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

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.

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.

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 system™ (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 PerfeCTa 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.

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.

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/non-tumor, 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.

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 (1-α) 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 Qiagen®, 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 (R² = 0.9715) and WIF1 (R² = 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.

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 (R² = 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.

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.

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 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 non-tumor 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.

Concerning the WIF1, its repression leads to an overexpression of the Wnt signaling pathway thus promoting cell transformation [24]. However, transcriptomic analysis from the Cancer Genome Atlas (TCGA) shows overexpression of WIF1 in CRCs (Fig. 5B). The dPCR target is entirely contained within the 5’UTR of WIF1 but does not overlap the promoter of WIF1 (Fig. 1A). Thus, WIF1 is a tumor suppressor gene whose expression should be rather inhibited in tumor tissues. Therefore, regulatory sequences have to be hypermethylated. We could suggest that sequences analysed by dPCR are not implicated in the regulation of WIF1 expression and could result from systematic methylation. Hypermethylation of dPCR targeted WIF1 region would therefore not lead to repression of the gene. In CISTROME database, WIF1 hypermethylated region is experimentally associated with some proteins such as EP300, EZH2, HDAC2, HIC1, JARID2, KDM2B, MBD2, PRDM11, RYBP, SMAD2, SPDEF, SRF, SUZ12, ZMYND8, ZNF180, ZNF189 and ZNF483 [20]. In silico analysis shows that CpG region targeted by dPCR could interact with several TF such as CREB1, CTCF, EGR2, EZH2, GLIS2, HIC1, KDM1A, KLF9, KLF16, PATZ1, POLR2A, TBP, ZBTB family (ZBTB8A, ZBTB17, ZBTB26, ZBTB33, ZBTB48), ZEB2, ZFHX2, ZFP69B, ZFP37ZSCAN21 and ZNF family (ZNF398, ZNF335, ZNF341, ZNF501, ZNF513, ZNF600, ZNF692, ZNF777, ZNF792) [20]. Feng et al demonstrates that miR-590-3p regulates colon cancer progression via WIF1 which suggests that miR-590-3p may be a promising candidate for therapeutic applications in colon cancer treatment [25]. In nasopharyngeal cancers and gastric carcinoma cell lines the promoters of WIF1 is hypermethylated, and his expression is regulated by miR-BART19-3p [26]. Also the miR-552-5p promoted osteosarcoma development and progression by inhibiting WIF1 [27]. These studies show that WIF1 could be highly regulated by post-transcriptional factors. These data only provide information on the level of mRNA expression but not the protein functionality.

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.

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.