Comparative Analysis of Two Digital PCR Platforms for Detecting DNA Methylation in Patient Samples

ABSTRACT

The choice of appropriate analytical methods for determining methylation patterns at specific loci across the genome is essential for identifying novel diagnostic and prognostic markers for subsequent clinical implementation. Various methods exist for determining methylation status using different technologies. In this study, we compared two distinct digital polymerase chain reaction (PCR) platforms: the nanoplate‐based Qiagen QIAcuity Digital PCR (dPCR) System and the droplet‐based Bio‐Rad QX‐200 Droplet Digital PCR (ddPCR) System. By assessing their efficacy and other attributes, we aimed to elucidate each platform's strengths and limitations in the sensitive detection of DNA methylation, thus contributing valuable insights to the field of molecular diagnostics. We analyzed the methylation status of the CDH13 gene in 141 formalin‐fixed, paraffin‐embedded breast cancer tissue samples using our in‐house developed methylation‐specific labeled assay. The specificity and sensitivity of the CDH13 assay evaluated by dPCR were 99.62% and 99.08%, respectively; ddPCR analysis reached a specificity of 100% and a sensitivity of 98.03%. In addition, our data revealed a strong correlation between the methylation levels measured by both methods (r = 0.954). Although both methods are based on different technologies, they yielded comparable, highly sensitive experimental data in our study. Consequently, the main criteria for selecting an optimal digital PCR platform for methylation analysis may lie in other factors such as workflow time and complexity, instrument requirements, the possibility of temperature gradient, reanalysis, or offline options.

Summary

• Digital polymerase chain reaction (PCR) is an innovative technology enabling ultrasensitive nucleic acid detection and absolute quantification.

• Compared to real‐time PCR, it offers several advantages, including absolute quantification without external references and greater robustness to PCR efficiency variations, thereby enhancing assay precision.

• These attributes make digital PCR particularly well‐suited for detecting and quantifying methylated DNA.

• In this study, we compared two digital PCR platforms: a nanoplate‐based system and a droplet‐based system.

• By evaluating their performance and characteristics, we aimed to clarify each platform's strengths and limitations in the sensitive detection of DNA methylation, providing valuable insights for molecular diagnostics.

• Notably, there is a lack of comparable studies directly examining different digital PCR platforms in the context of DNA methylation quantification.

1. Introduction

Epigenetic modifications represent comprehensive events describing reversible changes in gene expression without altering primary nucleotide sequences [1]. Epigenetics, as the concept “beyond genetics,” includes heritable changes such as the chemical alteration of histones, expression of noncoding RNAs, and DNA methylation [2]. Epigenetics plays a fundamental role in the development and differentiation of tissues and organisms via the modulation of gene expression [3]. On the other hand, disruption of gene expression mediated by aberrantly changed epigenetic machinery can lead to various diseases, from neurodevelopmental and cardiovascular disorders to initiation of carcinogenesis [4]. DNA methylation at the fifth position of cytosine (5mC) at the CpG sequence is a well‐investigated epigenetic mechanism essential for gene regulation [5]. Specific methylation patterns within promoter areas, as well as the region downstream of the transcription start site (TSS), provide significant insights into gene activity. Generally, increased DNA methylation is characterized as a repressive mark with the capacity to silence genes [6]. Hypermethylation of CpG in promoters is frequently observed in tumor‐suppressor genes, in which transcriptional silencing is associated with cancer development. Early detection and accurate identification of aberrant DNA methylation patterns offer a promising chance for curative therapy and can reduce the mortality of patients suffering from different malignancies [7]. Up to now, numerous different methods have been proposed for DNA methylation detection. Overall, current methods used for DNA methylation detection could be divided into single‐gene techniques and whole‐genome techniques. Single gene techniques are subsequently divided into three major approaches: restriction enzyme‐based (e.g., methylation‐sensitive restriction enzyme polymerase chain reaction [PCR], methylation‐specific multiplex ligation‐dependent probe amplification), bisulfite‐based (e.g., methylation‐specific PCR, methylation sensitive‐high resolution melting, pyrosequencing, and Sanger sequencing), and affinity enrichment‐based techniques (e.g., antibody‐based hydroxymethyl DNA immunoprecipitation, hydroxymethyl selective chemical labeling) [8]. The choice of the appropriate methods depends on the target of the analysis. The main goals in decision processes for choosing appropriate DNA analysis methods include questions about the quality and quantity of DNA input, cost‐effectivity, time, and availability of required laboratory equipment [9]. It is important to keep in mind that the selected methods must demonstrate high specificity and sensitivity to eliminate false negative or false positive results. Routinely, bisulfite conversion is regarded as the gold standard for DNA methylation analysis, enabling the quantitative and qualitative detection of methylcytosine and, thus, distinguishing methylated from unmethylated DNA at single‐base resolution [10]. However, treatment with sodium bisulfite has several limitations, such as harsh reaction conditions, subsequent DNA fragmentation, and the absence of DNA complementarity, resulting in its instability. These limitations can significantly reduce the ability of various analytic procedures to detect DNA methylation abnormalities; thus, further investigation and development of novel diagnostic methods for estimating methylated CpGs pose a major challenge. Moreover, the biological source of DNA plays an important role in downstream analysis. Formalin‐fixed and paraffin‐embedded (FFPE) samples are one of the most common ways to preserve clinical specimens for clinical practice as well as research investigation [11]. DNA extracted from FFPE is typically characterized by a high degradation rate due to the harsh process of fixation and long‐term storage [12]. In recent years, digital PCR has gained the interest of the scientific community due to the precise and efficient DNA detection with a wide range of applications in basic and clinical research [13]. This method represents a highly sensitive tool for rare mutation detection and an optimal approach for detecting methylated DNA in clinical samples from various biological sources, including those routinely preserved as FFPE. In this study, we compare two distinct digital PCR platforms employed for detecting DNA methylation within breast cancer FFPE tissue samples.

2. Materials and Methods

The analyzed sample set consisted of 141 FFPE cancer tissue samples provided by the Department of Pathological Anatomy, Jessenius Faculty of Medicine in Martin, Comenius University and University Hospital in Martin.

2.1. DNA Isolation and Bisulfite Modification

Genomic DNA from tissue deparaffinized with xylene was isolated using a DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany) according to the manufacturer's recommended protocol. The concentration of obtained DNA was determined by Qubit 3.0 using a dsDNA BR Assay kit (Thermo Fisher Scientific, Waltham, MA, USA). For further methylation analysis, one µg of isolated DNA was modified with an EpiTect Bisulfite kit (Qiagen) following the manufacturer´s instructions.

2.2. Methylation Analysis

Methylation levels of three CpG sites in the CDH13 promotor region (chr16:82,626,843; chr16:82,626,845; chr16:82,626,859) were detected by two different digital PCR platforms: nanoplate‐based QIAcuity Digital PCR (dPCR) System (Qiagen) and droplet‐based QX‐200 Droplet Digital PCR (ddPCR) System (Bio‐Rad Laboratories, Hercules, CA, USA). Primers and probes were previously used in our other study [14] and were designed in online programs, MethPrimer [15] and Primer3Plus [16]. The sequences are provided in Table 1. The assay was optimized for the simultaneous detection of methylated and unmethylated DNA in a single reaction mix. Each primer/probe set contained an internal probe specific for the methylated (M‐probe, FAM‐labeled) or unmethylated (UnM‐probe, HEX‐labeled) regions, while primers (forward/reverse) were the same for both sets.

Table 1.

Primers and probe sequences for CDH13 methylation detection.

Primer/probe	Primer/Probe Sequence 5′ → 3′
Forward primer	AAAGAAGTAAATGGGATGTTATTTTC
Reverse primer	ACCAAAACCAATAACTTTACAAAAC
M‐Probe (FAM)	TCGCGAGGTGTTTATTTCGT
UnM‐Probe (HEX)	TTTTGTGAGGTGTTTATTTTGTATTTGT

Abbreviations: M, methylated; UnM, unmethylated.

Each dPCR reaction was prepared according to the manufacturer's instructions in a 12 μL volume per well containing 3 µL of QIAcuity 4× Probe PCR master mix, 0.96 µL of forward/reverse primer, 0.48 µL of each probe, 2.5 µL of DNA template, and RNase‐free water. Reaction mixtures were pipetted into 24‐well nanoplate, with 8500 partitions per well, and processed in QIAcuity One, 2plex Device. The instrument automatically generated partitions, ran PCR, and detected fluorescence in all partitions. Cycling consisted of initial heat activation (95°C for 2 min), followed by 40 cycles of denaturation (95°C for 15 s), and by a combined annealing/extension step (57°C for 1 min). The exposure duration for both analyzed channels (FAM/HEX) was 500 ms. The QIAcuity Software Suite, version 2.1.7, was utilized to analyze individual partitions regarding the presence/absence of FAM or HEX signals. The threshold was manually set at a value of 45, taking into account the signal amplitude of positive controls (fully methylated and unmethylated DNA) in the optimal concentration, the overall count of positive droplets, and binding specificity. The acceptance criteria for the sample were over 7000 valid partitions and at least 100 positive partitions. The analysis of the sample that did not meet the acceptance criteria was repeated. The methylation level was expressed as a ratio of positive FAM‐detected partitions to the sum of all positive partitions detected on both channels.

The reaction mixture for ddPCR contained 10 μL of Supermix for Probes (No dUTP) (Bio‐Rad Laboratories), 0.45 μL of forward/reverse primer, 0.45 μL of each probe, 2.5 μL of DNA template adjusted with RNase‐free water up to a final volume of 20 μL. After transfer to the DG8 cartridge and the addition of 70 μL of Droplet Generation Oil for Probes, approximately 20,000 droplets per sample were generated in the QX200 Droplet Generator. The resulting droplet emulsion (40 μL) was transferred into a 96‐well PCR plate and submitted to endpoint PCR in the T100 thermal cycler (Bio‐Rad Laboratories). The optimized PCR protocol consisted of an initial denaturation (95°C for 10 min), 40 cycles of denaturation (94°C for 30 s), annealing/extension (57°C for 1 min) with a ramp rate decreased to 1°C/s, and enzyme deactivation (98°C for 10 min). The product was incubated overnight at 12°C following recommendations by Rowlands et al. [17], and fluorescence was detected by QX200 Droplet Reader. The QuantaSoft v.1.7 software (Bio‐Rad Laboratories) was utilized to analyze and categorize each droplet based on the fluorescence emission detected in HEX or FAM channels. The threshold amplitudes for the FAM and HEX signals were determined based on similar criteria as in the previous method, with values of 2600 and 1745, respectively. Exclusion criteria were applied to the samples with fewer than 10,000 accepted droplets, as suggested by Rashid et al. and Arroyo et al. [18, 19], and those with fewer than 100 positive droplets. One repeated analysis was conducted on the sample that did not meet the quality acceptance criteria. Similar to dPCR, the methylation level was calculated as a ratio of positive FAM‐detected droplets to the sum of all positive droplets.

2.3. Statistical Analysis

Two‐fold dilutions of fully methylated and unmethylated EpiTect DNA controls (Qiagen) with a range of 8000–62 copies per reaction were performed to determine the linearity. The specificity and sensitivity were determined in each run and then averaged. Specificity was calculated as the ratio of copies of the unmethylated control DNA detected by the UnM‐Probe to all copies of the unmethylated control DNA (detected by both probes); sensitivity as the ratio of copies of the methylated control DNA detected with the M‐Probe to all copies of the methylated control DNA.

The correlation between the two detection methods was assessed using linear regression and Pearson correlation. The Bland‐Altman analysis was conducted in an online web tool (https://huygens.science.uva.nl/BA-plotteR/). All statistical analyses and graphical outputs were performed using GraphPad Prism 8.1.1 (GraphPad, San Diego, CA, USA).

3. Results

Using dPCR, we successfully analyzed 123 samples. After retesting 22 samples, 18 samples again failed to meet the quality criteria and had to be excluded. The number of successfully analyzed samples by ddPCR was 127; 29 samples required retesting, and 14 were subsequently excluded. The reasons for exclusion, along with the number of excluded samples, are summarized in Table 2. In some cases, we encountered difficulty analyzing sample methylation using one method but had no problem using the other one. Five samples were successfully analyzed only by dPCR and did not meet the quality criteria set in ddPCR. On the other hand, nine samples could be analyzed successfully only by using ddPCR. We were unable to analyze nine samples altogether using either method.

Table 2.

Exclusion criteria.

Exclusion criterium	Excluded samples
	dPCR	ddPCR
< 7000 partitions/< 10,000 droplets	3	4
< 100 partitions/droplets	9	10
Ambiguous result	6	0
Total	18	14

Abbreviations: dPCR, digital polymerase chain reaction; ddPCR, droplet digital polymerase chain reaction.

The average specificity of dPCR across all runs was 99.62%, the sensitivity reached 99.08%. As no copies of unmethylated control DNA were detected by the M‐Probe in ddPCR in any run, the specificity reached 100%. The average sensitivity across all runs was 98.03%.

The regression curve describing the linearity of dPCR and ddPCR, respectively, is presented in Figure 1.

Figure 1.

Dilution linearity of two digital methylation‐specific polymerase chain reaction (PCR) methods. The methods show strong linearity for both methylated and unmethylated probes, with R ² values indicating high correlation.

The number of accepted partitions in each dPCR run ranged from 7062 to 8281, representing an average acceptance rate of 94.53%. The number of accepted droplets in ddPCR ranged from 11,592 to 17,932, with an average acceptance rate of 72.84%. The methylation levels of individual samples detected by both methods (n = 118) are compared in Figure 2D. Figure 2A shows simple linear regression and Pearson correlation between detected data (correlation coefficient r = 0.9537). Figure 3 also includes a Bland‐Altman plot and a 2D boxplot to illustrate the agreement and distribution between the two methods. In Figure 3 is presented an illustrative data output of three samples obtained from dPCR and ddPCR analysis.

Figure 2.

Comparison of DNA methylation detected using digital polymerase chain reaction (PCR) (dPCR) and droplet digital PCR (ddPCR). (A) Pearson correlation plot showing a strong positive correlation between dPCR and ddPCR methylation levels. (B) Bland–Altman plot displaying the agreement between the two methods. The dotted line indicates the mean difference, and the dashed lines ±1.96 standard deviation (SD) limits of agreement. The slope of the difference suggested proportional bias; using a regression‐based limits of agreement was recommended. (C) Boxplots comparing the distribution of methylation levels; boxes represent the interquartile range (IQR), whiskers represent 1.5× IQR, circles denote outliers, and “×” indicates the mean. (D) Bar plots of methylation levels across individual samples measured by ddPCR (light gray) and dPCR (dark gray), showing overall agreement with some variability.

Figure 3.

Representative one‐dimensional (1D) fluorescence plots from green channel (methylated signal) and yellow channel (unmethylated signal) detected by digital polymerase chain reaction (PCR) (top) and droplet digital PCR (bottom). Colored dots represent positive partitions (particles or droplets), separated from negative (gray) ones by a threshold line. The methylation levels of sample A were 2% by both methods; sample B 13% (dPCR) and 9% (ddPCR), and sample C exhibited methylation levels of 26% and 31%, respectively.

4. Discussion

The detection and monitoring of DNA methylation status using molecular methods provide valuable insights into the biology of various pathologies, including malignant transformation. Several methods are available for detecting DNA methylation in biological samples [1]. Choosing appropriate high‐throughput methods with a single‐base resolution to evaluate methylated CpG dinucleotides, localized particularly in gene promoter areas and regions downstream of the TSS, is essential for subsequent experimental research and potential applications in clinical practice [20].

In this study, we aimed to compare two digital PCR platforms: the nanoplate‐based Qiagen QIAcuity Digital PCR and the Bio‐Rad QX‐200 Droplet Digital PCR. For this purpose, we used the CDH13 assay targeting CpG methylation in FFPE tissue samples. Both platforms allow the absolute quantification of the target DNA template without the need for calibrators and standard curves, thus overcoming the shortcomings of quantitative PCR (qPCR) [21]. To detect methylation levels at specific CpG sites, we optimized our in‐house developed methylation‐specific labeled assays. The tumor suppressor gene CDH13 was selected due to its increased methylation in various types of malignant tissues, as documented in a number of studies [22, 23, 24, 25]. Recently, several studies have focused on DNA methylation analysis using digital PCR platforms [26, 27, 28, 29, 30]. All mentioned studies used ddPCR platforms to estimate the level of DNA methylation without any comparison with any other digital PCR system. Both digital PCR platforms tested in our comparative study offer highly precise, absolute quantification of nucleic acid obtained from various types of biological samples. However, differences exist in methodical, technical, and analytical aspects. We have summarized relevant factors for consideration in Table 3, comparing selected criteria between dPCR and ddPCR systems.

Table 3.

Comparison of selected criteria between two digital PCR platforms.

	Digital PCR	Droplet Digital PCR
Workflow time	+	+++
Workflow complexity	+	+++
Instruments requirements	+	++
Personnel training requirements	+	+++
Contamination risk	+	++
Susceptibility to errors	+	++
Temperature gradient	−	+
Droplet/partition loss	+	++
Off‐line functionality	−	+
Reanalysis possibility	+	−
User‐friendly software	−	+
Cost	++	+++

Abbreviation: PCR, polymerase chain reaction.

In our study, we observed remarkable conformity between the platforms regarding CDH13 methylation. To ensure the reliability of the tested assay evaluated by both dPCR and ddPCR and to establish the thresholds, we first determined the Limit of Blank and Limit of Detection using commercial methylated and unmethylated DNA. Our results showed similar outcomes in the specificity and sensitivity of the assay evaluated by dPCR and ddPCR systems (99.62% and 99.08% vs. 100% and 98.03%). This data indicates the enormous discriminative power of our designed assay to distinguish methylated from unmethylated DNA. Both methods provide partitioning of PCR reactions into thousands of independent reactors, significantly reducing the risk of PCR inhibition from contaminants compared to standard qPCR analysis [31]. When comparing dPCR and ddPCR, the risk of contaminations and susceptibility to errors are more probable when using ddPCR due to its workflow involving four separate devices: droplet generator, heat sealer, PCR cycler, and automated reader. For example, the process of transferring reaction mixes into DG8 cartridges for generating oil droplets and subsequent transfer into 96‐well plates presents multiple opportunities for sample contamination or errors, a risk mitigated by the dPCR platform where all workflow steps are performed in one device. On the other hand, the all‐in‐one approach of dPCR can also pose a disadvantage. An essential aspect of PCR optimization involves empirically determining the optimal annealing temperature through temperature gradients. In various digital PCR platforms, poorly optimized assays may manifest as “rain”, characterized by droplets appearing between the cluster of negative and positive droplets. This issue often arises from inadequate annealing temperatures that tend to be higher than initially predicted [32]. Due to the separation of the thermal cycler from the reader instrument, the ddPCR platform enables detached temperature optimization. Conversely, in integrated dPCR platforms where all components are consolidated into a single instrument, the absence of a separate device for temperature gradient analysis poses a complication. To overcome this limitation, users may resort to employing multiple nanoplates, each dedicated to a specific tested annealing temperature. However, this approach invariably increases the costs associated with the analysis.

The next critical factor affecting the final analysis result lies in the number of particles usable for subsequent detection. As mentioned above, ddPCR technology allows the partitioning of the reaction mix into approximately 20,000 oil‐encapsulated nanodroplets to produce data. According to our results, the number of accepted droplets in each ddPCR ranged from 11,592 to 17,932, with an average of 72.84% acceptance rate. These data are consistent with a previously published study by Pinheiro et al., who assessed a number of positive droplets in the range from 11,000 to 18,000 [33]. In the dPCR platform, each well of the nanoplate is divided into 8,500 partitions with a volume of nL [34]. The number of accepted partitions was between 7062 and 8281, representing an average acceptance rate of 94.53%. Despite the different number of partitions/droplets per well, as well as the lower acceptance rate in ddPCR, both methods showed similar percentages of methylation in all analyzed samples (Pearson r = 0.954).

In conclusion, our research uncovered that both tested platforms offer precise analysis of DNA methylation status, and selecting the suitable digital PCR platform for methylation analysis, therefore, relies on various additional criteria. Considering the workflow time and complexity, instrument requirements, susceptibility to errors and contamination, the possibility of reanalysis, or user‐friendliness of accompanying software can significantly aid in the decision‐making process. While both platforms hold promise for future clinical diagnostics, cost currently stands out as a significant limitation for broader applications.

Ethics Statement

The study was conducted according to the ethical principles of the Helsinki Declaration and approved by the Ethics Committee of Jessenius Faculty of Medicine (EK1822/2016).

Consent

Informed consent was obtained from all participants.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.