Molecular Human Reproduction, Vol. 9, No. 8, 437-448,
August 2003
© 2003 European Society of Human Reproduction and Embryology
Article |
Molecular biology of prostate cancer
Submitted on March 14, 2003; accepted on May 8, 2003
Department of Urology, Heinrich-Heine-University Düsseldorf, Moorenstr. 5, 40225 Düsseldorf, Germany
1 To whom correspondence should be addressed. e-mail: wolfgang.schulz{at}uni-duesseldorf.de
 |
ABSTRACT |
|---|
 Top ABSTRACT IntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
In spite of progress in diagnosis and treatment, prostate cancer has become one of the most frequent lethal cancers in males in many Western industrialized countries. Research on the molecular biology of prostate cancer is expected to reveal those aspects of Western lifestyle contributing to its high incidence with the aims of improving prevention, distinguishing slow-growing from aggressive clinically relevant cancers, and providing targets for treatment, particularly of locally advanced and of metastatic disease. Traditionally, prostate cancer research focused on androgens. More recently, tumour suppressors and proto-oncogenes important in other human cancers have been intensely investigated. Current approaches include the search for genes mutated in familial cases, identification of recurrent chromosomal alterations and their associated potential tumour suppressor genes, determination of gene expression profiles characterizing tumour stages and subclasses, and elucidation of the importance of epigenetic alterations. Results from such studies have begun to be translated into the clinic. Further successful transfer of results from molecular biology to the clinic will, however, require integration of the amassed molecular data into a biological framework model of prostate carcinoma.
Key words: androgens/epigenetics/expression profiling/oncogene/tumour suppressor
|
Introduction |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Prostate cancer has become a major public health problem in Western industrialized countries. It is predominantly a disease of elderly men, its incidence increasing steeply in the 7th decade of life. Partly, the rising incidence is caused by ageing of the population, but the age-adjusted incidence has also increased. Lately, a reversal may have set in, perhaps as a consequence of improved detection by screening for prostate-specific antigen (PSA) and better treatment.
The aetiology of prostate cancer is insufficiently understood. There is a striking >10-fold gradient in the incidence of clinically significant prostate cancer between Western industrialized and East Asian countries. Immigrant studies show genetic factors to be only partially responsible for this gradient. Therefore, some aspect of the Western lifestyle appears to promote prostate cancer development. Since prostate cancer is probably not caused by exposure to exogenous chemical carcinogens, the Western diet is usually implicated. A large amount of literature (Giovannucci, 1999; Schmitz-Drager et al., 2001; Greenwald, 2002; Jankevicius et al., 2002) indicates that the Western diet may be low in protective factors, i.e. micronutrients such as selenium, vitamins such as folate, phytoestrogens, and antioxidants such lycopene. Conversely, high fat, dairy product and red meat consumption have emerged as risk factors in epidemiological studies. Although the data are not conclusive, achieving a decrease of prostate cancer incidence by lifestyle changes and chemoprevention seems a realistic aim, but it would be helpful to understand by which molecular mechanisms protective and risk factors act.
Nowadays, an increasing number of prostate cancers are detected through elevated serum PSA levels. Before PSA assays became routine, many cases were recognized by clinical syndromes or digital rectal examination (DRE). Typically, tumours were advanced and extended beyond the organ capsule or had metastasized. In such cases, PSA levels as a rule are much higher than 10 ng/ml. In contrast, many cases today are detected by PSA levels in the 2.5–10 ng/ml range and are confirmed by histology of biopsies. PSA detection is very sensitive and advances in specificity have been achieved by determining several protein forms (Balk et al., 2003), but it remains imperfect. Thus, many men have to undergo unnecessarily the diagnostic procedure with its physiological and psychological stress and associated costs. Moreover, with current techniques, imaging and biopsies can be ambiguous.
These diagnostic problems are complicated by the extreme variability in the clinical course. Prostate cancer can be an indolent, latent disease that will not result in clinical symptoms during the life-time of an elderly patient. Other prostate cancers take an aggressive course, spreading into the seminal vesicles, bladder and rectum, and metastasizing to lymph nodes, bone, lung and other organs. Currently, there are several treatment options for prostate cancer which need to be adapted to the individual patient. Watchful waiting can be appropriate for slow-growing carcinomas, especially in older patients. Radical prostatectomy or radiation can cure organ-confined cancers. Androgen ablation therapy by orchiectomy, GnRH feedback signalling and androgen receptor (AR) antagonists can be used adjuvantly or to alleviate metastatic disease. Recently, chemotherapy with taxols, metaxanthrone and estramucin phosphate has been found to be remarkably efficacious in some cases (Gilligan and Kantoff, 2002). To aid in treatment decisions, nomograms have been introduced, which combine information from digital rectal examination, biopsy histology and PSA assays (Partin et al., 1997).
Given this background, expectations are high for the molecular biology of prostate cancer. It is expected to reveal insights into the molecular mechanisms of carcinogenesis to aid in prevention, provide molecular markers for definitive diagnosis and for tumour staging and classification. New therapeutic targets are also sought, in particular for locally advanced or metastasized cancers.
For a long time, androgen dependence and insensitivity were the main issues in prostate cancer research and prostate cancer was conceptualized as a prototypic hormone-dependent cancer (Denmeade and Issacs, 2002; Isaacs et al., 2002). This is understandable, since androgen depletion by orchiectomy had been shown already in the 1940s to improve the symptoms of prostate cancer. A second period of prostate cancer research gaining ground in the 1990s was stimulated by the recognition that human cancers are typically caused by accumulation of multiple genetic alterations, detectable as point mutations or chromosomal alterations, which activate proto-oncogenes and inactivate tumour suppressor genes, with colon cancer as a paradigm. This led to screening for recurrent chromosomal alterations and mutational analysis of almost any oncogene and tumour suppressor gene known from other carcinomas. Both lines of research have yielded many insights, but may have suffered from not considering prostate cancer sufficiently in its own right. Indeed, a third period of prostate cancer research may have begun that better addresses the peculiarities of this cancer. In the following sections, we will emphasize this line of research.
|
Prostate structure and development |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
The development of the male reproductive tract is dependent upon mesenchymal-epithelial interactions and fetal androgens (Cunha, 1994). In the fetus, testosterone stimulates budding of the prostate epithelium from the urogenital sinus which then produces growth factors such as sonic hedgehog to activate the underlying mesenchyme (Podlasek et al., 1999). During branching morphogenesis of the prostate, the urogenital mesenchyme expresses high levels of the AR, which is almost undetectable in the epithelium at that time. Paracrine growth factors from the mesenchyme are responsible for glandular morphogenesis and epithelial cell growth in the developing prostate. During maturation of the gland, the AR appears in the secretory cell layer (McNeal, 1981). In the adult prostate, androgens act on stromal and secretory epithelial cells.
The normal adult prostate is composed of a glandular epithelial and a fibromuscular stroma compartment. Both compartments turn over slowly with balanced levels of proliferation and cell death. In the 1980s McNeal (1981) developed the concept of zonal anatomy of the prostate which forms the current basis for describing the location of neoplastic processes in the prostate. According to this concept, the glandular part of the gland is composed of a large peripheral zone and a small central zone which together constitute 
95% of the gland. The remainder is composed of the transition zone and the peri-urethral glands. Although peripheral and transition zones can be distinguished on the basis of their anatomical relationships and the composition of their stromal elements, they share a similar acinar structure reflecting the common origin from the urogenital sinus. The glands from the central zone are morphologically distinct, possibly due to their embryonic origin from the Wolffian duct. About 60–70% of prostatic cancers occur in the peripheral zone and 10–20% in the transition zone. Only 5–10% of the cancers originate from the central zone. Benign prostatic hyperplasia (BPH) develops mainly from the peri-urethral stroma and glands of the transition zone.
|
Natural history of prostate cancer |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Most prostate cancers start in the glands of the peripheral zone and retain some glandular structure. They are therefore classified as adenocarcinoma. The earliest precursor detected histologically is prostatic intraepithelial neoplasia (PIN) characterized by thickening of the epithelial layer and loss of distinct basal and secretory layers. This is often interpreted as a loss of the basal layer, but prostate carcinoma cells in fact carry markers both of basal cells such as specific cytokeratins and of secretory cells such as the AR and PSA. High grade PIN is accepted as a precursor of carcinoma from which it is distinguished by preservation of the basal membrane. Rates of cell proliferation in the prostate are low and often remain so in PIN and in prostate carcinomas during much of their development. The abnormal increase in cell number largely results from inappropriately low apoptosis. Prostate carcinoma is frequently multifocal. Several distinct foci of carcinoma and PIN can be found in one prostate varying in the degree of cellular dysplasia, tissue disorganization and genetic alterations. Moreover, even contiguous carcinomas may be heterogeneous. As a practical consequence of this heterogeneity, histological grading (G1–G3) has been largely replaced by Gleason grading which considers the degree of tissue disorganization in the two main components of the carcinoma (Gleason, 1966).
While many prostate carcinomas retain an indolent growth pattern, an estimated third become locally invasive, spreading beyond the tissue capsule, or metastasizing to local lymph nodes and distal organs, most frequently bone, liver and lung. Prostate cancer metastases are rarely accessible to study, but usually appear undifferentiated. Throughout progression, most prostate cancers express the AR and androgen depletion decreases the tumour mass, probably by induction of apoptosis and growth arrest. However, the carcinomas inevitably contain or develop cells that continue to grow during hormone depletion.
|
Androgens in prostate cancer |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Today, androgen depletion is part of many prostate cancer treatments. However, androgen withdrawal is rarely curative. Locally recurrent tumours and metastases often regress, but the effect on overall survival is at best moderate, even by modern drugs targeting feedback receptors in the hypothalamus and AR in the tumour cells. There is an extensive literature, including excellent reviews (Eder et al., 2001; Feldman and Feldman, 2001), on the mechanisms responsible for this failure. Several mechanisms seem to be involved (Figure 1). First, some prostate carcinomas do not express the AR, sometimes because the gene is silenced by promoter hypermethylation (Jarrard et al., 1998). Proliferation and survival signals in these tumours are thought to be conferred by peptide growth factors. Of note, in this respect such tumours resemble basal cells of the prostate epithelium. Second, several peptide growth factors and cytokines, such as fibroblast growth factor 7 (FGF7), epidermal growth factor (EGF) and interleukin-6 (IL-6), can activate the AR synergistically with, or independent of, a steroid ligand (Culig et al., 1994; Hobisch et al., 1998). Third, in some prostate carcinomas somatic AR mutations alter the specificity of the receptor, making it responsive to estrogens, progesterone, dehydroepiandrosterone, or even synthetic antiandrogens (Veldscholte et al., 1990). Fourth, amplification of the AR gene may occur in up to 30% of prostate carcinomas growing under androgen depletion leading to increased sensitivity towards minimal levels of androgens and other signals activating the receptor (Visakorpi et al., 1995). Fifth, numerous coactivator proteins have been identified which mediate the AR effects on chromatin structure and transcriptional initiation and its interaction with other signalling pathways (Yeh and Chang, 1996; Muller et al., 2000; Gregory et al., 2001; McKenna and O’Malley, 2001; Sharma and Sun, 2001). Changes in the expression level and perhaps structure of these proteins may act by themselves or synergize with other mechanisms to confer ‘androgen independence’. Another interacting protein is p53 which at high levels represses AR function (Shenk et al., 2001).
|
In consequence, each mechanism leads to altered responsiveness of prostate carcinoma cells to androgens, ranging from complete independence to hypersensitivity. The diversity of the mechanisms makes it difficult to find a therapeutic solution to the problem of ‘androgen independence’. It is now even asked heretically, whether androgen depletion may act only ‘cosmetically’ by decreasing PSA levels and by killing those cells within the tumour that critically depend on androgens, but not the more aggressive androgen-independent fraction. Alternatively, some aspect of androgen signalling may be essential for survival of prostate carcinoma cells, and mechanisms of androgen-independence may arise as a consequence of a strong selection pressure. In fact, down-regulation of the AR in a prostate carcinoma cell line caused growth inhibition (Eder et al., 2001), indicating that at least a survival signal is required through androgen signalling. On the other hand, AR-deficient prostate carcinoma cell lines do not tolerate re-expression (Heisler et al., 1997).
|
Hereditary prostate cancer |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Identification of genes mutated in hereditary human cancers has helped in understanding the genetic changes in the corresponding sporadic cases. A well-known example is familial adenomatous polyposis coli (FAP) with the APC gene located at 5q21 in colorectal cancer (Vogelstein and Kinzler, 2002). Hereditary cancers are typically distinguished from sporadic cancers by familial clustering and autosomal-dominant inheritance, multifocality and earlier onset. In most hereditary cancers, these features are explained by an inherited mutation in one allele of a tumour suppressor gene. Somatic mutation of the second allele is then sufficient to start off tumour development. In sporadic cases, two successive mutations in the same cell line are required to inactivate the same gene.
The search for genes associated with inherited forms of prostate cancer has been more difficult than in colorectal cancer. No syndrome obviously predisposes to prostate carcinoma. Moreover, since multifocality is common, the search had to rely on the criteria of familial clustering and comparatively early onset. Both are not very stringent in a highly prevalent disease. Nevertheless, the investigation of prostate carcinoma families including one or several cases with early onset has yielded several candidate genes and markers that co-segregate with prostate carcinoma.
At least seven regions in the genome have been suggested to contain an autosomal-dominant hereditary prostate cancer gene, including 1q24-1q25 (HPC1), 1q42-q43 (PCAP), Xq27-q28 (HPCX), 1p36 (CAPB), 20q13 (HPC20), 17p11 (ELAC2) and 16q23 (Nwosu et al., 2001; Bratt, 2002; Simard et al., 2002). None of the regions has been consistently identified in different populations and none is as consistent a target of allelic loss in prostate carcinomas as 5q21 in colorectal cancer. This suggests that genetic predisposition to prostate cancer may not follow the standard scheme. Two candidate genes cloned to date compound this suspicion. ELAC2 was implicated by a frameshift mutation in the gene segregating with the disease in a large pedigree (Tavtigian et al., 2001). Further inactivating mutations were found in other populations, but are rare. More common polymorphisms in the gene, notably Ser217Leu, appear to be associated with a risk of prostate cancer only in certain populations (Simard et al., 2002). The function of ELAC2 is unclear; it may interact with 
-tubulin (Koerver et al., 2003). In the most consistent candidate region, 1q24-25, for hereditary prostate cancer (hence: HPC1), the relevant gene is probably RNASEL (Carpten et al., 2002) which encodes an antiviral, proapoptotic, and interferon-activated RNase. However, the presumed causative mutation, Glu265X, segregated with prostate cancer in only one of eight ‘HPC1 families’ and incompletely in a second. No mutations in the gene were found in the other families, in which the disease segregated with markers in the region, and Glu265X occurs also in controls with appreciable frequency. Male relatives of prostate cancer patients are clearly at enhanced risk compared to the general population, but considerable mathematical acrobatics must be invoked to ascribe this increase to the genes and mutations identified so far.
Rather than from mutations in specific genes inferring a high risk, genetic susceptibility to prostate cancer may result from prevalent polymorphisms, moderately increasing the risk for the disease in heterozygotes or homozygotes. Polymorphisms in genes related to hormone response and metabolism, cell protection, DNA repair and nucleotide metabolism have been implicated. In the first group, polymorphisms in the AR and vitamin D receptor (VDR) are best studied. In the N-terminal transcriptional activation domain of the AR, polymorphic CAG and GGN repeats encode glutamine and glycine repeats respectively. Overall, shorter repeats seem to be associated with a small increase in risk, although results from different studies vary (Simard et al., 2002). Of several polymorphisms in VDR, different lengths of a polyA-repeat in the 3'-region correlate best with altered susceptibility (Ingles et al., 1997; Correa-Cerro et al., 1999). Another well-studied polymorphism is Ala49Thr in the SRD5A2 5
-reductase, which in the prostate converts testosterone to the more active dihydrotestosterone. This polymorphism may carry an increased risk in specific populations such as African-Americans, but seems irrelevant in European populations (Makridakis et al., 1999). In cancers caused by chemical carcinogens, polymorphisms in xenobiotic metabolism genes can substantially modulate the risk. However, enzymes such as glutathione transferases (GST) also protect against potentially mutagenic endogenous compounds and oxidative stress. Indeed, polymorphisms in GSTT and GSTP have been linked to prostate cancer susceptibility (Steinhoff et al., 2000; Gsur, et al., 2001; Kote-Jarai et al., 2001). Among DNA repair genes, mutations in the breast cancer susceptibility genes BRCA1 and BRCA2 also seem to confer an increased risk for prostate cancer (Rosen et al., 2001; Edwards et al., 2003). Finally, a highly prevalent polymorphism in the MTHFR (methylene tetrahydrofolate reductase) gene, Ala677Val, may be related to prostate cancer susceptibility (Kimura et al., 2000). The enzyme catalyses a key reaction in nucleotide biosynthesis, channelling methyl groups towards synthesis of either thymidine or S-adenosylmethionine. Polymorphisms in MTHFR become most relevant in combination with nutritional deficiencies, notably in folate or vitamin B12 (Kim, 2000). This case illustrates that polymorphisms often modulate cancer risk depending on other factors, such as nutrition or exposure to specific toxic compounds. Since lifestyle factors seem so important in determining prostate cancer risk, their interaction with the genotype should be considered for other polymorphisms as well. For instance, the correlation between increased IGF-1 and decreased IGFBP-3 serum levels and prostate carcinoma (Chan et al., 1998) might reflect another genotype–environment interaction.
|
Chromosomal alterations |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
While early stage prostate carcinomas often remain euploid, numerical and structural chromosomal alterations accumulate at advanced stages. Deletions of chromosomal segments predominate, while gains of chromosomal segments and amplifications become more frequent in advanced cases (Nupponen and Visakorpi, 1999; Dong, 2001). Cytogenetic methods such as comparative genomic hybridization and fluorescence in-situ hybridization yield slightly different results from molecular genetic analyses for loss of heterozygosity (LOH) or allelic imbalance using polymorphic markers such as microsatellites. Cytogenetic methods better detect numerical changes, whereas molecular analyses also identify recombination events that do not change gene copy numbers. Moreover, results depend on tumour stage and, probably, populations. Overall, the most frequently altered autosomes in prostate carcinoma are 8, 13, 7, 10, 16, 6 and 17 in this approximate decreasing order. In addition, gains or amplification of parts of X and loss of Y are often observed. Decreased copy numbers and LOH of chromosome 8p are the most consistent finding throughout all studies, detected in generally more than half the cases. Likewise, deletion and LOH of chromosome 13q are frequent (Hyytinen et al., 1999; Latil et al., 1999). The issue with chromosome 7 is more complex, since both gains and losses have been reported (Jenkins et al., 1998; Alers et al., 2000). Specifically, it is unclear whether the allelic imbalances on 7q reflect increased copy numbers pointing to an oncogene in the region, loss of alleles suggesting a tumour suppressor gene, or instability of a fragile site located at 7q31. The evidence for allelic loss is more convincing for chromosome 16 (Elo et al., 1997; Latil et al., 1997). For chromosome 10, cytogenetic and molecular genetic methods suggest allelic loss on the long arm to be moderately frequent in progressive cancers, but precise mapping of the deleted region has proven difficult (Leube et al., 2002). Chromosomal regions on 5q, 6q and 17p appear to be lost in a subfraction of cases. The most consistently gained region is chromosome 8q. Amplifications involving 8q24.1 are frequent in metastatic disease (Jenkins et al., 1997; Nupponen et al., 1998). Other gains occur with lower frequencies, usually in more advanced tumours. Nevertheless, prostate carcinomas do not regularly display pronounced chromosomal instability and the number of changes is often limited. This suggests that the observed changes have indeed been selected for their functional impact on the tumour phenotype and indicate tumour suppressor genes or oncogenes (in the broadest sense) within the affected regions.
|
Tumour suppressors |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Tumour suppressors in human cancers have been identified in three major ways. Hereditary cancers are often caused by mutations in tumour suppressor genes. As mentioned above, this approach has not been productive in prostate cancer. The second approach, the investigation of tumour suppressors important in other types of cancers, has been moderately successful in prostate cancer. Therefore, the major approach pursued in prostate cancer is the third one, seeking tumour suppressor genes in regions of the genome that are consistently deleted.
Among established tumour suppressor genes, p53 and PTEN are clearly involved in the progression of prostate carcinoma. Losses of chromosome 17p and 10q, where they are located, occur with moderate frequencies in advanced cancers. Loss of one allele is accompanied by point mutations in the remaining copy of p53, leading to its functional inactivation (Navone et al., 1993; Heidenberg et al., 1995). Point mutations have also been reported for PTEN (Cairns et al., 1997; Li et al., 1997). Alternatively, the second PTEN copy is transcriptionally silenced, by an unidentified mechanism (Whang et al., 1998). The RB1 gene is located at 13q14 within one of the most commonly deleted regions in prostate carcinoma. Nevertheless, it is not established as a decisive tumour suppressor in prostate cancer. The remaining copy of RB1 usually seems to remain intact. Down-regulation at the protein level has been observed by immunohistochemistry, but the results vary between studies. In fact, the smallest common region of deletion on 13q is located centromeric to RB1, but telomeric to BRCA2 (Tricoli et al., 1996; Latil et al., 1999; Yin et al., 1999; Schmidt et al., 2001). A similar controversy concerns CDKN2A on 9p21 which encodes p16INK4A, an important regulator of RB1, and p14ARF1, a protein activating p53 in response to inappropriate proliferation signals. Deletions on chromosome 9p are not as prevalent in prostate carcinoma as in many other cancers and CDKN2A mutations and promoter hypermethylation are both rare (Jarrard et al., 1997).
Taken together, the limited involvement of established tumour suppressors and the consistent loss of specific chromosomal regions strongly suggests that tissue-specific tumour suppressor genes are crucial in prostate carcinoma. Several candidates have been identified in recent years (Table I). Obviously, those on chromosome 8p are of greatest interest. The most interesting candidate is NKX3A (also NKX3.1) at 8p21 which encodes a transcription factor almost exclusively expressed in the prostate (reviewed by Abate-Shen and Shen, 2000). It is induced by androgens and stimulates the transcription of prostate-specific genes. Inactivation of the gene in knockout mice inhibits maturation of the prostate and induces hyperproliferation. In spite of these convincing functional data and the consistent allelic losses, NKX3A is not established as a tumour suppressor. Neither mutations nor promoter hypermethylation have been detected in the retained allele, although down-regulation of its expression may occur (Korkmaz et al., 2000; L.L.Xu et al., 2000; Ornstein et al., 2001).
|
Similar conclusions apply to other tumour suppressor candidates in prostate carcinoma. All candidates in Table I have been found in regions of allelic loss, but convincing data for inactivation of the second allele are sparse. Mutational inactivation of the remaining allele has only been reported in a few instances, such as the transcription factor gene KLF6 at 10p (Narla et al., 2001). However, KLF6 mutations have not been confirmed independently and allelic loss at 10p was much rarer in other studies. Deletions of the second allele, i.e. homozygous deletions, have not been observed for candidate tumour suppressor genes in prostate cancer. Even within the rare homozygously deleted regions on chromosome 8p, no tumour suppressor could be confirmed (van Alewijk et al., 1999). Other candidate genes affected by deletions reside at fragile sites (Bednarek et al., 2000), raising the question whether deletions only reflect chromosomal instability. Thus, the most usual mechanism for inactivation of retained alleles of tumour suppressor genes in prostate carcinoma seems to be down-regulation, sometimes by promoter hypermethylation.
These findings create a conceptual dilemma. The established model calls for functional inactivation of both tumour suppressor alleles by deletion or mutation. Promoter hypermethylation is now recognized as a third mechanism which can lead to stable inactivation of transcription. However, decreased expression as such is not an accepted mechanism of tumour suppressor inactivation, because it is by definition secondary and potentially unstable. Conceivably, a diminished gene dosage as such may be sufficient to promote tumour development. Such ‘haploinsufficiency’ could explain the findings with NKX3A, but is very difficult to ascertain.
|
Oncogenes |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Altered expression of several proto-oncogenes contributes to prostate cancer development and progression, whereas activating mutations, e.g. of RAS genes, are much less frequent. A very consistent finding in metastatic cancers is overexpression of MYC, usually associated with increased gene copy numbers by chromosomal gains or amplification (Jenkins et al., 1997; Nupponen et al., 1998). Further genes on chromosome 8q are concomitantly overexpressed and may contribute to the tumour phenotype. For instance, amplification of EIF3S3 encoding a translation initiation factor may support increased protein synthesis (Saramaki et al., 2001). The antiapoptotic protein BCL2 is overexpressed in about half of all prostate cancers, particularly in androgen-independent cases (McDonnell et al., 1992). In normal cells, the cell growth signal conferred by MYC expression is limited by the pro-apoptotic action of the protein mediated through p14ARF1, p53, and BAX. A combination of deregulated MYC and altered BCL2/BAX ratio could well account for the obdurate growth of prostate carcinoma cells. A further stimulus may result from increased expression of several growth factor receptors, including the tyrosine kinase receptors EGFR, MET, FGFR2c and ERBB2 (Djakiew, 2000) as well as the ET1A receptor for endothelins (Nelson and Carducci, 2000). Specifically, increased responsiveness to endothelins and local paracrine signalling is discussed as a mechanism for the predominance of bone metastases in prostate carcinoma (Koeneman et al., 1999). Chromosomal gain may contribute to overexpression of EGFR and MET, but does not account for all cases, in particular, the almost invariable overexpression of EGFR in metastatic cases (Schwartz et al., 1999; Skacel et al., 2001; Di Lorenzo et al., 2002).
|
Cancer pathways |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Hanahan and Weinberg (2000) have summarized the central concepts of current cancer research. Alterations in a limited number of ‘cancer pathways’, which normally regulate cell proliferation, differentiation, survival, senescence, and cell–cell interactions in a tissue, are regarded as fundamental for the development of human cancers. This concept implies that alterations in different components of one pathway can lead to very similar outcomes. Thus, proliferation signals for a tumour cell may come from an autocrine growth factor, an overexpressed tyrosine kinase receptor, a RAS transducer mutation, or overexpression of a growth factor-dependent transcription factor such as MYC. As discussed above, this type of pathway may be important in advanced prostate cancers.
Another cancer pathway is WNT signalling, whose constitutive activation is crucial in colorectal carcinoma. This pathway can be activated by several alterations. In colon cancers, mutations of the negative regulator APC prevail, complemented by mutations leading to constitutive activation of the co-transcription factor ß-catenin. These mutations have also been found at a low frequency in prostate carcinoma (Chesire et al., 2000; Gerstein et al., 2002). Allelic loss or hypermethylation of APC have also been reported (Gerstein et al., 2002; Maruyama et al., 2002). In addition, ßTRCP, a protein involved in the degradation of ß-catenin, may be mutated in rare cases (Gerstein et al., 2002). In colon cells, WNT signalling not only stimulates proliferation, but also confers a tissue stem cell phenotype (van de Wetering et al., 2002). It is not clear whether this also occurs in prostate carcinoma cells with mutations in this pathway. Furthermore, ß-catenin interacts with the AR (Truica et al., 2000).
Proliferation signals by the WNT pathway are relayed to the cell cycle by Cyclin D1 and MYC. In other cancers, increased proliferation and blocked differentiation are caused by alterations in proximate cell cycle regulators such as RB1 and p16INK4A. Conceivably, their inactivation in prostate cancer is not obligatory for continuous proliferation, because proliferation signals are provided by other means. However, like p53, RB1 is also involved in the control of cellular senescence and genomic stability. Therefore, inactivation or bypassing of both regulator proteins may be required during progression of prostate cancer. In this context, down-regulation of p27KIP1, which is expressed in normal prostate epithelium and retained in some carcinomas, is associated with a significantly higher risk of progression (Cheville et al., 1998; Cordon-Cardo et al., 1998; Yang et al., 1998), as is gain of chromosome 8q (Jenkins et al., 1998). The p27KIP1 protein, an inhibitor of cyclin-dependent kinases, coordinates cell cycle progression and helps to establish terminal differentiation in normal cells. It is mainly regulated at the post-transcriptional level. Phosphorylation, which is stimulated by MYC, leads to its degradation. Thus, the association of p27KIP1 down-regulation with tumour progression may reflect the inactivation of a cell cycle checkpoint.
Diminished apoptosis is a salient feature of prostate cancer. In the adult prostate, androgens are involved in balancing cell proliferation and cell death. Androgen withdrawal causes a massive involution of the prostate gland by apoptosis. Of note, only the secretory epithelial cells, but not the basal epithelial cells or stromal cells, are eliminated. Similarly, androgen ablation activates apoptosis in androgen-dependent prostatic cancer cells (Kyprianou et al., 1990). Therefore, changes in androgen signalling in androgen-independent cancers (Figure 1) are likely to contribute to tumour cell survival.
As discussed above, a second factor in diminished apoptosis in prostate cancers is BCL2. BCL2 is not expressed in normal secretory epithelial cells of the prostate, but starting from PIN, BCL2 is frequently expressed throughout the epithelium (McDonnell et al., 1992). Interestingly, in the androgen-dependent LNCaP prostate carcinoma cell line BCL2 expression is androgen-dependent (Berchem et al., 1995).
Growth factors not only increase cell proliferation, but also decrease apoptosis. Those relevant in prostate cancer may largely correspond to those that mediate glandular morphogenesis and epithelial cell growth during prostate fetal development. For instance, FGF-7, which is produced by prostatic mesenchymal cells, decreases apoptosis and prolongs cell survival in prostate carcinoma cells, probably by increasing BCL2 expression (Crescioli et al., 2002). Similarly, IGF-1 possesses antiapoptotic as well as mitogenic properties in the prostate (Grimberg and Cohen, 2000; Yu and Rohan, 2000). IGF-1 is known to activate the antiapoptotic PI3K/AKT pathway and to stimulate the expression of BCL-like proteins and suppress BAX. Moreover, IGF-binding proteins, which also modulate apoptosis, show altered expression in prostate carcinoma (Grimberg and Cohen, 2000; Chan et al., 2002).
Immortalization allows expansion of tumours beyond the limits of replicative senescence. It requires inactivation of p53 and RB1 and the emergence of telomerase activity. These changes have all been observed in advanced prostate cancers (Jarrard et al., 1999).
Further crucial transitions during tumour progression in the Hanahan and Weinberg model are acquisition of invasive properties and the ability to metastasize. These properties require multiple changes in gene expression leading to decreased adhesion, increased motility, altered interaction with the extracellular matrix, enhanced angiogenesis and ability to survive under hypoxic conditions. Such changes occur in aggressive prostate carcinomas. Decreased expression of cell adhesion molecules, such as E-Cadherin (Richmond et al., 1997), CD44 (Gao et al., 1997), and KAI-1 (Dong et al., 1996; Kawana et al., 1997), increased expression of metalloproteinases (Lichtinghagen et al., 2002) as well as altered expression of angiogenic factors (Hrouda et al., 2003) have been documented and provide reasonably good prognostic markers. Changes in the interaction between prostate cancer cells and their environment could also result from altered patterns of cell-surface carbohydrates and secreted glycoproteins (Schut et al., 2003), which in turn are influenced by enzymes such as the STM sialyltransferase which is down-regulated in prostate carcinoma cells (Sotiropoulou et al., 2002). Overall, the acquisition of invasive properties by tumour cells often occurs in a coordinate manner as if by activation of a transcriptional program for invasion following a primary genetic change. In some cancers, this underlying change may be provided by RAS mutations or inactivation of the TGFß response. In prostate cancer, the primary cause is not clear, although loss of TGFßRII has been reported (Kim et al., 1996).
|
Expression profiling |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Analysis of gene expression patterns by cDNA or oligonucleotide microarrays, designated ‘expression profiling’, is now an important technique for the analysis of human tumours. Expectations for its application to prostate carcinoma have been particularly high. One hopes that the biological heterogeneity of the disease might be reflected in expression profiles, yielding classifications for prognosis and treatment and new diagnostic markers and therapeutic targets. Another hope, probably originating from the moderate success in identifying primary genetic changes in prostate carcinoma, is that these elusive changes will emerge from analysis of altered gene expression patterns. Specifically, one might pinpoint transcriptional programs for invasion, for androgen independence, or indicating the activation of a particular ‘cancer pathway’.
More than a dozen expression profiling studies have now been published on clinical specimens and preclinical models (Bubendorf et al., 1999; J.Xu et al., 2000; Dhanasekaran et al., 2001; Luo et al., 2001a;b; 2002; Mousses et al., 2001; Stamey et al., 2001; Waghray et al., 2001; Welsh et al., 2001; DePrimo et al., 2002; Ernst et al., 2002; LaTulippe et al., 2002; Rhodes et al., 2002; Shou et al., 2002; Singh et al., 2002; Varambally et al., 2002). Studies using clinical samples have focused on classification of prostate carcinomas and on identification of marker genes. All studies have yielded clear distinctions between metastatic prostate carcinomas, localized prostate carcinomas and benign tissues each, which could be separated by clustering algorithms. A crucial question is how reliable the patterns are and whether they can provide critical distinctions such as organ-confined versus progressive cancers or cases with only lymph-node metastases versus systemic disease. Two main strategies are now pursued. The first involves construction of new, smaller arrays with defined sets of genes that appear to be useful after the first round of profiling experiments and testing these on larger numbers of cases in retrospective and prospective studies. The second strategy aims at implementing single gene markers that stand out from the mass of data as particularly strongly and reproducibly altered. Two such markers have already been developed. The cell surface protease hepsin was identified as strongly overexpressed in prostate cancers in several independent expression profiling studies. Having been confirmed by immunohistochemistry, this marker is expected to rapidly enter routine diagnostics (Luo et al., 2001b). Hepsin may not represent a good therapeutic target, since down-regulation increased the tumorigenicity of a prostate carcinoma cell line (Srikantan et al., 2002). An even better marker seems to be the P504S protein identical to -methyl-acyl-coA racemase (AMACR), a peroxisomal enzyme involved in the metabolism of branched chain amino acids. Its utility also appears to lie mostly in differentiating prostate carcinoma from benign changes such as atrophy and hyperplasia during initial diagnosis (Jiang et al., 2001; Rubin et al., 2002).
|
Epigenetics |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Stable changes in gene expression can be caused by alterations in the DNA sequence such as point mutations, gain or loss. However, during normal development, stably inherited states of gene expression are almost exclusively established by epigenetic mechanisms, i.e. without alterations in the DNA sequence. Many mechanisms of epigenetic inheritance involve establishment of specific chromatin states which can be maintained during cell proliferation.
A well-studied epigenetic mechanism is DNA methylation, in mammals restricted to cytosines in the dinucleotide CpG. It does not alter the coding potential of DNA, but alters its interaction with transcription factors and often gives rise to inactive chromatin states. Conversely, inactive chromatin tends to attract DNA methylation. Therefore, gene silencing is often associated with increased methylation of regulatory sequences. Significant changes in DNA methylation patterns occur in prostate cancer. Several genes are hypermethylated in >70% of carcinomas, notably GSTP1 encoding the glutathione transferase 
isozyme, which is accordingly down-regulated (Lee et al., 1994; Millar et al., 1999; Santourlidis et al., 1999; Lin et al., 2001). GSTP1 hypermethylation assays can be used for detection of prostate carcinoma cells in biopsies, urine, ejaculate, and serum (Goessel et al., 2001; Jeronimo et al., 2001). As for gene expression profiles, it is hoped that patterns of methylation changes can be used to classify prostate carcinomas (Adorján et al., 2002; Maruyama et al., 2002). If so, the stability of DNA compared to RNA would be an advantage. Moreover, inhibition of DNA methylation might elicit re-expression of genes silenced by this mechanism including several presumed tumour suppressors.
Paradoxically, while DNA methylation is increased at specific sites in the genome, that of repetitive sequences, which in normal cells contain the bulk of methylcytosine, can be decreased. In prostate carcinoma this ‘global hypomethylation’ appears to be closely related to tumour progression (Bedford and van Helden, 1987; Santourlidis et al., 1999). It is more frequent in high-stage and lymph-node positive carcinomas and ubiquitous in metastatic carcinomas recurring during androgen depletion therapy. Its causes are unknown, but the decondensation of chromatin ensuing from diminished methylation of repetitive sequences may contribute to chromosomal instability during tumour progression (Ehrlich, 2002). Accordingly, global hypomethylation in prostate cancer correlates with the number of chromosomal losses and gains, and particularly well with alterations of chromosome 8 (Schulz et al., 2002).
Other epigenetic mechanisms are more difficult to observe in cancer tissues. Still, stable down-regulation of genes such as NKX3A and PTEN and stable up-regulation of others, e.g. EGFR, in prostate cancer may be caused by epigenetic mechanisms. Indeed, treatment of prostate cancer cell lines with inhibitors of histone deacetylation, which cause chromatin decondensation, very efficiently induces cell death, usually by apoptosis (Carducci et al., 1996; Maier et al., 2000). Moreover, several chromatin proteins are differentially expressed in prostate cancer, including coactivators and corepressors interacting with the AR (Yeh and Chang, 1996; Muller et al., 2000; Gregory et al., 2001; Sharma and Sun, 2001). The EZH2 protein, a member of the polycomb family mediating gene silencing, emerged from a microarray study as a gene overexpressed in metastatic prostate cancers (Varambally et al., 2002).
|
Prospects |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
In recent years, knowledge on the molecular biology of prostate cancer has vastly expanded and research has moved far beyond androgen signalling and oncogenes and tumour suppressors important in other cancers. Expression profiling has further increased the mass of available data. Other high-throughput methods for genotyping, DNA methylation analysis and proteomics are applied to prostate cancers as well. Genes, RNA, and protein markers derived from these analyses are being introduced to improve detection of prostate cancer and treatment selection in a pragmatic manner and inspire novel therapeutic strategies.
However, these important steps should not detract from the recognition that fundamental questions regarding prostate cancer are unanswered. Most of all, a comprehensive model integrating the growing bulk of molecular data into a biological context is lacking. This is urgently needed to understand the widely different clinical behaviour of individual carcinomas. As argued throughout the present review, this model cannot simply be borrowed from another cancer type, but has to consider the peculiarities of prostate carcinoma. Some of the open questions are as follows.
Are there tumour stem cells in prostate carcinoma? A stem cell model proposed by Isaacs (1999) elegantly accounts for the reaction of prostate carcinomas to androgen depletion by assuming that the tumour—such as normal prostate epithelium—contains a mixture of androgen-independent stem cells and differentiated androgen-dependent derivatives. Androgen depletion kills most of the latter, but they are substituted by new progeny from the stem cells, often mutated to be more resistant to depletion. Genes, proteins and pathways involved in stem cell maintenance and expansion are becoming better understood. What is their state in prostate carcinoma? Are the mutations observed in the WNT pathway significant in this regard? In this context, it is interesting to consider the success of Gleason scoring, which has proven to be one of the best predictors of the clinical course of prostate cancer, even from biopsies. What does this rough estimate of tissue disorganization reflect? Figure 2 illustrates a very hypothetical explanation integrating Isaacs’ ideas.
|
What drives carcinogenesis in the prostate? Clearly, lifestyle factors are important, most likely diet. But how do they act on prostate cancer? There is little evidence for chemical mutagens to be involved, but much more for an endogenous process. Many factors that protect against oxidative stress also seem to protect against prostate cancer. Is this the decisive process? Do genetic alterations in prostate cancer bear the fingerprint of mutagenesis by oxidative stress? If so, the road to prevention might be straightforward. A widely discussed alternative hypothesis is that hormonal changes in the ageing male make the prostate susceptible to carcinoma. Accordingly, differences in the incidence between different populations are ascribed to the protective effect of phytoestrogens. This hypothesis needs an explanation why prostate carcinomas appear at a stage of life when androgen levels decline.
The growth characteristics of prostate cancer suggest tumour growth to be initially caused by decreased apoptosis. While some candidate proteins and pathways have been proposed that might stimulate proliferation and inhibit apoptosis in late stage prostate carcinoma, such as following androgen depletion, it is not clear what leads to decreased apoptosis in the initial stages.
Why is prostate cancer so often multifocal, even with different genetic alterations in individual foci? It is possible that carcinogenesis in the ageing prostate is very efficient, perhaps because of insufficient protection against carcinogens. The inactivation of GSTP1 has been discussed in this context. However, while detectable in high grade PIN, GSTP1 inactivation does not take the form of a ‘field change’ and, in fact, distinguishes tumour tissue from adjacent normal prostate. Still, a pre-neoplastic state in the ageing prostate might result from other epigenetic changes.
There is in fact substantial evidence that many initial changes in prostate carcinoma are epigenetic. So, for how much of prostate carcinoma development do epigenetic changes account? How efficacious is treatment aiming at epigenetic changes? Is the sensitivity of prostate carcinoma cell lines to inhibitors of DNA methylation and histone deacetylases representative of tumour tissues?
An important role of epigenetic mechanisms in prostate cancer would obviously solve the dilemma on the non-classical behaviour of tumour suppressors detailed above and may help to explain some of the oncogene up-regulation phenomena. Nevertheless, progression of prostate carcinoma to a metastatic phenotype is clearly associated with aneuploidy and chromosomal alterations. As in many other cancers it is unclear what drives chromosomal instability. As indicated above, the onset of chromosomal instability may coincide with that of loss of cell cycle coordination which has been shown to account for chromosomal instability in model experiments.
Finally, provocatively, one may ask whether metastatic prostate cancer and typical localized cancers are really the same disease? Again, it is remarkable how well the degree of tissue disorganization detected by Gleason scoring in biopsies predicts systemic disease. In addition, substantial differences between metastatic and localized cancers have been identified by studies on individual genes such as p53, PTEN and EGFR, as well as expression profiling. Should these differences be considered as reflecting a sequence of events or separate classes of prostate cancers? For instance, one class may establish genomic instability, cell cycle deregulation and invasive gene expression programs early on, whereas a second class may acquire these changes successively and slowly, if at all. These classes may correspond to tumours with high and low Gleason scores respectively.
With many findings on prostate carcinoma already at hand and a wealth of novel methods for analysing the molecular biology of human tumours, there are good prospects that these open questions will be answered over the coming years.
|
REFERENCES |
|---|
|
TopABSTRACTIntroductionProstate structure and...Natural history of prostate...Androgens in prostate cancerHereditary prostate cancerChromosomal alterationsTumour suppressorsOncogenesCancer pathwaysExpression profilingEpigeneticsProspectsREFERENCES |
|---|
Abate-Shen, C. and Shen, M.M. (2000) Molecular genetics of prostate cancer. Genes Dev., 14, 2410–2434.
Adorján, P., Distler, J., Lipscher, E., Model, F., Müller, J., Pelet, C., Braun, A., Florl, A.R., Gütig, D., Grabs, G. et al. (2002) Tumour class prediction and discovery by microarray-based DNA methylation analysis. Nucleic Acids Res., 30, e21.
Alers, J.C., Rochat, J., Krijtenburg, P.-J., Hop, W.C.J., Kranse, R, Rosenberg, C., Tanke, H.J., Schröder, F.H. and van Dekken, H. (2000) Identification of genetic markers for prostatic cancer progression. Lab. Invest., 80, 931–941.[ISI][Medline]
Balk, S.P., Ko, Y.-J. and Bubley, G.J. (2003) Biology of prostate-specific antigen. J. Clin. Oncol., 21, 383–391.
Banham, A.H., Beasley, N., Campo, E., Fernandez, P.L., Fidler, C., Gatter, K., Jones, M., Mason, D.Y., Prime, J.E., Trougouboff, P. et al. (2001) The FOXP1 winged helix transcription factor is a novel candidate tumor suppressor gene on chromosome 3p. Cancer Res., 61, 8820–8829.
Bedford, M.T. and van Helden, P.D. (1987) Hypomethylation of DNA in pathological conditions of the human prostate. Cancer Res., 47, 5274–5276.
Bednarek, A.K., Laflin, K.J., Daniel, R.L., Liao, Q., Hawkins, K.A. and Aldaz, C.M. (2000) WWOX, a novel WW domain-containing protein mapping to human chromosome 16q23.3–24.1, a region frequently affected in breast cancer. Cancer Res., 60, 2140–2145.
Berchem, G.J., Bosseler, M., Sugars, L.Y., Voeller, H.J., Zeitlin, S. and Gelmann, E.P. (1995) Androgens induce resistance to bcl-2-mediated apoptosis in LNCaP prostate cancer cells. Cancer Res., 55, 735–738.
Bratt, O. (2002) Hereditary prostate cancer: clinical aspects. J. Urol., 168, 906–913.[CrossRef][ISI][Medline]
Bubendorf, L., Kolmer, M., Kononen, J., Koivisto, P., Mousses, S., Chen, Y., Mahlamaki, E., Schraml, P., Moch, H., Willi, N. et al. (1999) Hormone therapy failure in human prostate cancer: analysis by complementary DNA and tissue microarrays. J. Natl Cancer Inst., 91, 1758–1764.
Cabeza-Arvelaiz, Y., Thompson, T.C., Sepulveda, J.L. and Chinault, A.C. (2001) LAPSER1: a novel candidate tumor suppressor gene from 10q24.3. Oncogene, 20, 6707–6717.[CrossRef][ISI][Medline]
Cairns, P., Okami, K., Halachmi, S., Halachmi, S., Esteller, M., Herman, J.G., Jen, J., Isaacs, W.B., Bova, G.S. and Sidransy, D. (1997) Frequent inactivation of PTEN/MMAC1 in primary prostate cancer. Cancer Res., 57, 4997–5000.
Carducci, M.A., Nelson, J.B., Chan-Tack, K.M., Ayyagari, S.R., Sweatt, W.H., Campbell, P.A., Nelson, W.G. and Simons, J.W. (1996) Phenylbutyrate induces apoptosis in human prostate cancer and is more potent than phenylacetate. Clin. Cancer Res., 2, 379–387.[Abstract]
Carpten, J., Nupponen, N., Isaacs, S., Sood, R., Robbins, C., Xu, J., Faruque, M., Moses, T., Ewing, C., Gillanders, E. et al. (2002) Germline mutations in the ribonuclease L gene in families showing linkage with HPC1. Nat. Genet., 30, 181–184.[CrossRef][ISI][Medline]
Chan, J.M., Stampfer, M.J., Giovannucci, E., Gann, P.H., Wilkinson, P., Hennekens, C.H. and Pollak, M. (1