Examination of polymorphic glutathione S-transferase (GST) genes, tobacco smoking and prostate cancer risk among Men of African Descent: A case-control study

Even though prostate cancer (PCA) ranks the highest in incidence and second mortality among all cancers affecting American men, its etiology and ethnic disparities are largely unknown[1]. For instance, age and family history are the strongest risk factors for PCA, but African-American (AA) men are more than twice as likely to develop the disease compared to other racial or ethnic groups[1]. AA men are also more likely to be diagnosed with PCA at a younger age, with more aggressive disease and poorer prognosis [1–3]. Despite increases in five year survival rates for AA men over the last few decades; their rates still lag far behind other races[1]. While, the reasons for this disparity are largely speculative; risk, incidence, and mortality rates suggest that genetic factors play an important role in PCA initiation and progression[1]. However, lifestyle habits (e.g., cigarette smoking, diet) have also been implicated to increase risk; indicating that environmental factors may contribute to PCA[1, 2, 4]. Therefore, prostate carcinogenesis and its disparity most likely involve a complex interplay between genetic and environmental factors. More specifically, variations in carcinogen metabolism genes may play a critical role in PCA development due to their activation or detoxification functions.

The glutathione S-transferase (GST) gene superfamily encodes enzymes that catalyze the conjugation of glutathione to electrophilic compounds[5, 6]. These enzymes generally detoxify endogenous and exogenous agents, but also participate in the activation and inactivation of oxidative metabolites of carcinogenic compounds associated with prostate cancer[7, 8]. Variant GST alleles have been identified within the general population. The most extensively studied variant GSTs include two GST deletion alleles (i.e., GSTM1*0/*0 [GenBank: BC024005.2] and GSTT1*0/*0 [GenBank: BC007065.1]) and the GSTP1 Val allele which is characterized by an adenine to guanine substitution at position -313 (A^-313G) in exon 5 [GenBank: BC010915.1; dbSNP: rs1695][9]. The functional consequences of the GSTM1 and GSTT1 (*0/*0) genotypes are obvious in terms of enzyme activity; gene deletion results in loss of conjugation potential. The GSTP1 polymorphism, resulting from an isoleucine to valine substitution within the active site of the enzyme at codon 105 (I¹⁰⁵V), is linked to altered substrate-specific thermostability and conjugation activity [10–12]. For instance, the GSTP1 105 Val variant has been associated with lower efficiency for diol epoxides of some polycyclic aromatic hydrocarbons, therefore resulting in decreased detoxification of these compounds compared to the Ile allele[10].

Genetic variations in polymorphic GST genes have been implicated in the etiology of numerous cancers [13–18]. Some studies indicate GST polymorphisms are associated with prostate cancer; however, this association is not accepted across all observational studies [18–25]. Discrepancies may be partially attributed to failure to consider gene combinations or interactions with tobacco smoking. Furthermore, few studies have involved a sufficient number of African-Americans in order to adequately investigate the role of GST polymorphisms and environmental factors in PCA development[12, 26, 27].

To clarify the role of GST genes in PCA risk within a high-risk sub-group, we evaluated the individual and joint modifying effects of three commonly studied sequence variants in a case-control study of 208 cases and 665 disease-free controls among a population of men of African descent. In an exploratory analysis, we evaluated gene-gene interactions using an available non-parametric statistical model, namely Multifactor Dimensionality Reduction (MDR). This advanced statistical tool readily overcomes sample size limitations often encountered by parametric statistical methods (e.g., logistic regression analysis). MDR has greater than 80% statistical power and rigor to evaluate gene-gene interactions even in the presence of small sample sizes (i.e., 200 cases and 200 controls). This computationally sound statistical tool enabled evaluations of single and multi-locus genetic and environmental markers as indicators of PCA susceptibility within our population sample.

The patient and tumor characteristics in the current study are summarized in Table 1. Cases were significantly older than controls and had higher PSA levels. Although there was a small portion of controls (5.6%) who had PSA levels that exceeded 4.0 ng/ml, these individuals did not have an abnormal DRE or irregular biopsy. There was a significant difference in median West African genetic ancestry estimates comparing cases and controls (p = 0.020). Cigarette smoking data was available for approximately 25% of our study subjects (n = 216, 136 cases and 80 controls). These subjects were classified by cigarette smoking: current (n = 38), former (n = 71) and never (n = 107). Cigarette smoking did not differ significantly (p > 0.05) between cases and controls.

We evaluated the independent effects of genetic variations in highly variant GST genes in relation to PCA susceptibility using unconditional LR models, as detailed in Table 2. Inheritance of at least one GSTP1 105 Val allele (linked with decreased capacity to conjugate electrophilic compounds) was associated with a 1.6-fold increase in prostate cancer risk (OR = 1.56; 95%CI = 0.95-2.58; p = 0.049). The GSTP1 Val loci was also identified as the best single factor predictor of PCA based on a cross-validation consistency of 100%, an average testing accuracy of 53% and a 1000-fold permutation testing p-value of 0.001 (Table 3). Upon evaluation of gene-gene combination effects, LR modeling revealed a 2.1-fold increase in PCA risk associated with men possessing the GSTP1 (Val/Val) and GSTM1 (*1/*1 + *1/*0) alleles (OR = 2.11; 95%CI = 1.07-4.16; p-value for interaction = 0.062). Initially, it appeared as though this interaction was a significant PCA predictor following MDR analysis with a statistically significant 1000-fold permutation testing p-value of 0.01. However, this gene combination effect was primarily driven by the GSTP1 loci attributed to the lack of additional information gain comparing the GSTM1-GSTP1 axis (IG = 0.04%) to either GSTP1 or GSTM1 loci alone (IG = 0.58% and 0.03%, respectively) (Figure 1). Lastly, no significant risk effects were detected among the GST deletion polymorphisms (p = 0.622) or the GSTP1-GSTP1 and GSTM1-GSTT1 combinations (p = 0.557 and 0.814, respectively). For our exploratory analysis of gene-environment interactions associated with risk, LR modeling indicated that smoking status did not modify the relationship between the GST SNPs and PCA risk (p for interaction ≥ 0.135), as shown in Table 4.

Previous studies suggest that oxidative stress and reactive species are becoming increasingly important in prostate carcinogenesis[4, 11, 12, 42–45]. If these electrophilic compounds are not removed or reduced to less active forms, they may result in damage to biomolecules (e.g., DNA, proteins, etc.) and ultimately lead to cellular dysfunction or transformation. Consequently, the role of antioxidant enzymes has been commonly investigated in relation to PCA development because of their function in the detoxification of potentially damaging carcinogens and/or reactive oxygen species (ROS)[4, 46–48]. The GSTs in particular have been extensively studied since they are able to conjugate a wide range of oxidative substrates[6, 11, 43, 49, 50]. Variations within these genes can cause a loss or reduction in enzymatic activity and have been associated with increased risk of prostate as well as several other cancers (e.g., colon, breast, and lung)[6, 12, 51–53].

In the case of PCA, reports have observed that PCA cases possess lower antioxidant enzyme levels in prostate tissues compared to both healthy controls and men with benign prostatic hyperplasia (BPH)[48]. In addition, PCA tissues appear to contain higher amounts of ROS and oxidative DNA damage [46–48]. Some investigators speculate these effects may be attributed to GST genetic polymorphisms (i.e., GSTM1*0/*0, GSTT1*0/*0, and GSTP1 Val) linked with compromised oxidative repair capacity[5, 12, 26, 41, 43, 54]. The goal of this study was to evaluate GSTM1 and GSTT1 gene deletions, variant GSTP1 105 Val, and cigarette smoking as predictors of PCA risk among 873 men of African descent (208 cases and 665 controls). We hypothesized that individuals who possess one or more genotypes associated with reduced detoxification would have a higher risk of developing PCA. In an exploratory analysis, we investigated the effects of cigarette smoking combined with these polymorphisms in relation to PCA risk. This increased risk may be partially attributed to exposure to tobacco carcinogens and altered capacity to metabolically detoxify hazardous compounds. Logistic regression analysis revealed a moderately significant association between PCA susceptibility and GSTP1 105 Val. These findings were confirmed using a robust data-mining tool specifically designed to evaluate main effects and higher order interactions. Although we did not detect any significant single gene effects among GSTM1 and T1 deletions, we did observe a moderately synergistic interaction between GSTP1-GSTM1 and PCA susceptibility. We found a 2.1-fold increase in PCA risk associated with men possessing 2-3 high risk alleles with the GSTP1 GG (Val/Val) &GSTM1 *1/*1 + *1/*0 genotype according to logistic regression analysis.

Numerous previous studies have investigated the role of these GST polymorphisms in relation to PCA susceptibility. For example, Agalliu et al. 2006 reported a moderate increase in PCA risk associated with the GSTM1 gene deletion for Caucasian men (OR = 1.54; 95%CI = 1.19-2.01); however, GSTT1 deletion and GSTP1 105 Val polymorphisms were not significantly associated with prostate cancer[5]. The null findings for GSTT1 and GSTP1 were corroborated in a recent meta-analysis of pooled data from ≥ 3,837 cases and ≥ 4,552 controls[12]. This same report, however, observed a 1.3 fold increase in PCA risk among all study participants was associated with the GSTM1*0 loci using pooled data from 4,564 cases and 5,464 controls (OR = 1.33; 95%CI = 1.15-1.55)[12]. Stratification based on self-identified race indicated that risk estimates were comparable between Caucasians and Asians; however, the GSTM1 and GSTT1 deletion polymorphisms did not modify risk among Africans or African-Americans[12]. Interestingly, only two published studies investigated the role of the GSTP1 105 Val SNP in relation to PCA among men of African descent[5, 12, 54]. The null findings in these two reports may be partially attributed to the failure to utilize rigorous statistical tools that permit the evaluation of main effects even in the presence of small sample sizes (i.e., at least 200 cases and 200 controls) such as MDR. The MDR selection of the GSTP1 105 Val loci as a significant PCA predictor in the current study is promising and will undergo subsequent validation within future studies among men of African descent.

We have considered the strengths and limitations of the current study. Although MDR doesn't allow for adjustments of covariates, it does control for multiple comparisons and spurious risk estimates by using a cross validation and permutation testing scheme as a built-in feature. Misclassification by case status is also a potential limitation for this study. There is a slight possibility that some of the controls with PSA less than 4 ng/ml may actually have prostate cancer that remains undetected. Controls who had an abnormal PSA (>4.0 ng/ml) and/or irregular DRE underwent multiple core needle biopsies. Patients with an abnormal biopsy were reclassified as cases. Participants with a PSA (i.e., ≤ 4.0 ng/ml) but an abnormal DRE were excluded abnormal DRE were excluded from participating in the current study. These individuals were excluded because we could not predict with any level of certainty whether these individuals would develop prostate cancer. We also excluded individuals who were: (1) diagnosed with BPH following a biopsy; or (2) had an abnormal PSA and/or an irregular DRE. Since one cannot predict with any level of certainty whether individuals diagnosed with BPH would develop prostate cancer, these individuals were excluded from the current study. As implied within Lobe et al., 2006, even after close inspection of prostate cancer tissue, it is possible to miss a microscopic nodule that can later develop into cancer[55]. If controls in our study population were still misclassified after undergoing a PSA test, DRE, and/or multiple core needle biopsies, then we may expect our calculated risk estimates to underestimate the relationship between the selected GST polymorphisms and prostate cancer susceptibility. But this issue plaques all cancer epidemiology studies. Unfortunately, it is impractical to subject all patients to a radical prostatectomy to permit an extensive evaluation and more accurate classification of case status.

Another challenge for genetic epidemiology studies involving study participants of African descent is their unique population history of gene flow from divergent populations[56, 57]. The current study adjusted single loci models for genetic heterogeneity (i.e., population admixture). This also helps to circumvent misclassification of study participants related to self-identified race/ethnicity (SIRE). Our findings suggest that inclusion of WAA did not significantly change the risk estimates relative to unadjusted models; if anything it makes them more precise. Also, this study may be limited due to lack of consideration of the genetic heterogeneity of selected targets. This could be the reason we did not observe any statistical significant effects among GST polymorphisms and smoking interactions in relation to PCA risk (p = 0.105).

In summary, our findings indicate a moderately significant association between risk and among men possessing at least one variant GSTP1 105 Val allele (p = 0.049) among men of African descent. MDR analysis validated the logistic regression findings and identified the GSTP1105 Val allele (p = 0.001) as the best single factor model for predicting PCA risk. The ability of MDR to evaluate main effects and remain effective with relatively small sample sizes of at least 200 cases and 200 controls strengthens the power of this study. Statistically significant single and multi-locus models maybe adjusted for potential confounders by placing the MDR output variables into a LR model. Furthermore, this study is robust since it utilizes multiple statistical analysis tools to examine the single gene as well as gene-gene or gene-cigarette smoking combination effects in relation to PCA risk.