The detection of specific hypermethylated WIF1 and NPY genes in circulating DNA by crystal digital PCR™ is a powerful new tool for colorectal cancer diagnosis and screening

Background In oncology, liquid biopsy is of major relevance from theranostic point of view. The searching for mutations in circulating tumor DNA (ctDNA) in case of colorectal cancers (CRCs) allows the optimization of patient care. In this context, independent of mutation status biomarkers are required for its detection to confirm the presence of ctDNA in liquid biopsies. Indeed, the hypermethylation of NPY and WIF1 genes appear to be an ideal biomarker for the specific detection of ctDNA in CRCs. The objective of this work is to develop the research of hypermethylation of NPY and WIF1 by Crystal Digital PCR™ for the detection of ctDNA in CRCs. Methods Detection of hypermethylated NPY and WIF1 was developed on Cristal digital PCR™. Biological validation was performed from a local cohort of 22 liquid biopsies and 23 tissue samples from patients with CRC. These patients were treated at the University Hospital of Besancon (France). Results The local cohort study confirmed that NPY and WIF1 were significantly hypermethylated in tumor tissues compared to adjacent non-tumor tissues (WIF1 p < 0.001; NPY p < 0.001; non-parametric Wilcoxon paired-series test). Histological characteristics, tumor stages or mutation status were not correlated to the methylation profiles. On the other hand, hypermethylation of NPY or WIF1 in liquid biopsy had a 95.5% [95%CI 77–100%] sensitivity and 100% [95%CI 69–100%] specificity. Conclusion Using Crystal digital PCR™, this study shows that hypermethylation of NPY and WIF1 are constant specific biomarkers of CRCs regardless of a potential role in carcinogenesis. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-021-08816-2.


Introduction
Colorectal cancer (CRC) is the third most common cancer worldwide with more than one million new cases diagnosed every year. The development of new chemotherapies, especially cancer personalized therapies, has improved outcomes of patients with CRC. The effectiveness of targeted therapies is based on mutational profiles of RAS/MAPK pathway genes [1]. These mutations are typically sought at the time of diagnosis from a cancer tissue biopsy. However, in cases of a non-feasible biopsy, this search can be performed on circulating tumor DNA (ctDNA). Liquid biopsy is an increasingly common oncology test for the diagnosis of cancer and follow-up of treatments. The search for ctDNA mutations is mainly used in a theranostics approach, particularly in lung and colorectal cancers [2]. The liquid biopsy is a non-invasive approach and can be repeated over time to perform dynamic monitoring of tumors. Unfortunately, this theranostic approach depends on the fluctuant release of tumoral DNA in the vascular compartment [3].
Next Generation Sequencing (NGS) strategies are now commonly used for the detection of mutations on ctDNA [4]. Advanced quantitative technologies, such as the digital PCR (dPCR), have been developed to increase sensitivity of detection [5]. The dPCR amplifies millions of individual DNA fragments using thousands of waterin-oil droplets. This compartmentalization increases the detection sensitivity, and is especially adapted for mutations present in low concentration. If no mutations are detected in liquid biopsy, discrimination between an unmutated profile and an absence of ctDNA in the plasma sample is not possible. To address this issue, high sensitivity tumor-specific epigenetic biomarkers have been identified to assert presence of ctDNA [6,7]. DNA methylation is the most studied epigenetic mechanism in this respect [8].
Indeed, the tumor cell epigenome associates global hypomethylation [9,10] interspersed with hypermethylated specific regions such as promoters of tumor suppressor genes, and correlated with a decreased expression [11]. These modifications happen in tumorigenesis and many studies are looking at these new epigenetic markers [11,12].
Based on Roperch and al [13]. and Garrigou and al [14]. studies, DNA hypermethylation of NPY (Neuropeptide Y) and WIF1 (Wnt inhibitory factor 1) appears as a specific marker of ctDNA in CRCs. Hypermethylation of NPY and WIF1 is found in 100% of CRCs [13,14], while the presence of a defined mutation is inconstant in CRCs. For instance, a mutation of KRAS is only found in 40% of CRCs and therefore cannot be used as a biomarker for the presence of CRCs ctDNA. Thus, hypermethylation of NPY and WIF1 appears to be a better biomarker [6].
In the present study, we adapted a droplet-based dPCR protocol previously described [13,14] in the Naica Crystal Digital PCR system™ (Stilla Technonologies, Villejuif, France) in order to investigate the hypermethylation of the NPY and WIF1 in CRC tissues or ctDNA. The objective is to determine whether the hypermethylation of NPY and WIF1 is a specific biomarker of CRC in liquid biopsy by Crystal Digital PCR™ and could be used for routine diagnosis, recurrence and treatment follow-up.

Patients
Digital PCR analyses were conducted on 22 blood and 23 tissue samples from patients with CRC treated at the University Hospital of Besancon (France) ( Table 1). Before inclusion all patients provided written informed consenting to the use of their clinical, biological and demographic data for research purposes. Samples were preserved in the framework of the "Tumorothèque Régionale de Franche-Comté". This scientific board has an authority to approve human studies. And blood samples from patients without any oncologic background (considered as control group) were collected at the "Etablissement Français du sang". These samples were blood donations.

DNA isolation and bisulfite modification
Tumors DNA was extracted from frozen biopsies and FFPE samples. In EDTA collected blood samples were pre-treated to obtain supernatants which were stored at − 80°C. Circulating cell-free DNA (cfDNA) was extracted from 4 mL to 6 mL of plasma using the QIAamp® Circulating Nucleic Acid kit (Qiagen®, Hilden, Germany) according to the manufacturer's protocol and resuspended in 50 μL of buffer. The quantity of DNA was mesured by Qubit 2.0 fluorometer (Invitrogen®, Life Technologies) and were obtain between 1 to 240 ng/μL (mean = 22.6 ng/μL).
For all samples, bisulfite treatment was performed to transform unmethylated cytosine into thymidine without changing methylated cytosine, by EZ DNA Methylation kit® (Zymo Research) for DNA concentration of 1 ng/μL. Mutation status Analysis by NGS, microsatellite phenotype and MLH1 promoter methylation by pyrosequencing were performed as part of patient management. After bisulfite treatment of tumor DNA, 5 μL of the bisulfite-treated DNA solution was analyzed by a pyrosequencing technique according to the PyroMark™ Q24 CpG MLH1 procedure (Qiagen®, Hilden, Germany). The analyzed promoter sequences correspond to the proximal region, − 209 to − 181, relative to the transcription start site of hMLH1 gene.

Development of the digital PCR analysis
An aberrant hypermethylation of NPY and WIF1 genes has been described in CRCs. We developed a (previously described) 2-panel assay targeting these biomarkers previously described on a Naica Crystal Digital PCR sys-tem™ (Stilla Technonologies, Villejuif, France). After bisulfite conversion, a volume of 5 μL of DNA extract was assembled in 20 μL PCR mixtures using 1 X Per-feCTa Multiplex qPCR ToughMix (Quanta Biosciences, Gaithersburg, MD, USA), 100 nM Fluorescein, 1 μM each primer, 250 nM each hydrolysis probe. In Table 2, probe and primer sequences were previously designed by Garrigou et al. [14] (Fig. 1). But we developed Crystal digital PCR™ specific conditions by 95°C for 5 min, followed by 50 cycles of 95°C for 15 s and 57°C for 10 s.

Data analysis
The droplet identification and fluorescence measurements in each detection channel were performed using Stilla's Crystal Miner® software. Spill-over compensation was defined and applied. Gating of positive and negative droplet clusters was performed. Transcriptional impact of hypermethylation of NPY and WIF1 was evaluated on the TCGA-COAD data. Transcription data were normalized with Deseq2 [15] and gene expression was compared between non-tumor (n = 41) and tumor samples (n = 480) using bilateral Student test.

Statistical analysis
An analysis of the difference in methylation between tumor and healthy tissue was conducted using the nonparametric Wilcoxon test. The difference in methylation between the 2 groups was determined using a nonparametric Mann-Whitney test; the analysis of more than 2 groups was performed using the non-parametric Kruskal-Wallis test. Spearman's nonparametric test was used for correlation research. Statistical analyses were performed using GraphPad (GraphPad Software Inc., San Diego, CA). An uncertainty of 5% was defined for each of the tests and a p-value < 0.05 was considered statistically significant.
The required numbers of subjects (RNS) per groups were computed with the software R v4.0.2 using the observed means and standard deviations in our cohort and the usual statistical parameters (a significance level of 0.05 and a power of 0.90). Grouping by tumor/nontumor, tstandard deviations were significantly different, therefore the ANOVA test was used. This estimation shows that an important difference of the positive droplets number between tumor and non-tumor. For the tumoral status (tumor vs non-tumor), the estimated RNS was 10 samples for WIF1 and 16 samples for NPY. Using a Student test for the liquid biopsies analyzes, the RNS were 10 samples for NPY and 9 samples for WIF1.
In our study, 23 tumor tissues and 22 bloods samples were analysed with powerful significativity (p < 0.001).

Detection of NPY and WIF1 methylation by crystal digital PCR™
A digital PCR technique has been developed on the Naica Crystal Digital PCR system™ (Stilla Technonologies, Villejuif, France) for the detection of NPY and WIF1 genes' methylation.
A limit of blank (LOB) was calculated for the two detection channels allocated to Cyanine 5 and Cyanine 3 for NPY and WIF1 respectively. A total of 12 experiments were performed with unmethylated DNA quantified at 0.2 ng/μL. Garrigou et al. [14] showed that the rate of false positive droplets is independent of the total amount of DNA. The LOB with the confidence level (1α) was defined as the maximum number of false positive events that are plausible with a 1-α level probability (95% for risk α = 5%). The number of false positive droplets was recorded to targeted channel detection. The LOB was set as one false positive droplet for NPY and five for WIF1 promoters' methylation.
A dilution test was performed in order to assess the detection sensibility of the technique. Five concentrations of a fully methylated control DNA (EpiTect Qia-gen®, Hilden, Germany) at 10% have been tested: 0.5 ng/ μL, 0.25 ng/μL, 0.1 ng/μL, 0.05 ng/μL, and 0.01 ng/μL (Fig. 2). For a concentration of 0.05 ng/μL with a percentage of DNA methylated at 10%, the developed technique was able to detect the methylation of the NPY genes (R 2 = 0.9715) and WIF1 (R 2 = 0.9775). These results show the high sensibility of this method. The difference in the level of detection between the two targets can be explained by the difference in the numbers of CpG analyzed by our technique (Fig. 1). Indeed, the methylation profile of NPY is more restrictive because 11 CpG must be methylated for their detection. The detection of WIF1 applies to a region of five CpG.

Validation on local cohort
Characteristics of the patient population A cohort of 45 patients (22 blood and 23 tissue samples) with CRC was included in the study (  (Fig. 3). With nonparametric Spearman's test, the number of positive droplets was also correlated with the concentration of  DNA extracted for WIF1 (R 2 = 0.436, p = 0.0377) and for NPY (R 2 = 0.809, p < 0.0001). For control group, 11 CRC adjacent non-tumor tissues were used. A comparison of the number of positive droplets was performed according to histology (partially or well differentiated adenocarcinoma) and tumor stage (I and II, III or IV). No significant difference was found between the methylation profile and tumor histology (p = 0.6950; for NPY p = 0.6319 for WIF1; non-parametric Kruskal-Wallis test, n = 21) (Fig. 4A). The same absence is observed with the tumor stage (p = 0.2873 for NPY; p = 0.0517 for WIF1; non-parametric Kruskal-Wallis test, n = 22) (Fig. 4B).
Importantly, significant hypermethylation of both genes was demonstrated in tumor tissues compared to adjacent non-tumor tissues (NPY p = 0.001; WIF1 p = 0.002; nonparametric Wilcoxon paired-series test) (Fig. 4D). The As shown by multivariate Anova analyzes, the NPY and WIF1 methylation are powerful biomarkers of all types of CRCs independently of mutations, MSI and MLH1 methylation status. Comparing tumor samples and non-tumor colonic tissues, TCGA data were analyzed for WIF1 and NPY transcripts (Fig. 5). The NPY transcripts in CRCs are lower than in non-tumor tissues (Fig. 5A). However, transcriptomic analysis shows overexpression of WIF1 in CRCs (Fig. 5B).

CRC patient liquid biopsies detection of WIF1 and NPY DNA methylation
In the cohort of 22 total circulating DNA samples, hypermethylation of NPY or WIF1 in liquid biopsy had 95.5% of sensitivity [95% CI, 77 to 100%] and 100% of specificity [95% CI, 69 to 100%]. And hypermethylation of NPY or WIF1 was observed for 95.5% of the extracts, of which 77.3% with methylation for both genes. All patients with stage IV disease were detected (Fig. 6). For one stage I CRC patient, the extract showed no methylation for both the genes. This extract was characterized by a low concentration of total circulating DNA (1.1 ng/ μL) and no mutation was detected by NGS either on the plasma extract or on the tissue biopsy, suggesting a disease in the early stages of carcinogenesis. The ten ctDNA in the control group were analyzed and the methylation of the NPY and WIF1 genes was negative (Supplementary data).
A significant correlation was observed between the concentration of total circulating DNA in the plasma extracts and the number of positive droplets (p < 0.0001 for NPY and WIF1; non-parametric Spearman's test). As expected, the correlation between the number of positive droplets and the percentage of mutation found in NGS was not observed (p = 0.0703 for NPY; p = 0.0787 for WIF1; non-parametric Spearman's test).

Methylation of WIF1 and NPY in colorectal tissues
In the cohort, 23 primary colorectal tumor tissues were analyzed associated with 11 adjacent non-tumor tissues to measure DNA hypermethylation of NPY and WIF1, as potential new biomarkers of CRC in liquid biopsies. A significant hypermethylation of NPY and WIF1 in the tumor tissues was demonstrated (p < 0.001 for NPY; p < 0.001 for WIF1) (Fig. 4D). However, a higher than LOB positive droplet counts was found in healthy tissues. This result could be explained by the very high sensitivity of dPCR. Indeed, tissues are considered healthy by microscopic analyzes but some tumor cells, or pre-tumor cells, would be present without any microscopical characteristic. In addition, biopsies of adjacent non-tumor tissues may contain tumor cells and/or ctDNA within their vascularization. This tumor "contamination" of healthy tissue was already found with low positivity [13,14]. These results are consistent with those of Roperch et al. who set a methylation threshold above this, and considered the threshold of positive result, 25% for NPY and 7% for WIF1, because non-tumor adjacent tissues also had a low percentage of methylation level [13].
Furthermore, a significant correlation between the number of positive droplets for NPY and WIF1 was found (R 2 = 0.62, p = 0.0016), suggesting that methylation of both genes is early and concomitant in carcinogenesis. On the other hand, no significant correlation was observed with the MLH1 promoter methylation used as a predictive factor of sporadic forms of CRC, but on the other hand found in MSI CRCs. Thus, methylation of the NPY and WIF1 genes appears to be a process independent of mismatch repair genes methylation and suggests mechanisms of systematic methylation, concerning low significance genes for carcinogenesis (as NPY or WIF1). And a conditioned methylation of some tumor suppressor genes whose repression is necessary for carcinogenesis process. This is confirmed in tumor DNA extracted from a biopsy for which no methylation of the MLH1 gene promoter was found, whereas methylation of the NPY and WIF1 genes was observed. Significant difference in the level of methylation was not found according to the stage of the tumor. These results are consistent with constant methylation of both genes in tumor DNA. NPY and WIF1 DNA methylation are powerful specific biomarkers of all types of CRCs.

Methylation of NPY and WIF1 on circulating tumor DNA
The method has been validated on tissues and then on ctDNA from CRC patients. All analyzed plasma samples from CRC patients were positive for WIF1 and/or NPY methylation except one stage I patient sample (Fig. 6). This negative DNA extract was very low in total DNA (1.1 ng/μL). The absence of ctDNA in this sample is probably the cause of the negative result [6]. These results were also confirmed in preliminary data on Stilla Application Note [17]. Nevertheless, the presence of methylation of NPY and WIF1 genes in all other samples suggests that methylation process occurs is constant in carcinogenesis. Therefore, the detection of this epigenetic process could be a relevant marker for CRCs screening. Thus, this sensitive and non-invasive technique can be an interesting screening tool for CRCs exploration, and especially in advanced stages that require rapid treatment.

NPY and WIF1's role in colorectal carcinogenesis
The hypermethylation of NPY promoter in CRCs leads to a strong repression of its transcription (Fig. 1A). The region targeted by our dPCR protocol partially overlaps with the 5'UTR of NPY and is entirely contained within the promoter of NPY (Fig. 1B). The genomic colocation with the promoter could explain the negative correlation between the methylation of the dPCR target and the expression of NPY. Nevertheless, the role of NPY in the tumorigenesis process is not fully elucidated. In vitro, NPY appears to  (stage I and II, III or  IV). Non-parametric Kruskal-Wallis test. C) The mutated or non-mutated status of the tumor. Non-parametric Mann-Whitney test. D) Tumor tissues compared to the adjacent healthy tissues. Non-parametric Mann-Withney test in paired series. ADC-WD: well-differentiated adenocarcinoma, ADC-PD: partially differentiated adenocarcinoma promote tumorigenesis, probably in a neoneurogenesis context in which tumor cells exploit neurotransmitters to generate a pro-tumor environment [18]. NPY repression should thus inhibit tumor proliferation. Paradoxically, NPY appears to reduce the invasive potential of tumor cells in vitro [19]. In the CISTROME database with experimental data, we observed that EP300, EZH2, JARID2, RYBP, PAX5, and SUZ12, might bind the NPY targeted region [20]. Also, in silico analysis shows that this CpG island could interact with several transcription factors (TF) such as CTCF, EZH2, GLIS2, RAD21, ZFP37, ZBT family (ZBTB20, ZBTB26, ZBTB17, ZBTB11) and ZNF family (ZNF777, ZNF335) [20]. The specific methylation of the dPCR targeted region could inhibit the transcription of NPY enhanced by those TF. Currently, only in vitro data are available and the role of NPY in CRCs is still to be defined. By the way, Alshalalfa et al have shown, that in the case of prostate cancer, the decrease of NPY appears to be associated with aggressive phenotype and with a high risk of developing metastasis [21].
Amlal et al. showed that estrogen up-regulates NPY receptor (Y1R) expression through estrogen receptor alpha [22] in breast cancer cell lines. Estrogen plays an important role in the up-regulation of Y1R, which in turn regulates estrogen-induced cell proliferation in breast cancer cells. In another model, estrogen significantly decreased NPY secretion in both the mHypoE-42 and mHypoA-2/12 neurons [23]. These findings indicate that the central anorexigenic action of estrogen occurs at least partially through hypothalamic NPY-synthesizing neurons. Estrogen actions on NPY receptor might affect NPY signaling according to genders. In our study, no signicant differences were observed between female and male patients concerning methylation of NPY (p = 0.055 for non-tumor tissues and p = 0.13 for tumor tissues) (Supplementary data 3A). These observations were confirmed by TCGA-databases analyzes (p = 0.89 for nontumor tissues and p = 0.69 for tumor tissues) (Supplementary data 3B). We can suppose that gender does not affect the methylation of NPY in CRC carcinogenesis.
Thus, the hypermethylated promoters of NPY and WIF1 are specific early markers of colorectal cancers but their roles in CRCS carcinogenesis are not clearly established.

Limitations of the study
The size of our cohort is sufficient to demonstrate the efficiency of our technique for the detection of NPY and WIF1 methylation status. However, it is difficult to make subgroup comparisons. Nevertheless, our study confirms previous studies results suggesting that methylation of one or both genes seems to be a relevant biomarker to detect the presence of ctDNA in plasma liquid biopsies. Our study is robust and highlights an original and powerful technique in the detection of specific methylation profile of CRC. Roperch and al [13]. tested 161 sera from patients with normal colonoscopy using Methylation Specific PCR. They showed a specificity of 80 and 95% for NPY and WIF1 respectively. Garrigou et al. analyzed 46 plasmas from non-cancer patients with their dPCR technique. Only 3 patients had a higher than the LOB droplet count for the NPY gene, i.e. a specificity of 93%. The specificity of the WIF1 gene was 100% [14].
Many exogenous factors known to modulate DNA methylation were not included in our study. Indeed, many links between lifestyle and epigenetic modifications have been shown [28]. For example, tobacco [28] or alcohol [29] consumption have been shown to modulate DNA methylation. The comparison between consumers and non-consumers could help to understand if by modifying the methylation profiles, the consumption of tobacco or alcohol, could generate false positive results in our technical approach.