Association between cadmium and androgen receptor protein expression differs in prostate tumors of African American and European American men

Abstract

Cadmium is a known carcinogen that has been implicated in prostate cancer, but how it affects prostate carcinogenesis in humans remains unclear. Evidence from basic science suggests that cadmium can bind to the androgen receptor causing endocrine disruption. The androgen receptor is required for normal prostate development and is the key driver of prostate cancer progression. In this study, we examined the association between cadmium content and androgen receptor protein expression in prostate cancer tissue of African American (N = 22) and European American (N = 30) men. Although neither overall tumor cadmium content (log transformed) nor androgen receptor protein expression level differed by race, we observed a race-cadmium interaction with regard to androgen receptor expression (P = 0.003) even after accounting for age at prostatectomy, smoking history, and Gleason score. African American men had a significant positive correlation between tumor tissue cadmium content and androgen receptor expression (Pearson correlation = 0.52, P = 0.013), while European Americans showed a non-significant negative correlation between the two (Pearson correlation = −0.19, P = 0.31). These results were unchanged after further accounting for tissue zinc content or dietary zinc or selenium intake. African American cases with high-cadmium content (> median) in tumor tissue had more than double the androgen receptor expression (0.021 vs. 0.008, P = 0.014) of African American men with low-cadmium level. No difference in androgen receptor expression was observed in European Americans by cadmium level (high 0.015 vs. low 0.011, P = 0.30). Larger studies are needed to confirm these results and if upheld, determine the biologic mechanism by which cadmium increases androgen receptor protein expression in a race-dependent manner. Our results suggest that cadmium may play a role in race disparities observed in prostate cancer.

Introduction

Cadmium (Cd) is a human carcinogen (i.e. lung cancer) and a potential risk factor for prostate cancer [1]. Findings from epidemiological studies [2–4], however, are inconsistent and have resulted in Cd rarely being listed among the more accepted prostate cancer risk factors, which include older age, African American race, family history of the disease, and living in a westernized nation[5]. A better understanding of the potential pathways through which Cd may act in prostate carcinogenesis could lead to improvements in epidemiological study design that may produce more consistent findings between Cd exposure and prostate cancer in humans.

Evidence from basic science studies suggests that Cd may play a role in prostate cancer through disruption of the androgen receptor (AR). AR, a hormone-activated transcription factor, is the key driver of prostate cancer progression [6]. Ironically, the AR is also required for normal prostate growth and development. Work in cell lines has shown changes in AR activity or AR expression when cells are exposed to Cd [7–10]. Studies have also shown that Cd can bind directly to the ligand binding domain of AR and can compete with dihydrotestosterone, the natural ligand of AR [8]. Binding of Cd has been shown to modify the conformation of the receptor [11], potentially changing its transcriptional potential. Epidemiological studies evaluating the relationship between Cd and AR have largely employed peripheral measures of Cd exposure (in blood, toenails, or urine) and proxy measures of AR function such as serum prostate specific antigen (PSA) level [12–15], a downstream target of the AR. Such approaches have yielded mixed results and have not confirmed a clear interaction between Cd exposure and AR signaling in the human prostate.

In the current study, we evaluated the association between prostate tissue Cd content and AR protein expression measured directly in tumors of African and European American men who underwent prostatectomy for treatment of prostate cancer.

Materials and Methods

Subjects

A subset of 59 men enrolled in the Gene-Environment Interaction in Prostate Cancer Study (GECAP) [16], which was conducted at Henry Ford Health System between 1999 and 2004, were included in this study. As part of the parent GECAP study [17, 18], cases completed a food frequency questionnaire [19] and a face-to-face interview to capture health behavior information, such as smoking history [20, 21]. Comprehensive clinical data were abstracted from medical records. Race was self-reported. Census tract median household income was derived from patient address at time of diagnosis using MapInfo Professional^®v7.0 and MapMarker^®v.8.1 software programs (MapInfo Corporation, Troy, NY) as previously reported. [22]. The 59 cases included in this sub-study had tumor tissue embedded on a single tissue microarray and had previous assessment of AR expression conducted at Columbia University (R.B.M., C.C.C., and A.R.). For the present study, tumor tissue was excised at Henry Ford Health System from the same formalin-fixed, paraffin-embedded tissue blocks that were used to produce the tissue microarray. The parent study and all sub-studies were approved by the human subjects committees of Henry Ford Health System, Columbia University, and Wayne State University.

Prostate Tissue Metal Measurement by Inductively Coupled Plasma Mass Spectrometry

For metals assessment, prostate tissue samples were re-reviewed by a urologic pathologist (O.N.K) using contemporary Gleason grading [23] and were prepared according to the methods reported by Sarafanov et al [24]. Of the 59 cases in this study (on the single tissue microarray), 52 had adequate remaining tumor tissue for evaluation of Cd content. For each tissue block, tumor tissue was macrodissected with a titanium knife, using metal-free methods and placed in tubes that had been prewashed with Optima grade nitric acid (Thermo Fisher Scientific, Waltham, MA) and MilliQ water and dried. Tissue was deparaffinized using hexane (Optima grade, Thermo Fisher) at 20°C for 1 week with frequent changes of solvent.

Prostate tissue metal levels were measured by inductively coupled plasma mass spectrometry (ICP-MS) in the laboratory of B.M. at Wayne State University School of Medicine. In short, samples were dried until a constant tissue weight was achieved. Approximately 5-10 mg of each sample was digested with concentrated nitric acid (high purity Optima Grade, Thermo Fisher) in prewashed and dried polypropylene tubes overnight at room temperature. The samples were then heated at 90°C for 20-30 minutes. They were then diluted with MilliQ pure water containing 1% Triton X-100 (Sigma-Aldrich, St. Louis, MO) to the appropriate range of metal concentrations. Metal measurements were carried out in a PE Sciex Elan 9000 ICP-MS (PerkinElmer, New York, NY) with a cross flow nebulizer and Scott type spray chamber. The radio-frequency power was set at 1000 W and the argon flow optimized at 0.92 L/min. The optimum lens voltage was centered on rhodium sensitivity. Standard external calibration curves were generated from stock solutions purchased from VWR (Radnor, PA). All samples included internal standards, Yttrium and Indium (VWR). The ICP-MS measurements were repeated on duplicate samples and reported in parts per million.

AR Immunofluorescence and Imaging

Commercially available primary antibodies and fluorescent secondary labels were used to measure AR protein expression within the tumor tissue specimens. The tissue labeling was optimized for the visualization of AR (AR441/M3562, Dako North America, Carpinteria, CA) positive epithelial nuclei, and incorporated labels for Cytokeratin-18 (CK18, Novocastra NCL-CK18) and 4′,6-diamidino-2-phenylindole (DAPI) to mark the epithelial and nuclear compartments, respectively. AR expression, as mentioned above, was previously measured in tissue microarrays produced for the entire GECAP prostatectomy cohort using the multiplex IF assay, which produces immunofluorescent profiles that can be studied quantitatively. This work was conducted at Columbia University (by R.B.M.). The method allows visualization and quantitation of AR levels in distinct microanatomical compartments, namely stroma cells, normal prostate epithelial cells, and tumor cells. Florescence imaging was done using combinations of 4 different emission and bandpass filters optimized to illuminate the different fluorescent secondary labels. Spectral segmentation was performed using a Nikon D90 microscope equipped with a CRI Nuance FX series multispectral camera and filter (Cambridge Research and Instrumentation, Inc., Woburn, MA). By applying spectral optics to separate the respective fluorochromes, individual gray-scale images were created for nuclei, epithelial cells, and AR. The primary focus was AR expression in tumor cells. A full description of the sensor resolution of the Nuance, the signal-to-noise ratios, and capacity to successfully distinguish similar spectral signals has been previously described [25–27].

For each tumor spot, a set of measurements were collected for each of the following signals (AR, CK18 and DAPI): average pixel intensity (1-256), standard deviation, total signal intensity, max intensity, and total pixel area. For each of the pixel area measures, we used threshold mapping to reduce background and improve specificity. Fixed thresholds for each signal were ascertained by taking the average auto-threshold level obtained from a random sample of 50 spots, spanning the entire dataset. The auto-thresholding was performed using the CRI Nuance software using an adaptation of the Otsu method that transforms the grey-scale TIFFs into binary images (signal vs. background), in which the optimum threshold between the 2 is where the intraclass variance is minimized [28, 29]. We also measured areas in which multiple signals were overlapping, or co-localized. The ability to capture information from overlapping signals enabled the evaluation of specific microanatomical compartments. The pixel areas corresponding to the overlap between AR CK18 and DAPI represent AR expression within epithelial nuclei (CK18+, DAPI+). This capability is particularly important in histological analyses of prostate cancer, because the tissues have a high degree of granularity and mixture between tumor epithelium and stromal cells, as well as differences in the amount of tumor present across different spots, which can introduce error into the expression measures.

Statistical Analysis

Tissue Cd and zinc (Zn) content and dietary Zn and selenium (Se) were log transformed. T-test or Mann-Whitney U test were used to assess between group differences for normally distributed (means) and non-normally distributed (medians) data. Pearson-correlations were used initially to assess relationships between exposures and AR expression. Stratified analyses were performed to evaluate association by race. Multivariable linear regression was performed to determine whether Cd content was associated with AR protein expression in prostate tumor tissue. Models included age at prostatectomy, race, dietary Zn or Se intake or tissue Zn content, ever smoking, and the Gleason score of the tumor foci. Tissue Zn and dietary Zn and Se were included in models as potential modifiers of Cd toxicity [13] and were log transformed. We also assessed whether race modifies the association between Cd and AR expression by including the interaction term race*Cd.

Results

Of the 52 cases, 22 self-identified as African American and 30 as European American. Age, dietary intake of Zn and Se, smoking status and Gleason score did not differ by race (Table 1). Tumor tissue Cd, Pb, and Zn content were determined in replicate samples and did not differ by race. African Americans had significantly lower Census tract median household income (p <0.001) compared with European Americans. In the Detroit metropolitan area, neighborhood poverty is positively associated with greater numbers of polluting industrial sites[30]. As part of a prior study, AR protein expression was measured in tissue from the same tumor foci in each specimen. Tumor AR expression did not differ between African American and European American prostate cancer cases.

Table 1.

Prostate Cancer Case Characteristics

	African-American (N=22)	European-American (N=30)	P value
Age, years, mean (SD)	57.7 (7.5)	60.7 (7.2)	0.15
Dietary zinc intake (mg/day), median (range)	12.74 (2.14, 35.01)	15.93 (4.98, 32.9)	0.78
Dietary selenium intake (mcg/day), median (range)	125.8 (27.0, 292.4)	125.7 (28.0, 252.3)	0.70
Ever smoker (yes)	63.6 %	46.7 %	0.27
Census Tract Median Household Income	34,850 (21705, 137720)	61,944 (37,924, 130,666)	0.001
Gleason ≥ 4+3	22.7 %	26.7 %	0.75
Tumor Cd, ppb, median (range)	236.5 (44. 7, 1095.2)	243.1 (68.2, 1337.9)	0.97
Tumor Pb, ppb, median (range)	229.9 (59.2, 1054.7)	149.1 (61.60, 739.2)	0.20
Tumor Zn, ppb, median (range)	389,101.2 (211,255.3, 1721672.7)	438,577.6 (214,582.2, 2,078,192.7)	0.21
AR expression, mean (SD)a	0.015 (.013)	0.013 (.010)	0.63

AR, androgen receptor; Cd, cadmium; Zn, zinc.

^aAndrogen receptor expression is the overlapping pixel area of AR and 4′,6-diamidino-2-phenylindole (DAPI) divided by the total area expressing DAPI (Area AR + DAPI/total nuclear area DAPI).

P-values for difference between races. Student’s t-test used to test difference in means, Mann-Whitney U test used for nonparametric variables, and Pearson chi-square test used to test differences between categorical variables.

We evaluated the association of Cd and AR expression overall and by race. Overall, there was no significant association between Cd and AR protein expression. However, as shown in Figure 1, the direction of the correlation between Cd and AR differed by race. A significant positive correlation between the log of tumor Cd content and AR expression was evident for African Americans (Pearson correlation, r = 0.52, 95% CI 0.13 – 0.91, P = 0.013); whereas a non-significant negative correlation was observed in European Americans (r = −0.19, 95% CI −0.57 – 0.19. P = 0.31). In multivariable analyses (Table 2) that accounted for age, ever smoking and Gleason score, a significant race-by-Cd content interaction (P = 0.003) was observed, indicating that the relationship between tumor AR expression and Cd content was dependent upon self-identified race. Adjusting for additional factors including tumor tissue Zn or Pb content, dietary Zn or Se intake, or median household income did not modify these results (Table 3).

Figure 1.

Correlation between prostate tumor tissue Cd content and androgen receptor (AR) protein expression in prostate tumor tissue of African-American and European-American men.

Table 2.

Multivariable analysis of predictors of androgen receptor protein expression in prostate tumor tissue

Variable	B(±SE)	p-value	B(±SE)	p-value
Age	6.62e-5 ± 1.47e-4	0.65	3.53e-4 ± 1.64e-4	0.036
Race (African-American)	0.004 ± 0.003	0.24	−0.054 ± 0.019	0.006
Ever smoke (Yes)	−0.006 ± 0.003	0.06	−0.007 ± 0.003	0.021
Gleason score (≥ 4+3)	0.003 ± 0.004	0.44	0.002 ± 0.003	0.511
log tumor Cd	0.005 ± 0.004	0.21	−0.003 ± 0.004	0.539
Race*log tumor Cd			0.025 ± 0.008	0.003
Model summary	R2 0.64	<0.001	R2 0.70	<0.001

B, beta, SE, standard error, Cd, cadmium.

Table 3.

Multivariable analysis of predictors of androgen receptor protein expression in prostate tumor tissue

Variable	B ± SE	p	B ± SE	p	B ± SE	p	B ± SE	p	B ± SE	p
Age	3.42e-4 ± 1.78e-4	0.060	3.61e-4 ± 1.53e-4	0.051	3.25e-4 ± 1.62e-4	0.051	2.62e-4 ± 1.70e-4	0.131	2.87e-4 ± 1.96e-4	0.151
Race (African-American)	−0.062 ± 0.02	0.004	−0.058 ± 0.20	0.006	−0.066 ± 0.020	0.051	−0.070 ± 0.021	0.001	−0.066 ± 0.020	0.002
Ever smoke (Yes)	−0.006 ± 0.003	0.036	−0.007 ± 0.003	0.021	−0.008 ± 0.003	0.002	−0.008 ± 0.003	0.014	−0.006 ± 0.003	0.044
log tumor Cd	−0.002 ± 0.006	0.74	−1.98e-5± 0.005	0.99	−0.005 ± 0.004	0.238	−0.007 ± 0.005	0.145	−0.004 ± 0.006	0.510
Log tumor Zn	−1.30e-4 ± 0.003	0.96
Log tumor Pb			−0. 003 ± 0.004	0.48
Log dietary Zn intake					0.008 ± 0.006	0.165
Log dietary Se intake							0.008 ± 0.005	0.098
Log of Census Tract Median Household Income									0.002±0.004	0.671
Race*log tumor Cd	0.027± 0.008	0.002	0.026 ± 0.008	0.003	0.030 ± 0.008	0.001	0.031 ± 0.009	0.001	0.029 ± 0.008	0.001
Model summary	R2 0.71	<0.001	R2 0.73	<0.001	R2 0.71	<0.001	R2 0.72	<0.001	R2 0.71	<0.001

B, beta, SE, standard error, p, p-value, Cd, cadmium, Pb, lead, Se, Selenium, Zn, zinc

Comparing mean AR expression by high (> median) and low Cd level showed that among African-American men with high Cd content, AR expression is more than double the AR expression in men with low Cd tissue levels (mean AR expression (±standard error): Cd high 0.021 (± 0.005) vs. Cd low 0.008 (± 0.003), P = 0.014). AR expression did not differ in European Americans by Cd content (mean AR expression (±standard error): Cd high 0.011 (±0.002) vs. Cd low 0.015 (±0.003), P = 0.30).

Discussion

In this study, we found that the association between Cd and AR protein expression in prostate tumor tissue was dependent upon self-identified race-ethnicity (African American). These preliminary findings (N = 52) support the hypothesis that Cd may be a direct or indirect endocrine disruptor of AR in human prostate tumor tissue and that race (African American) may modify the association between Cd and AR. This finding is provocative since African American men have 74% higher incidence of prostate cancer and are more than twice as likely to die of the disease than European American men (42.8 vs. 18.7 deaths per 100,000).[31]

Several mechanisms of Cd-induced carcinogenicity have been reported and are comprehensively summarized by Hartwig [32]. As early as 1980, basic science studies supported a role for endocrine disruption of AR by Cd. Donovan et al. [33], using extracts of mouse prostate cytosol, showed Cd to be the most effective divalent metal ion to inhibit dihydrotestosterone binding to the AR. Ye et al. [7] and Martin et al. [8] have both shown that Cd increases AR activity in LNCaP cells, and the latter also reported lower AR protein expression with increasing content of Cd (10⁻¹⁰ to 10⁻⁶ M). Originally derived from a lymph node of a European American prostate cancer patient, the LNCaP cell line is known to contain a functional mutation in the hormone-binding domain of AR at amino acid 877. The mutation alters steroid specificity and antiandrogen sensitivity [34]. Therefore, Martin et al. [8] also transfected wild-type AR into Cos-1 cells and showed that Cd affects the normal receptor as well. Using Wistar rats given Cd via drinking water, Lacorte et al. [9] showed increased AR protein expression and proliferation, as measured by Ki-67, in both the ventral and dorsal lobes of the prostate in treatment versus control rats. As mentioned previously, Wilson et al. [11] used cytosol AR from Dunning rats to show that Cd and other divalent metal ions alter the molecular size of the receptor and may, therefore, play a role in altering the state of transformation of the receptor. These studies all support Cd as an endocrine disruptor of AR. However, these cell culture studies are often criticized for using higher, non-physiological concentrations of Cd than what would be reasonably expected in humans. There may be other and less direct pathways through which Cd may influence the AR, as well.

Epidemiological prostate cancer studies to date are equivocal in terms of translating experimental observations of Cd as an endocrine disruptor of the AR. Human studies have used proxy measures of prostate Cd content, such as blood, urine and seminal fluid Cd concentrations, and have used PSA as a measure for AR activity. In a National Health and Nutrition Examination Survey analysis of 422 men over 40 years of age, van Wijngaarden et al. [13] found no overall association between urinary Cd and PSA concentration in blood samples but noted a potential protective interaction between Cd and Zn with regard to PSA concentrations (P = 0.09). Among husbands/male partners (> 18 years of age) of pregnant women living in Poland, Ukraine, and Greenland, Andreucci et al. [15] reported a significant inverse trend (log(β) = −0.121, P = 0.01) between blood Cd and seminal fluid PSA only in Greenlandic men. The concentration of Zn in semen significantly modified the association between Cd and PSA levels (overall P value for interaction = 0.01, Greenland p-value for interaction P = 0.04); associations were slightly stronger in men with semen Zn concentrations below the median values. In addition, Andreucci et al. [15] noted an inverse association between Cd and PSA concentrations in men with AR CAG repeat length of 24 (37% of subjects, log(β) = −0.23, P = 0.0009). The association was not significant for other AR CAG strata. In our study, we observed a non-significant inverse relationship between Cd and AR for European American cases, but a positive association for African-American cases. We were able to account for tumor tissue Zn content as well as dietary Zn or Se intake. Our limited sample size did not allow us to examine the influence of AR CAG repeat length. However, our observed difference by race suggests AR CAG should be considered further as a potential modifier of Cd and AR associations, as African American men on average carry shorter CAG repeat length than European American men [35–37].

Strengths and Weaknesses

This study revealed a correlation between tumor tissue Cd-content and AR protein expression in prostate tissues of African American men which was not found in European American men. This work, however has a number of limitations. We measured metals in just a small number of formalin-fixed, paraffin-embedded tissues. Moreover, the process of embedding tissue has been shown to reduce metal content [24]. As this is a substudy of an earlier large research study focused in part on occupational metal exposures (1999-2004), the tissues were stored for nearly 10 years before tissue metal analysis began. However, all of our prostatectomy specimens were originally processed and archived in the same room and manner in a single pathology facility at Henry Ford Health System, which would reduce the chance of variations in metals loss across subjects. Accounting for tissue Zn or Pb-content, dietary Zn or Se, or median household income did not modify our results. However, we are unable to account for all of these factors simultaneously due to our small sample size. Also tissue Se-content was not assessed and dietary nutrient intake was only assessed after diagnosis of prostate cancer and in most cases after prostatectomy. Patients may have changed their dietary habits after diagnosis, making it difficult to accurately assess Zn and Se intake at the time of prostatectomy. An additional concern that we were unable to address is the difference in use of menthol versus non-menthol cigarettes that has been observed between African Americans and European Americans [38], although our sample included only three current smokers, one African American and two European Americans. Menthol in general has been shown to decrease prostate cell proliferation [39] although it may increase Cd content associated with smoking[40]. We did not have comprehensive exposure information for potentially important windows of susceptibility (e.g. in-utero, neo-natal, adolescent and pubertal exposure). Finally, due to our sample size, we were also unable to account for genetic factors that may explain our observed difference by race.

Conclusions

In summary, we report a novel preliminary finding of a potential racial difference in the association of Cd-content and AR expression in prostate cancer that will need to be confirmed in larger studies. Although the association of Cd and AR protein expression is not strong in tumor tissue of adult African American men, it is the first direct evidence of an association between the two to be reported for human prostate tissue and it is unknown whether this association exists in normal prostate tissue (in men without cancer or BPH) or is present and important during development. Future studies should focus on direct measurements of metals and AR in the prostate or development of more accurate surrogate measures for estimating prostate Cd-content and AR expression. The later may be difficult since Minguez-Alarcon et al. [41] have shown that Cd-concentrations in blood and seminal plasma have low correlation with one another. Correlations between blood or urine Cd-concentrations and prostate tissue Cd-content may also be low. Capturing exposures across the life course may help identify windows of opportunity in which Cd exposure has the greatest influence on AR activity and prostate carcinogenesis. If this African American-specific Cd-AR association is real, then further exploration into any role Cd plays in racial differences in tumor aggressiveness or disease recurrence is also warranted.

Abbreviations

AR

androgen receptor

Cd

cadmium

CK18

cytokeratin-18

DAPI

4′,6- diamidino-2-phenylindole

ICP-MS

inductively coupled plasma mass spectrometry

PSA

prostate specific antigen

Se

selenium

Zn

zinc