Diagnostic and prognostic potential of circulating cell-free genomic and mitochondrial DNA fragments in clear cell renal cell carcinoma patients

Clinica Chimica Acta 452 (2016) 109–119

Contents lists available at ScienceDirect

Clinica Chimica Acta journal homepage: www.elsevier.com/locate/clinchim

Diagnostic and prognostic potential of circulating cell-free genomic and mitochondrial DNA fragments in clear cell renal cell carcinoma patients Hongbiao Lu a,b,c, Jonas Busch a, Monika Jung a, Silke Rabenhorst a, Bernhard Ralla a, Ergin Kilic d, Steffen Mergemeier e, Nils Budach f, Annika Fendler a,b,g, Klaus Jung a,b,⁎ a

Department of Urology, University Hospital Charité, Berlin, Germany Berlin Institute for Urological Research, Berlin, Germany Department of Urology, Changzhou No.2 People's Hospital, Jiangsu, China d Institute of Pathology, University Hospital Charité, Berlin, Germany e CONGEN Biotechnologie GmbH, Berlin, Germany f Department of Radiology, University Hospital Charité, Berlin, Germany g Department of Signal Transduction, Invasion and Metastasis of Epithelial Cells, Max Delbrück Center of Molecular Medicine, Berlin, Germany b c

a r t i c l e

i n f o

Article history: Received 22 October 2015 Received in revised form 9 November 2015 Accepted 9 November 2015 Available online 10 November 2015 Keywords: Clear cell renal cell carcinoma Circulating cell-free DNA Diagnostic Prognostic markers Metastasis Predictive models

a b s t r a c t Background: There is inconsistent information about the clinical usefulness of circulating cell-free DNA (cfDNA) in plasma from clear cell renal cell cancer (RCC) patients. This is attributed to preanalytical, analytical, and clinical factors that were considered as far as possible in this study. Methods: cfDNA was extracted from EDTA plasma of healthy people (n = 40), non-metastatic (n = 145) and metastatic (n = 84) RCC patients using the QIAamp Circulating Nucleic Acid Kit. Genomic and mitochondrial cfDNA concentrations were determined using qPCR of different cfDNA fragments (67–306 bp). Their diagnostic and prognostic potential was estimated using receiver operating characteristics (ROC) and Cox regression analyses. Results: The 67 bp and 180 bp genomic cfDNA fragments did not differ between the three study groups while the 306 bp fragment was lower in RCC patients than in controls. The mitochondrial cfDNA was higher in metastatic than in non-metastatic patients and controls. The cfDNA integrity indices decreased from controls to metastatic patients. Models built by logistic regression and Cox regression resulted in area under the ROC curves N0.75 and concordance indices of N 0.800 in predicting recurrence-free survival and overall survival. Conclusion: The study suggests that combinations of cfDNA markers have promising diagnostic and prognostic potential in RCC patients and are worth for further validation in future prospective multicenter studies. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Renal cell cancer (RCC) is with 2.4% of all adult malignancies one of the ten most frequent cancers worldwide. In 2012, RCC accounted for 338,000 new cases and 144,000 deaths [1]. Three main histological RCC morphotypes are distinguished, while the clear cell RCC is with approximately of 80–90% the most frequent one, followed by the

Abbreviations: APP-1, APP-2, and APP-3, fragments of 67, 180, and 306 bp of the gene amyloid beta (A4) precursor protein (APP); AUC, area under the ROC curve; bp, base pairs; ccRCC, clear cell renal cell carcinoma; cfDNA, circulating cell-free DNA; CI, confidence interval; C-index, concordance index; DCA, decision curve analysis; G, histopathological grading according to Fuhrman; Mito-1 and Mito-2, mitochondrial DNA sequences of 65 and 175 bp; pT, pathological tumor classification; R, surgical margin classification; ROC, receiver operating characteristics; rS, Spearman rank correlation coefficient; SINE-1 and SINE-2, Alu sequences with of 79 and 248 bp. ⁎ Corresponding author at: Department of Urology, CCM, University Hospital Charité, Schumannstr. 20/21, 10117 Berlin, Germany. E-mail address: [email protected] (K. Jung).

http://dx.doi.org/10.1016/j.cca.2015.11.009 0009-8981/© 2015 Elsevier B.V. All rights reserved.

papillary and chromophobe RCC with 15 and 5%, respectively. For localized, early-stage tumors, partial or total nephrectomy in curative intent represents the gold standard. However, approximately 30% of the RCC patients present evidence of distant metastasis, which is associated with poor prognosis [2]. After nephrectomy, one-third of the patients with initially localized RCC develop metastasis [3]. The diagnostic and prognostic indicators for patients with RCC are mainly based on traditionally clinicopathological and radiological examinations. The lack of non-invasive blood or urine markers and the inherent limitation of the diagnostic and prognostic models, which were built with the help of these conventional data, are essential short-comings in the management of these patients [4]. Therefore, new molecular markers are urgently needed for higher diagnostic accuracy and predicting the clinical outcome of patients with RCC. In this respect, circulating cell-free DNA (cfDNA) might be a promising option. Most cfDNAs are double-stranded molecules (approximately 0.18 kB to 21 kB), which circulate as nucleoprotein complexes in blood. Detection of cfDNAs in plasma or serum is possible not only in cancer

110

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

patients but also in healthy individuals [5]. DNA released from tumor cells by different molecular processes like cell apoptosis, necrosis, micrometastasis, and secretion has been considered to be an important cancer biomarker [6]. The size distribution of cfDNA from cell apoptosis and necrosis is different. Apoptotic cells release cfDNAs of 180 to 200 bp or multiples of this size, while necrotic cells generally generate longer DNA fragments [7,8]. The ratio of long to short DNA fragments, termed as integrity index, was reported to be a useful tool to differentiate these processes [9]. Increased concentrations of cfDNA and altered integrity indices were reported in patients with breast, liver, lung cancer, ovarian, prostate, and gastric cancer (reviewed in [10]). In rare reports about RCC, the level of plasma cfDNA has also been demonstrated as a useful diagnostic and prognostic marker [11–14]. However, as already criticized by de Martino et al. [12], the studies included generally few patients and were therefore underpowered and limited in predicting the outcome of patients. In addition, most studies used serum as sample. However, serum is generally considered a confounder variable as part of serum DNA does not correspond to the cfDNA because of the released DNA from blood cells during the clotting process [15]. Moreover, new data on cfDNA proved that fragments b100 base pairs (bp) and also mitochondrial DNA were more relevant for tumor cfDNA [16–19]. These new aspects, in addition with a higher analytical sensitivity using Alu sequences as new measurement tools [20], have prompted us to re-assess these factors of cfDNA measurements in RCC patients with regard to their diagnostic and prognostic validity. Alu sequences are repetitive DNA sequences that occur in multiple copies in the human genome. These elements have approximately 300 bp and belong to the group of the short interspersed elements (SINEs) with less than 500 bp [21]. Therefore, the aims of our study including controls and patients with non-metastatic and metastatic clear cell RCC were (a) to quantify plasma cfDNA with different fragments based on the determination of the gene amyloid beta (A4) precursor protein (APP), Alu sequences, and mitochondrial DNA, (b) to estimate the association between cfDNA and clinicopatholological variables, (c) to evaluate the diagnostic usefulness of these single markers, their ratios, and combinations regarding their discrimination ability between the study groups, and (d) to assess the validity of these markers in predicting the outcome of patients regarding recurrence-free survival and overall survival after nephrectomy. 2. Materials and methods 2.1. Patients and samples The study was approved by the local University Hospital Ethics committee and informed patients consent was obtained. The study was carried out in accordance with the Declaration of Helsinki. The REMARK and STARD guidelines were correspondingly applied [22,23]. In this retrospective study, a total of 269 subjects were investigated during 2005–2012 and followed up to 2014. The control group consisted of 40 healthy subjects with no evidence of malignancies, infections or gastrointestinal, hepatic, immunologic, renal or other serious diseases. Only patients suffering from a clear cell RCC as the most frequent RCC was included in this study. This RCC cohort was subdivided into 145 patients without metastases at the time of partial or radical nephrectomy and 84 patients with metastases at the time of nephrectomy, at initiation or during targeted therapy. The clinical and clinicopathological data of these study groups are summarized in Table 1. Blood samples of the RCC patients were collected in K2EDTA Vacutainer (Becton Dickinson, Heidelberg, Germany) one to 24 days before nephrectomy or before/during targeted therapy and centrifuged (2000 × g, 10 min, 4–8 °C) within 30 min after venipuncture. To prevent a cellular DNA contamination [24] the plasma supernatants were

carefully removed and re-centrifuged (16,000 × g; 10 min at 4–8 °C). The prepared plasma samples were archived at −80 °C until analysis. 2.2. DNA extraction and quantitative real-time PCR All methods were performed according to the MIQE guidelines and were documented in a corresponding checklist found in Supplementary Data [25]. Total DNA was isolated with QIAamp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany) according to the Qiagen Blood and fluids protocol. Each column was loaded with 1 ml plasma and the extracted DNA was eluted with AVE buffer in final volumes of 30 μl. Aliquots were stored at −20 °C until analysis. Two plasma pools were used for intra-run and between-run precision controls. DNA was quantified using Quant-iT™ PicoGreen® dsDNA Reagent (ThermoFisher, Invitrogen, Darmstadt, Germany) in Greiner 384 well plates (Sigma-Aldrich, Taufkirchen, Germany) on the microplate reader Mithras LB 940 (Berthold Technologies, Bad Wildbad, Germany). Quantitative real-time PCR (qPCR) was performed on the Light-Cycler 480 instrument (Roche Diagnostics GmbH, Mannheim, Germany). Human Genomic DNA (Roche; cat.no. 11691112001; 200 μg/ml) was used as traceable standard for all measurements. APP (HGNC id: 620) was used as target for the quantification of cfDNA fragments with the amplicons of 67, 180, and 306 bp according to Pinzani et al. [26]. These assays use the same forward primer and hydrolysis probe but specific reverse primers. Final reaction volumes of 10 μl were used consisting of 5 μl of LightCycler® 480 Probes Master, 1 μl DNA sample, 100 pmol/ml of the 5′-FAM-and 3′-TAMRA labeled probe ACCCCAgAggAgCgCCACCTg, and 250 pmol/ml forward and reverse primers (forward APP F: 5′-TCAggTTgACgCCgCTgT; reverse APP R1: 5′-TTCgTAgCCgTTCTgCTgC, size of PCR product is 67 bp, in the following termed APP-1; reverse APP R2: 5′-TCTATAAATggACACCgATgggTAgT, size of PCR product is 180 bp, in the following termed APP-2; reverse APP R3: 5′-gAgAgATAgAATACATTACTgATgTgTggAT, size of PCR product is 306 bp, in the following termed APP-3). The preincubation step was at 95 °C for 10 min, followed by 45 cycles of the amplification program (95 °C for 10 s, 57 °C for 20 s, 72 °C for 40 s), and then 1 min cooling to 40 °C. The qPCRs of SINEs and mitochondrial DNA were performed using Human SINE Screen 3plex Kits and Human Mitochondrion Screen 3plex Kits (CONGEN Biotechnologie, Berlin-Buch, Germany) according to the manufacturer's protocol on the same Instrument. As determined by gel electrophoresis, the lengths of the fragments for SINE-1 and SINE2 were 79 and 248 bp while the mitochondrial DNA fragments Mito-1 and Mito-2 had lengths of 65 and 175 bp, respectively (Supplementary Data, Fig. S1). The kits contain the ready reaction solutions with all components and a separate Taq Polymerase without giving further details on the primer sequences etc. Before analysis, a reagent mixture of one volume fraction of Taq Polymerase with 199 volume fractions of the ready reaction solution was prepared. The final reaction mixture consisted of 20 μl of this reagent mixture and 5 μl of 1:10 diluted DNA sample. Different detection channels (FAM and CY5) were used in each reaction for the long and short fragments. Calibrators, controls, and samples were analyzed in triplicate using LightCycler® 480 software, release 1.5.0; and water blanks were included in every run. Standard curves were performed using serial dilutions of DNA. PCR product specificities were confirmed by gel electrophoretic separation of the PCR products. All further methodical details are compiled within the Supplementary Data. 2.3. Data analysis and statistical analysis The differences between runs were normalized with the inter-plate calibrators by qbasePLUS software (Biogazelle, Gent, Belgium). Statistical analyses were performed using SPSS 23 (SPSS Inc., Chicago, IL, USA), GraphPad Prism 6.07 (GraphPad Software, La Jolla, CA, USA), and MedCalc 15.8 (MedCalc Software bvba, Ostend, Belgium). Non-

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

parametric statistical tests (Mann–Whitney U-test, Kruskal–Wallis test, and Spearman rank correlation) were used. Analysis of receiver operating characteristics (ROC) curves and binary logistic regression was performed to evaluate the diagnostic performance of cfDNA levels. Kaplan–Meier survival analysis approach and Cox proportional hazard regression analysis were used for analyses of recurrence-free survival and overall survival. The decision curve analysis (DCA) [27] for models and corresponding indicated markers were constructed with MATLAB software as previously described [28]. The C-index was calculated as a global measure for validating the predictive reliability of survival models [29]. A p value less than 0.05 (two-sided) was considered significant. MedCalc and G*Power 3.1.9 [30] were used for sample size determinations and power calculations. For sample size calculations, the thresholds of α = 5% (significance level) and β = 20% (1-power; power of 80%) with a medium effect size were selected to avoid both type I and type II errors and to detect already relatively low differences (Supplementary Data). 3. Results 3.1. Patient characteristics and analytical performance data A total of 229 clear cell RCC patients without (n = 145) and with (n = 84) metastases were recruited; their detailed clinicopathological characteristics are summarized in Table 1. 40 healthy individuals (17 males and 23 females) with a median age of 57 years were enrolled as controls. Global cfDNA measurements using the PicoGreen method resulted in a median concentration of 20.5 ng/ml in all samples (95% CI 19.7–23.4, range 6.52–397 ng/ml). The performance for all specific qPCR analyses based on the inter-assay precision of Cq values was between 0.41 and 1.81%. Detailed performance data of the individual cfDNA measurements using the APP gene (APP-1, APP-2, APP-3), the Alu sequences (SINE-1 and SINE-2), and the mitochondrial DNA fragments (Mito-1 and Mito-2) were compiled according to the MIQE guidelines and summarized in Supplementary Data (Supplemental Tables S1-3 and Supplemental Fig. S1). All these data confirmed the analytical specificity of measurements and also excluded analytical interferences due to possible cross-reactivities in the multiplexed assays. 3.2. cfDNA markers in association to clinicopathological characteristics and among each other There were no significant associations between the single cfDNA markers, sex (Mann–Whitney test, p values of 0.231 to 0.931), and age (rS between 0.105–0.001, p values of 0.086 to 0.993) except for APP-1 and APP-2 with very low rS of 0.174 and 0.213 that only explain about 4% of the variations of these markers in relation to age. With regard to the pathological variables tumor stage, tumor grade, and marginal status, there were no significant correlations of cfDNA markers (rS with p values of 0.087 to 0.938) except for the relationships of Mito-1 and Mito-2 with the Fuhrman grade (rS of 0.209 and 0.206, p values of 0.0121 and 0.014, respectively). In Table 2, the correlation coefficients between the individual cfDNA markers are given for all 269 subjects of the study cohort, since the coefficients of the three subgroups did not generally differ among the three subgroups (Supplementary Data: Supplemental Table S4). All fragments of APP and SINE were strongly correlated between each other with correlation coefficients between 0.545 and 0.998 (p b 0.0001), but they were only occasionally correlated with rS b 0.15 to the two mitochondrial DNA fragments. Moreover, the correlations between APP-1 and APP-2 (rS = 0.926) significantly (p b 0.0001) differed between the correlations of APP-1 to APP-3 (rS = 0.605) as well as of APP-2 to APP-3 (rS = 0.738). All these data allow the conclusion that similar but also various aspects could be assessed with this

111

Table 1 Characteristics of the patients suffering from clear cell renal cell carcinoma evaluated in the study. Characteristics

Primary ccRCCa non-metastatic (n = 145)

Metastatica ccRCC (n = 84)

p-Valueb

Age, yrs., median (95% CI) Gender, male/female Pathological stage, n (%) pT1 pT2 pT3 pT4 Unknown Fuhrman grade, n (%) G1 G2 G3 G4 Unknown Surgical margins, n (%) R0 R1 Unknown Patients followed,c n (%) Metastasis at follow-up, n (%) Death at follow-up, n (%) Follow-up, mo, median (95% CI) Recurrence-free survival Overall Plasma collection timepoint Before surgery Before targeted therapy During targeted

65 (62–67) 94/51

68 (64–70) 70/14

0.365 0.004

97 (67) 9 (6) 38 (26) 1 (1) 0 (0)

12 (14) 10 (12) 56 (67) 1 (1) 5 (6)

b0.0001

16 (11) 108 (74) 20 (14) 1 (1) 0 (0)

3 (4) 50 (60) 21 (25) 2 (2) 8 (10)

0.001

138 (95) 7 (4) 0 (0) 132/138 (86/95) 18 (14) 21 (15)

52 (62) 12 (14) 20 (24)

b0.0001

42.4 (33.7–49.1) 45.5 (40.9–51.9) 145 (100) – –

22 (26) 20 (24) 42 (50)

Abbreviations: ccRCC: clear cell renal cell carcinoma; CI: confidence interval; pT: pathological tumor classification, G: histopathological grading according to Fuhrman, R: surgical margin classification. a Imaging techniques were used to provide evidence of presence/non-presence of metastases. b p-Values from Fisher's exact test, Chi-square test or Mann–Whitney U-test. c Available information with regard to tumor recurrence and overall survival after surgery differed (132 and 138, respectively out of 145 patients) and explains the discrepant event data.

approach of using different measurement conditions of genomic and mitochondrial cfDNA in RCC patients. Fig. 1 displays scatterplots of the individual concentrations of the genomic and mitochondrial cfDNA fragments and their median values of controls in comparison to non-metastatic and metastatic RCC patients. The patients with metastases at the time of nephrectomy, at the initiation or during a systemic therapy were combined into one metastatic group (Table 1), since all the cfDNA marker concentrations did not differ between the subgroups (Kruskal–Wallis test, p = 0.338–0.607 with the Jonckheere–Terpstra trend test that the medians are not significantly ordered, p = 0.154–0.643). Using the fragments APP-1, APP-2, SINE-1, and SINE-2 for the measurements, there were no significantly different cfDNA concentrations among the three study groups. The control group showed a higher median APP-3 concentration than the two RCC groups (p b 0.0001) that had no different median APP-3 values between each other (p = 0.442). In contrast, higher Mito-1 and Mito-2 concentrations were found in the metastatic RCC group compared to non-metastatic RCC patients and controls (p b 0.0001), while there were no statistically significant differences of Mito-1 and Mito-2 between non-metastatic RCC patients and controls (p = 0.196 and 0.237). Since cfDNA fragment sizes are indicators of the integrity of cfDNA molecules, it was of special interest to compare the DNA concentrations within the same study groups measured by using the APP fragments of 67, 180, and 306 bp as well as the SINE- and Mito-DNA fragments. These data are summarized in Fig. 2. In the control group (Fig. 2a), there were no significantly different DNA concentrations for all genomic and mitochondrial fragments (all p N 0.100). However, in both RCC groups

112

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

(Fig. 2b and c), the concentrations of APP-1 fragments as well as the SINE-1 and Mito-1 fragments were significantly higher than APP-2 and APP-3 and the corresponding SINE-2 and Mito-2 fragments, respectively. The cfDNA integrity index [9] was calculated as ratio of longer to shorter fragments for all fragments in the three study groups (Fig. 3). All five genomic and mitochondrial integrity indices were significantly higher in controls compared with those in metastatic RCC patients. These differences were also observed in non-metastatic patients for the ratios APP-3/APP1 and APP-2/APP-1. In addition, the ratios of the SINE and Mito fragments as well as the ratio of APP-3 to APP-2 were higher in non-metastatic compared with metastatic patients. The integrity index for the genomic and mitochondrial DNA showed a decreased trend from controls to metastatic RCC patients and was be typical for RCC. It reflects the situation of higher DNA fragmentation in the development and progression of RCC.

3.3. cfDNA markers as discriminators between controls and RCC patients To evaluate the diagnostic discrimination potential of the seven single DNA markers and the five ratios, ROC analyses were performed with regard to the differentiation between (a) controls and all RCC patients, (b) controls and only non-metastatic patients, and (c) between non-metastatic and metastatic RCC patients. The approach of binary logistic regression was additionally used to filter the most discriminative markers and set up valid models by combining the most useful markers. In order to facilitate the overview for the reader, the detailed data for the individual markers are given in the Supplementary Data (Supplemental Tables S6-8) and the final results are compiled in Table 3 and Fig. 4 as the models with the combination of the most useful markers. The ROC data prove that (a) combinations of markers in comparison to the use of single markers improve the discrimination between the study groups and (b) a reduced number of markers instead of all markers is possible by establishing meaningful models without affecting the discrimination capability between the groups. These results are supported in Fig. 4 by the decision curve analyses (DCA). Fig. 4b and d illustrate the net benefit of the marker combination of Model C in comparison to the ratio of APP-2/APP-1 in the DCA curve that was shown in the corresponding ROC curves. On the other hand, the ROC and DCA curves in Fig. 4e and f show that another combination of markers, shown as Model D was needed for a successful discrimination between non-metastatic and metastatic RCC patients. However, acceptable AUC values of 0.78 could be obtained in all cases with the combination of four markers in the Model C (Mito-1, Mito-2, APP-1, and APP-2) and Model D (Mito-1, Mito-2, SINE-1, and SINE-2). As Mito-1 and Mito-2 as well as SINE-1 and SINE-2 were measured using

multiplexed assays, three or two assays were only necessary for these discrimination results between the different groups. 3.4. cfDNA markers as prognosticators for tumor recurrence and overall survival The different behavior of the various cfDNA markers in the study groups suggested the use of these markers as prognostic tools in RCC patients. We assessed therefore their predictive capacity regarding the two outcome criteria recurrence-free survival and overall survival in non-metastatic patients after nephrectomy with available follow-up data (Table 1). The time from the date of surgery to tumor recurrence or to the last follow-up and the time of death, respectively, were used to calculate the clinical endpoints recurrence-free survival and overall survival. For an initial overview, Kaplan–Meier curves were calculated to assess the association between these two outcome criteria and dichotomized cfDNA markers based on the cutoffs obtained as the point of maximal accuracy (Youden index) in the ROC analysis (Supplementary Data, Figs. S1 and S2). The recurrence-free interval significantly decreased with increasing pT stage and Fuhrman grade, indicating the representativeness of the study cohorts. The prognostic performance of the individual cfDNA markers was assessed by Cox regression analyses. In univariate analyses, the hazard ratios of cfDNA markers reflected the results of the Kaplan–Meier curves (Supplementary Data, Table S4). In fully and stepwise reduced multivariate Cox regression analyses, the prognostic potential of the significant individual cfDNA variables in univariate analyses were always tested with the clinicopathological variables age, sex, tumor stage and grade, and marginal status. Three models were established (Table 4). The C-indices of the models showed distinctly improved prognostic capabilities if selected cfDNA markers were combined with conventional clinicopathological factors in comparison to models based only on clinicopathological variables (Table 4). To enhance the reliability of the results, internal validations in the final models were performed by bootstrapping. 4. Discussion In this study, the results of the simultaneous measurement of genomic and mitochondrial cfDNA fragments revealed their potential as diagnostic and prognostic biomarkers in clear cell RCC patients. In this respect, the current study represents, to the best of our knowledge, the most comprehensive investigation of this patient group. Our data generally confirm and complement the estimation of previous studies regarding the usefulness of these markers [11–14,16]. However, definite differences emerge between our results and the heterogeneous data situation of previous quantitative PCR studies which may be taken

Table 2 Spearman rank correlation coefficients between the single cfDNA fragments in plasma samples of the combined groups of healthy controls and ccRCC patients (n = 269).

APP-1 APP-2 APP-3 SINE-1 SINE-2 Mito-1 Mito-2

rS p-Value rS p-Value rS p-Value rS p-Value rS p-Value rS p-Value rS p-Value

APP-2

APP-3

SINE-1

SINE-2

Mito-1

Mito-2

0.926 b0.0001

0.602 b0.0001 0.738 b0.0001

0.926 b0.0001 0.886 b0.0001 0.545 b0.0001

0.913 b0.0001 0.888 b0.0001 0.567 b0.0001 0.952 b0.0001

0.041 0.505 0.041 0.505 −0.095 0.119 0.125 0.040 0.071 0.244

0.049 0.420 0.046 0.455 −0.095 0.119

0.738 b0.0001 0.886 b0.0001 0.888 b0.0001 0.041 0.505 0.046 0.456

0.545 b0.0001 0.567 b0.0001 −0.095 0.119 −0.095 0.119

0.952 b0.0001 0.125 0.040 0.132 0.030

0.071 0.244 0.080 0.192

0.080 0.192 0.998 b0.0001

0.998 b0.0001

Abbreviations: APP-1, APP-2, and APP-3: fragments of 67, 180, and 306 bp of the gene amyloid beta (A4) precursor protein (APP); ccRCC: clear cell renal cell carcinoma; Mito-1 and Mito-2: mitochondrial DNA sequences of 65 and 175 bp; p-Value: significance level with p b 0.01 highlighted in bold/italics figures; rS: Spearman rank correlation coefficient; SINE-1 and SINE-2: Alu sequences with of 79 and 248 bp.

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

113

Fig. 1. Plasma concentrations of cfDNA markers in healthy controls and patients with clear cell renal cell carcinoma without (RCC non-metastatic) and with metastases (RCC metastatic). Medians are indicated as horizontal lines. Significant differences between the study groups were estimated by the Kruskal–Wallis test with Dunn's post test adjusted to account for multiple comparisons.

into account for comparison purposes. Interestingly, the design of our study considered critical shortcomings of cfDNA analyses in cancer and especially RCC patients as summarized in previous studies or reviews and was briefly explained in the introduction of this manuscript [12,15,18,31]. These deficiencies particularly relate, among others, to the partly low number of patients included in the studies, the heterogeneity of RCC patients with regard to the TNM stages and/or histological tumor morphotypes, the general determination of fragments N 100 bp that are recently estimated as rather untypical for tumor cfDNA, and the use of serum instead of plasma. All these clinical, pre-analytical, and analytical inconsistencies between the studies might contribute to these differences and should be discussed in relation to our data in the following.

All the measurements were performed with isolated cfDNA from EDTA plasma samples. This kind of sample and the following two-step centrifugation process avoid the interference by the background DNA released from hematopoietic cells when serum is used and the contamination of DNA from platelets and cell debris after the first low speed centrifugation [15,24]. In addition, using the QIAamp Circulating Nucleic Acid Kit for the cfDNA isolation an optimal recovery of fragmented DNA as short as 64 bp is achieved in comparison to other isolation techniques [32]. In a recent external quality assessment of 62 different cfDNA extraction methods performed in 21 European countries, the QIAmp kit was classified in the best group of “dedicated” methods [33]. The concentrations of the different genomic cfDNA fragments were determined using established assays with the same forward primer

114

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

Fig. 2. Separate comparison of the associated genomic and mitochondrial cfDNA markers of different fragment lengths in (a) controls and (b) patients with clear cell renal cell carcinoma without and (c) with metastases. Medians are indicated as horizontal lines. Significant differences between the corresponding fragments were estimated by the Kruskal–Wallis test with Dunn's post test adjusted to account for multiple comparisons.

and hydrolysis probe but specific reverse primers for the gene APP [26]. This gene is found on chromosome 21q21.3. Mutations of this location in clear cell RCC are not yet known [34] so that cfDNA measurements could be assumed to unaffected using this assay principle. Based on

these strictly controlled pre-analytical and analytical conditions, the assay performance data summarized in the Supplementary Data can be considered as evidence for the reliability of the gained information by this study.

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

115

Fig. 3. Ratios of long to short cfDNA fragments as integrity indices in controls and patients with clear cell renal cell carcinoma without (RCC non-metastatic) and with metastases (RCC metastatic). Medians are indicated as horizontal lines. Significant differences between the study groups were estimated by the Kruskal–Wallis test with Dunn's post test adjusted to account for multiple comparisons.

Table 3 Receiver operating characteristic (ROC) curve analyses of cfDNA marker based models to discriminate between healthy persons (n = 40) and ccRCC patients without (n = 145) and with (n = 84) metastases. Model

AUC (95% CI)

p-Value different to AUC = 0.5

Differentiating ability at the Youden indexa Sensitivity (95% CI)

Specificity (95% CI)

Differentiation between healthy persons and non-metastatic and metastatic RCC patients Model A (all variables)b Model B (APP-3, Mito-1, APP-3/APP-1, APP-2/APP-1, SINE-2/SINE-1, Mito-2/Mito-1)c Model C (Mito-1, APP-2/APP-1, Mito-2/Mito-1)d

0.84 (0.79–0.88) 0.81 (0.76–0.86) 0.80 (0.75–0.85)

b0.0001 b0.0001 b0.0001

70 (64–76) 76 (70–81) 58 (51–65)

88 (73–96) 78 (72–89) 90 (76–97)

Differentiation between healthy persons and non-metastatic RCC patients Model A (all variables)b Model C (Mito-1, APP-2/APP-1, Mito-2/Mito-1)c

0.82 (0.77–0.88) 0.78 (0.71–0.84)

b0.0001 b0.0001

70 (62–77) 55 (46–63)

88 (73–96) 90 (76–97)

Differentiation between non-metastatic and metastatic RCC patients Model A (all variables)b Model C (Mito-1, APP-2/APP-1, Mito-2/Mito-1)c Model D (Mito-1, Mito-2, SINE-1, SINE-2/SINE1)e

0.80 (0.75–0.85) 0.71 (0.64–0.76) 0.78 (0.73–0.84)

b0.0001 b0.0001 b0.0001

85 (75–91) 67 (56–77) 77 (67–86)

65 (57–73) 75 (67–81) 71 (63–78)

Abbreviations: AUC: area under the ROC curve; ccRCC: clear cell renal cell carcinoma; CI: confidence interval; abbreviations of the cfDNA markers: see Table 2. a The Youden index as a measure of overall diagnostic effectiveness is calculated by (sensitivity + specificity) − 1. When equal weight is given to sensitivity and specificity of a test, the cutoff at the maximum value of this index, which graphically corresponds to the maximum vertical distance between the ROC curve and the diagonal line, is referred to as optimal criterion. b Calculated by binary logistic regression using all markers and ratios. Detailed data for the single cfDNA markers and the ratios are given in Supplementary Data (Supplemental Tables S6–8). c Calculated by binary logistic regression using both significant single markers and ratios and if highly correlated (rS N 0.800) only with the inclusion of the markers with the highest AUC values. d Calculated by the backward elimination approach (p b 0.050 for entry and p N 0.100 for removal) of binary logistic regression using the markers of Model B. e Calculated by the backward elimination approach of binary logistic regression using all variables.

116

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

Fig. 4. Receiver-operating characteristics (ROC) curves and decision curve analyses (DCA) by using different models for the discrimination between healthy controls and clear cell renal cell carcinoma patients. Models were built by binary logistic regression using either all cfDNA variables including the ratios, the reduced number of cfDNA variables using only significant, but uncorrelated markers according to the procedure described in Table 3 or the markers obtained after the backward elimination approach of logistic regression (p b 0.05 for entry and p N 0.100 for removal). (a) ROC and (b) DCA curves regarding the discrimination between controls (n = 40) and all RCC patients (n = 229): Model built with all cfDNA variables; Model B built with the six remaining variables selected on the basis of the significant AUC values and the correlation approach (Table 3; APP-3, Mito-1, ratios of APP-3/APP-1, APP-2/ APP-1, SINE-2/SINE-1, and Mito-2/Mito-1); Model C built with the three remaining variables (Mito-1 and the ratios APP-2/APP-1 and Mito-2/Mito-1) after backward elimination with all the variables used in Model B. The integrity index AAP-2 to AAP-1 as the best “single” indicator was used for comparison. (c) ROC and (d) DCA curves regarding the discrimination between controls and patients without metastasis at the time of nephrectomy (n = 145): Model A built with all variables, Model C and APP-2/APP-1 as described above. (e) ROC and (f) DCA curves regarding the discrimination between patients without metastasis at nephrectomy (n = 145) and patients with metastases at nephrectomy, before and under systemic therapy (n = 84): Models A and C as described above, model D was built with the combination of Mito-1, Mito-2, SINE-2, and the ratio SINE-2/SINE-1 obtained as the significantly remaining variables after backward elimination of logistic regression with all variables in this study cohort.

Genomic cfDNA fragments of APP with 67 bp (APP-1) and 180 bp (APP-2) as well as of Alu sequences with 79 bp (SINE-1) and 248 bp (SINE-2) were not different between the controls and non-metastatic RCC patients, but metastatic RCC patients showed lower concentrations of the long 306 bp APP-3 fragment compared to the controls. These data

are in contrast to the results of the four mentioned previous studies in RCC patients that could be used for comparison despite their different study design [11–14]. These studies described higher cfDNA concentrations in RCC patients regardless of the metastatic status. Two of these four studies used serum [11,12] and EDTA plasma [13,14], respectively

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

but distinctly higher cfDNA values in plasma than in serum were surprisingly reported. In addition, there were no significant associations between cfDNA concentration and tumor stage or Fuhrman grade as we also found [11,12]. On the other hand, as shown in Fig. 3, the integrity index of the genomic cfDNA was characterized by a decreased trend from controls to metastatic RCC patients as typical indicator of higher DNA fragmentation in the development and progression of RCC. This effect is underlined by the cfDNA concentrations that depend on the assays with the different APP fragment sizes in the three study groups (Fig. 2a–c). The cfDNA concentrations in the controls measured with the assays for the short APP-1 and longer APP-2 and APP-3 fragments did not differ (Fig. 2a). These results confirm recently published experiences that the used cfDNA QIAmp isolation kit is dedicated to fulfill the cfDNA quality parameter “integrity” [33]. In contrast, significantly higher concentrations of the short APP-1 in comparison to APP-2 and APP-3 were found in the RCC groups (Fig. 2b,c) because of the higher fragmentation of cfDNA in the patients. The data allows the conclusion that measurements of total cfDNA concentrations in RCC patients require assays with gene fragment sizes below 100 bp. A similar recommendation was also given for other cancers [18]. The phenomenon of decreased cfDNA integrity indices in RCC patients agrees with the results of Hauser et al. [11]. Moreover, the increased mitochondrial cfDNA concentrations in metastatic RCC in comparison to controls and non-metastatic RCC patients (Fig. 1f,g) as well their decreased integrity indices (Fig. 3e) were concordant to a previous study with 33 RCC patients [16].

117

This short overview about the concordant and discordant results obtained in the compared studies cannot explain the reason for all these discrepancies. As already mentioned, a summary of the various clinical, pre-analytical, and analytical conditions applied in the studies might be responsible for this. However, it should be noted that similar different results have been reported for various studies of the same type of cancer (breast, lung, and prostate) [10]. Thus, further studies should draw attention to these various critical points. The evaluation of our study data with regard to their diagnostic potential was primarily focused to find out the most useful single markers and combinations (Supplementary Data with Tables S6-S8). ROC analyses showed their discriminatory capability. However, particular importance was placed to reduce the number of seven single markers and five ratios by meaningful combinations preferably without any restrictions of diagnostic accuracy. The final results presented in Table 3 and Fig. 4 show appropriate models constructed on the basis of the single markers and their ratios by binary logistic regression analysis. For example, by using Model C or D for the corresponding clinical issues only two or three assays with four markers were necessary to reach AUC values N0.75 (Table 3). The decision curve analyses confirmed not only their benefit in comparison with the ratio APP-2/APP-1 as one of the best single marker combination but also its equivalence in comparison with all markers and ratios (Fig. 4). Thus, these two essential aspects of reduced markers without any impairment of the diagnostic information would be possible to apply these markers in clinical practice. The highest AUC values reported by Hauser et al. [11]

Table 4 Two models for predicting overall survival and recurrence-free survival in non-metastatic ccRCC patients after nephrectomy based on multivariate Cox proportional hazard regression analyses of cfDNA markers and with their C-indexes. Variable

Overall survival analysis (n = 138)a

Recurrence-free survival analysis (n = 132)a

HR (95% CI)

p-Value

HR (95% CI)

p-Value

Model Ab Age (years, cont.) Sex pT1 + 2/pT3 + 4 G1 + 2/G3 + 4 R0/R1 + 2 APP-1 APP-2 APP-3 SINE-1 SINE-2 Mito-1 APP-3/APP-1 APP-3/APP-2 APP2/APP-1 C-Indexc with all variables C-Indexc only clinical data

1.02 (0.96–1.07) 0.38 (0.11–1.30) 1.19 (0.43–3.27) 0.86 (0.24–3.10) 1.83 (0.20–17.2) 1.86 (0.26–13.1) 0.14 (0.01–1.52) 2.18 (0.47–10.1) 39.1 (3.12–89.7) 0.21 (0.03–1.66) N/A N/A 0.36 (0.13–1.00) 0.50 (0.13–2.00) 0.829 0.701

0.542 0.125 0.741 0.824 0.598 0.535 0.109 0.323 0.005 0.142 N/A N/A0.052 0.332

0.95 (0.90–1.00) 0.44 (0.11–1.73) 3.83 (1.05–13.9) 0.47 (0.09–2.23) 4.04 (0.63–25.8) 0.40 (0.02–8.35) N/A N/A 3.51 (0.11–112) 2.81 (0.11–70.2) 4.41 (0.89–21.9) 0.23 (0.02–2.28) 0.42 (0.15–1.15) 0.21 (0.04–1.02) 0.853 0.725

0.056 0.240 0.042 0.345 0.142 0.558 N/A N/A 0.479 0.531 0.071 0.213 0.093 0.104

Model Bd Age (years, cont.) Sex pT1 + 2/pT3 + 4 G1 + 2/G3 + 4 R0/R1 + 2 Mito-1 APP-3/APP-2 SINE-1 C-Indexc of the model

1.03 (0.98–1.08) 0.42 (0.13–1.39) 1.02 (0.39–2.67) 1.25 (0.40–3.90) 1.83 (0.20–16.5) N/A 0.42 (0.17–1.07) 4.69 (1.57–14.0) 0.836

0.203 0.156 0.971 0.704 0.591 N/A 0.068 0.006

0.98 (0.94–1.02) 0.38 (0.10–1.47) 3.65 (1.12–11.9) 1.10 (0.34–3.57) 5.18 (0.90–29.8) 5.43 (1.17–25.2) 0.31 (0.12–0.84) 4.82 (1.19–19.6) 0.827

0.308 0.160 0.032 0.872 0.065 0.031 0.021 0.028

Abbreviations: ccRCC: clear cell renal cell carcinoma; cfDNA: circulating cell-free DNA; CI: confidence interval; C-Index: concordance index; G: histopathological grading according to Fuhrman; HR: hazard ratio; N/A: not applicable as only cfDNA markers with p b 0.05 in the univariate analysis were included in the full multivariate models or excluded in the multivariate model after backward elimination; pT: pathological tumor classification; R: surgical margin classification; abbreviations of the cfDNA markers: see Table 2. a This study group included the non-metastatic patients after nephrectomy listed in Table 1. Out of the 145 patients, follow-up data from 138 and 132 patients, respectively were available for the analysis of overall survival and recurrence-free survival. b Model A: All clinicopathological data and only the significant individual cfDNA markers from univariate analysis (p b 0.050; Supplementary Data) dichotomized according to the cutoffs obtained in ROC analyses at the point of maximal accuracy to discriminate between dead and alive as well as between recurrence and recurrence-free situation were used for multivariate Cox regression analysis. The cutoffs corresponded to those shown in the Kaplan–Meier analyses in Supplementary Data, Figs. S2 and S3. c C-indices were calculated according to Harrell [29]. d Model B: All clinicopatholological data and the cfDNA markers obtained after backward elimination approach (p b 0.050 for entry and p N 0.100 for removal) of Cox regression analysis of the significant individual cfDNA markers from the univariate Cox regression analysis. The p values and the 95% CI of the hazard ratios of the final models were obtained after bootstrapping (2000 resamples using the bias corrected accelerated algorithm).

118

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119

and de Martino et al. [12] in their studies were 0.725 and 0.755, respectively. Such AUC levels were also described for cfDNA in patients suffering from breast, lung or prostate cancer [35–38]. The second important clinical objective of this study was the assessment of the prognostic potential of the cfDNA markers and their ratios. For this purpose, we investigated only non-metastatic RCC patients at the time of nephrectomy to estimate the significance of cfDNA markers as prognostic indicators for recurrence-free survival and overall survival. In this respect, we aimed to integrate potential markers together with classical prognostic clinicopathological factors to increase the predictive accuracy of conventional prognostic models [39]. We established three models (Table 4) using the approach of fully and stepwise reduced multivariate Cox regression analyses. Recent prognostic models are generally based only on clinicopathological and imaging data and show limited accuracy for risk stratification following nephrectomy for clear cell RCC [4]. The C-indices of our constructed models based on combinations of the conventional clinicopathological variables with cfDNA markers distinctly increased in comparison to C-indices obtained together with models based only on clinicopathological data (Table 4). Certain cfDNA markers, for example Mito-1, SINE-1, and the ratio APP-3 to APP-2 appear as independent factors in the models in contrast to the clinicopathological variables (Table 4). These characteristics of cfDNA markers being uncorrelated to the conventional progression variables as also previously discussed reflect their potential as orthogonal biomarkers [40]. This specific feature of cfDNA markers results in an additional gain of information. It is obvious that improved prognostic information could be achieved using cfDNA markers. Moreover, it is advantageous that similar cfDNA markers could be used for the prediction of the probability of tumor recurrence and overall survival. That would be helpful not only for the treating clinician to counsel patients and recommend individualized surveillance protocols but also for the patient, if desired, in its decision-making regarding life planning. The REMARK and STARD guidelines were considered in conducting the study and presenting the data [22,23]. However, in addition to the already discussed critical points of the general use of cfDNA markers in cancer studies, some apparent limitations of our study need to be commented. It was a retrospective study and the validation of the results in a future prospective study is necessary, although a nonrandom sample selection bias was largely excluded as the consecutive available EDTA plasma samples and follow-up data were study inclusion criteria. The validity of the study was additionally supported by the fact that all analyses were performed in a blinded manner. Moreover, as explained in detail, the sample size of the study groups included 10–15% more subjects as calculated on the conventional basis of α = 5% and ß = 80%. Thus, the risk of type II error as the problem with small studies and the probability of a type I error could be excluded as far as possible. This was confirmed by the internal validation of data using the bootstrapping approach as also applied in our previous biomarker studies [41,42]. However, it should be always considered that changed cfDNA concentrations are unspecific from the diagnostic point of view as changes are also observed with other malignancies, infections or gastrointestinal, hepatic, and immunologic diseases [10]. It remains the task of the physician to take into account such an impairment of diagnostic specificity if he orders such a test. In our study, we excluded such possible interferences by other diseases to obtain an unambiguous result what could be expected by ccRCC. This aspect must also be considered in the above-mentioned future prospective studies. In conclusion, the present study results suggest that cfDNA in clear cell RCC patients has promising diagnostic and prognostic potential. Adapted combinations of genomic cfDNA fragments and mitochondrial cfDNA are helpful tools for the detection of clear cell RCC and for the differentiation between patients with and without metastases. The utilization of the prognostic capability of cfDNA in predicting recurrence-free survival and overall survival of patients after tumor nephrectomy combined with clinicopathological data could distinctly

improve the risk stratification process of patients in comparison to strategies based only on clinicopathological data. We suggest that the most useful diagnostic and prognostic markers summarized here in the presented diagnostic and prognostic models are worth for further clinical validation in prospective multicenter studies. Conflict of interest statement Dr. Steffen Mergemeier is a shareholder of the diagnostic manufacturer CONGEN Biotechnologie GmbH, which produces the kits for the determination of SINEs and human mitochondrial DNA used in the present study. The other authors declare that they have no conflict of interest and no direct or indirect commercial incentive associated with publishing the manuscript. Acknowledgments This work is part of the doctoral thesis of Hongbiao Lu who was supported by a fellowship (U2012/01) from the Changzhou No.2 People's Hospital, Jiangsu, China. The sponsor had no involvement in the study design, in the collection, analysis, and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication. Appendix A. Supplementary data Supplementary data to this article with a list of Supplemental Figures and Tables can be found online at http://dx.doi.org/10.1016/j. cca.2015.11.009. References [1] J. Ferlay, I. Soerjomataram, R. Dikshit, S. Eser, C. Mathers, M. Rebelo, et al., Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012, Int. J. Cancer 136 (2015) E359–E386. [2] K. Gupta, J.D. Miller, J.Z. Li, M.W. Russell, C. Charbonneau, Epidemiologic and socioeconomic burden of metastatic renal cell carcinoma (mRCC): a literature review, Cancer Treat. Rev. 34 (2008) 193–205. [3] H.T. Cohen, F.J. McGovern, Renal-cell carcinoma, N. Engl. J. Med. 353 (2005) 2477–2490. [4] M. Sun, S.F. Shariat, C. Cheng, V. Ficarra, M. Murai, S. Oudard, et al., Prognostic factors and predictive models in renal cell carcinoma: a contemporary review, Eur. Urol. 60 (2011) 644–661. [5] M. Fleischhacker, B. Schmidt, Circulating nucleic acids (CNAs) and cancer—a survey, Biochim. Biophys. Acta 2007 (1775) 181–232. [6] H. Schwarzenbach, D.S. Hoon, K. Pantel, Cell-free nucleic acids as biomarkers in cancer patients, Nat. Rev. Cancer 11 (2011) 426–437. [7] M.B. Giacona, G.C. Ruben, K.A. Iczkowski, T.B. Roos, D.M. Porter, G.D. Sorenson, Cellfree DNA in human blood plasma: length measurements in patients with pancreatic cancer and healthy controls, Pancreas 17 (1998) 89–97. [8] S. Jahr, H. Hentze, S. Englisch, D. Hardt, F.O. Fackelmayer, R.D. Hesch, et al., DNA fragments in the blood plasma of cancer patients: quantitations and evidence for their origin from apoptotic and necrotic cells, Cancer Res. 61 (2001) 1659–1665. [9] B.G. Wang, H.Y. Huang, Y.C. Chen, R.E. Bristow, K. Kassauei, C.C. Cheng, et al., Increased plasma DNA integrity in cancer patients, Cancer Res. 63 (2003) 3966–3968. [10] K. Jung, M. Fleischhacker, A. Rabien, Cell-free DNA in the blood as a solid tumor biomarker—a critical appraisal of the literature, Clin. Chim. Acta 411 (2010) 1611–1624. [11] S. Hauser, T. Zahalka, J. Ellinger, G. Fechner, L.C. Heukamp, R.A. von, et al., Cell-free circulating DNA: diagnostic value in patients with renal cell cancer, Anticancer Res. 30 (2010) 2785–2789. [12] M. de Martino, T. Klatte, A. Haitel, M. Marberger, Serum cell-free DNA in renal cell carcinoma: a diagnostic and prognostic marker, Cancer 118 (2012) 82–90. [13] G. Feng, X. Ye, F. Fang, C. Pu, H. Huang, G. Li, Quantification of plasma cell-free DNA in predicting therapeutic efficacy of sorafenib on metastatic clear cell renal cell carcinoma, Dis. Markers 34 (2013) 105–111. [14] J. Wan, L. Zhu, Z. Jiang, K. Cheng, Monitoring of plasma cell-free DNA in predicting postoperative recurrence of clear cell renal cell carcinoma, Urol. Int. 91 (2013) 273–278. [15] S. El Messaoudi, F. Rolet, F. Mouliere, A.R. Thierry, Circulating cell free DNA: preanalytical considerations, Clin. Chim. Acta 424 (2013) 222–230. [16] J. Ellinger, D.C. Muller, S.C. Muller, S. Hauser, L.C. Heukamp, R.A. von, et al., Circulating mitochondrial DNA in serum: a universal diagnostic biomarker for patients with urological malignancies, Urol. Oncol. 30 (2012) 509–515.

H. Lu et al. / Clinica Chimica Acta 452 (2016) 109–119 [17] M. Yu, Circulating cell-free mitochondrial DNA as a novel cancer biomarker: opportunities and challenges, Mitochondrial DNA 23 (2012) 329–332. [18] F. Mouliere, A.R. Thierry, The importance of examining the proportion of circulating DNA originating from tumor, microenvironment and normal cells in colorectal cancer patients, Expert. Opin. Biol. Ther. 12 (Suppl. 1) (2012) S209–S215. [19] T.M. Gorges, J. Schiller, A. Schmitz, D. Schuetzmann, C. Schatz, T.M. Zollner, et al., Cancer therapy monitoring in xenografts by quantitative analysis of circulating tumor DNA, Biomarkers 17 (2012) 498–506. [20] J.A. Nicklas, T. Noreault-Conti, E. Buel, Development of a real-time method to detect DNA degradation in forensic samples, J. Forensic Sci. 57 (2012) 466–471. [21] M.A. Batzer, P.L. Deininger, Alu repeats and human genomic diversity, Nat. Rev. Genet. 3 (2002) 370–379. [22] D.G. Altman, L.M. McShane, W. Sauerbrei, S.E. Taube, Reporting recommendations for tumor marker prognostic studies (REMARK): explanation and elaboration, PLoS Med. 9 (2012), e1001216. [23] P.M. Bossuyt, J.B. Reitsma, D.E. Bruns, C.A. Gatsonis, P.P. Glasziou, L.M. Irwig, et al., Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Standards for Reporting of Diagnostic Accuracy, Clin. Chem. 49 (2003) 1–6. [24] D.W. Swinkels, E. Wiegerinck, E.A.P. Steegers, J.B. de Kok, Effects of blood-processing protocols on cell-free DNA quantification in plasma, Clin. Chem. 49 (2003) 525–526. [25] S.A. Bustin, V. Benes, J.A. Garson, J. Hellemans, J. Huggett, M. Kubista, et al., The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments, Clin. Chem. 55 (2009) 611–622. [26] P. Pinzani, F. Salvianti, S. Zaccara, D. Massi, G.V. De, M. Pazzagli, et al., Circulating cell-free DNA in plasma of melanoma patients: qualitative and quantitative considerations, Clin. Chim. Acta 412 (2011) 2141–2145. [27] A.J. Vickers, E.B. Elkin, Decision curve analysis: a novel method for evaluating prediction models, Med. Decis. Mak. 26 (2006) 565–574. [28] C. Stephan, K. Jung, A. Semjonow, K. Schulze-Forster, H. Cammann, X. Hu, et al., Comparative assessment of urinary prostate cancer antigen 3 and TMPRSS2:ERG gene fusion with the serum [−2]proprostate-specific antigen-based prostate health index for detection of prostate cancer, Clin. Chem. 59 (2013) 280–288. [29] F.E. Harrell, Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis, Springer, New York, 2001. [30] F. Faul, E. Erdfelder, A. Buchner, A.G. Lang, Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses, Behav. Res. Methods 41 (2009) 1149–1160.

119

[31] B. Ralla, C. Stephan, S. Meller, D. Dietrich, G. Kristiansen, K. Jung, Nucleic acid-based biomarkers in body fluids of patients with urologic malignancies, Crit. Rev. Clin. Lab. Sci. 51 (2014) 200–231. [32] K. Page, D.S. Guttery, N. Zahra, L. Primrose, S.R. Elshaw, J.H. Pringle, et al., Influence of plasma processing on recovery and analysis of circulating nucleic acids, PLoS One 8 (2013), e77963. [33] F. Malentacchi, S. Pizzamiglio, P. Verderio, M. Pazzagli, C. Orlando, C.M. Ciniselli, et al., Influence of storage conditions and extraction methods on the quantity and quality of circulating cell-free DNA (ccfDNA): the SPIDIA-DNAplas External Quality Assessment experience, Clin. Chem. Lab. Med. (2015)http://dx.doi.org/10.1515/ cclm-2014-1161 (Epub ahead of print). [34] The Cancer Genome Atlas Research Network, Comprehensive molecular characterization of clear cell renal cell carcinoma, Nature 499 (2013) 43–49. [35] A. Altimari, A.D. Grigioni, E. Benedettini, E. Gabusi, R. Schiavina, A. Martinelli, et al., Diagnostic role of circulating free plasma DNA detection in patients with localized prostate cancer, Am. J. Clin. Pathol. 129 (2008) 756–762. [36] C. Kohler, R. Radpour, Z. Barekati, R. Asadollahi, J. Bitzer, E. Wight, et al., Levels of plasma circulating cell free nuclear and mitochondrial DNA as potential biomarkers for breast tumors, Mol. Cancer 8 (2009) 105. [37] E. Papadopoulou, E. Davilas, V. Sotiriou, E. Georgakopoulos, S. Georgakopoulou, A. Koliopanos, et al., Cell-free DNA and RNA in plasma as a new molecular marker for prostate and breast cancer, Ann. N. Y. Acad. Sci. 1075 (2006) 235–243. [38] G. Sozzi, D. Conte, M. Leon, R. Ciricione, L. Roz, C. Ratcliffe, et al., Quantification of free circulating DNA as a diagnostic marker in lung cancer, J. Clin. Oncol. 21 (2003) 3902–3908. [39] H.M. Bovelstad, S. Nygard, O. Borgan, Survival prediction from clinico-genomic models—a comparative study, BMC Bioinformatics 10 (2009) 413. [40] R.E. Gerszten, T.J. Wang, The search for new cardiovascular biomarkers, Nature 451 (2008) 949–952. [41] J. Busch, B. Ralla, M. Jung, Z. Wotschofsky, E. Trujillo-Arribas, P. Schwabe, et al., Piwiinteracting RNAs as novel prognostic markers in clear cell renal cell carcinomas, J. Exp. Clin. Cancer Res. 34 (2015) 61. [42] Z. Wotschofsky, J. Busch, M. Jung, C. Kempkensteffen, S. Weikert, K.D. Schaser, et al., Diagnostic and prognostic potential of differentially expressed miRNAs between metastatic and non-metastatic renal cell carcinoma at the time of nephrectomy, Clin. Chim. Acta 416 (2013) 5–10.