Skip to content

Advertisement

Clinical Epigenetics

What do you think about BMC? Take part in

Open Access

Comparison of quantification algorithms for circulating cell-free DNA methylation biomarkers in blood plasma from cancer patients

Clinical EpigeneticsThe official journal of the Clinical Epigenetics Society20179:125

https://doi.org/10.1186/s13148-017-0425-4

Received: 29 April 2017

Accepted: 20 November 2017

Published: 1 December 2017

Abstract

Background

SHOX2 and SEPT9 methylation in circulating cell-free DNA (ccfDNA) in blood are established powerful and clinically valuable biomarkers for diagnosis, staging, prognosis, and monitoring of cancer patients. The aim of the present study was to evaluate different quantification algorithms (relative quantification, absolute quantification, quasi-digital PCR) with regard to their clinical performance.

Methods

Methylation analyses were performed in a training cohort (141 patients with head and neck squamous cell carcinoma [HNSCC], 170 control cases) and a testing cohort (137 HNSCC cases, 102 controls). DNA was extracted from plasma samples, bisulfite-converted, and analyzed via quantitative real-time PCR. SHOX2 and SEPT9 methylations were assessed separately and as panel [mean SEPT9/SHOX2 ] using the ΔCT method for absolute quantification and the ΔΔCT-method for relative quantification. Quasi-digital PCR was defined as the number of amplification-positive PCR replicates. The diagnostic (sensitivity, specificity, area under the curve (AUC) of the receiver operating characteristic (ROC)) and prognostic accuracy (hazard ratio (HR) from Cox regression) were evaluated.

Results

Sporadic methylation in control samples necessitated the introduction of cutoffs resulting in 61–63% sensitivity/90–92% specificity (SEPT9/training), 53–57% sensitivity/87–90% specificity (SHOX2/training), and 64–65% sensitivity/90–91% specificity (mean SEPT9/SHOX2 /training). Results were confirmed in a testing cohort with 54–56% sensitivity/88–90% specificity (SEPT9/testing), 43–48% sensitivity/93–95% specificity (SHOX2/testing), and 49–58% sensitivity/88–94% specificity (mean SEPT9/SHOX2 /testing). All algorithms showed comparable cutoff-independent diagnostic accuracy with largely overlapping 95% confidence intervals (SEPT9: AUCtraining = 0.79–0.80; AUCtesting = 0.74–0.75; SHOX2: AUCtraining = 0.78–0.81, AUCtesting = 0.77–0.79; mean SEPT9/SHOX2 : AUCtraining = 0.81–0.84, AUCtesting = 0.80). The accurate prediction of overall survival was possible with all three algorithms (training cohort: HR SEPT9  = 1.23-1.90, HR SHOX2  = 1.14-1.85, HRmeanSEPT9/SHOX2  =1.19-1.89 ; testing cohort: HR SEPT9  =1.22-1.67, HR SHOX2  = 1.15-1.71, HRmeanSEPT9/SHOX2  = 1.12-1.77).

Conclusion

The concordant clinical performance based on different quantification algorithms allows for the application of various diagnostic platforms for the analysis of ccfDNA methylation biomarkers.

Keywords

Free-circulating cell-free DNADNA methylation SHOX2 SEPT9 Diagnostic accuracyRisk stratificationQuantitative real-time PCRΔΔCT-methodΔCT-methodQuasi-digital PCR

Background

Cancer is the second most common cause of death in the USA, therefore posing a major health burden. Approximately 1,688,780 newly diagnosed cases are expected in 2017 [1]. Consequently, there is a pressing need to identify biomarkers that might help to address key clinical questions and improve patient outcome. Biomarkers derived from liquid biopsies appear to be particularly suitable, since they can be obtained minimally invasive. Circulating cell-free DNA (ccfDNA), which can be detected in the bloodstream of patients with advanced but also early stages of various malignancies [2], constitutes a promising liquid biopsy cancer biomarker [3] and is suitable for diagnosis, prognosis, and the identification of occult tumor recurrence [46]. Cancer-specific changes, e.g., mutations or aberrant methylation patterns, facilitate the robust discrimination between non-tumorous and tumorous ccfDNA [7].

Aberrant DNA methylation, a hallmark of cancer, frequently occurs in defined gene regions in most tumor entities and is chemically stable [710]. The methylation biomarkers with the highest level of validation in bodily fluid are SHOX2 and SEPT9. In lung cancer, the reliability of SHOX2 as a biomarker has been confirmed in bronchial aspirates and plasma [1114]. Hypermethylated SEPT9, in addition, has been validated as a highly sensitive and specific biomarker for the detection of colorectal cancer in plasma and has recently received approval as a screening marker by the Food and Drug Administration (FDA) [1519]. Furthermore, both SEPT9 and SHOX2 are diagnostic and prognostic tools for the distinction between malignant and paramalignant ascites and pleura effusions [20, 21]. Recently, a large prospective cohort study showed that both SEPT9 and SHOX2 are clinically valuable for diagnosis, staging, prognosis, and monitoring for patients suffering from head and neck squamous cell carcinoma (HNSCC) [3]. Since not all tumors presented with elevated methylation levels in both genes, the paneling of SEPT9 and SHOX2 turned out to be advantageous compared to the separate analysis of each marker [3]. Especially with regard to the diagnostic application, the combined analyses allowed for a reduction of false-positive results [3].

The application of methylated ccfDNA as a cancer biomarker, however, poses significant challenges. Firstly, ccfDNA concentrations in body fluids are considerably low, and the methylated portion is usually in the range of single genome copies. Hence, relatively high bodily fluid volumes are required, and DNA analysis needs to be highly sensitive and specific [2224]. Methylated ccfDNA can be obtained either from plasma or from serum. The latter shows higher concentrations of ccfDNA, presumably due to the clotting process by which most leucocytes are lysed and release DNA [23, 25]. Secondly, an unspecific and low sporadic methylation in healthy individuals can occur, necessitating a cutoff above which samples are defined as methylation positive [21, 26]. While some healthy individuals might have elevated biomarker levels (false-positives), some cancer patients might not (false-negatives). Not only the underlying biological process but also the evaluation algorithm has an influence on the clinical performance of methylation biomarkers. In previous analyses of SEPT9, for instance, the sensitivity and specificity were heavily influenced by the evaluation algorithm [14, 27]. Accordingly, the algorithm has been discussed in the past in order to increase the performance of the test.

Quantitative real-time PCR is well established for DNA methylation analyses [3, 1114, 20, 21, 24]. Quantification can be carried out either relatively referring to the total amount of DNA in the sample or absolutely by quantifying the number of genome equivalents in a given sample volume [28]. In order to detect the exact amount of methylated ccfDNA from the sample, a known quantity of reference DNA, which is compared to the unknown quantity from the sample, is required (reference sample). The absolute value can be calculated by the ΔCT-method. Estimating that a diploid genome weights about 6.6 pg, the absolute values can be converted into genome equivalents. For relative quantification, the methylated portion of ccfDNA is referred to the total DNA amount in the sample and can be calculated by the ΔΔCT-method [29]. Furthermore, a qualitative method based of the counts of positive PCRs can be applied (digital PCR). In accordance with a prior study, in which rare DNA variants were detected, this digital counting of PCR replicates with methylated ccfDNA can be called “quasi-digital PCR” [30]. On the basis of the distribution of positive and negative reactions, Poisson statistics can predict the amount of methylated ccfDNA in the sample [31]. Most studies on SEPT9 in colorectal cancer employed a qualitative method, especially the triplicate PCR method [15, 1719, 27]. Here, the sample is divided into three real-time PCR reactions, with some aliquots containing methylated gene copies while others do not. The number of positive reactions is then counted and represents the total amount of methylation in the sample. In different prior studies, the sample was considered positive, if either one out of three (1/3 algorithm) or two out of three (2/3 algorithm) replicates showed a reaction [27, 3234].

Currently, no explicit recommendation exists as to whether quantitative or qualitative analyses are supposed to be applied when using methylated ccfDNA as a biomarker for cancer. Depending on the study, different evaluation methods, mostly relative quantification and quasi-digital PCR, were used [3, 1115, 1721, 27, 29, 31]. The aim of the present study was to compare the three different algorithms in order to improve the clinical performance of methylated ccfDNA as a biomarker with the example of SEPT9 and SHOX2. This study was carried out based on a reevaluation of recently published data [3] from a prospective cohort study including HNSCC patients. According to the original study, analysis was performed in a training/testing design in order to avoid overfitting. Methylation of SEPT9 and SHOX2 in plasma was analyzed with three different algorithms (absolute quantification, relative quantification, and quasi-digital PCR), and the diagnostic accuracy of all three algorithms was compared. Furthermore, the patients’ clinical follow-up was updated allowing for an accurate prognostic analysis.

Methods

Patient cohort and clinical endpoints

A training cohort of 137 HNSCC patients and 170 control patients with benign diseases was included in the present study. Of these, 137 HNSCC patients and 122 control patients had been enrolled in the course of a previous study [3]. Forty-eight additional control patients were included in the training cohort, and the clinical follow-up was updated. For validation of the results, the same testing cohort as in the previous study was included (141 HNSCC patients, 102 control patients), and overall survival data were updated. For detailed patients’ characteristics see Additional file 1. All patients were treated at the University Hospital Bonn between 2012 and 2017. Control patients had no history of SCC and were free of any other malignancy in at least 3 years [3]. A plasma sample was taken prior to any treatment [3].

The study protocol was approved by the ethics committee of the University Hospital Bonn (vote no. 224/12). All patients provided written informed consent. The clinical study endpoint was death by any cause, and time-to-event (time-to-death) was defined as overall survival.

Plasma preparation and SHOX2 and SEPT9 methylation quantification

Blood samples were taken in EDTA-stabilized collection tubes. A total of 3 ml of plasma was prepared within 2 h as previously described [3]. Bisulfite DNA was split into six parts, and each part was applied to the SHOX2/SEPT9/ACTB triplex quantitative methylation-specific real-time PCR. A CpG-free region within the gene actin beta (ACTB) was chosen to quantify methylation independently ccfDNA, representing the total DNA [20]. A calibrator sample (bisulfite converted artificially methylated human DNA) was analyzed in triplicate (3 ng each) within each PCR run. Quantitative methylation levels for SEPT9 and SHOX2 were calculated with the ΔCT and ΔΔCT methods.

Relative quantification was carried out using the ΔΔCT method as follows: ΔΔCtSample = ΔCtSample − ΔCtCalibrator with ΔCtSample = CtSample(ACTB) − CtSample(SEPT9 or SHOX2) and ΔCtCalibrator = CtCalibrator(ACTB) − CtCalibrator(SEPT9 or SHOX2). Percent methylation levels were calculated (MethylationSample = 100% • 2ΔΔCTSample [29]) for each of the six PCR replicates and the mean was computed.

Absolute quantification was carried out with the ΔCT-method: ΔCtSample = CtSample(ACTB) − CtSample(SHOX2 or SEPT9). Based on the quantity of the calibrator, absolute values were calculated: Methylated DNA [ng] = 3 ng × 2ΔCT. The number of haploid genome equivalents was calculated: \( \mathrm{Genome} \mathrm{equivalent}\mathrm{s}=\mathrm{DNA}\ \left[ ng\right]/\frac{3.3\times {10}^{-3}\ \mathrm{DNA}\ \left[ ng\right]\ }{\mathrm{Genomeequivalent}} \).

For qualitative analysis (quasi-digital PCR), the number of replicates showing a reaction was counted.

Statistical analyses

The area under the curve (AUC) of the receiver operating characteristic (ROC) was computed in order to describe the diagnostic accuracy. Sensitivity and specificity were calculated. Kaplan-Meier analyses and univariate Cox proportional hazard analysis were conducted to investigate the overall survival. In order to improve comparability between quantitative and qualitative analyses, a logarithmic scale was performed for relative methylation, absolute methylation, and positive counts. Methylation levels below 0.01% in quantitative analyses were set to 0.01% for a logarithmic illustration. In quasi-digital PCR, values below one positive replicate were set to one by adding one to all six values. Logarithmic data were used to perform Cox analysis. P values < 0.05 were considered statistically significant. P values refer to the log-rank (Kaplan-Meier) and Wald tests (Cox proportional hazards analysis), respectively. Ninety-five percent confidence intervals were reported. For the combination of SEPT9 and SHOX2 (mean SEPT9/SHOX2 ), methylation levels were mean averaged: mean SEPT9/SHOX2  = [(SEPT9 + SHOX2)/2] [3].

Results

Evaluation of the diagnostic accuracy: Training cohort

As sporadic methylation also occurs in blood plasma from patients without a malignant disease, cutoffs are required in order to classify samples as test-positive and test-negative. In the present study, cutoffs were defined that resulted in test-negativity in 90% of control patients (specificity). According to Schröck et al. [3], SEPT9 and SHOX2 were analyzed separately and in combination (averaged, mean SEPT9/SHOX2 ). Methylation levels below 0.055% (SEPT9), 0.218% (SHOX2), and 0.118% (mean SEPT9/SHOX2 ) were considered test-negative (Fig. 1). As seen in Fig. 1, more than 1.4 (4.68 pg) gene copies of SEPT9, 6.9 (22.67 pg) of SHOX2, and on average 4.3 (14.27 pg) copies of both markers per sample were considered test-positive. Since the PCR analysis was performed in six single reactions per sample, the quasi-digital PCR only took on seven distinct states (0–6 PCR replicates positive). Accordingly, a determination of specificity of exactly 90% was not achievable. Specificities of 92% for SEPT9, 87% for SHOX2, and 91% for mean SEPT9/SHOX2 were accomplished by > 1 (SEPT9), > 4 (SHOX2), and > 2 (mean SEPT9/SHOX2 ) out of 6 positive PCR replicates.
Fig. 1

Training cohort. Methylation levels of SEPT9, SHOX2, and mean SEPT9/SHOX2 in plasma of HNSCC (n = 137) and control (n = 170) patients in relative quantification, genome equivalents (absolute quantification), and quasi-digital PCR with sensitivity (sens.), specificity (spec.), and cutoffs. Receiver Operating Characteristics with AUCs of HNSCC and control patients

SEPT9 showed a low basal methylation level in blood from individuals without a malignant disease. This was indicated by the low number of positive PCR reactions when analyzing blood from the control patients (n = 115 with zero; n = 41 with one; and only n = 14 with more than one positive PCR). Due to the low background methylation, similar performances of all three algorithms were to be expected. Accordingly, sensitivity (63% for relative, 61% for absolute quantification, 62% for quasi-digital PCR), specificity (90, 90, and 92%, respectively), and AUCs (AUC = 0.80/0.80/0.79) were comparable with identical 95% confidence intervals (all: 95%CI [0.74–0.85]; Fig. 1). Quasi-digital PCR showed the same AUC, but in fact, the specificity was slightly higher (92%).

SHOX2, in contrast, had higher levels of background methylation indicated by the high number of positive PCR reactions (n = 48 with zero; n = 45 with one; and n = 77 with more than one positive PCR) in control patients. Therefore, a higher cutoff was required (more than four out of six PCR reactions positive). Because of the increased number of positive PCR reactions in control patients, a lower sensitivity (53%) and specificity (87%) were seen in the qualitative analysis (AUC = 0.78, 95% CI [0.72–0.83]). However, all algorithms showed similar AUCs with highly overlapping 95% confidence intervals. Relative quantification showed a slightly higher sensitivity (sensitivity = 57%) compared to absolute quantification (sensitivity = 55%) at the same defined specificity (90%).

As shown by Schröck and co-workers [3], mean SEPT9/SHOX2 gave best results in all three algorithms compared to the particular algorithm for either biomarker separately. As a result of the higher sensitivity for SHOX2 in relative quantification, best performance of relative quantification for the mean SEPT9/SHOX2 was to be expected. In fact, absolute and relative quantification showed equal results (both: sensitivity = 64%, specificity = 90%, AUCrelative = 0.84; 95% CIrelative [0.79–0.88], AUCabsolute = 0.83, 95% CIabsolute [0.78–0.88]). Both biomarkers evaluated by quasi-digital PCR showed highest performance with a sensitivity of 65% at 91% specificity (AUC = 0.81, 95%CI [0.76–0.87]).

Validation of the diagnostic accuracy: testing cohort

The high performance of the biomarkers was validated in the testing cohort (Fig. 2). In concordance with the results from the training cohort, all algorithms gave similar results with similar AUCs and overlapping confidence intervals. Relative quantification resulted in a slightly higher sensitivity (56%) for SEPT9 compared to quasi-digital PCR (55%) at lower specificity (88% for relative quantification compared to 90% for quasi-digital PCR). Relative quantification also appeared to be the algorithm with highest accuracy for SHOX2 (sensitivity = 48%, specificity = 93%, AUC = 0.79, 95% CI [0.73–0.84]). Compared to both biomarkers separately, the mean SEPT9/SHOX2 showed highest sensitivity and specificity in relative quantification (sensitivity = 58%, specificity = 94%, AUC = 0.80, 95% CI [0.75–0.86]).
Fig. 2

Testing cohort. Methylation levels of SEPT9, SHOX2, and mean SEPT/SHOX2 in plasma of HNSCC (n = 141) and control (n = 102) patients in relative quantification, genome equivalents (absolute quantification), and quasi-digital PCR with sensitivity (sens.), specificity (spec.), and cutoffs. ROC curves with AUCs of HNSCC and control patients

Prognostic value: training cohort

In concordance with the previous study, 129 HNSCC patients were included in the survival analysis [3]. Both biomarkers were used as continuous values that were converted to a logarithmic scale prior to analysis in order to improve the comparability between the quantitative scales (relative and absolute quantification) and the qualitative scale (quasi-digital PCR). For example, methylation values in plasma varied from 0% to more than 20%, whereas quasi-digital PCR, in contrast, only varied in a range of 0–6 positive replicates. All data were therefore transformed into a logarithmic scale, and Cox proportional hazards analysis was carried out with the converted data.

All algorithms for the evaluation of SEPT9, SHOX2, and mean SEPT9/SHOX2 turned out to be valuable for the prognosis of patients suffering from HNSCC and all with a similarly high power. Patients with elevated biomarker levels had a significantly increased hazard of death (Fig. 3, Table 1). All algorithms showed similar results with overlapping confidence intervals. Hazard ratios for SEPT9, SHOX2, and mean SEPT9/SHOX2 using relative quantification (SEPT9: HRrelative = 1.23, 95% CI [1.11–1.35]; SHOX2: HRrelative = 1.14, 95% CI [1.03–1.26]; mean SEPT9/SHOX2 : HRrelative = 1.19, 95% CI [1.07–1.32]) were slightly smaller compared to absolute quantification (SEPT9: HRabsolute = 1.27, 95% CI [1.14–1.42]; SHOX2: HRabsolute = 1.16, 95% CI [1.03–1.31]; mean SEPT9/SHOX2 : HRabsolute = 1.22, 95% CI [1.08–1.37]). Highest hazard ratios for all biomarkers were found with quasi-digital PCR (SEPT9: HRquasi-digital = 1.90, 95% CI [1.34–2.69]; SHOX2: HRquasi-digital = 1.85, 95% CI [1.18–2.91]; mean SEPT9/SHOX2 : HRquasi-digital = 1.89, 95% CI [1.28–2.79]). However, this was not followed by the corresponding p values. All p values showed similar results for SEPT9 (all: p < 0.001), SHOX2 (p relative = 0.009; p absolute = 0.013, p quasi-digital = 0.008), and mean SEPT9/SHOX2 (all: p = 0.001) indicating a similar prognostic performance for all algorithms.
Fig. 3

Training cohort. Kaplan-Meier analysis of overall survival in HNSCC patients (n = 129). Patients are stratified according to SHOX2 and SEPT9 plasma methylation levels. Plasma methylation levels were quantified using relative, absolute quantification, and quasi-digital PCR and dichotomized based on cutoffs that resulted in specificities and sensitivities as seen in Figs. 1 and 2. Cutoff values for positive (above cutoff) and negative (below cutoff) classification: SEPT9, SHOX2, and mean SEPT9/SHOX2 . Cutoffs for relative quantification were SEPT9 = 0.055%, SHOX2 = 0.281%, and mean SEPT9/SHOX2  = 0.118%; for absolute quantification SEPT9 = 4.68 pg, SHOX2 = 22.7 pg, and mean SEPT9/SHOX2  = 14.3 pg; for quasi-digital PCR for SEPT9 > 2, for SHOX2 > 4 and for mean SEPT9/SHOX2  > 2 positive PCR reactions

Table 1

Diagnostic and prognostic performance of SHOX2 and SEPT9 hypermethylation

Cohort

Biomarker

Quantification method

Sens. (%)

Spec. (%)

AUC

Hazard ratio

P value

95% CI

Training

 

SEPT9

Relative

63

90

0.80

1.23

< 0.001

1.11–1.35

Absolute

61

90

0.80

1.27

< 0.001

1.14–1.42

Quasi-digital PCR

62

92

0.79

1.90

< 0.001

1.34–2.69

SHOX2

Relative

57

90

0.81

1.14

0.009

1.03–1.26

Absolute

55

90

0.80

1.16

0.013

1.03–1.31

Quasi-digital PCR

53

87

0.78

1.85

0.008

1.18–2.91

Mean SEPT9/SHOX2

Relative

64

90

0.84

1.19

0.001

1.07–1.32

Absolute

64

90

0.83

1.22

0.001

1.08–1.37

Quasi-digital PCR

65

91

0.81

1.89

0.001

1.28–2.79

Testing

 

SEPT9

Relative

56

88

0.74

1.22

0.002

1.07–1.38

Absolute

54

89

0.74

1.35

0.001

1.45–1.59

Quasi-digital PCR

55

90

0.75

1.67

0.010

1.13–2.45

SHOX2

Relative

48

93

0.79

1.15

0.045

1.00–1.31

Absolute

43

94

0.78

1.24

0.007

1.06–1.44

Quasi-digital PCR

43

95

0.77

1.71

0.032

1.05–2.80

Mean SEPT9/SHOX2

Relative

58

94

0.80

1.12

0.006

1.06–1.39

Absolute

49

90

0.80

1.30

0.001

1.11–1.53

Quasi-digital PCR

58

88

0.80

1.77

0.009

1.15–2.72

Clinical performance was evaluated in HNSCC patients included in a training and a testing cohort. Methylation levels were calculated using three different algorithms (relative, absolute quantification, and quasi-digital PCR). Sensitivities (sens.), specificities (spec.), Area Under the Curve (AUC) of the Receiver Operating Characteristic, and Cox Proportional Hazards with univariate analysis of SEPT9, SHOX2, and mean SEPT9/SHOX2 , corresponding p values, and 95% confidence intervals are reported

Validation of the prognostic value: testing cohort

In accordance with the training cohort, all markers were significantly prognostic in all algorithms and all gave similar results (Fig. 4, Table 1). In Cox proportional hazards analysis, quasi-digital PCR showed highest hazard ratios for SEPT9, SHOX2, and mean SEPT9/SHOX2 with only small differences in confidence intervals (SEPT9: HRquasi-digital = 1.67, 95% CI [1.13–2.45]; SHOX2: HRquasi-digital = 1.71, 95% CI [1.05–2.80]; mean SEPT9/SHOX2 : HRquasi-digital = 1.77, 95% CI [1.15–2.72]). P values again did not concur. The smallest p value was found for SEPT9 (p = 0.001), SHOX2 (p = 0.007), and mean SEPT9/SHOX2 (p = 0.001) in absolute quantification.
Fig. 4

Testing cohort. Kaplan-Meier analysis of overall survival in HNSCC patients (n = 137). Patients are stratified according to SHOX2 and SEPT9 plasma methylation levels. Plasma methylation levels were quantified using relative, absolute quantification and quasi-digital PCR and dichotomized based on cutoffs established in the training cohort. Cutoff values for positive (above cutoff) and negative (below cutoff) classification: SEPT9, SHOX2, and mean SEPT9/SHOX2 . Cutoffs for relative quantification were SEPT9 = 0.055%, SHOX2 = 0.281%, and mean SEPT9/SHOX2  = 0.118%; for absolute quantification SEPT9 = 4.68 pg, SHOX2 = 22.7 pg, and mean SEPT9/SHOX2  = 14.3 pg; for quasi-digital PCR for SEPT9 > 2, for SHOX2 > 4, and for mean SEPT9/SHOX2  > 2 positive PCR reactions

Discussion

Hypermethylation of the SHOX2 and SEPT9 gene loci in ccfDNA has been known as valuable diagnostic and prognostic tests for patients suffering from HNSCC and other cancers [3, 7, 1121, 27, 29, 3234]. The performance of methylation biomarkers, however, is strongly dependent on the algorithm employed [15, 18, 19, 27, 33, 34]. The aim of the present study was to compare qualitative (quasi-digital PCR) and quantitative (absolute and relative quantification) evaluation methods in order to determine the most suitable algorithm. Analysis of both SEPT9 and SHOX2 in different algorithms led to the finding that differences in sensitivity and specificity examined with relative quantification, absolute quantification, and quasi-digital PCR were only marginal.

SEPT9 and SHOX2 both have been shown to be strong and valid biomarkers in several cancer entities. Many studies have been carried out in order to examine their potential in the clinical management of cancer patients [3, 1421]. In all studies, methylation analysis was based on PCR. While some studies used quantitative methods like the 1/3 or the 2/3 algorithms [19, 27, 32, 33], others quantified both biomarkers relatively in reference to the amount of total DNA [3, 20, 21]. This led to the following questions: (a) to what extend can the clinical performance of biomarkers be compared between studies using quasi-digital PCR and studies using relative quantification and (b) which of the algorithms is more suitable. It has been shown previously that the clinical performance of methylation biomarkers is heavily influenced by the algorithm employed for analysis. In a prospective cohort study of SEPT9 in colorectal cancer screening, for instance, sensitivity could be increased from 48.2 to 63.9% while specificity simultaneously decreased from 91.5 to 88.4% by using three instead of only two PCR replicates [15]. Not only the number of PCR replicates but also the number of positive PCR reactions to call the sample positive had an influence in prior studies. In reports comparing the 1/3 algorithm to the 2/3 algorithm, the 1/3 algorithm resulted in a higher sensitivity (75%) compared to the 2/3 algorithm (57%) but also lead to more false-positive results (specificity: 87% for the 1/3 algorithm vs. 98% for the 2/3 algorithm) [19]. This finding was further confirmed in other studies [27, 3133]. In the past, such comparisons have been made with algorithms all based on quasi-digital PCR but not between quasi-digital PCR and quantitative methods. The present study revealed that both SHOX2 and SEPT9 show an equally strong performance using absolute quantification, relative quantification, and quasi-digital PCR and can therefore be considered extraordinary robust biomarkers. This aspect allows for the application of different analytical platforms that all exhibit advantages and disadvantages. One major advantage of quasi-digital PCR is its simplicity and feasibility. The digital algorithm with three PCR replicates was implemented in most studies of SEPT9. In the past, it was found that an increase in PCR replicates substantially enhanced the sensitivity at the expense of a slight reduction of specificity [15]. This led to the consideration that clinical performance (accuracy) of methylation biomarkers could further be increased by the application of additional PCR replicates. Thus, six replicates were used in the present study, assuming that the representation of the overall methylated ccfDNA in the sample may be more accurate.

Even though samples are retrieved under highly standardized conditions, ccfDNA can increase in vitro. In an earlier study evaluating patients with bone marrow transplants, most of the ccfDNA in the bloodstream was shown to originate from the hematopoietic system. The patients under investigation had received bone marrow transplants from individuals of the other gender, and ccfDNA was related by analyzing the y-chromosome. Additionally, serum and plasma samples were compared, and serum contained 14-fold higher concentrations of ccfDNA compared to plasma, indicating that the additional ccfDNA was derived from lysed leucocytes. Besides, long storage times also increase concentrations of ccfDNA in plasma [35]. Samples with long times of storage before plasma preparation could therefore benefit from absolute quantification rather than from relative quantification, since it is independent of in vitro increased total DNA values. Especially if sample quality cannot be reproduced, evaluation of methylation could benefit from absolute quantification. In contrast to relative quantification in reference to the total DNA present in the sample, absolute quantification is not influenced by increased total DNA originating from leucocytes. Because methylated ccfDNA might predominantly originate from the tumor rather than from leucocytes, absolute quantification could be an ideal algorithm for the analysis of methylation biomarkers from serum. As the present study shows that absolute quantification, quasi-digital PCR, and relative quantification perform equally, one perspective for the future could be the application of absolute quantification to serum analysis of methylated ccfDNA. However, because the short half-life of 10–15 min of ccfDNA in blood allows for no biomarker accumulation in vivo [36, 37], concentrations, especially of tumorous ccfDNA, are always low. This dynamic might limit the performance of absolute quantification in contrast to relative quantification. In the future, an internal control might help to overcome such issues, for example as recently described [38]. This internal control, when chosen with a chemical quality similar to the analyzed ccfDNA, could improve validating the analytical sensitivity, which is especially important for analytes with low target concentrations like methylation [38]. The internal control could also serve as a reference standard for relative quantification, being independent of increased ccfDNA from leucocytes occurring during plasma preparation.

In the present study, SEPT9, SHOX2, and mean SEPT9/SHOX2 proved to be significant prognostic factors for overall survival in HNSCC patients. Elevated biomarker levels were associated with a worse overall survival in both the training and the testing cohort. Transferring methylation levels and the count of positive PCR replicates onto a logarithmic scale made them comparable. For all biomarkers, quasi-digital PCR showed highest hazard ratios. However, even in a logarithmic scale, relative quantification showed values between 0% and about 22% methylation (training cohort) and 25% methylation (testing cohort), whereas in quasi-digital, PCR results only ranged from zero to six positive replicates. This limited the comparability of the methods applied. Because all algorithms were tested in plasma of the same patients, comparing p values had a higher explanatory power compared to the hazard ratios. The comparison of the p values did not show a clear direction towards which algorithm performed best for prognostic purposes. In conclusion, prognosis of patients with HNSCC, SEPT9, SHOX2, and mean SEPT9/SHOX2 can be analyzed either with relative or absolute quantification or with quasi-digital PCR.

In the present study, only HNSCC patients who have been enrolled and analyzed within a previous study were included [3]. Hence, it needs to be emphasized that the current study is not suited and does not aim to validate the clinical performance of the biomarkers that have been described by Schröck and colleagues earlier [3]. This study rather showed the robustness of the methylation biomarkers. In the future, methylation analysis of SEPT9 and SHOX2 can be applied to various different platforms, for example real-time PCR platform, digital PCR, or next generation sequencing. It could even be applied to end-point PCR or a point of care platform.

Conclusions

In summary, quantitative and qualitative evaluation methods are equally suitable for both SEPT9 and SHOX2 confirming the robustness and reliability of both biomarkers. The concordance of the evaluation algorithms further allows for the application of different platforms when working with methylation biomarkers.

Abbreviations

ACTB

Actin beta

AUC: 

Area under the curve

ccfDNA: 

Circulating cell-free DNA

CI: 

Confidence interval

CT: 

Threshold cycle

HNSCC: 

Head and neck squamous cell carcinoma

HR: 

Hazard ratio

PCR: 

Polymerase chain reaction

ROC: 

Receiver operating characteristic

SEPT9

Septin 9

SHOX2

Short stature homeobox 2

Declarations

Acknowledgements

Not applicable.

Availability of data and materials

The datasets used for the current study are available from the corresponding author on reasonable request.

Authors’ contributions

LdV drafted the manuscript and coordinated the study. AS, AF, GK, and FB provided sample material. HG revised the manuscript. DD designed and supervised the study and revised the manuscript. All authors read and approved the final version of the manuscript.

Ethics approval and consent to participate

The study protocol was approved by the ethics committee of the University Hospital Bonn (vote no. 224/12). All patients provided signed informed consent.

Consent for publication

Not applicable.

Competing interests

DD has been an employee and is a consultant of Epigenomics AG, a company that aims to commercialize the DNA methylation biomarkers SEPT9 and SHOX2. DD is co-inventor and owns patents on methylation biomarkers and related technologies. DD receives inventor’s compensation from Epigenomics AG.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
Department of Otolaryngology, University Hospital Bonn, Head and Neck Surgery
(2)
Institute of Pathology, University Hospital Bonn (UKB)

References

  1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2017. CA Cancer J Clin. 2017;67:7–30.View ArticlePubMedGoogle Scholar
  2. Bettegowda C, Sausen M, Leary RJ, Kinde I, Wang Y, Agrawal N, et al. Detection of circulating tumor DNA in early- and late-stage human malignancies. Sci Transl Med. 2014;6:224ra24.View ArticlePubMedPubMed CentralGoogle Scholar
  3. Schröck A, Leisse A, de Vos L, Gevensleben H, Dröge F, Franzen A, et al. Free-circulating methylated DNA in blood for diagnosis, staging, prognosis and monitoring of head and neck squamous cell carcinoma patients: an observational prospective cohort study. Clin Chem. 2017;63:1288–96.View ArticlePubMedGoogle Scholar
  4. Marzese DM, Hirose H, Hoon DS. Diagnostic and prognostic value of circulating tumor-related DNA in cancer patients. Expert Rev Mol Diagn. 2013;13:827–44.View ArticlePubMedGoogle Scholar
  5. Crowley E, Di Nicolantonio F, Loupakis F, Bardelli A. Liquid biopsy: monitoring cancer-genetics in the blood. Nat Rev Clin Oncol. 2013;10:472–84.View ArticlePubMedGoogle Scholar
  6. Heitzer E, Ulz P, Geigl JB. Circulating tumor DNA as a liquid biopsy for cancer. Clin Chem. 2015;61:112–23.View ArticlePubMedGoogle Scholar
  7. Warton K, Samimi G. Methylation of cell-free circulating DNA in the diagnosis of cancer. Front Mol Biosci. 2015;2:13.View ArticlePubMedPubMed CentralGoogle Scholar
  8. Baylin SB, Jones PA. A decade of exploring the cancer epigenome—biological and translational implications. Nat Rev Cancer. 2011;11:726–34.View ArticlePubMedPubMed CentralGoogle Scholar
  9. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nat Rev Genet. 2002;3:415–28.View ArticlePubMedGoogle Scholar
  10. Jones PA, Baylin SB. The epigenomics of cancer. Cell. 2007;128:683–92.View ArticlePubMedPubMed CentralGoogle Scholar
  11. Schmidt B, Liebenberg V, Dietrich D, Schlegel T, Kneip C, Seegebarth A, et al. SHOX2 DNA methylation is a biomarker for the diagnosis of lung cancer based on bronchial aspirates. BMC Cancer. 2010;10:600.View ArticlePubMedPubMed CentralGoogle Scholar
  12. Kneip C, Schmidt B, Seegebarth A, Weickmann S, Fleischhacker M, Liebenberg V, et al. SHOX2 DNA methylation is a biomarker for the diagnosis of lung cancer in plasma. J Thorac Oncol. 2011;6:1632–8.View ArticlePubMedGoogle Scholar
  13. Dietrich D, Kneip C, Raji O, Liloglou T, Seegebarth A, Schlegel T, et al. Performance evaluation of the DNA methylation biomarker SHOX2 for the aid in diagnosis of lung cancer based on the analysis of bronchial aspirates. Int J Oncol. 2012;40:825–32.PubMedGoogle Scholar
  14. Schmidt B, Beyer J, Dietrich D, Bork I, Liebenberg V, Fleischhacker M. Quantification of cell-free mSHOX2 plasma DNA for therapy monitoring in advanced stage non-small cell (NSCLC) and small-cell lung cancer (SCLC) patients. PLoS One. 2015;10:e0118195.View ArticlePubMedPubMed CentralGoogle Scholar
  15. Church TR, Wandell M, Lofton-Day C, Mongin SJ, Burger M, Payne SR, et al. Prospective evaluation of methylated SEPT9 in plasma for detection of asymptomatic colorectal cancer. Gut. 2014;63:317–25.View ArticlePubMedGoogle Scholar
  16. U.S. Food and Drug Administration. PMA Number P130001. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfPMA/pma.cfm?id=P130001. Accessed 27 Apr 2017.
  17. Warren JD, Xiong W, Bunker AM, Vaughn CP, Furtado LV, Roberts WL, et al. Septin 9 methylated DNA is a sensitive and specific blood test for colorectal cancer. BMC Med. 2011;9:133.View ArticlePubMedPubMed CentralGoogle Scholar
  18. Grützmann R, Molnar B, Pilarsky C, Habermann JK, Schlag PM, Saeger HD, et al. Sensitive detection of colorectal cancer in peripheral blood by septin 9 DNA methylation assay. PLoS One. 2008;3:e3759.View ArticlePubMedPubMed CentralGoogle Scholar
  19. de Vos T, Tetzner R, Model F, Weiss G, Schuster M, Distler J, et al. Circulating methylated SEPT9 DNA in plasma is a biomarker for colorectal cancer. Clin Chem. 2009;55:1337–46.View ArticleGoogle Scholar
  20. Jung M, Pützer S, Gevensleben H, Meller S, Kristiansen G, Dietrich D. Diagnostic and prognostic value of SHOX2 and SEPT9 DNA methylation and cytology in benign, paramalignant, and malignant ascites. Clin Epigenetics. 2016;8:24.View ArticlePubMedPubMed CentralGoogle Scholar
  21. Dietrich D, Jung M, Puetzer S, Leisse A, Holmes EE, Meller S, et al. Diagnostic and prognostic value of SHOX2 and SEPT9 DNA methylation and cytology in benign, paramalignant and malignant pleural effusions. PLoS One. 2013;8:e84225.View ArticlePubMedPubMed CentralGoogle Scholar
  22. Gold B, Cankovic M, Furtado LV, Meier F, Gocke CD. Do circulating tumor cells, exosomes, and circulating tumor nucleic acids have clinical utility? A report of the association for molecular pathology. J Mol Diagn. 2015;17:209–24.View ArticlePubMedPubMed CentralGoogle Scholar
  23. Lee TH, Montalvo L, Chrebtow V, Busch MP. Quantitation of genomic DNA in plasma and serum samples: higher concentrations of genomic DNA found in serum than in plasma. Transfusion. 2001;41:276–82.View ArticlePubMedGoogle Scholar
  24. Holmes EE, Jung M, Meller S, Leisse A, Sailer V, Zech J, et al. Performance evaluation of kits for bisulfite-conversion of DNA from tissues, cell lines, FFPE tissues, aspirates, lavages, effusions, plasma, serum, and urine. PLoS One. 2014;9:e93933.View ArticlePubMedPubMed CentralGoogle Scholar
  25. Warton K, Lin V, Navin T, Armstrong NJ, Kaplan W, Ying K, et al. Methylation-capture and next-generation sequencing of free circulating DNA from human plasma. BMC Genomics. 2014;15:476.View ArticlePubMedPubMed CentralGoogle Scholar
  26. Dietrich D. Current status and future perspectives of circulating cell-free DNA methylation in clinical diagnostics. LaboratoriumsMedizin. 2016. https://doi.org/10.1515/labmed-2016-0039.
  27. Song L, Jia J, Yu H, Peng X, Xiao W, Gong Y, et al. The performance of the mSEPT9 assay is influenced by algorithm, cancer stage and age, but not sex and cancer location. J Cancer Res Clin Oncol. 2017;143:1093-101.Google Scholar
  28. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the comparative C(T) method. Nat Protoc. 2008;3:1101–8.View ArticlePubMedGoogle Scholar
  29. Dietrich D, Hasinger O, Bañez LL, Sun L, van Leenders GJ, Wheeler TM, et al. Development and clinical validation of a real-time PCR assay for PITX2 DNA methylation to predict prostate-specific antigen recurrence in prostate cancer patients following radical prostatectomy. J Mol Diagn. 2013;15:270–9.View ArticlePubMedGoogle Scholar
  30. Sundberg SO, Wittwer CT, Zhou L, Palais R, Dwight Z, Gale BK. Quasi-digital PCR: enrichment and quantification of rare DNA variants. Biomed Microdevices. 2014;16:639–44.View ArticlePubMedGoogle Scholar
  31. Whale AS, Huggett JF, Cowen S, Speirs V, Shaw J, Ellison S, et al. Comparison of microfluidic digital PCR and conventional quantitative PCR for measuring copy number variation. Nucleic Acids Res. 2012;40:e82.View ArticlePubMedPubMed CentralGoogle Scholar
  32. Tóth K, Sipos F, Kalmár A, Patai AV, Wichmann B, Stoehr R, et al. Detection of methylated SEPT9 in plasma is a reliable screening method for both left- and right-sided colon cancers. PLoS One. 2012;7:e46000.View ArticlePubMedPubMed CentralGoogle Scholar
  33. Nian J, Sun X, Ming S, Yan C, Ma Y, Feng Y, et al. Diagnostic accuracy of Methylated SEPT9 for blood-based colorectal cancer detection: a systematic review and meta-analysis. Clin Transl Gastroenterol. 2017;8:e216.View ArticlePubMedPubMed CentralGoogle Scholar
  34. Song L, Yu H, Jia J, Li Y. A systematic review of the performance of the SEPT9 gene methylation assay in colorectal cancer screening, monitoring, diagnosis and prognosis. Cancer Biomark. 2017;18:425-32.Google Scholar
  35. Lui YY, Chik KW, Chiu RW, Ho CY, Lam CW, Lo YM. Predominant hematopoietic origin of cell-free DNA in plasma and serum after sex-mismatched bone marrow transplantation. Clin Chem. 2002;48:421–7.PubMedGoogle Scholar
  36. Norton SE, Lechner JM, Williams T, Fernando MR. A stabilizing reagent prevents cell-free DNA contamination by cellular DNA in plasma during blood sample storage and shipping as determined by digital PCR. Clin Biochem. 2013;46:1561–5.View ArticlePubMedGoogle Scholar
  37. Lo YM, J Zhang J, Leung TN, Lau TK, Chang AM, Hjelm NM. Rapid clearance of fetal DNA from maternal plasma. Am J Hum Genet. 1999;64:218–24.View ArticlePubMedPubMed CentralGoogle Scholar
  38. Zhang R, Peng R, Li Z, Gao P, Jia S, Yang X, et al. Synthetic circulating cell-free DNA as quality control materials for somatic mutation detection in liquid biopsy for cancer. Clin Chem. 2017;63:1465–75.Google Scholar

Copyright

© The Author(s). 2017

Advertisement