Rubin MA, Zhou M, Dhanasekaran SM, Varambally S, Barrette TR, Sanda MG, Pienta KJ, Ghosh D, Chinnaiyan AM. α-Methylacyl Coenzyme A Racemase as a Tissue Biomarker for Prostate Cancer. JAMA. 2002;287(13):1662-1670. doi:10.1001/jama.287.13.1662
Author Affiliations: Departments of Pathology (Drs Rubin, Zhou, Dhanasekaran, Varambally, and Chinnaiyan and Mr Barrette), Urology (Drs Rubin, Sanda, Pienta, and Chinnaiyan), Internal Medicine (Dr Pienta), and Biostatistics (Dr Ghosh) and Comprehensive Cancer Center (Drs Rubin, Sanda, Pienta, and Chinnaiyan), University of Michigan Medical School, Ann Arbor.
Context Molecular profiling of prostate cancer has led to the identification
of candidate biomarkers and regulatory genes. Discoveries from these genome-scale
approaches may have applicability in the analysis of diagnostic prostate specimens.
Objectives To determine the expression and clinical utility of α-methylacyl
coenzyme A racemase (AMACR), a gene identified as
being overexpressed in prostate cancer by global profiling strategies.
Design Four gene expression data sets from independent DNA microarray analyses
were examined to identify genes expressed in prostate cancer (n = 128 specimens).
A lead candidate gene, AMACR, was validated at the
transcript level by reverse transcriptase polymerase chain reaction (RT-PCR)
and at the protein level by immunoblot and immunohistochemical analysis. AMACR levels were examined using prostate cancer tissue
microarrays in 342 samples representing different stages of prostate cancer
progression. Protein expression was characterized as negative (score = 1),
weak (2), moderate (3), or strong (4). Clinical utility of AMACR was evaluated using 94 prostate needle biopsy specimens.
Main Outcome Measures Messenger RNA transcript and protein levels of AMACR; sensitivity and specificity of AMACR as
a tissue biomarker for prostate cancer in needle biopsy specimens.
Results Three of 4 independent DNA microarray analyses (n = 128 specimens) revealed
significant overexpression of AMACR in prostate cancer
(P<.001). AMACR up-regulation
in prostate cancer was confirmed by both RT-PCR and immunoblot analysis. Immunohistochemical
analysis demonstrated an increased expression of AMACR
in malignant prostate epithelia relative to benign epithelia. Tissue microarrays
to assess AMACR expression in specimens consisting
of benign prostate (n = 108 samples), atrophic prostate (n = 26), prostatic
intraepithelial neoplasia (n = 75), localized prostate cancer (n = 116), and
metastatic prostate cancer (n = 17) demonstrated mean AMACR protein staining
intensity of 1.31 (95% confidence interval, 1.23-1.40), 2.33 (95% CI, 2.13-2.52),
2.67 (95% CI, 2.52-2.81), 3.20 (95% CI, 3.10-3.28), and 2.50 (95% CI, 2.20-2.80),
respectively (P<.001). Pairwise comparisons demonstrated
significant differences in staining intensity between clinically localized
prostate cancer compared with benign prostate tissue, with mean expression
scores of 3.2 and 1.3, respectively (mean difference, 1.9; 95% CI, 1.7-2.1; P<.001). Using moderate or strong staining intensity
as positive (score = 3 or 4), evaluation of AMACR protein expression in 94
prostate needle biopsy specimens demonstrated 97% sensitivity and 100% specificity
for detecting prostate cancer.
Conclusions AMACR was shown to be overexpressed in prostate
cancer using independent experimental methods and prostate cancer specimens. AMACR may be useful in the interpretation of prostate needle
biopsy specimens that are diagnostically challenging.
Prostate cancer affects 1 of 9 men older than 65 years and is a leading
cause of cancer-related death in men, second only to lung cancer.1,2 While the advent of prostate-specific
antigen (PSA) screening has led to earlier detection of prostate cancer,3 the impact of PSA screening on cancer-specific mortality
is unknown, pending the results of prospective randomized screening studies.4- 6 A major limitation of
the serum PSA test is a lack of prostate cancer sensitivity and specificity,
especially in the intermediate range of PSA detection (4-10 ng/mL). Coincident
with increased serum PSA testing, there has been a significant increase both
in the number of prostate needle biopsies performed7
and in the number of equivocal prostate needle biopsy specimens.8
Thus, development of additional serum and tissue biomarkers to supplement
PSA screening is needed.
Molecular profiling establishes a new perspective in the study of cancer.9 Recent studies have suggested that it is possible
to study cancer with a global perspective, taking advantage of DNA10- 13 and
protein microarrays14 as "windows" into the
expressed human genome. Several groups recently implemented DNA microarrays
to analyze prostate cancer specimens.15- 18
Different laboratories, using distinct microarray platforms and reference
controls, identified and validated hepsin as a serine protease up-regulated
in prostate cancer.15- 18
Using high-density tissue microarrays (TMAs), our group15
reported that both hepsin and another protein, pim-1 kinase, are down-regulated
in a sample of aggressive, localized prostate cancer specimens, making these
biomarkers potentially useful for predicting prognosis but less applicable
for diagnosing prostate cancer in tissue sections.
In this study, we examined α-methylacyl coenzyme A racemase (AMACR),
a peroxisomal and mitochondrial enzyme that was found to be up-regulated in
Although the precise physiologic roles of AMACR are not clear, this enzyme
has an important role in bile acid biosynthesis and β-oxidation of branched-chain
fatty acids19,20 and mediates
the interconversion of (R)- and (S)-2-methyl-branched-chain fatty acyl coenzyme
As.19 Mutations of the AMACR gene have been shown to cause adult-onset sensory motor neuropathy,21 but, to our knowledge, a link to prostate cancer
has not been made.
Our overall approach of discovery and validation of candidate biomarkers
and regulatory genes in prostate cancer is shown in Figure 1.22 By using DNA microarrays,
we can examine thousands of genes in the context of prostate cancer (Figure 1B-D). Combining this approach with
tissue microarrays, we can simultaneously examine hundreds of clinically stratified
patient specimens effectively (Figure 1F-H).
Thus, the development and coordination of these technologies allows the survey
of thousands of genes in many patients in a high-throughput fashion.
First, we examined 4 independent gene expression data sets, including
To investigate the statistical significance associated with the differential
expression of AMACR across 3 independent microarray studies, we used standard
methods from meta-analysis23 to combine the
results. For each of the studies, we computed a t
statistic (with the 2 groups being benign tissue compared against localized
prostate cancer tissue) and transformed the associated P values using a negative logarithmic transformation. These numbers
were then doubled and added to arrive at a summary measure of differential
gene expression across the 3 studies. To assess the statistical significance
associated with this summary measure, a permutation-based approach was used.23 Tissue types were permutated within studies, the
summary measure was computed for the permutated data, and a P value was computed using the permutation distribution of the summary
measure. One million resamplings from the permutation distribution were implemented.
Two of the 3 studies used the spotted complementary DNA (cDNA) microarray
technology, whereas Affymetrix oligonucleotide microarrays (GeneChip) were
used for the third study. To address whether t statistics
from the 3 studies are comparable, we used the Wilcoxon-Mann-Whitney statistic
separately for the 3 studies.
To examine the widest range of prostate cancer specimens, clinical samples
were obtained from the radical prostatectomy series at the University of Michigan
and from the Rapid Autopsy Program.
Prostatectomy cases for the TMA outcomes array were selected from a
cohort of 632 patients, who underwent radical retropubic prostatectomy at
the University of Michigan as a primary therapy (ie, no preceding hormonal
or radiation therapy) for clinically localized prostate cancer from 1994 to
1998. Consecutive cases were taken from 1995 and 1996 to ensure sufficient
clinical follow-up. Clinical and pathologic data for all patients were acquired
with approval from the Institutional Review Board at the University of Michigan.
Detailed clinical, pathologic, and TMA data are maintained on a secure relational
The prostate specimens were transported to the frozen section room located
adjacent to the operating rooms and processed within approximately 15 to 20
minutes after surgical resection. The prostate specimens were partially sampled
with approximately 50% of the tissue used for research. This protocol has
been evaluated in a formal study to ensure that partial sampling does not
impair accurate staging and evaluation of the surgical margins.25
The Rapid Autopsy Protocol has been described previously.26
In brief, patients with advanced hormone-refractory prostate cancer are asked
to participate in this posthumous, institutional review board–approved
tumor donor program, and permission for autopsy is obtained before death,
with consent provided by the patient or next of kin. Twenty-three complete
autopsies have been performed as part of this program, with a median time
from death to autopsy of 3 hours. Hormone-refractory primary and metastatic
prostate cancer samples are collected. Half of each specimen is snap frozen
in liquid nitrogen and the other half is placed in 10% buffered formalin.
The fixed samples are embedded in paraffin and used for the development of
TMAs. As with the prostatectomy samples, the study pathologist (M.A.R.) reviewed
the glass slides, circled areas of viable prostate cancer, avoiding areas
of necrosis, and used these slides as a template for TMA construction.
Prostate gland specimens were inked before pathological assessment of
surgical margins. Surgical margins from the apex and base were cut perpendicular
to the prostatic urethral axis. The seminal vesicles were cut perpendicular
to their entry into the prostate gland and submitted as the seminal vesicle
margin. The prostate specimens for this study were all partially embedded,
taking alternate full sections from the apex, mid, and base. Detailed prostatectomy
pathology reports included the presence or absence of surgical margin involvement
by tumor (surgical margin status), the presence of extraprostatic extension,
and seminal vesicle invasion. Tumors were staged using the TNM system, which
includes extraprostatic extension and seminal vesicle invasion but does not
take into account surgical margin status.27
Tumors were graded using the Gleason grading system.28,29
As preparation for the construction of the TMAs, all glass slides were reexamined
to identify areas of benign prostate, prostatic atrophy, high-grade PIN, and
prostate cancer. To optimize the transfer of these designated tissues to the
arrays, the area of tumor involvement was encircled on the glass slide template
as tightly around each lesion as possible. Areas with infiltrating tumor adjacent
to benign glands were avoided.
Total RNA integrity was judged by denaturing-formaldehyde agarose gel
electrophoresis. Complementary DNA was prepared using 1 µg of total
RNA isolated from prostate tissue specimens. Primers used to amplify specific
gene products were as follows: AMACR sense, 5′-CGTATGCCCCGCTGAATCTCGTG-3′; AMACR antisense, 5′-TGGCCAATCATCCGTGCTCATCTG-3′;
glyceraldehyde-3-phosphate dehydrogenase (GAPDH) sense, 5′-CGGAGTCAACGGATTTGGTCGTAT-3′;
and GAPDH antisense, 5′-AGCCTTCTCCATGGTGGTGAAGAC-3′. Reverse transcriptase
polymerase chain reaction (RT-PCR) conditions for AMACR and GAPDH were 94°C for 5 minutes, 28 cycles of 95°C for 1
minute, 60°C for 1 minute (annealing), 72°C for 1 minute, and a final
elongation step of 72°C for 7 minutes. The RT-PCR reactions used a volume
of 50 µL, with 1 unit of Taq DNA polymerase (Invitrogen, Carlsbad, Calif).
Amplification products (5 µL) were separated by 2% agarose gel electrophoresis
and visualized by UV light.
Representative prostate tissue specimens that were previously profiled
using DNA microarrays15 were used for Western
blot analysis. Tissues were homogenized in NP-40 lysis buffer containing 50
mM Tris hydrochloride, pH 7.4, 1% Nonidet P-40 (Sigma, St Louis, Mo), and
complete proteinase inhibitor cocktail (Roche, Indianapolis, Ind). Fifteen
micrograms of protein extracts was mixed with sodium dodecyl sulfate sample
buffer and electrophoresed onto a 10% sodium dodecyl sulfate–polyacrylamide
gel under reducing conditions. The separated proteins were transferred onto
nitrocellulose membranes (Amersham Pharmacia Biotech, Piscataway, NJ). The
membrane was incubated for 1 hour in blocking buffer (Tris-buffered saline
with 0.1% Tween and 5% nonfat dry milk). The AMACR antibody (a gift of Ronald
J. A. Wanders) was applied at 1:10 000 diluted in blocking buffer overnight
at 4°C. After washing 3 times with Tris-buffered saline with 0.1% Tween
buffer, the membrane was incubated with horseradish peroxidase–linked
donkey anti–rabbit IgG antibody (Amersham Pharmacia Biotech) at 1:5000
for 1 hour at room temperature. The signals were visualized with the ECL detection
system (Amersham Pharmacia Biotech) and autoradiography.
Standard indirect biotin-avidin immunohistochemical analysis was performed
to evaluate AMACR protein expression using a polyclonal anti-AMACR antibody
(1:5000 dilution, a gift of Ronald J. A. Wanders). Protein expression was
scored as negative (score = 1), weak (2), moderate (3), and strong (4).
Five TMAs were used for this study. Three contained tissue from consecutive
prostatectomies from 1996 and 2 contained hormone-refractory prostate cancer
from the Rapid Autopsy Program. The TMAs were assembled using the manual tissue
arrayer (Beecher Instruments, Silver Spring, Md) as previously described.30,31 Tissue cores from the circled areas
(as described herein) were targeted for transfer to the recipient array blocks.
Five replicate tissue cores were sampled from each of the selected tissue
types. The 0.6-mm-diameter TMA cores were each spaced at 0.8 mm from core
center to core center. After construction, 4-µm sections were cut, and
hematoxylin-eosin staining was performed on the initial slide to verify the
AMACR protein expression was evaluated using the BLISS Imaging System
(Bacus Lab, Lombard, Ill) in a blinded manner. All images were scored for
AMACR protein expression intensity. In addition, all TMA samples were assigned
a diagnosis (eg, benign, atrophy, PIN, or prostate cancer), with verification
required at each step. The TMA slides were evaluated for proliferation index
using the CAS200 Cell Analysis System (Bacus Lab). Selected areas were evaluated
under the ×40 objective. Measurements were recorded as the percentage
of total nuclear area that was positively stained. All positive nuclear staining,
regardless of the intensity, was measured. Sites for analysis were selected
to minimize the presence of stromal and basal cells; only tumor epithelium
was evaluated. Specimens were evaluated for Ki-67 expression as previously
described.31 Each measurement was based on
approximately 50 to 100 epithelial nuclei. Because of the fixed size of TMA
samples, 5 repeat nonoverlapping measurements were the maximum attainable.
Differences in AMACR protein expression were evaluated statistically
using the mean score results from each case for each prostate tissue type
(eg, benign, atrophy, high-grade prostatic intraepithelial neoplasia, localized
prostate cancer, and hormone-refractory prostate cancer). To test for significant
differences in the mean AMACR protein expression among all tissue types, we
performed a 1-way analysis of variance (ANOVA) test. To determine differences
between all pairs (eg, localized prostate cancer vs benign), we performed
a post hoc analysis using the Scheffe method.32
The mean expression scores for all examined cases were presented in a graphical
format using error bars with 95% confidence intervals (CIs). For the clinically
localized prostate cancer samples, AMACR protein expression was evaluated
for associations with pathology parameters (eg, advanced tumor stage and surgical
margin involvement) and clinical outcome as measured by postsurgical PSA progression
(PSA >0.2 ng/mL) by Cox hazards regression analysis.
To evaluate the usefulness of AMACR expression in diagnostic, 18-gauge
needle biopsy specimens, we collected 100 consecutive biopsy specimens with
prostate cancer or atypia that required further workup from one calendar year
(2001). All cases were immunostained using 2 basal cell–specific markers
(34βE12 and p63) and AMACR. Twenty-six of these cases were seen in consultation
by one of the authors (M.A.R.) and are considered diagnostically difficult,
requiring expert review and additional characterization. Six cases did not
have sufficient tissue for further evaluation by AMACR.
Specificity, sensitivity, and positive and negative predictive values
were determined. The criterion standard for accurately diagnosing prostate
cancer was reviewed by an experienced genitourinary pathologist (in this case,
M.A.R.) using a combination of routine hematoxylin-eosin–stained slides
and a basal cell–specific marker (eg, p63 or 34βE12). All diagnoses
made on needle biopsy specimens have been confirmed by pathologic review of
By examining our gene expression data set,15
we identified an interesting candidate gene, AMACR,
that was consistently overexpressed in prostate cancer. Evaluation of AMACR transcript levels as determined by DNA microarray
analysis of 57 prostate cancer specimens showed that in relation to benign
prostate tissues, localized prostate cancer and metastatic prostate cancer
were 3.1-fold (Mann-Whitney test, P<.001) and
1.67-fold (Mann-Whitney test, P<.004) up-regulated,
respectively (represented as Cy5/Cy3 ratios) (Figure 2).
Summary of findings from 3 studies that used high-density DNA microarrays
to analyze prostate cancer and measured AMACR gene
expression showed a statistically significant differential expression of AMACR between benign prostate and prostate cancer (Table 1).
Using AMACR-specific primers, RT-PCR performed on the various RNA samples
from 28 prostate tissue specimens and 6 prostate cell lines (Figure 3A) showed that an RT-PCR product was observed in the 20
localized prostate cancer samples but not in the benign samples examined.
Metastatic prostate cancer and prostate cell lines displayed varying levels
of AMACR transcript compared with localized prostate
cancer. Immunoblot analysis on selected prostate tissue extracts showed overexpression
of AMACR protein in malignant prostate tissue relative to benign prostate
tissue (Figure 3B).
In separate prostate samples than those used in the cDNA expression
array analysis, high-density TMAs revealed moderate-to-strong AMACR protein
expression in clinically localized prostate cancer samples with predominately
cytoplasmic localization (Figure 4).
Levels of AMACR protein expression in malignant epithelia were greater than
in adjacent benign epithelia (Figure 3B).
Both PIN and some atrophic lesions, which are thought to be potentially precancerous
lesions,33,34 demonstrated cytoplasmic
staining of AMACR. High-grade prostate cancer (Gleason score, 4 + 4 = 8) and
low-grade prostate cancer (not shown) demonstrated strong cytoplasmic staining
(Figure 3C). However, no association
was identified with AMACR staining intensity and Gleason (tumor) score (data
not shown). Many hormone-refractory, metastatic prostate cancer samples (from
the Rapid Autopsy Protocol) demonstrated only weak AMACR expression (Figure 3D) but uniform PSA immunostaining
(data not shown), confirming the immunogenicity of these samples.
Quantitation of TMA data in benign prostate tissue (n = 108 samples),
atrophic prostate (n = 26), PIN (n = 75), localized prostate cancer (n = 116),
and metastatic prostate cancer (n = 17) demonstrated mean AMACR protein staining
intensity of 1.31 (SE, 0.041; 95% CI, 1.23-1.40), 2.33 (SE, 0.096; 95% CI,
2.13-2.52), 2.67 (SE, 0.071; 95% CI, 2.52-2.81), 3.20 (SE, 0.046; 95% CI,
3.10-3.28), and 2.50 (SE, 0.14; 95% CI, 2.20-2.80), respectively (1-way ANOVA, P<.001). Pairwise comparisons demonstrated differences
in staining intensity between clinically localized prostate cancer with respect
to benign prostate tissue, with mean expression scores of 3.2 and 1.3, respectively
(ANOVA post hoc analysis using Scheffé method; mean difference, 1.9;
95% CI, 1.7-2.1; P<.001). Significant differences
were also seen between benign tissue and PIN (1.4; 95% CI, 1.1-1.6; P<.001) and benign and prostatic atrophy (1.02; 95%
CI, 0.70-1.36; P<.001). No significant differences
in AMACR protein expression were identified between prostatic intraepithelial
neoplasia and prostatic atrophy (0.34; 95% CI, –0.017 to 0.698; P = .12). A significant decrease in AMACR protein expression
was observed in the metastatic hormone-refractory prostate cancer samples
compared with clinically localized prostate cancer (0.699; 95% CI, 0.292-1.107; P<.001)
There was no significant association between AMACR expression and Ki-67
expression (a marker of tumor proliferation)30
(r = 0.13; P = .22). There
were no significant associations between AMACR protein expression and pathology
parameters, such as radical prostatectomy Gleason score, tumor stage, tumor
size (in centimeters), or surgical margin status (data not shown). α-Methylacyl
coenzyme A racemaseprotein levels were not associated with PSA recurrence
following surgery in 120 prostatectomy cases within a median follow-up time
of 3 years (data not shown). α-Methylacyl coenzyme A racemase demonstrated
uniform moderate-to-strong expression in clinically localized prostate cancer
with high tumor sensitivity and specificity.
In evaluation of the clinical utility of AMACR immunostaining on 94
prostate needle biopsy specimens (Figure 6), we considered moderate or strong staining intensity to be positive.
We were not convinced that weak staining intensity could be reliably distinguished
from negative staining with faint background staining. Of the 94 specimens,
70 were diagnosed as prostate cancer by pathologic review. Of these, AMACR
protein was considered positive in 68, was falsely negative in 2, and was
negative in 24. The sensitivity and specificity were calculated as 97% and
100%, respectively. The positive predictive value was 100% and the negative
predictive value was 92%. These results included 26 cases for which the final
diagnosis required expert pathologic review and the use of a basal cell–specific
immunohistochemical marker (eg, 34βE12 or p63). These 26 cases were particularly
difficult diagnostically for several reasons. For example, the atypical focus
in question often consisted of only 4 or 5 glands. Furthermore, some cases
demonstrated bland nuclear features, making the diagnosis more challenging,
whereas some required distinguishing a cancerous lesion from PIN.
This approach to discovery and validation of biologically and clinically
relevant genes in prostate cancer involves DNA microarrays serving as a discovery
"engine" to identify novel candidate genes and TMAs to validate and characterize
genes in patient specimens.
In this study, we examined gene expression data sets from 3 independent
groups (including our own) and identified AMACR as
a gene consistently up-regulated in prostate cancer across all 3 studies.
Grouping tissue samples into benign prostate, benign prostatic hyperplasia,
localized prostate cancer, and metastatic prostate cancer allowed us to evaluate AMACR expression along a line of prostate cancer progression.
In this context, TMA analysis provided important data on the in situ protein
expression of AMACR, suggesting that it was prostate cancer specific.
α-Methylacyl coenzyme A racemase encodes an enzyme that catalyzes
the racemization of α-methyl–branched carboxylic coenzyme A thioesters19 and is localized in peroxisomes and mitochondria.19,20 Deficiency of AMACR has been linked
to certain adult-onset sensory motor neuropathies; however, clinical symptoms
are often mild and the condition presumably remains undiagnosed in many affected
of the overexpression of AMACR in prostate cancer may have clinical implications
beyond potential diagnostic uses. For example, AMACR activity may be required
for prostate cancer growth and, by virtue of its specificity, may serve as
an attractive therapeutic target. Presumably, specific inhibition of AMACR
activity may have minimal toxic effects, because individuals with AMACR deficiency
exhibit only mild clinical manifestations.21,35
Screening tests for prostate cancer include serum PSA level, digital
rectal examination, and transrectal ultrasound. A positive screening test
result, in combination with other clinical factors (eg, patient age), is an
indication for prostate needle biopsy. Based on the architectural pattern
of the cancerous glands, pathologists assign a Gleason grade, which is associated
with clinical outcome.36 In diagnostically
challenging cases, pathologists often use the basal cell markers 34βE1237- 39 or p63,40,41
which stain the basal cell layer of benign glands, which is not present in
malignant glands (Figure 6). Thus,
in some biopsy specimens, the pathologist must rely on absence of staining
to make the final diagnosis of prostate cancer.
Unlike AMACR, immunohistochemical stains for
PSA, a prototypic cancer biomarker, highlight both normal and malignant prostatic
epithelium. Thus, in this context, AMACR may have
potential clinical utility for prostate needle biopsy specimens. The results
of this study suggest that AMACR may be a useful
addition to current diagnostic tools for detecting prostate cancer, although
these findings require further evaluation in larger studies. We did note that
the precancerous lesion, high-grade PIN, demonstrated AMACR protein expression,
preventing us from being able to use AMACR alone in diagnostically challenging
prostate needle biopsy specimens. However, in combination with basal cell–specific
markers, such as 34βE12 or p63, we suspect that few cases will be diagnosed
as "atypical without a definitive diagnosis." This positive marker also may
be helpful for general surgical pathologists, because the absence of a basal
staining may be due to lack of basal cells or failure of the antibody to work
in the area critical for diagnosis.
In this study, we used high-throughput molecular and tissue technologies
to identify AMACR as a biomarker for prostate cancer.
Further research regarding AMACR overexpression in
prostate cancer may lead to improved diagnostic and therapeutic modalities
for this common disease.