs to be evaluated.
possibility. Recently, our group has shown that increased GOLM1
transcript in urine, along with SPINK1 and PCA3 transcript expres-
sion, and TMPRSS2:ERG fusion status were able to predict prostate
cancer outcome. Multivariate regression analysis showed that a multi-
plex of these biomarkers outperformed serum PSA or PCA3 alone in
detecting prostate cancer .
Although its function remains uncharacterized, the differential ex-
pression of GOLM1 seems to be a common feature of malignancy
and has great potential to be developed as a cancer biomarker. For
example, Lu et al.  identified GOLM1 as one of 187 genes that
were consistently dysregulated in 20 common types of cancers, in-
cluding bladder, breast, colon, endometrium, kidney, liver, lung, mel-
anoma, lymphoma, pancreatic, prostate, and thyroid cancers. In
cultured cells, adenovirus infection induces GOLM1 expression
through the CtBP domain of the adenoviral E1A protein . Inter-
estingly, recent cell culture studies have demonstrated estrogen-
induced changes in GOLM1 expression in the mouse uterus 
and in androgen-resistant prostate cancer cell lines , suggesting
a possible hormone regulation in prostate neoplasia. A recent study
indicated that there is cancer cell–specific alterations of GOLM1 and
MYO6 in prostate cancer .
In the absence of any obvious protein domains that would indicate
the normal biologic function of GOLM1, gene knock-out studies
may be necessary to determine its role in cellular processes. Prelim-
inary in vivo studies in mouse model with a C-terminal truncation
suggest a decreased survival and severe renal abnormalities in ho-
mozygotes, but no obvious defects in prostate morphology .
Cell-specific knockout studies of GOLM1 are currently in progress
(Fimmel, in preparation) and may shed light on the function of the
protein and its role in prostate and other cancers.
The membrane localization/secretory form of GOLM1 makes it
an ideal target for immunotherapy approaches. Further, a better un-
derstanding of its function and processing may provide new insights
into the transport, sorting, and processing of resident Golgi and se-
cretory proteins. Of particular interest will be the possible role, if any,
played by dysregulated GOLM1 expression in PSA secretion through
the ER-Golgi-membrane transport pathway.
The authors thank Venkatesh Basrur and for helpful discussions.
The authors thank the staff of the Microscopy and Image Analyses
laboratory at the University of Michigan for their assistance in the
microscopic analyses used in this study.
Figure 5. Evaluation of GOLM1 protein and transcripts in urine. (A) Detection of GOLM1 in urine of prostate cancer patients. Urine from
healthy normal controls and confirmed prostate cancer patients were analyzed by SDS-PAGE and immunoblot analysis using anti–
GOLM1 antibody to test for the presence of GOLM1 protein in the urine. Twenty representative samples are shown out of a total of
102 that were analyzed. LNCaP, prostate cancer cell line extract, was used as a positive control for GOLM1 immunoreactivity. (B) Plot
based on the mean score for urine-associated GOLM1 reactivity. (C) Receiver-operating characteristic curve for the detection of GOLM1
in urine as assessed by immunoblot analysis (n = 52 sera from clinically localized prostate cancer patients and n = 50 control subjects).
Area under the curve for GOLM1 is 0.785. (D) Quantitative PCR for the detection of GOLM1 transcripts in urine. Quantitative PCR was
performed on WTA cDNA from urine obtained from patients presenting for needle biopsy or prostatectomy. GOLM1 expression in pa-
tients with negative needle biopsies (green) or patients with prostate cancer (positive needle biopsy or prostatectomy; red) are shown.
Normalization was performed using −ΔCt, with PCA3 normalized to the average of urine sediment PSA and GAPDH expression. (E)
Receiver-operating characteristic curves for GOLM1 for the diagnosis of prostate cancer. Areas under the curve for GOLM1 and serum
PSA are 0.622, and 0.495, respectively.
Neoplasia Vol. 10, No. 11, 2008GOLM1 Is a Tissue Biomarker of Prostate Cancer Varambally et al.
 Abate-Shen C and Shen MM (2000). Molecular genetics of prostate cancer.
Genes Dev 14, 2410–2434.
 Gronberg H (2003). Prostate cancer epidemiology. Lancet 361, 859–864.
 Dhanasekaran SM, Barrette TR, Ghosh D, Shah R, Varambally S, Kurachi K,
Pienta KJ, Rubin MA, and Chinnaiyan AM (2001). Delineation of prognostic
biomarkers in prostate cancer. Nature 412, 822–826.
 Emmert-Buck MR, Gillespie JW, Paweletz CP, Ornstein DK, Basrur V, Appella
E, Wang QH, Huang J, Hu N, Taylor P, et al. (2000). An approach to proteo-
mic analysis of human tumors. Mol Carcinog 27, 158–165.
 Rubin MA, Zhou M, Dhanasekaran SM, Varambally S, Barrette TR, Sanda MG,
Pienta KJ, Ghosh D, and Chinnaiyan AM (2002). alpha-Methylacyl coenzyme A
racemase as a tissue biomarker for prostate cancer. JAMA 287, 1662–1670.
 Rubin MA, Varambally S, Beroukhim R, Tomlins SA, Rhodes DR, Paris PL,
Hofer MD, Storz-Schweizer M, Kuefer R, Fletcher JA, et al. (2004). Overex-
pression, amplification, and androgen regulation of TPD52 in prostate cancer.
Cancer Res 64, 3814–3822.
 Tomlins SA, Rhodes DR, Perner S, Dhanasekaran SM, Mehra R, Sun XW,
Varambally S, Cao X, Tchinda J, Kuefer R, et al. (2005). Recurrent fusion of
TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 310,
 Mertz KD, Setlur SR, Dhanasekaran SM, Demichelis F, Perner S, Tomlins S,
Tchinda J, Laxman B, Vessella RL, Beroukhim R, et al. (2007). Molecular char-
acterization of TMPRSS2-ERG gene fusion in the NCI-H660 prostate cancer
cell line: a new perspective for an old model. Neoplasia 9, 200–206.
 Varambally S, Dhanasekaran SM, Zhou M, Barrette TR, Kumar-Sinha C, Sanda
 Tomlins SA, Laxman B, Varambally S, Cao X, Yu J, Helgeson BE, Cao Q,
Prensner JR, Rubin MA, Shah RB, et al. (2008). Role of the TMPRSS2-ERG
gene fusion in prostate cancer. Neoplasia 10, 177–188.
 Menschikowski M, Hagelgans A, Gussakovsky E, Kostka H, Paley EL, and
Siegert G (2008). Differential expression of secretory phospholipases A2 in nor-
mal and malignant prostate cell lines: regulation by cytokines, cell signaling
pathways, and epigenetic mechanisms. Neoplasia 10, 279–286.
 Dong Z, Saliganan AD, Meng H, Nabha SM, Sabbota AL, Sheng S, Bonfil RD,
and Cher ML (2008). Prostate cancer cell–derived urokinase-type plasminogen
activator contributes to intraosseous tumor growth and bone turnover. Neoplasia
 Rogers CG, Yan G, Zha S, Gonzalgo ML, Isaacs WB, Luo J, De Marzo AM,
Nelson WG, and Pavlovich CP (2004). Prostate cancer detection on urinalysis
for alpha methylacyl coenzyme a racemase protein. J Urol 172, 1501–1503.
 Sreekumar A, Laxman B, Rhodes DR, Bhagavathula S, Harwood J, Giacherio
D, Ghosh D, Sanda MG, Rubin MA, and Chinnaiyan AM (2004). Humoral
immune response to alpha-methylacyl-CoA racemase and prostate cancer. J Natl
Cancer Inst 96, 834–843.
 Kladney RD, Bulla GA, Guo L, Mason AL, Tollefson AE, Simon DJ, Koutoubi
Z, and Fimmel CJ (2000). GP73, a novel Golgi-localized protein upregulated
by viral infection. Gene 249, 53–65.
 Bachert C, Fimmel C, and Linstedt AD (2007). Endosomal trafficking and pro-
protein convertase cleavage of cis Golgi protein GP73 produces marker for he-
patocellular carcinoma. Traffic (Copenhagen, Denmark) 8, 1415–1423.
 Marrero JA, Romano PR, Nikolaeva O, Steel L, Mehta A, Fimmel CJ, Comunale
 Lapointe J, Li C, Higgins JP, van de Rijn M, Bair E, Montgomery K, Ferrari M,
Egevad L, Rayford W, Bergerheim U, et al. (2004). Gene expression profiling
identifies clinically relevant subtypes of prostate cancer. Proc Natl Acad Sci USA
 Laxman B, Morris DS, Yu J,SiddiquiJ, Cao J, Mehra R, Lonigro RJ, TsodikovA,
WeiJT,TomlinsSA, etal.(2008).Afirst-generationmultiplexbiomarker analysis
of urine for the early detection of prostate cancer. Cancer Res 68, 645–649.
 Strous GJ, van Kerkhof P, van Meer G, Rijnboutt S, and Stoorvogel W (1993).
Differential effects of brefeldin A on transport of secretory and lysosomal pro-
teins. J Biol Chem 268, 2341–2347.
 Ivessa NE, Gravotta D, De Lemos-Chiarandini C, and Kreibich G (1997).
Functional protein prenylation is required for the brefeldin A–dependent retro-
grade transport from the Golgi apparatus to the endoplasmic reticulum. J Biol
Chem 272, 20828–20834.
 Kleer CG, Cao Q, Varambally S, Shen R, Ota I, Tomlins SA, Ghosh D, Sewalt
RG, Otte AP, Hayes DF, et al. (2003). EZH2 is a marker of aggressive breast
cancer and promotes neoplastic transformation of breast epithelial cells. Proc
Natl Acad Sci USA 100, 11606–11611.
 Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, and
Speleman F (2002). Accurate normalization of real-time quantitative RT-PCR
data by geometric averaging of multiple internal control genes. Genome Biol 3,
 Laxman B, Tomlins SA, Mehra R, Morris DS, Wang L, Helgeson BE, Shah RB,
Rubin MA, Wei JT, and Chinnaiyan AM (2006). Noninvasive detection of
TMPRSS2: ERG fusion transcripts in the urine of men with prostate cancer.
Neoplasia 8, 885–888.
 Specht K, Richter T, Muller U, Walch A, Werner M, and Hofler H (2001).
Quantitative gene expression analysis in microdissected archival formalin-fixed
and paraffin-embedded tumor tissue. Am J Pathol 158, 419–429.
 TomlinsSA, MehraR,Rhodes DR,CaoX, WangL, Dhanasekaran SM,Kalyana-
Sundaram S, Wei JT, Rubin MA, Pienta KJ, et al. (2007). Integrative molecular
concept modeling of prostate cancer progression. Nat Genet 39, 41–51.
 Manley S, Mucci NR, De Marzo AM, and Rubin MA (2001). Relational data-
base structure to manage high-density tissue microarray data and images for pa-
thology studies focusing on clinical outcome: the prostate specialized program of
research excellence model. Am J Pathol 159, 837–843.
 Mehra R, Varambally S, Ding L, Shen R, Sabel MS, Ghosh D, Chinnaiyan AM,
and Kleer CG (2005). Identification of GATA3 as a breast cancer prognostic
marker by global gene expression meta-analysis. Cancer Res 65, 11259–11264.
 Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, Barrette
T, Pandey A, and Chinnaiyan AM (2004). ONCOMINE: a cancer microarray
database and integrated data-mining platform. Neoplasia 6, 1–6.
 Chandran UR, Ma C, Dhir R, Bisceglia M, Lyons-Weiler M, Liang W,
Michalopoulos G, Becich M, and Monzon FA (2007). Gene expression profiles
of prostate cancer reveal involvement of multiple molecular pathways in the
metastatic process. BMC Cancer 7, 64.
 Hessels D, Klein Gunnewiek JM, van Oort I, Karthaus HF, van Leenders GJ,
van Balken B, Kiemeney LA, Witjes JA, and Schalken JA (2003). DD3(PCA3)-
based molecular urine analysis for the diagnosis of prostate cancer. Eur Urol 44,
8–15; discussion 15–16.
 Johnstone RM (2006). Exosomes biological significance: a concise review. Blood
Cells Mol Dis 36, 315–321.
 MacKenzie A, Wilson HL, Kiss-Toth E, Dower SK, North RA, and Surprenant
A (2001). Rapid secretion of interleukin-1beta by microvesicle shedding. Immu-
nity 15, 825–835.
 Lu Y, Yi Y, Liu P, Wen W, James M, Wang D, and You M (2007). Common
human cancer genes discovered by integrated gene-expression analysis. PLoS
ONE 2, e1149.
 Kladney RD, Tollefson AE, Wold WS, and Fimmel CJ (2002). Upregulation of
the Golgi protein GP73 by adenovirus infection requires the E1A CtBP inter-
action domain. Virology 301, 236–246.
 Moggs JG, Tinwell H, Spurway T, Chang HS, Pate I, Lim FL, Moore DJ,
Soames A, Stuckey R, Currie R, et al. (2004). Phenotypic anchoring of gene
expression changes during estrogen-induced uterine growth. Environ Health Per-
spect 112, 1589–1606.
 Coleman IM, Kiefer JA, Brown LG, Pitts TE, Nelson PS, Brubaker KD,
Vessella RL, and Corey E (2006). Inhibition of androgen-independent prostate
cancer by estrogenic compounds is associated with increased expression of
immune-related genes. Neoplasia 8, 862–878.
 Wei S, Dunn TA, Isaacs WB, De Marzo AM, and Luo J (2008). GOLPH2 and
MYO6: putative prostate cancer markers localized to the Golgi apparatus. Pros-
tate 68, 1387–1395.
 Wright LM, Yong S, Picken MM, Rockey D, and Fimmel CJ (2009). Decreased
survival and hepato-renal pathology in mice with C-terminally truncated GP73
(GOLPH2). Int J Clin Exp Pathol 2, 34–47.
 Varambally S, Yu J, Laxman B, Rhodes DR, Mehra R, Tomlins SA, Shah RB,
Chandran U, Monzon FA, Becich MJ, et al. (2005). Integrative genomic and
proteomic analysis of prostate cancer reveals signatures of metastatic progression.
Cancer Cell 8, 393–406.
GOLM1 Is a Tissue Biomarker of Prostate CancerVarambally et al.Neoplasia Vol. 10, No. 11, 2008
Table W1. Clinical and Pathology Characteristics for the Patients Presenting for Prostate Biopsy or Download full-text
Radical Prostatectomy for the Urines Used for Detection of GOLM1 Transcripts.
Age, mean ± SE (n = 332)
Gland weight, mean ± SE (n = 70)
Tumor size, mean ± SE (n = 69)
PSA, mean ± SE (n = 271)
0-2.5 ng/ml (%)
2.6-10 ng/ml (%)
4-10 ng/ml (%)
>10 ng/ml (%)
Gleason Sum (n = 169)
Stage (n = 151)
62.7 ± 8.7
33.6 ± 19.3
1.2 ± 0.7
8.6 ± 10.2
Table W2. Clinical and Pathologic Characteristics for Prostate Cancer Patients Used for Urine
Analysis of GOLM1 Protein.
Age,* mean ± SEM (years)
Gland weight,* mean ± SEM (g)
Max dimension of largest tumor,* mean ± SEM (cm)
PSA,†mean ± SEM (ng/ml)
0-2.5 ng/ml (%)
2.6-10 ng/ml (%)
4-10 ng/ml (%)
Pathologic tumor staging§
58.1 ± 0.7
49.1 ± 2.4
1.4 ± 0.1
8.51 ± 1.7
*Values available for 50 samples only.
†Values available for 45 samples only.
‡Values available for 48 samples only.
§Values available for 50 samples only.