Serum Free Circulating DNA Is a Useful Biomarker to Distinguish Benign versus Malignant Prostate Disease

Prostate cancer is the second leading cause of death in men in the United States, with an estimated 192,280 new cases and 27,360 prostate cancer related deaths for 2009 (1). Although prostate-specific antigen (PSA) test along with digital rectal examination is the currently used prostate cancer screening method in men >50 years of age, it is still unclear whether the benefits of PSA screening outweigh its limitations. The predictive value of elevated serum PSA levels (>4 ng/mL) is greater than that of digital rectal exam. However, PSA screening alone results in a large number of false positives and false negatives because elevated PSA levels do not represent the presence of prostate cancer alone but may also reflect certain nonmalignant conditions such as benign prostatic hyperplasia (BPH), as well as prostatitis (2, 3). In men undergoing prostate biopsy for elevated PSA levels between 4.1 and 9.9 ng/mL, 25% to 35% were diagnosed with prostate cancer, indicating that the remaining 65% to 75% of the men were subjected to unnecessary prostate biopsy (4). This high rate of false positives results in additional medical procedures with potential health risks, significant healthcare expenses, and anxiety to the patients and their family. On the contrary, prostate cancer was detected in 22% of men who underwent prostate biopsy with PSA level between 2.6 and 4 ng/mL (5). Moreover, 30% of men with PSA levels between 4.1 and 10 ng/mL who are diagnosed with prostate cancer are likely to have an extraprostatic extension of prostate cancer. Based on these findings, it has been suggested that the PSA cut point for prostate cancer screening be lowered to 2.5 ng/mL (6, 7). Although this reduction in PSA cutoff value has improved the detection rate of prostate cancer, it has also led to increased number of false positives who are then subjected to unnecessary prostate biopsy (8). A noninvasive test with higher specificity is required to eliminate unnecessary prostate biopsies.

The ratio of free PSA to total PSA has been shown to improve the diagnostic accuracy in the group with slightly elevated serum PSA values (9-11). However, the clinical significance of free to total PSA ratio remains controversial. Hence, there is a desperate need for novel biomarkers that, when used independently or in combination with PSA, are capable of increasing the specificity and sensitivity of prostate cancer detection.

The presence of circulating cell-free nucleic acids in cancer patients has been well documented, and the correlation between elevated levels of free circulating DNA (fcDNA) and cancer has been established. In 1977, Leon et al. (12) reported increased levels of serum DNA in cancer patients compared with healthy controls. This was followed by other studies showing elevated levels of circulating serum or plasma DNA in patients with malignant tumors compared with benign controls (13-19). Sozzi et al. (17) examined plasma fcDNA levels as a marker for lung cancer and found that the median fcDNA levels in plasma were 8 times greater in cancer than that in control samples (area under the curve, 0.94). The use of fcDNA as a biomarker for prostate cancer has been evaluated in several studies (20-25). Jung et al. (23) described no difference in fcDNA levels between patients with localized prostate cancer and BPH, although fcDNA levels were found to be significantly elevated in metastatic prostate cancer patients. Another study examined levels of a noncancer gene prostaglandin-endoperoxidase synthase 2 in 168 prostate cancer, 42 BPH, and 11 healthy individuals and found significantly elevated fcDNA levels in prostate cancer (median, 70.2 ng/mL) compared with that of BPH and healthy controls (10.5 and 7.1 ng/mL, respectively; ref. 26). Altimari et al. (27) showed that plasma fcDNA quantitation distinguished between patients with prostate cancer and healthy controls and correlates with pathologic tumor stage. Chun et al. (22) prospectively assessed plasma fcDNA in 142 men with localized prostate cancer and 19 BPH controls and found that fcDNA is an accurate and informative predictor of prostate cancer independent of the established clinical variables such as age and PSA. Bastian et al. (28) showed a significant association of increased preoperative serum fcDNA levels in men with localized prostate cancer with PSA recurrence after radical prostatectomy, indicating that serum fcDNA levels may be a useful prognostic biomarker in patients undergoing radical prostatectomy. Although the above reports indicate the potential clinical application of fcDNA as a diagnostic and prognostic marker of prostate cancer, contrary results have been published, showing that fcDNA levels are higher in BPH controls compared with prostate cancer patients (21, 23).

However, there have been no studies evaluating the role of fcDNA as a biomarker in a “representative” patient group who have undergone prostate cancer screening. In the present study, we prospectively collected blood samples and assessed fcDNA levels before prostate biopsy in patients referred for the procedure based on PSA and digital rectal exam screening. We analyzed fcDNA levels in relation to the demographic and clinicopathologic features and report results of this analysis.

Patient group consisted of men with elevated PSA levels >4 ng/mL and/or abnormal digital rectal exam, undergoing prostate biopsy at the Urology Clinic of University of Miami/Jackson Memorial Hospitals between 2004 and 2008 under an Institutional Review Board protocol. Patients with other malignancies and patients not willing to sign an informed consent form were excluded. This comprised of a racially diverse population that included non-Hispanic Blacks, Hispanic Whites, non- Hispanic Whites, and Hispanic Blacks. Two hundred fifty-two men who fit the study inclusion criteria were enrolled in the study before their prostate biopsy. Baseline demographics and medical history were collected. The pathologic and clinical staging information was obtained from the hospital charts.

Blood was collected in 10 mL serum plus blood collection tubes (BD Vacutainer) before patients undergoing prostate biopsy. Serum was separated by centrifugation at 3,500 rpm for 15 minutes at room temperature, after which DNA was extracted immediately. DNA was extracted from 1 mL serum using QIAamp UltraSens virus kit (Qiagen) following the manufacturer's protocol and stored at −20°C until further analysis.

Serum DNA was quantitated by real-time PCR for glutathione S-transferase, pi (GSTP1), based on the previous study by Bastian et al (28). We used the amplification primers (forward, 5′ AGG CCT TCG CTG GAG TTT C 3′; reverse, 5′ CCA TGC TGG GAG CTC TGA G 3′) and an amplicon-specific fluorogenic hybridization probe (6FAMCGC CGC AGT CTT CGC CAC CTAMRA). The PCR was carried out in an iCycler (Biorad). The PCR mixture consisted of 12.5 μL of Taqman Universal master mix (Applied Biosystems), 5 pmoles of probe, and 5 pmoles each of the forward and reverse primer in a 25 μL reaction volume. Each sample was analyzed in triplicate. All PCR runs included a negative control using water blanks. A standard curve was generated for each PCR run using serial dilutions of human placental DNA (Sigma) at concentrations ranging between 160 ng/μL to 160 pg/μL. For calculation of DNA concentrations, the standard curve was interpolated with the threshold cycle (C_T) of unknown target samples.

Data were tabulated by biopsy result: 89 prostate cancer, 59 prostatitis, and 104 BPH patients. Groups were compared using Fisher's exact test or χ² test for categorical variables and ANOVA and pairwise comparison by contrasts with Bonferroni's adjustment for continuous variables. Given the large variability of PSA and fcDNA data, corresponding data were log 2 transformed for better normal approximation and stabilization of variance. Univariate and multivariate logistic regression was used to examine the effect of fcDNA (log 2–transformed data and binary indicators of various cut points) on risk for disease comparing prostate cancer versus a combined benign group of prostatitis and BPH patients. Multivariate models were adjusted for age (years), race (Black or White), and PSA (log 2 PSA and binary indicator of PSA > 10 ng/mL). Bivariate interactions were tested. Results for selected logistic models are presented as age- and race-adjusted odds ratio estimates with corresponding 95% confidence intervals (95% CI) and P values for the association between fcDNA or PSA and risk for disease [prostate cancer versus benign (prostatitis + BPH)]. In addition, for selected models, we report area under the curve of receiver operating characteristic, which is a popular measure of model classification accuracy, and predictive probability cut point associated with sensitivity about 95%, as well as the corresponding estimates of specificity, positive predictive value, and negative predictive value. All analyses were done in SAS 9.2.

Characteristics of patients grouped according to prostate biopsy are reported in Table 1. Prostate cancer was diagnosed in 35.3% of the patients. The mean age was 66 years in prostate cancer, 63 in prostatitis, and 63 in BPH patients. The prostate cancer group had a significantly higher percentage of Blacks (48%) than prostatitis (17%) and BPH (34%) groups. Most Blacks (75%) were non-Hispanic, and most Whites (96%) were Hispanic. Comparison of log 2–transformed PSA in prostate cancer versus BPH and prostate cancer versus benign (BPH + prostatitis) showed statistically significant mean differences, whereas there was no significant difference in mean of log 2–transformed PSA between BPH and prostatitis. A similar pattern was observed with respect to the proportion of patients with PSA >10 ng/mL, which was significantly higher in prostate cancer (57%) than in prostatitis (37%) and in BPH patients (31%). Most prostate cancer patients (58.4%) had a Gleason score of 6 (3 + 3).

The median fcDNA levels were 149.4 ng/mL in prostate cancer, 96.9 ng/mL in BPH, and 81.8 ng/mL in prostatitis. There was no statistically significant difference between prostatitis and BPH about fcDNA original measurements or log 2–transformed PSA. Figure 1 shows the distribution of fcDNA levels in different patient groups. There was significant difference in fcDNA distributions for prostate cancer versus benign (prostatitis + BPH), considering all patients (P = 0.0129) and in the subgroup of patients with PSA ≤ 10 ng/mL (P = 0.023); however, there was no statistically significant difference in fcDNA distributions for prostate cancer versus benign (prostatitis + BPH) in patients with PSA > 10 ng/mL(P = 0.485).There was no significant association between fcDNA levels and Gleason score among prostate cancer patients (data not shown).

Univariate logistic models were computed as a first step in examining the effect of fcDNA (log 2 fcDNA or fcDNA high versus low), PSA (log 2 PSA or PSA > 10 versus ≤10 ng/mL), age (in years), and race (Black versus White) on the risk for prostate cancer in comparison with benign (prostatitis + BPH) patients. Based on the area under the curve criteria, the best discrimination between prostate cancer and benign (prostatitis + BPH) using fcDNA was found to be a cut point of 180 ng/mL. This characterization of fcDNA was then included in a multivariate model together with indicator of PSA > 10 ng/mL, age, and race. All two-way interactions were tested, and only fcDNA and PSA interaction was found to be significant (P = 0.003). Odds ratios for fcDNA or PSA comparing prostate cancer cases versus controls (prostatitis + BPH) based on various fitted logistic models, with adjustment for age and race, are reported in Table 2. Model 4 including age, race, fcDNA (>180 versus ≤180 ng/mL), PSA (>10 versus ≤10 ng/mL), and the interaction between fcDNA and PSA provides the best fit to data by the area under the curve criteria (area under the curve, 0.742). Based on this model, patients with PSA ≤ 10 ng/mL with fcDNA > 180 ng/mL were at increased risk for prostate cancer compared with those with fcDNA ≤ 180 ng/mL (odds ratio, 4.27; 95% CI, 2.05-8.88; P < 0.001), whereas there was no significant effect of fcDNA among patients with PSA > 10 ng/mL (odds ratio, 0.66; 95% CI, 0.25-1.78; P = −0.417). Subset analyses of low and high PSA patients confirmed estimates obtained from model 4 (data not shown). Table 3 shows finding from analyses carried out in each of the three subsets defined by PSA ≤ 4, 4 to 10, and >10 ng/mL. The corresponding odds ratios were 3.12 (95% CI, 0.56-17.22; P = 0.192), 4.32 (95% CI, 1.83-10.21; P < 0.001), and 0.65 (95% CI, 0.24-1.79; P = 0.406). Estimates in the low and intermediate PSA groups were similar, although statistical significance was not achieved in the low PSA group, which included only 12 cases and 30 controls. Furthermore, inclusion of fcDNA and the interaction of fcDNA and PSA to a model with PSA, age, and race significantly improved predictivity (area under the curve, 0.742 versus 0.687; P = 0.029; Fig. 2).

Although, in terms of area under the curve, the overall gain is modest, there is considerable gain in specificity at cut points defining high sensitivity. As shown in Fig. 2, based on the model including fcDNA, PSA, interaction of fcDNA and PSA, race, and age, a predicted probability ≥ 0.158 yields sensitivity and specificity of 95.5% and 33.1%. The model without fcDNA at a predictive probability ≥ 0.175 yields similar sensitivity of 95.5% but a much smaller specificity of 6.7%. Furthermore, we observed a considerable increase in negative predictive value in the model, including fcDNA (93.1%) compared with the model without fcDNA (73.3%).

We designed this prospective study to evaluate the role of fcDNA as a biomarker to distinguish prostate cancer patients from those with benign prostatic conditions. The patients included in this study were suspected to have prostate cancer based on their PSA values (≥4 ng/mL) and/or abnormal digital rectal exam. Serum fcDNA levels were evaluated in 252 patients (89 prostate cancer, 59 prostatitis, and 104 BPH). Our results indicate that at a cut point of 180 ng/mL, fcDNA levels can differentiate prostate cancer from benign prostatic conditions such as prostatitis and BPH in patients with PSA ≤ 10 ng/mL. After adjustment for age and race, patients with PSA ≤ 10 ng/mL with fcDNA > 180 ng/mL were at increased risk for prostate cancer compared with those with fcDNA ≤ 180 ng/mL (odds ratio, 4.27; 95% CI, 2.05-8.88; P < 0.001). The multivariate model with fcDNA, at a predicted probability ≥ 0.158, yields sensitivity of 95.5% and specificity of 33.1%, whereas the multivariate model without fcDNA, at a predicted probability ≥ 0.175, yields sensitivity of 95.5% and specificity of 6.7%. We also show that inclusion of fcDNA yields a large increase in negative predictive value as compared with when fcDNA was not included (93.3% versus 73.3%).The role of fcDNA as a biomarker of prostate cancer has been evaluated in about 10 studies (20-29). Of these, only three studies have examined the usefulness of fcDNA quantitation as a diagnostic marker in prospective settings (20-22). The results are summarized in Table 4. The common drawback of these studies is the lack of a “representative patient distribution” in a prostate cancer screening scenario. Our study represents a typical distribution of patients referred for prostate biopsy based on elevated PSA and/or abnormal digital rectal exam (DRE) because 35% were diagnosed with prostate cancer, similar to results obtained in other studies (4). In addition, we maintained stringent patient inclusion criteria. Our control cohort comprised of BPH and prostatitis patients. Because elevated fcDNA levels are not prostate cancer specific and may be related to a variety of cancers, we excluded patients who were diagnosed with other malignancies.

Median fcDNA levels have been found to vary across different studies. These variations can be attributed to the source of fcDNA, as well as variation in methods of sample processing and fcDNA quantitation. A critical factor for determining a standard fcDNA cutoff value is to establish uniform methodologies for sample collection, processing, DNA extraction, and fcDNA quantitation and analysis (30). To rule out differences in DNA concentrations arising from variation in sample processing and DNA extraction, we uniformly processed all samples. DNA was extracted immediately after serum separation to avoid variations in fcDNA levels because of the sample storage (31). Among different methods of fcDNA quantitation, PCR based assays for quantitation of target genes have been found to be highly sensitive and thus offer a simple and fast method for fcDNA analysis (32).

Cell lysis and apoptosis have been postulated to be the predominant mechanisms of release of DNA into circulation (19, 33). Ellinger et al. (26) examined levels of a noncancerous gene, prostaglandin-endoperoxidase synthase 2, as a fcDNA marker to distinguish between prostate cancer and BPH. In this study, GSTP1 hypermethylation was used as a marker of tumor derived DNA, and it was found that tumor DNA accounted only for about 1.9% of the total DNA. These data supported the theory that a large proportion of fcDNA was derived from noncancerous cells because of the induction of apoptosis by proapoptotic cytokines released from prostate cancer cells (34). Therefore, levels of non-cancer-specific gene for quantitation of fcDNA levels may be more useful for cancer detection. Genetic heterogeneity between tumor cells hinders the use of a cancer-specific gene mutation as a fcDNA marker. As stated previously, Bastian et al. (28) previously assayed three genes, GSTP1, EDNRB, and MDR1 that were present in the genome of all cells to quantitate fcDNA and found that means levels of fcDNA were similar for the three genes. Because GSTP1 has not been shown to be deleted or mutated in prostate cancer, it is a suitable non-cancer-specific marker for quantitation of fcDNA.

Thus far, fcDNA analysis has not been done in patients stratified according to PSA in a prospective study. Total PSA above 10 ng/mL is associated with a high likelihood of prostate cancer, and a prostate biopsy is commonly recommended in this situation. In contrast, PSA levels between 4.1 and 9.9 ng/mL is associated with prostate cancer in only 25 to 35% of men (4). We compared fcDNA analysis in patients with PSA levels ≤ 10 ng/mL and those with PSA > 10 ng/mL. In patients with PSA ≤ 10 ng/mL, fcDNA levels ≤ 180 ng/mL was associated with reduced risk for prostate cancer. Although there is a relatively high incidence of prostate cancer diagnosis in the high PSA group, the proportions with high fcDNA were comparable in prostate cancer versus benign (prostatitis + BPH; 31% versus 36%). Our results indicate that fcDNA levels could act as a discriminating marker in patients with PSA ≤ 10 ng/mL.

In our study cohort, 27% of men with PSA ≤ 10 ng/mL were diagnosed with prostate cancer by prostate biopsy. Hence, a test with high negative predictive value is required to identify men who can forgo biopsy in this group of patients. Our results indicate that, with a high negative predictive value of 93.1%, adding fcDNA can spare 33% of patients with PSA ≤ 10 ng/mL of unwanted prostate biopsies. This finding is of major clinical significance because of the fact that ∼65% to 75% of patients with intermediate PSA levels between 4 and 10 ng/mL are subjected to unnecessary prostate biopsies (4). If validated in independent studies, fcDNA assay could potentially improve the specificity of prostate cancer screening in this group of patients. Future prospective studies using a standardized technique of isolation and quantitation of fcDNA are needed to evaluate the role of fcDNA as a prostate cancer biomarker.

No potential conflicts of interest were disclosed.