Serum cell‐free DNA in renal cell carcinoma

Currently, there are no established diagnostic and prognostic serum markers for renal cell carcinoma (RCC). The objective of this study was to evaluate the putative significance of serum cell‐free DNA.

surgery for clear cell RCC by conventional PCR of different glyceraldehyde 3-phosphate dehydrogenase (GAPDH) fragments. Those authors observed a different fragment distribution between patients and controls and within different RCC tumor (T) classifications. Hauser et al 10 detected higher concentrations of actin beta (ACTB) fragments in 35 patients with RCC compared with 54 controls. In total, relatively few patients were included, methylation of CpG islands was not analyzed, and associations with survival were not assessed.
Thus, the overall role of cell-free DNA as a diagnostic and prognostic RCC biomarker remains poorly understood. Therefore, the objective of the current study was to evaluate the putative significance of total cell-free DNA and CpG island methylation in diagnosing RCC and to associate these markers with pathologic variables and survival.

Study Population
This prospective, institutional review board-approved study was designed to test the hypothesis that total serum cell-free DNA concentrations and cell-free DNA CpG island methylation can serve as diagnostic and prognostic biomarkers for RCC. In total, 200 consecutive patients were enrolled who underwent surgery for a unilateral, sporadic, solid renal tumor suspicious for RCC at our institution between 2008 and 2009. None of the patients had concomitant malignancy, autoimmune disease, trauma, or myocardial infarction within the last 12 months before surgery, because these diseases may impact cell-free DNA levels. 11 All patients provided informed consent.
Clinical and pathologic features were obtained for each patient. Clinical features included age and sex. All pathologic features were evaluated by 1 expert uropathologist (A.H.) and included subtype according to the World Health Organization classification of tumors 12 ; tumorlymph node-metastasis (TNM) classification 13 ; Fuhrman grade 14 ; the presence of microscopic coagulative tumor necrosis; and pathologic tumor size according to the greatest tumor dimension. After surgery, patients with N0M0 disease were followed according to a risk-stratified surveillance protocol 15 with intermittent imaging of the chest, abdomen, and pelvis.
Sample Collection and Isolation of Cell-Free DNA Whole blood was drawn from patients before surgery. Serum was obtained immediately through centrifugation at Â1800 g for 10 minutes and stored at À80 C until DNA extraction. Cell-free DNA was isolated from 1 mL of serum using the QIAamp Ultrasens Virus Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. In a subgroup of 20 patients who underwent curative surgery for a clinically N0M0 tumor, serum also was obtained 4 weeks after surgery.

Real-Time Quantitative PCR
Real-time quantitative PCR (q-PCR) was performed to assess the total amount of cell-free DNA and CpG island methylation status of the genes Ras association domain family member 1A (RASSF1A), prostaglandin-endoperoxidase synthase 2 (PTGS2), P16 (cyclin-dependent kinase inhibitor 2A), and von Hippel-Lindau (VHL) using an Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems, Carlsbad, Calif). Each sample was assayed in triplicate, and each run included water blanks and an external control.

Determination of total cell-free DNA
The total quantity of cell-free DNA was determined using hydrolysis probes for the ring finger protein 185 gene RNF185 (TaqMan gene expression assay RNF185; assay identification no. Hs01565008_m1 [Applied Biosystems]; amplicon length, 249 base pairs [bp]). This gene is located on chromosome 22q12, which typically is not mutated in RCC, so that concentrations in serum are not lowered by deletions in tumor DNA. We used 10 lL of TaqMan Universal PCR Master Mix (Applied Biosystems), 1 lL of Taqman Gene Expression Assay, and 30 ng of cell-free DNA, which were brought to a total volume of 20 lL by adding UltraPure DNase/RNase-Free Distilled Water (Invitrogen, Lofer, Austria). The PCR conditions were as follows: 50 C for 2 minutes, 95 C for 10 minutes, 40 cycles at 95 C for 15 seconds, and 60 C for 1 minute. A standard dilution curve of Human Male Genomic DNA (Novagen Merck Chemicals Ltd., Nottingham, United Kingdom) of known concentration was used to determine the copy numbers (genome equivalents [GE]) of cell-free DNA. A factor of 6.6 pg of genomic DNA was applied for conversion to GE. One GE-equivalent was defined as the amount of a specific target sequence in a single reference cell. 16,17 with restriction enzymes that recognize and cut restriction sites with suspected methylation. Unmethylated DNA is cut by the enzyme, and no products are detected after PCR amplification. In contrast, the enzymes do not cut the methylated form of the sequences. To accomplish this, 50 ng of cell-free DNA solution were incubated for 4 hours at 37 C with 3 lL Buffer Tango 10X (Fermentas, St. Leon-Rot, Germany), 15 U restriction enzyme, and UltraPure DNase/RNase-Free Distilled Water up to total volume of 30 lL. The restriction enzymes were Hin6I for PTGS2 and P16 and HpaII for RASSF1A and VHL (Fermentas), which cuts GCGC and CCGG, respectively. For a control, cell-free DNA was incubated with buffer and water without restriction enzyme according to the same protocol. After incubation, q-PCR was performed. The PCR mixture contained 5 lL SYBR2X (Applied Biosystems), 3 lL water, 0.5 lL of each reverse and forward primer, and 10 lL of the digested mix. Forward and reverse primers are provided in Table 1. The q-PCR conditions were: 10 minutes at 95 C, 40 cycles for 30 seconds at 95 C and for 30 seconds at annealing temperature specific for each primer pair (Table 1), and 40 seconds at 72 C.

Statistical Analysis
Design and analysis of this study were performed according to reporting recommendations for tumor marker prognostic studies (ReMARK) guidelines. 19 The chisquare test and the nonparametric Kruskal-Wallis test were used to compare categorical data and continuous data, respectively. The Wilcoxon signed-rank test was used to compare preoperative and postoperative levels of total cell-free DNA. Methylation status was coded as a categorical variable. 6 Correlations were determined using the Pearson coefficient. The diagnostic performance (sensitivity and specificity) of serum cell-free DNA and methylated DNA was analyzed with a receiver operating characteristic curve (ROC). The area under the ROC curve (AUC) was applied to evaluate diagnostic accuracy. AUCs were compared as described previously. 20 Disease-specific survival was calculated from the date of surgery to the date of either death from RCC or last follow-up. Survival analyses were performed with univariate Cox proportional hazards models. For visualization of Kaplan-Meier survival estimates, a cutoff point for subcategorization was detected with recursive partitioning-based tree analysis. Because there should be at least 10 to 15 events for each variable that is included in a multivariate model to avoid over fitting, 21 multivariate Cox models that included all pathologic variables were not fit. Instead, we used the tumor stage, size, grade, and necrosis (SSIGN) score, which incorporates TNM classifications, tumor size, nuclear grade, and tumor necrosis. 22 The statistical package R 2.10.1 (R Foundation for Statistical Computing, Vienna, Austria; http://cran.r-project.org. [Accessed on March 1, 2011]) was used for all analyses. The level of statistical significance was set at P < .05.
The mean total serum cell-free DNA level, represented by Association With RCC Total cell-free DNA levels were higher in patients with RCC than in patients with benign tumors (mean AE standard deviation: 3319 AE 3181 GE/mL vs 1288 AE 913 GE/ mL; P < .001) (Fig. 1). In patients who underwent curative surgery, mean total cell-free DNA levels dropped significantly after surgery (from 3089 AE 1964 GE/mL to 1446 AE 1053 GE/mL; P ¼ .003; n ¼ 20).
Of the genes that we analyzed for methylation, VHL and RASSF1A were methylated more frequently in patients with RCC: Methylated VHL was detected in 79 patients (50.3%) with RCC but in only 4 patients (9.3%) with benign tumors (P < .001). Similarly, RASSF1A was methylated in 72 patients with RCC (45.9%) versus 3 patients (7%) with benign tumors (P < .001). P16 methylation and PTGS2 methylation were observed in similar percentages among patients with benign tumors and RCC (P ¼ .788 and P ¼ .689, respectively).
Association With Pathologic Variables RNF185 levels were higher in necrotic tumors (P ¼ .003) and in patients with lymph node metastasis (P ¼ .028) and distant metastasis (P < .001), whereas no significant

Association With Disease-Specific Survival
The mean follow-up duration was 28 months (median, 28 months; IQR, 22-35 months), during which 24 patients (15.3%) died of RCC. There was a significant association of total cell-free DNA levels (RNF185) with survival (P ¼ .004); with each 1000-GE/mL increase in the level of total cell-free DNA, the risk of RCC-specific death increased by 10% (hazard ratio, 1.10; 95% CI, 1.03-1.17). In recursive partitioning-based tree analysis, the cutoff point for subcategorization (high/low) was 2400 GE/mL, which corresponded to the value obtained for RNF185 in the diagnostic ROC curve analysis. There were 76 patients (48.4%) with high total cell-free DNA levels and 81 patients (51.6%) with low total cell-free DNA levels. The difference in survival between both groups was statistically significant (hazard ratio, 15.03; 95% CI, 3.52-64.21; P < .001) (Fig. 2). Methylation of VHL (P ¼ .184), RASSF1A (P ¼ .232), P16 (P ¼ .236), and PTGS2 (P ¼ .263) was not associated with survival. In multivariate analyses, SSIGN score (P < .001) and categorized total cell-free DNA levels (P ¼ .028), but not continuous total cell-free DNA levels (P ¼ .192), were identified as independent prognostic factors (Table 4).

DISCUSSION
In the current study, we evaluated the diagnostic and prognostic role of serum cell-free DNA concentration and methylation of RASSF1A, VHL, PTGS2, and P16 in RCC. Our data indicate that total cell-free DNA and methylation of VHL and RASSF1A may serve as diagnostic biomarkers for patients with RCC. Furthermore, total serum cell-free DNA appears to predict disease-specific survival.
It has been noted for decades that patients with cancer have elevated cell-free circulating DNA levels compared with controls. 3,10 These elevated levels may be  Cancer caused by higher degrees of necrosis, apoptosis, and spontaneous active release of DNA in cancer patients, but they also may be caused by therapy. 11 It is noteworthy that nonmalignant diseases, such as trauma, systemic inflammatory disorders, pulmonary embolism, and acute myocardial infarction, also can elevate levels of cell-free circulating DNA. 11 It has been documented previously and confirmed in the current study that a subset of healthy individuals has methylation of cell-free DNA, 8 which may be related to these nonmalignant conditions. Our current results further confirm the role of cell-free circulating DNA as a diagnostic RCC marker that may assist in the differential diagnosis of solid renal masses. Two previously published studies on this topic assessed different sized DNA fragments, thus, different fragmentation patterns, and the results indicated that larger DNA fragments are derived mainly from tumor necrosis, and smaller fragments (<200 bp) are derived from apoptosis. 23 Gang et al 4 observed a large 397-bp fragment in 71 of 78 patients with RCC and a 456-bp fragment in 69 of 78 patients with RCC, but no such fragments were observed in controls. In addition, larger fragments were detected less frequently after surgery. 4 Another study quantified 106-bp and 384-bp fragments of ACTB by q-PCR. 10 Levels of both fragments were higher in patients with RCC than in controls. The AUC values ACTB-106 and ACTB-384 were 0.686 and 0.725, respectively. The current study quantified total cell-free circulating DNA by amplification of a 249-bp fragment of RNF185. The fragment size suggests that it is derived from necrosis, which is supported further by the finding that higher levels were observed in patients who had necrotic RCCs. The AUC of 0.755 is very similar to the AUC of ACTB-384 reported by Hauser et al. 10 For the current study, we did not analyze small apoptotic fragments, which appear to be ubiquitous in all patients and seem to have lower power to diagnose RCC than the larger fragments; however, a combination of both may further increase diagnostic accuracy. 10 Because cell-free DNA is composed in part of tumor DNA, 11 tumor-specific genetic aberrations like methylation may be detected and, thus, may improve diagnostic sensitivity and specificity. Therefore, we studied CpG island methylation of RASSF1A, VHL, PTGS2, and P16, which are methylated frequently in several cancer entities. We further hypothesized that VHL methylation is helpful to diagnose the clear cell subtype, in which inactivation is noted in approximately 50% to 60% of tumors. 24 In contrast to previous studies in patients with testicular, prostate, and bladder cancers, we did not observe any diagnostic or prognostic value of PTGS2 methylation or P16 methylation. 8,25,26 RASSF1A methylation, however, was detected in 45.9% of patients with RCC and in 7% of patients with benign tumors, which corresponded to an AUC of 0.694. This frequency of RASSF1A methylation is in agreement with a report on patients with testicular cancer. 8 It is noteworthy that VHL methylation was detected in 50.3% of patients with RCC but in only 9.3% of controls. The AUC was 0.705 but increased to 0.744, when only clear cell tumors were analyzed. The observed frequency of VHL methylation was higher than that reported in previous studies of RCC tissue samples. 27,28 There are several factors that may account for this variation. Most researchers have used methylation-specific PCR to detect VHL methylation, and it appears to be more specific but less sensitive than restriction-endonuclease q-PCR. 8 Furthermore, the methylation pattern of serum and tumor DNA has not always had a large overlap. 29 Finally, cell-free DNA consists of portions other than tumor DNA, which has an impact on the serum methylation pattern. VHL methylation was detected more frequently in clear cell RCC than papillary or chromophobe RCCs, as hypothesized and mirroring results obtained in tumors. Surprisingly, a proportion of papillary and chromophobe tumors had VHL methylation, which may have been caused by the existence of hybrid tumors that had genetic alterations similar to those observed in as clear cell RCC. 30 Papillary RCC with clear cell features accounts for approximately 30 to 40% of papillary RCCs. In this entity, genetic alterations similar to those observed in clear cell RCC have been reported. 30 The prevalence of papillary RCC with clear cell features correlates with the frequency of VHL methylation observed in our papillary RCC patients. Methodological differences, such as the use of restriction-endonuclease q-PCR, also may contribute to the finding that VHL methylation was observed in a subset of patients who had nonclear cell RCC. For future studies, tumor material should be available for analysis of somatic VHL status. Thus, VHL status both in cell-free DNA and in the tumor can be evaluated. Taken together, detection of methylation status may increase the sensitivity and specificity of cellfree DNA for diagnosing RCC in the clinical setting of a solid renal mass. Furthermore, VHL methylation may be a novel, noninvasive tool for pretherapeutic subtyping of renal tumors. However, the specificity of total cell-free DNA levels and the methylation of cell-free DNA was high, but the sensitivity was very low. Because of the biologic heterogeneity of RCC, a single marker may not be sufficient to assist in clinical decision making. Instead, a combination of multiple markers, such as total-cell free DNA levels and VHL methylation, likely will increase sensitivity and overall predictive accuracy.
The current results demonstrate that cell-free circulating DNA is associated with TNM classification and may serve as a prognostic marker. Gang et al 4 noted that the detection of large fragments was associated with higher T classification. Hauser et al 10 reported an association of vascular invasion with DNA integrity. In both of those studies, however, survival was not assessed as an endpoint. We observed that higher total cell-free circulating DNA levels were associated with metastatic disease, tumor necrosis, and poorer survival. If these findings are confirmed by others in multivariate models, then cell-free circulating DNA levels may be integrated in prognostic models and may further enhance the predictive accuracy of established pathologic prognostic factors as it has for other biomarkers. [31][32][33][34][35] These models may assist in tailoring postoperative surveillance and therapy. High-risk populations that may benefit from adjuvant therapies may be defined; equally, low-risk groups may be identified in which repeated surveillance imaging and hospital visits can be avoided. Because our follow-up was of short duration and the number of deaths from RCC was relatively low, results from our multivariate analyses should be viewed with reservation. To avoid over fitting of the Cox model, TNM classification, tumor size, necrosis, and grade were not entered as separate variables but were summarized as the SSIGN score. In addition, cell-free DNA was significant only as a categorical variable, and prognostic information would be lost if it was categorized as a continuous variable. A categorical variable, however, may be more appropriate for routine clinical use. Taken together, our preliminary study generates the hypothesis that total cell-free DNA may play a role as a prognostic marker. This hypothesis has to be tested in external, prospective datasets. Our analyses also demonstrated that total cell-free DNA levels decrease after curative nephrectomy. It would be important to determine whether the persistence of cell-free DNA levels after nephrectomy correlates with the risk of recurrence. Our small subgroup cohort with relatively short follow-up, however, did not allow reliable statistical analyses regarding this issue. Future studies should address this, confirm our data in more patients, and focus on the prognostic value of cell-free DNA as a continuous variable.
Methodological inconsistencies regarding studies on cell-free DNA need to be acknowledged, including the source of DNA (plasma/serum), the impact of different kits on DNA purification, the mode of detection and the inconsistency of control populations. DNA concentrations in serum are about 6 times higher than in plasma, although the reason for this difference is unknown. 5 Second, different DNA extraction kits may offer different purification efficacies. 36,37 Third, different techniques have been applied to detect cell-free DNA and methylation, including fluorometric assays, spectrophotometry, bisulphite sequencing, conventional PCR, and q-PCR. 5 Thus, procedures are not standardized, and this limits our ability to compare studies. Furthermore, we want to emphasize that the composition of the control group is of paramount importance for interpreting results. Many studies use healthy individuals, 8 some studies use relevant control groups like those with elevated prostate-specific antigen levels and negative prostate biopsies, 18 and other studies do not specify the control group in detail. 4 This may explain in part the enormous range of cell-free DNA specificity (58%-98%), sensitivity (58%-85%), and AUC values (0.708-0.881) that have been reported for the diagnosis of prostate cancer. 2 A control group should consist of patients to whom the diagnostic test would be applied in clinical practice, 38 such as patients with renal masses of unknown histology. Healthy individuals, such as blood donors, are less appropriate.
In conclusion, the results from this study suggest that cell-free DNA represents a novel, serum-based diagnostic and prognostic biomarker for RCC: Total circulating cell-free DNA concentration and CpG island methylation of RASSF1A and VHL may be useful diagnostic biomarkers for RCC; VHL methylation may be a novel, noninvasive tool for pretherapeutic subtyping of Original Article renal tumors; and higher total circulating cell-free DNA concentrations may be associated with metastatic disease, tumor necrosis, and diminished survival. However, we must emphasize that this was a study on a limited number of patients with short follow-up duration. Thus, external validation of these data will be required before serum cellfree DNA can be implemented as routine biomarker. Furthermore, a longer follow-up will be necessary to evaluate the role of serum cell-free DNA as a prognostic factor.