Department of Medicine, Division of Endocrinology, Veterans Affairs Medical Center and Baylor College of Medicine, Houston, Texas 77030
Address correspondence and requests for reprints to: Glenn R. Cunningham, Department of Medicine, Division of Endocrinology, Veterans Affairs Medical Center, 2002 Holcombe Boulevard, Houston, Texas 77030.
THE PROSTATE is the site of two common diseases in aging men, benign prostatic hyperplasia (BPH) and prostate cancer (CaP). Although mortality rates due to BPH are less than 0.5 per 100,000 population, this disease affects the quality of life and health care costs exceed 4 billion dollars per year. If a man in the United States lives to be 80 yr of age, there is a 25% chance that he will require surgical treatment for BPH. In 1999, CaP will be diagnosed in 179,000 American men and it will cause 37,000 deaths. This represents the second most common cause of cancer deaths in men.
Epidemiology
Epidemiology of BPH and CaP: genetic vs. environmental causes
Histological BPH increases with age in all ethnic groups. By age 80, it is present in more than 80% of men. In contrast, mean prostate weight at autopsy or prostate volume in community-based studies increases after age 50 or 60. Comparing 7079-yr-old men, 93% of men in Olmstead County, Minnesota, but only 37% of Japanese men had volumes exceeding 20 mm3 (1). This suggests that there may be genetic differences among ethnic groups. However, greater prostate volumes in urban-dwelling Chinese than in rural-dwelling Chinese suggest that environmental factors contribute to prostate volume, and twin studies in Utah indicate that both genetic and environmental factors contribute to prostate volume.
Population-based studies of prostate enlargement in Africans and African-Americans are lacking. Most recent data using National Hospital Discharge Data Survey indicate that the percentage of African-American men undergoing prostatectomy for prostate enlargement is similar to the percentage of whites when adjusted for age (2). This suggests that the prevalence of prostate enlargement in African-American men is similar to that in white American men.
In 1999, CaP will account for 28.7% of the total number of new, nonskin cancers, but only 12.7% of cancer-related deaths for American males. These numbers imply that the large majority of affected men will die with this disease, rather than from it. This is, in part, due to the age at the time of diagnosis, to the low degree of malignancy of many CaP, and to the effectiveness of treatment. The epidemiology of CaP is complicated by the fact that 1530% of men over the age of 50 have histological evidence of CaP at autopsy (latent CaP) (3). The factors responsible for causing a latent CaP to become a clinical CaP are poorly understood.
The prevalence and mortality rates from CaP vary among different ethnic groups. They are higher in African-Americans (224 and 55/100,000 persons/yr), intermediate among whites (150 and 24/100,000), and lower among Orientals living in Asia (82 and 10/100,000) (4). This is in sharp contrast with the prevalence of latent CaP, which is roughly the same worldwide (5).
Thus, some ethnic differences in the prevalence of prostate enlargement and significant ethnic differences in the incidence of clinical CaP seem to provide opportunities for learning more about the pathogenesis of these important diseases. Potential differences in sex steroid levels have been the focus for many studies.
The development, growth, and differentiation of the normal prostate are
regulated by androgens, acting through the androgen receptor (AR). The
concepts that a normal androgen-signaling pathway is required for the
development of a normal prostate and for the development of BPH and CaP
come from the following observations: 1) abnormal androgen biosynthesis
and inactivating mutations of the AR and 5-reductase type II genes
are associated with absent or rudimentary development of the prostate
in 46-XY individuals; 2) there is anecdotal evidence that BPH and CaP
do not develop in men who underwent castration around the time of
puberty; 3) androgen can induce experimental prostate enlargement or
cancer in susceptible animals.
Although other hormones, such as insulin-like growth factor (IGF)-I, vitamin D, KGF, estradiol (E2), and PRL may affect the prostate, this review focuses mainly on the androgen-signaling pathway. Our aim is to critically assess the most important hormonal mechanisms related to the androgen signaling pathway that may be involved in the etiology and in the response to treatment of these two common disorders.
Sex steroids
It would be very convenient if one could correlate serum levels of sex steroids with risks of developing BPH or CaP. Although there are many studies, their significance is limited by several factors: 1) to establish linkage between development of a disease and certain steroid hormone levels, studies need to be prospective and should last until the disease of interest has developed; 2) circulating hormone levels vary significantly throughout the day and over the years. It is difficult to determine to what extent the level of a hormone measured once reflects a lifetime index of hormone status; and 3) the levels of circulating steroid hormones do not necessarily reflect their intraprostatic concentrations.
Studies have been performed to see whether there is a correlation between sex steroid levels at various ages and the prevalence of prostatic diseases in the various ethnic groups. It seems that African-American fetuses are exposed to somewhat higher levels of testosterone (T) than white fetuses. Additionally, young adult African-American men have higher circulating levels of T than age-matched Caucasian-American men do. However, levels in young Japanese men are similar to Caucasians.
Ethnic differences in male hair distribution suggest differences in
serum androgen levels or in androgen effects on androgen-sensitive hair
follicles and, perhaps, the prostate. Dihydrotestosterone (DHT) is the
principle androgen for both. Hair follicles are known to have high
levels of 5 reductase type 1 activity, and the prostate is known to
have high levels of 5
reductase type 2 activity. Lookingbill
et al. (6) evaluated 184 Caucasian and Chinese subjects.
Mean chest hair scores using a scale of 04 were 3.0 vs.
0.8 (P < 0.0001), Caucasian vs. Chinese.
Serum levels of total and bioavailable T did not differ, but serum
levels of precursor androgens, dehydroepiandrosterone sulfate, and
androstenedione were higher in Caucasians. Levels of 5
-reduced
androgen products, AAG and androsterone glucuronide, were higher in
Caucasians. A follow-up study sought to determine whether the lower
5
-reduced products in Chinese were caused by a reduction in
precursor steroids or decreased 5
-reductase activity (7). They found
no evidence of reduced 5
-reductase activity, but there was
significant reduction in T production rates, plasma T, and sex hormone
binding globulin (SHBG) levels in Chinese men living in Beijing
compared with Chinese men living in Pennsylvania. No differences in T
levels or metabolism were noted between Chinese and Caucasians
living in Pennsylvania. These studies indicate differences in sex
steroid levels among Chinese and Caucasians are due more to
environmental factors than genetic factors. It is likely that
differences in diets contribute to the differences because a diet
higher in polyunsaturated fat intake and fiber is known to decrease
serum total and free T levels.
If the differences in hair patterns in Orientals and Caucasians are not
due to differences in circulating levels of androgens or to differences
in 5-reductase activity, it would seem that they must be related to
differences in the AR or to postreceptor signaling. Differences in AR
or in postreceptor signaling also could contribute to the ethnic
differences in the incidence of BPH and CaP.
BPH
Another approach is to compare sex hormone levels in men who develop prostate enlargement or CaP with men from the same ethnic group who do not develop these diseases. Many investigators have evaluated sex steroid levels at the time when prostate enlargement was diagnosed and/or at the time of surgery for lower urinary tract symptoms and prostate enlargement. It should be noted that prostate size correlates poorly with symptoms and urine flow rates; however, men with prostate enlargement are three to four times more likely to undergo surgical intervention when followed for several years (8). Although some studies have noted differences in one or more hormonal parameters, consistent hormonal differences have not been observed within an ethnic group at the time when prostate enlargement was diagnosed.
The relationship between plasma hormone levels obtained before development of clinical symptoms and subsequent surgical treatment for BPH has been evaluated in 320 men participating in the Physicians Health Study (9). An additional 320 men who did not undergo prostatectomy served as controls. The plasma specimens were obtained up to 9 years before surgery. Using multivariate analyses, they found a strong trend for increasing risk across quintiles of E2 (Q5 vs. Q1 OR = 3.56). This excess risk was confined to men with lower androgen levels. Plasma T was not correlated with subsequent prostate surgery.
CaP
Sex steroid hormone levels and CaP have been correlated in two prospective studies. Serum levels for T, DHT, and AAG in 59 men who developed CaP between 1973 and 1994 were compared with those in 180 patients who were CaP-free in 1994 (10). No association was detected when the initial levels of these hormones were individually correlated with CaP risk. In another study, levels in 222 men who provided plasma in 1982 and developed CaP by 1992 were compared with those in 390 controls (11). When T and SHBG levels were adjusted simultaneously, CaP risk correlated positively with levels of T and negatively with levels of SHBG. Additionally, serum levels of DHT and AAG did not seem to be correlated to CaP risk.
Epidemiological observations suggest the following factors influence the development of CaP: diet, genetic susceptibility, sexually transmitted agents, environmental carcinogens and endocrine function. There is good evidence for inherited and genetic components (12); however, the fact that Japanese and Chinese males migrating to the United States experience a sharp increase in CaP prevalence in the first and second generation, indicates the importance of environmental factors. There is a strong correlation between per capita fat intake and CaP incidence and mortality; however, saturated fat is thought to account for only 1015% of the differences in CaP incidence. Perhaps, total calorie intake rather than fat intake, per se, should be considered. Restriction of energy intake in mice inhibits prostate tumor growth by inhibiting tumor angiogenesis. The model proposed by these authors predicts that total energy intake rather than fat may be a determinant for CaP growth.
Pathogenesis
Components of the AR signaling pathway
Many components of the androgen-signaling pathway have been
investigated to identify potential mechanisms responsible for the
development of macronodular BPH and CaP. Since the 1930s, it has been
recognized that T is the principal circulating androgen secreted by the
testis. However, DHT is the predominant androgen bound to the AR in the
nuclei of target cells in the prostate. In the prostate, T is converted
to DHT by the enzyme steroid 5-reductase type II (SRD5A2). The
effects of androgen are mediated by AR, a member of the
steroid-thyroid-retinoid superfamily of nuclear receptors.
Nonligand-bound AR is thought to be in the cytoplasm of target cells
complexed to a number of ancillary protein members of the heat shock
protein family. Binding of hormone is followed by dissociation from
heat shock proteins, phosphorylation, dimerization, and binding to
specific DNA sequences within or adjacent to androgen-responsive genes.
The ligand-activated receptor regulates transcription of
androgen-dependent genes. The AR contains two microsatellites
consisting of poly-Q and poly-G repeats in exon 1. The length of these
two repeats is highly polymorphic in the general population.
Investigators agree that an expanded poly-Q tract is associated with
reduced transcriptional activity (13). This modification is associated
with normal 3H-DHT binding ability. This initial
observation was followed by studies in which differences in the size of
the poly-Q repeat were correlated to the transcriptional activity of
AR. Some (14), but not all authors (15), have concluded that there is
an inverse correlation between poly-Q size and transcriptional
activity. The poly-Q repeats exert an inhibitory effect on
transcription, either directly (14), or indirectly by affecting AR mRNA
stability (16).
Pathogenesis of BPH
Investigators have proposed several hypotheses to explain the development of pathological prostatic enlargement. Sex steroids are invoked directly or indirectly in each. Berry et al. observed that pathological prostate enlargement is present only when there is histological BPH. Assuming that this observation is correct, one might consider what causes micronodular BPH and what is responsible for the development of macronodular BPH. The pathogenesis of each may be different. In this respect, micronodular BPH is like latent CaP and prostate enlargement or macronodular BPH is like clinical CaP. It is useful to note that macronodular disease primarily involves the stroma.
The increase in prostate volume seen with BPH is caused by cellular hyperplasia and reduced apoptosis. Stereological analysis indicates that the relative volume of stroma in macronodular BPH is increased 33% over that in the normal prostate. These findings have been confirmed by quantitative morphometry in men with symptomatic BPH (17). Apoptosis is reduced and the life span of BPH stromal cells is increased. Using estimates of superoxide dismutase activity, the average life span of BPH stromal cells has been estimated to be more than 30 yr (18). A recent study found the mean proliferation index in the epithelium (0.142 ± 0.097) and stroma (0.121 ± 0.082) to be similar (19). The apoptotic rate in epithelium (0.172 ± 0.156) was similar to the proliferative rate, but no apoptotic cells were detected in the stroma. These investigators conclude that stromal growth in BPH is due to cell proliferation in the absence of cell death.
Androgen hypothesis. Because androgen is required for prostate
development, it has been proposed that increased prostatic
concentrations of androgen or increased androgen activity causes
macronodular BPH. BPH tissue has increased 5-reductase activity
(20); however, the mRNA for both isoenzymes in BPH tissue were
decreased when compared with levels in normal prostates from younger
and older normal men. Initial studies indicated that BPH tissue
contained higher concentrations of DHT, but Walsh et al.
(21) demonstrated that differences were artifacts related to metabolism
of DHT in cadaveric tissue before the time of assay (21). DHT levels in
epithelium from normal and BPH tissue decreased with increasing age of
the subject (22); however, stromal DHT in both normal and BPH tissues
was not affected by age. T levels did not change with increasing age.
The periurethral zone is said to have higher concentrations of T, DHT,
and AAG than intermediate and subcapsular zones (23). Thus, current
evidence indicates that macronodular enlargement of the prostate is not
caused by an increase in DHT, but DHT may facilitate its
development.
AR mRNA, AR levels, and localization and AR transcriptional activity have been assessed in BPH tissue. Both higher and lower levels of AR mRNA have reported in BPH than in normal prostatic tissue. Using binding assays, AR levels in BPH have been reported to be unchanged or increased. Immunostaining demonstrates that AR is present in both BPH luminal epithelial cells and stromal cells. Nuclear concentrations of AR are higher in the periurethral zone (23). Using genomic DNA from 310 men who had lower urinary tract symptoms and an enlarged prostate by digital rectal examination and who underwent prostate surgery and genomic DNA from 1041 controls (24), risk of surgery decreased linearly with decreasing AR CAG repeat. Comparing CAG repeat length of 19 or less to 25 or more, the odds ratio for prostate surgery was 1.76 with a confidence interval of 1.16 to 2.65. Thus, there is evidence for increased AR concentration in the periurethral area (where BPH macronodules develop) and for shorter poly-Q repeats in patients who undergo surgery for BPH.
Estrogen hypothesis. Support for the role of estrogen is based mainly on observations that E2 potentiates the effects of androgen in inducing BPH in dogs, apparently by increasing AR levels and, perhaps, by other mechanisms (25), and the fact that the serum estrogen to T ratio increases with age in men. Prostatic conversion of T to E2 or of androstenedione to estrone also could contribute to higher prostate concentrations of estrogen. Aromatase mRNA can be detected by RT-PCR technique in most BPH specimens. Stromal, but not epithelial, tissue levels of E2 and estrone increase with age, resulting in a very significant increase in the estrogen/androgen ratio in the stroma (22). Estrogen receptor (ER) and its mRNA have been localized primarily in the stroma. The recent identification of a second ER subtype (ERß) in the rat prostate requires careful analysis of human pathology. It is of considerable interest that prostatic hyperplasia is observed in aging ERß knockout mice (26). Clinical trials with the aromatase inhibitor, atamestane, did not reduce prostate size. Thus, available data are suggestive, but they do not support a strong role for estrogen in the development and maintenance of human macronodular BPH.
Dysregulation of growth factors hypothesis. Autocrine and paracrine growth factors are considered to be the factors that directly mediate enlargement of the prostate. Many of these growth factors are at least partly regulated by sex steroids. Dysregulation of growth factor production and secretion has been compared with embryonic growth potential of prostatic stroma. Recent studies of the pathogenesis of BPH have focused on excessive growth factors such as bFGF, KGF, EGF, IGF-I, and IGF-II and their receptors (27). IGF-I, IGF-II, and IGF-IR increase with aging in the presence or absence of BPH. Therefore, they may not be a major stimulus for stromal proliferation, but there is evidence that this pathway inhibits apoptosis. At this time it is not clear whether the increases in these growth factors are dependent or just facilitated by the androgen-signaling pathway.
Stem cell hypothesis. The stem cell hypothesis suggests that hyperplasia results from an increase in the number of stem cells (28). Castration of dogs at an early age and replacement after adulthood reduces prostate size from its normal adult size. Early exposure to androgen or estrogen may affect the number of stem cells, and this could determine the number of potential clones and potential prostate size. Androgen chronically stimulates proliferation and prevents cell death of the luminal epithelial cells. An increased clonal expansion number, a decreased rate of apoptosis or a combination of the two would increase prostate size. Presumably, such changes would result in a diffuse enlargement of the prostate. Because BPH is localized to the transition zone and consists primarily of stromal nodules, it is not clear how relevant these observations are to the development of macroscopic BPH. Recent studies suggest that stem cell expansion can be caused by decreased expression of p27Kip1 and its protein (29). This, of course, could cause hyperplasia and/or neoplasia.
In summary, there is considerable evidence that the androgen-signaling pathway contributes to the development of BPH. Differences in androgen signaling and estrogen signaling during normal prostate development may be important. Androgen seems to be required for pathological prostate enlargement, and it is likely that its effects on cells are mediated by growth factors. The growth factors act by both autocrine and paracrine mechanisms to enhance cell proliferation and reduce apoptosis. Estrogen may contribute to this process. Whereas both epithelial and stromal cells contribute to prostate enlargement, the increase in stromal cells dominates in most glands with macronodular BPH.
Pathogenesis of CaP
SRD5A2. Population-based screening assays for the presence of SRD5A2 mutations in various ethnic groups have correlated CaP risk with the prevalence of certain SRD5A2 germ line mutations. At least two mutations of this gene have been identified: V89L (30) and A49T (31). Interestingly, the prevalence of V89L, which reduces activity by 33% from wild type, is more common among Asian men (in whom prevalence of CaP is the lowest) than among African-Americans and Caucasians (in whom prevalence of CaP is higher). On the contrary, A49T, which creates an enzyme with a 5-fold higher Vmax for T conversion, is more common among men with advanced CaP. Thus, these authors suggest that polymorphisms in the SRD5A2 can affect enzymatic activity, intraprostatic levels of DHT and risk of developing CaP.
AR expression and mutations. AR is expressed in all stages of CaP evolution, including prostatic intraepitheilial neoplasia, primary and metastatic disease both before and after androgen ablation therapy. Few CaP are AR negative, and, thus, even the androgen-independent tumors express the AR protein. In view of the immunodetection of AR in every stage of CaP, investigators have tried to determine whether mutations of this molecule correlate with the development of CaP. Mutations have been detected only in 5 of 231 cases (frequency 2.1%) of stage A-B CaP (32). In addition, in our unpublished series (Marcelli et al., unpublished data) of 99 stage B patients, no AR mutation was found. Thus, a 0.6% prevalence of AR mutations in the combined 330 cases excludes AR mutations as a common mechanism in the pathogenesis of CaP. Takahashi et al. (33) proposed that inactivating mutations of AR, which are frequent in latent CaP in Japanese men and absent in American men, would prevent the evolution of latent CaP into a clinically detectable entity in this ethnic group (33).
AR repeats. Changes in the size of the poly-Q repeat and altered transcriptional activity could result in increased AR activity. When the number of Q repeats was compared with the incidence of CaP in the general population, there was an association between an increased risk of developing the disease and a shorter poly-Q tract. In addition, some studies have also detected an association between a shorter poly-Q repeat and the presence of metastatic disease, high histological grade, and younger age of onset. Interestingly, there is also a correlation between the shorter number of Qs and the increased incidence, higher mortality and more aggressive nature of CaP in the African-American population.
A correlation between a shorter poly-G repeat and CaP risk has also been identified by some, but not other authors. However, because AR with a reduced poly-G tract has reduced transcriptional activity, it is not clear how such receptor may predispose to CaP.
In summary, the data are suggestive only for a permissive role of the androgen-signaling pathway in the pathogenesis of CaP. They identify a weak relationship between the prevalence of SRD5A2 mutations and of CaP in various ethnic groups. However, it is not clear whether this has any clinical consequences. They indicate that AR mutations do not play a significant role in the pathogenesis of CaP. Finally, they identify a correlation between changes in the size of the poly-Q repeat of the AR with CaP prevalence. However, experiments looking at the transcriptional competence of AR molecules with various Q repeats have been performed only in cell lines. It is not known if they have any significant effect in vivo. This very important question will be answered by the development of transgenic models carrying AR molecules with various length of the Q repeats.
Treatment based on hormonal manipulation
Antihormonal treatment of BPH
Available medical therapies are only modestly effective for mild
or moderately symptomatic BPH. Once macronodular BPH is established,
agents that decrease DHT levels (34), T and DHT levels (35), or
androgen effects reduce prostate volume only by 2030%, diminish
symptoms moderately and increase urine flow only by an average of less
than 2 mL/sec. Reduction of prostatic DHT levels can prevent
progression of the disease (36); however, lowering prostatic DHT does
not provide good therapy for moderate to severe BPH. The reasons for
the limited effectiveness of therapies that reduce DHT and/or T are
poorly understood. Although AR are present in the secretory epithelial
and stromal cells of the prostate, only the former appear to undergo
apoptosis in patients treated with a 5-reductase inhibitor (36).
Thus, 5
-reductase inhibitors may be more useful in preventing rather
than treating enlargement of the prostate.
Antihormonal treatment of CaP
Men with metastatic CaP have been treated with androgen ablation since the seminal observation that removal of androgen by castration induces temporary remission of the disease in as many as 80% of the patients. After a median response of 1218 months, CaP cells acquire the ability to proliferate in the presence of a very low concentration of androgen. Unfortunately, 70% of men with androgen-independent tumors eventually die of CaP. Thus, understanding the molecular basis of androgen independence in CaP is one of the most compelling problems in this field of research.
Attempts to explain the molecular basis of androgen independence have focused on the various components of the androgen-signaling pathway. In view of the immunochemical detection of AR in most samples of androgen-independent CaP, many of these hypotheses are centered on abnormalities of this molecule (for references see, Ref. 32): 1) amplification of the AR gene, which would facilitate tumor growth at very low concentrations of the ligand; 2) AR mutations associated with a phenotype showing: a) a superactive AR; b) a promiscuous receptor protein that is activated by ligands other than androgen; c) inactivating mutations of the AR, generating undifferentiated CaP cells with a malignant and aggressive phenotype; d) changes in the size of the polyglutamine repeat of AR that may increase response to stimulation with androgen; and e) activation of growth-stimulating pathways with the ability to bypass AR-regulated growth and differentiation; 3) increased (or decreased) intraprostatic bioavailability of DHT to activate AR; 4) androgen-independent activation of AR; 5) activation of AR by alternative signaling mediated by interaction with various coactivators; 6) overexpression of molecules mediating androgen insensitivity, such as caveolin; and 7) accumulation of antiapoptotic molecules after androgen ablation.
Most of these hypotheses are supported by significant scientific evidence. However, at present, it is not clear whether some of them, such as androgen-independent activation of AR, involvement of co-activators, molecules mediating androgen-independence or pathways bypassing AR-regulated growth, are clinically significant. AR amplification (37) or AR mutations have been detected in a cohort of patients with advanced CaP (38, 39). In our own series, we have detected mutations in 8 out 38 specimens of metastatic prostate cancer (Marcelli et al., unpublished data). Analyses of cell lines transfected with these mutant AR have elucidated the functional phenotypes of some of them and possible mechanisms of androgen-independent growth. Thus, based on these correlations, one could conclude that AR mutations might facilitate metastatic growth and/or prevent response to antihormonal treatment. Nevertheless, it remains to be demonstrated using in vivo models that these mutations play an active biological role in the progression of CaP.
Conclusions
Serum sex steroids within the normal range do not correlate with development of prostatic enlargement and CaP; however, BPH and CaP probably do not occur in males who develop androgen deficiency before age 20 if they do not receive androgen replacement. Although most investigators think that androgens facilitate the development of macronodular BPH and CaP, the evidence indicates that most of these patients have normal levels of androgens acting via a normal androgen-signaling pathway. The importance of environmental factors is suggested by differences in serum hormone levels between Chinese living in China and Chinese-Americans, differences in prostate size between rural- and urban-dwelling Chinese, differences in prostate size among older monozygotic twins, and the increased prevalence of CaP in second generation Chinese-American and Japanese-American men.
The possibility of an abnormal androgen-signaling pathway is more likely when CaP becomes androgen insensitive. Of the possible causes of androgen insensitivity, amplification or mutations of the AR gene and growth regulation by AR-independent pathways seem most likely.
Possibilities for effective antihormonal preventive therapy are better
than for antihormonal curative therapy. It is unlikely that a large
percentage of the male population will take a 5-reductase inhibitor
to prevent prostate enlargement. However, if the Prostate Cancer
Prevention Trial were to demonstrate that finasteride
caused a significantly lower incidence of both prostate
enlargement and CaP, many men would choose preventive therapy. Once
prostate enlargement or CaP is established, it is unlikely that
antihormonal therapy will be curative. A potentially better therapeutic
approach is to use the endogenous apoptotic pathway of the cell. We
have demonstrated the effectiveness of this approach in
vitro (40) (M. Marcelli et al., submitted for
publication). The challenge is to develop an effective targeting
system.
Acknowledgments
Space limitations precluded referencing all of the publications discussed.
Footnotes
1 Supported by a grant from the Veterans Affairs Medical
Center (to M.M. and G.R.C.) and NIH Grant CA68615 (to
M.M.).
Received July 21, 1999.
Revised August 12, 1999.
Accepted August 12, 1999.
References