Prostate cancer is the most common nonskin malignancy and the second leading cause of cancer deaths among men in the United States. Prostate cancer ([Mendelian Inheritance in Man 176807]) has a complex etiology; presently, age, ethnicity, and family history are the most consistently reported risk factors associated with disease. Other potential risk and protective factors have also been suggested. Androgen, acting through the androgen receptor (AR) is helpful in preserving the normal function and structure of the prostate. The AR ([Mendelian Inheritance in Man 313700]) is a structurally conserved member of the nuclear receptor superfamily of ligand-activated transcription factors. Androgens, such as testosterone, are strong tumor promoters, and work with the AR to augment the effect of any carcinogens present and stimulate cell division. The CAG repeats encode long glutamine homopolymeric amino acid chains in the amino-terminal domain of the AR gene. The authors focus on CAG repeat length because recent research suggests that men with shorter AR CAG lengths (e.g., < 22 repeats) are at a greater risk of developing prostate cancer than are those with longer variants. Among populations studied to date, African Americans appear to have the highest frequency of short CAG repeats. Several potential interactions have also been explored, including molecular interactions, androgen deprivation therapy, and prostate-specific antigen expression. CAG repeat length can be determined with high sensitivity and specificity. Presently, there is no recommended population screening for AR CAG repeat length.

Keywords: epidemiology, prostatic neoplasms, receptors, androgen, trinucleotide repeats
Abbreviations: AR, androgen receptor; BPH, benign prostatic hypertrophy; CI, confidence interval; PCR, polymerase chain reaction; SBMA, spinal and bulbar muscular atrophy

Androgens affect the human embryo in utero and lead to the development of male internal and external genitalia. During puberty, an increase in androgen levels leads to the start of spermatogenesis and growth of accessory sex organs such as the prostate gland¹. Androgen plays a role in prostate cancer growth. This is known because androgen is required in rodent induction models of prostate carcinogenesis because dogs and males castrated before puberty do not get prostate cancer and because androgen-ablative therapy inhibits prostate tumor growth during the time prior to the tumor reaching androgen independence². Unfortunately, within a short period of time, the tumor reaches a rapidly proliferating, hormone-independent state, resulting in adverse outcomes for the patient³. It is not known exactly how androgen is involved in prostate cancer etiology. If the androgen receptor (AR) had oncogenic potential, the androgen may play a role in initiation⁴. Alternatively, androgen may be involved in promotion or progression (i.e., clonal expansion) by enhancing androgen-regulated processes, for example, growth and cellular activity⁴.

The AR is made up of a C-terminal hormone-binding domain that helps with ligand specificity, a central DNA-binding domain that attaches to androgen-responsive target genes and an N-terminal domain that influences transcription efficiency⁵. The AR gene is located at Xq11.2–q12 (markers DXS991-DXS983) and is more than 90 kb in length^6,7. The open reading frame is separated over eight exons. The large amino-terminal domain is encoded by exon one, which includes the highly polymorphic CAG repeats⁸. Exons two and three encode the DNA-binding domain, and exons four to eight encode the information for the steroid hormone- (ligand) binding domain^8,9. Androgen, acting through the AR, helps to maintain the normal function and structure of the prostate^5,10,11. When androgens bind to the AR, it is activated, dimerizes, and localizes in the nucleus, where it attaches to specific sequences in the regulatory regions of target genes¹². Important functions of the product of the AR gene include activation of the expression of other genes¹¹ and the transport of the androgen hormone¹³.

It has been observed that shorter AR CAG repeats impose a higher transactivation activity on the receptor and have an increased binding affinity for androgens^6,14. This may make the prostate more vulnerable to chronic androgen overstimulation and increased proliferative activity, which, in turn, could increase the rate of somatic mutations among tumor suppressor genes (e.g., the AR). In contrast, the expansion of the CAG repeat (>40 repeats) leads to a below-normal AR level when measured by transient transfections assays with hormone-responsive-computerized axial tomography constructs^15,18 and by Scatchard analysis of AR in skin fibroblasts¹⁹. It has also been found that there is altered coactivator interaction with the AR³. Thus, the CAG microsatellite encodes a polyglutamine tract, which has a length that is inversely and linearly related to AR activity^20,21

Table 1 presents the frequency by ethnicity of AR CAG repeat lengths among relevant studies detected in a search of Medline between 1990 and September 2001. We linked the keywords "androgen receptor" and "trinucleotide repeats" for one search and the keywords "androgen receptor" and "prostatic neoplasms" for another. (This search strategy was also applied to the remaining sections of this paper.) Note that estimates of population-specific allele frequencies might be affected by misclassification of ethnicity. For example, Edwards et al.²² give estimates of CAG repeat frequency among European Americans, African Americans, Asian Americans, and Latinos using DNA extracted from a convenience sample of blood donated to Houston, Texas, blood banks. Blood-bank personnel noted European-American and African-American ethnicity based on visual appearances. Asian-American and Latino ethnicities were approximated based on surnames. Using visual or name identification could have led to severe misclassification of ethnicity, resulting in under- or overestimates of CAG repeat frequencies.

click to view

Table 1: Androgen receptor CAG repeat length and association with prostate cancer.

There is some evidence that CAG repeats are in linkage disequilibrium with other polymorphisms in the AR locus. CAG repeats were found to be in disequilibrium (i.e., allelic association) with Stu I mutations, and both have been associated with prostate cancer¹³. The disequilibrium may, however, simply reflect conserved haplotypes arising from the close physical distance between these polymorphisms, since they are both in the AR gene. Irving et al.²⁴ also found evidence of linkage disequilibrium between CAG and GGC/N alleles among prostate cancer patients. However, Stanford et al.¹¹ were unable to demonstrate such an association and speculate that a nearby gene may be in linkage disequilibrium with the AR. Finally, Xue et al.²⁵ speculated that short CAG alleles may be in linkage disequilibrium with the prostate-specific antigen G allele.

Incidence and mortality
Prostate cancer is the most common solid tumor and the second leading cause of cancer deaths among men in the United States: In 2001, approximately 198,100 men were diagnosed with prostate cancer, and about 31,500 will die of the disease. On the basis of data gathered during 1994–1996, a male has a one-in-six chance of developing prostate cancer at some point in his lifetime²⁶.

Ethnicity
Ethnicity is also one of the most widely accepted risk factors for prostate cancer. This helps to explain, in part, the wide geographic variation in the incidence of this disease. For instance, the United States has an incidence of prostate cancer that is eight times higher than that of Japan²⁷, and African Americans have the highest incidence and mortality in the world. For example, the Surveillance, Epidemiology, and End Results program data from the years 1988–1992 estimate age-adjusted incidence rates of 181/100,000 for African-American men versus only 135/100,000 among Caucasians; similarly, the age-adjusted prostate cancer mortality rate for African Americans is 54/100,000 versus only 24/100,000 in Caucasians²⁸. In all likelihood, racial and geographic differences are multifactorial in nature, with environmental, genetic, and possibly social factors^29,30. Differences in diet and access to improved detection methods could explain some of the ethnic variation³¹. Some have also suggested that varying levels of testosterone and its intraprostatic metabolite dihydrotestosterone across ethnic groups may be responsible for these differences³².

Family history
In addition, family history is a generally accepted risk factor for prostate cancer. For younger men, familial factors appear to be particularly influential, and the attributable risk of strong familial factors may be as high as 43 percent for men who are less than age 55 years³³. Nevertheless, family history of prostate cancer accounts for only about 9 percent of all cases³³. First-degree relatives of men diagnosed with prostate cancer have a two- to sixfold excess risk of the disease ^34-38. Spitz et al.³⁷ and Steinberg et al.³⁹ suggest that the more closely a man is related to an affected relative and the greater the number of affected relatives in his family, the higher is his risk of acquiring prostate cancer. Narod et al.⁴⁰ and Monroe et al.⁴¹ found that brothers of cases faced significantly higher risks of prostate cancer than did fathers.

Other risk factors
Age is an important risk factor for prostate cancer³⁰: Men are rarely diagnosed before age 40 years, and the rate of increase with aging—at approximately the ninth to tenth power of age—is greater than that for any other cancer¹³. A large number of other putative risk factors for prostate cancer have been identified, although whether these are causal remains controversial. Among these are obesity⁴², alcohol⁴³, sexual behavior³⁵, cigarette smoking⁴⁴, use of pesticides⁴⁵, and nuclear power⁴⁶. Other factors (for example, selenium⁴⁷; vitamins A, D, and E^48-51; and phytoestrogens^52-53 have been inversely associated with prostate cancer.

AR and prostate cancer
Prostate cancer is one of the primary diseases that have been associated with somatic and germline polymorphisms in the AR gene. In the "Associations" section below, we discuss in detail the relation between CAG repeats in the AR gene and prostate cancer. Other AR polymorphisms potentially associated with prostate cancer include GGC/N repeats, Stu I, and numerous base substitutions^{6,11,13,24,50,54-56}.

CAG polymorphisms in the AR
Shorter CAG repeats have been associated with benign prostatic hyperplasia (BPH)^57-58. Short CAG repeat lengths have also been associated with androgenetic alopecia^59,60, ankylosing spondylitis⁶¹, mental retardation⁶², younger age of male rheumatoid arthritis⁶³, and hepatitis B-related hepatocellular carcinoma⁶⁴. In contrast, extremely long AR CAG length (i.e., 38–52 repeats) has been associated with Kennedy's disease, spinal and bulbar muscular atrophy (SBMA)^65-66. SBMA is an adult-onset, motor neuron disease that is associated with reduced fertility, low virilization, testicular atrophy, and reduced sperm production¹³. In addition, it is of interest to note that men with SBMA have also been diagnosed with prostate cancer⁶⁷. Expansions of the CAG repeat may also be involved in the carcinogenesis of uterine endometrial cells⁶⁸. Finally, Rebbeck et al.⁶⁹ found that BRCA1 mutation carriers who carry long CAG repeats have an earlier age of breast cancer diagnosis. However, Spurdle et al.⁷⁰ found no association between CAG repeat length and breast cancer risk in women under age 40 years.

CAG repeat length and occurrence of prostate cancer
Table 1 also gives odds ratios for the association between CAG repeats in the AR gene and prostate cancer. These results suggest that the length of CAG repeats might be inversely associated with the risk of prostate cancer. Coetzee and Ross¹⁴ initially suggested that variations in CAG repeat length are associated with prostate cancer. One possible explanation for this is that short alleles may impose higher transactivation on the receptor due to the inverse relation between the number of glutamine residues in the polyglutamine tract and transcriptional activity^16,14,18. This potential association may be modified by stage/grade at diagnosis⁷¹, and men diagnosed at an older age appear to have longer CAG repeats⁹, suggesting that AR CAG repeat length may also be associated with prostate cancer aggressiveness. The use of a lower cutpoint shifts men from the short to the long category of trinucleotide repeat length. Thus, the proportion of short alleles is reduced, while that of longer alleles is increased. Other things being equal, this would bias the association between shorter alleles and the occurrence of prostate cancer. One would also expect that the use of a lower cutpoint would accentuate short-long allele differences between racial groups because a higher proportion of African Americans have a low number of CAG repeats^(9-17) compared with Caucasians or Asians. Reducing the short-long cutpoint should reclassify a greater proportion of Caucasians from the short to the long group compared with African Americans. The ratio of proportions for the shorter alleles for African Americans compared with Caucasians is higher in the study by Sartor et al.⁷² than it is for the other studies that would be expected since a lower cutpoint is used.

Two studies listed here^73,74 have not detected a statistically significant association between shorter CAG length and the occurrence of prostate cancer. Bratt et al.⁷⁵ concluded that CAG repeats are associated with young age at diagnosis, but not with higher risk of the disease. This study did not give the distribution of CAG repeat lengths among the controls in tabular form. As a result, the odds ratio could not be computed, and this study was not included in Table 1. Eeles et al.⁷⁶ did not find an association between CAG repeat length and prostate cancer development or aggressiveness among 178 Caucasian cases and matched controls. (Note that this has been published only in abstract form and does not give sufficient details for inclusion in Table 1).

Most studies reported to date suggest a positive association between shorter CAG repeat lengths and prostate cancer. However, the magnitude of any potential association appears to be relatively limited, and small sample sizes have restricted the precision of published results. Nevertheless, the large population frequency of short CAG repeat lengths suggests that, if causal, variants in CAG lengths could have substantial public health implications. More specifically, this potentially low-penetrance, high-frequency mutation could theoretically account for many more prostate cancer cases than a high-penetrance, low-frequency mutation.

CAG repeat length and aggressiveness of prostate cancer
With regard to prostate cancer aggressiveness, Giovannucci et al.⁷¹ found that comparing CAG repeat lengths among subjects with a high grade/stage (n = 269) gave an odds ratio of 1.64 (95 percent confidence interval (CI): 1.22, 2.22), whereas those with a low grade/stage (n = 309) had an odds ratio of 1.02 (95 percent CI: 0.77, 1.37). Stanford et al.¹¹ also looked at the impact of CAG repeat length on disease aggressiveness (i.e., stage C or D or Gleason score 8–10). Comparison of short versus long CAG repeats (cutpoint = 22) gave an odds ratio of 1.20 (95 percent CI: 0.79, 1.84) for more aggressive prostate cancer. These results suggest that shorter CAG repeat length may be involved not only with the development of prostate cancer but also the potential aggressiveness of the disease.

CAG repeat length and BPH
A recent study by Giovannucci et al.⁵⁷ found that shorter CAG repeats are associated with BPH. The data for this came from the Health Professionals Follow-up Study, an ongoing nationwide prospective cohort study of men aged 40–75 years. The subjects were first enrolled in 1986 and were employed in a variety of health professions. This study used 349 cases and 449 controls, and an odds ratio of 1.92 (95 percent CI: 1.22, 3.03) for BPH was observed for subjects with short alleles (< 19 repeats) compared with those with long alleles (> 25 repeats)⁵⁷.

The potential public health relevance and the biologic rationale supporting the relation between AR CAG and prostate cancer has helped generate a broad interest in CAG repeats. A recent paper in Epidemiologic Reviews provides additional information about CAG repeats and prostate cancer⁷⁷. When results from studies that are investigating the potential associations between trinucleotide repeats and prostate cancer are available, we will incorporate them into the electronic version of this Human Genome Epidemiology review. (Results can be sent directly to Dr. John Witte at witte@darwin.cwru.edu.) Since numerous different cutpoints have been used to distinguish between short and long AR CAG and GGC/N repeats, we suggest that future reports give crude data that allow for using a range of repeat cutpoints

Molecular and genetic interactions
Inappropriate RNA-binding protein interaction with specific messenger RNA may disturb cellular dynamics or change the regulation, transport, and expression of CAG containing RNA. This, in turn, may account for the observation of an inverse correlation between CAG repeat length and AR messenger RNA expression in studies involving lower primates. Furthermore, the CAG polyglutamine region may interact with some cellular proteins, and differences in the polyglutamine length could affect the tenacity of such interactions⁷⁹. There may also be interaction between CAG and GGN repeats. For example, when the subgroup in which both (CAG)n and (GGN)n alleles were short (CAG, < 22; GGN, < 16) was compared with those in which both alleles were long (CAG, > 22; GGN, > 16), an odds ratio of 2.05 (95 percent CI:1.09, 3.84) was observed¹¹. Finally, there is evidence that allelic differences in AR-driven, prostate-specific antigen expression may affect prostate cancer risk²⁵.

The study of interactions between CAG repeats and environmental and/or genetic mechanisms in prostate cancer is hampered by lack of power. These limitations have been highlighted by Smith and Day 80 and observed in another the Human Genome Epidemiology review by Cotton et al.⁸¹.

To determine AR CAG repeat length, extracted undigested DNA is amplified by polymerase chain reaction (PCR) in a series of two rounds using nested primers surrounding the repeat in exon 1²⁴. The final products can then be analyzed by electrophoresis on a 5 percent denaturing polyacrylamide gel and subjected to autoradiography. One can then obtain the number of CAG repeats from the size of the predominant PCR product relative to a series of previously determined CAG size standards. Subsequently, all samples are ranked by size and reanalyzed by electrophoresis and autoradiography, with each allele of an equivalent size next to each other to validate the original size estimates²⁴. Bharaj et al.⁸² improved on this technique somewhat by utilizing a fully automated system for the electrophoretic separation of the PCR product, utilizing fluorescent (nonisotopic) detection of standard-sized markers and using fragment analysis to establish the length of the PCR product. Estimates of sensitivity and specificity for this laboratory test have not been reported. Nevertheless, since the primers for this test are designed specifically for the CAG repeat, the chance of making an error in CAG number counts is very low, and the sensitivity and specificity are extremely high; the main source of error is likely to come from human origins, for example, incorrectly mixing or labeling samples (S. Ingles, personal communication, 1999). Wada et al.⁸³ describe a promising, but not widely used, alternative to the above method. This involves amplification via PCR, using a simple purification procedure and counting CAG repeats via matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.

On the basis of the evidence summarized here, testing for CAG trinucleotide repeat length in the general population as part of a population screening program is not presently warranted. This underscores the need for doing additional research on this topic. The association between AR CAG repeat length and prostate cancer is intriguing but is still somewhat limited. To understand prostate carcinogenesis better, we should work to fully explicate any relation among genes such as CAG repeats and environmental factors that may interact to increase prostate cancer risk. The resulting information may ultimately improve our ability to prevent the development of prostate cancer and provide us with better prognostic information following diagnosis.

Supported by grants from the National Institutes of Health (CA88164), DOD Prostate Cancer Research Program (DAMD17-98-1-8589), and the Urologic Research Foundation.

Table 1 - Androgen receptor CAG repeat length and association with prostate cancer

List of References

Androgen Receptor Gene and Prostate Cancer

Prostate Cancer

• American Cancer Society
• Prostate Cancer Foundation
• National Cancer Institute
• National Prostate Cancer Coalition
• University of Pennsylvania Cancer Center OncoLink (last accessed 01/2008)

Genetic Databases

• The Androgen Receptor Gene Mutations Database
• Online Mendelian Inheritance in Man (OMIM)
  (for PCA [MIM 176807], AR [MIM 313700], HPC1 [MIM 601518), HPC2 [MIM 602759], and HPCX [MIM 300147])
• The Genome Database (GDB)