SST gene hypermethylation acts as a pan-cancer marker for pancreatic ductal adenocarcinoma and multiple other tumors; toward its use for blood-based diagnosis.

Aberrant DNA methylation is often involved in carcinogenesis. Our initial goal was to identify DNA methylation biomarkers associated with pancreatic cancer. A genomewide methylation study was performed on DNA from pancreatic ductal adenocarcinoma (PDAC) and endocrine pancreas tumors. Validation of DNA methylation patterns and concomitant alterations in expression of gene candidates was performed on patient samples and pancreatic cancer cell lines. Furthermore, validation was done on independent data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). Finally, droplet digital PCR was employed to detect DNA methylation marks in cell‐free (cf) DNA isolated from plasma samples of PDAC patients and cancer‐free blood donors. Hypermethylation of the SST gene (encoding somatostatin) and concomitant downregulation of its expression were discovered in PDAC and endocrine tumor tissues while not being present in chronic pancreatitis (inflamed) tissues and normal pancreas. Fittingly, treatment with a somatostatin agonist (octreotide) reduced cell proliferation and migration of pancreatic cancer cells. Diagnostic performance of SST methylation in a receiver operating characteristic curve analysis was 100% and 89% for tissue and plasma samples, respectively. A large body of TCGA and GEO data confirmed SST hypermethylation and downregulation in PDAC and showed a similar effect in a broad spectrum of other tumor entities. SST promoter methylation represents a sensitive and promising molecular, pan‐cancer biomarker detectable in tumor tissue, and liquid biopsy samples.

Aberrant DNA methylation is often involved in carcinogenesis. Our initial goal was to identify DNA methylation biomarkers associated with pancreatic cancer. A genomewide methylation study was performed on DNA from pancreatic ductal adenocarcinoma (PDAC) and endocrine pancreas tumors. Validation of DNA methylation patterns and concomitant alterations in expression of gene candidates was performed on patient samples and pancreatic cancer cell lines. Furthermore, validation was done on independent data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). Finally, droplet digital PCR was employed to detect DNA methylation marks in cell-free (cf) DNA isolated from plasma samples of PDAC patients and cancer-free blood donors. Hypermethylation of the SST gene (encoding somatostatin) and concomitant downregulation of its expression were discovered in PDAC and endocrine tumor tissues while not being present in chronic pancreatitis (inflamed) tissues and normal pancreas. Fittingly, treatment with a somatostatin agonist (octreotide) reduced cell proliferation and migration of pancreatic cancer cells. Diagnostic performance of SST methylation in a receiver operating characteristic curve analysis was 100% and 89% for tissue and plasma samples, respectively. A large body of TCGA and GEO data confirmed SST hypermethylation and downregulation in PDAC and showed a similar effect in a broad spectrum of other tumor entities. SST promoter methylation represents a sensitive and promising molecular, pan-cancer biomarker detectable in tumor tissue, and liquid biopsy samples. late; one reason for this is the lack of appropriately informative diagnostic tools. In addition, metastasis occurs very early and aggressively in PDAC, leading to a rather dismal overall prognosis (Costello et al., 2012). Although there have been reports on potential biomarkers that might be helpful in diagnosis, prognosis, and prediction of PDAC, basically none has made it into clinical practice in recent years (Dutta et al., 2012;Fong and Winter, 2012;Nolen et al., 2014;Winter et al., 2013). Also epigenetic modifications, in particular DNA methylation, have been reported as biomarkers to detect pancreatic cancer in early stages of the disease (Bydoun et al., 2018;Kisiel et al., 2015;Nones et al., 2014;Yi et al., 2013). Aberrant alterations of DNA methylation patterns usually occur in CpG islands (CGI) and could play a role in the inactivation of tumor suppressor genes, activation of oncogenes, and altering the integrity of the genome (Irizarry et al., 2009). Several genes have been reported to be differentially methylated in PDAC tissue compared to that of normal pancreas (Lee et al., 2019;Nones et al., 2014;Vincent et al., 2011), suggesting utility for diagnostic purposes.
In this study, we performed a genomewide DNA methylation study to identify DNA methylation marks that could be applicable for PDAC diagnosis. Candidates were identified by analyzing clinical tissue samples of PDAC, pancreatic neuroendocrine tumor (NET) for comparison, chronic pancreatitis (CP), noncancerous pancreas tissue, and noncancerous peritumoral tissue surrounding PDAC (N.PDAC). NETs originate from neuroendocrine cells, while PDAC arises from exocrine cells, and are relatively rare with about 1-2% of all pancreatic neoplasms. Their incidence has increased during recent decades (Zhou et al., 2012), however. The peritumoral tissue surrounding PDAC has been described as being more similar at the transcript level to PDAC than to healthy noncancerous tissue (Bauer et al., 2018). This may actually be a reason for the miserable prognosis of PDAC as compared to cystic tumors (CT), whose peritumoral tissues are basically very similar to healthy tissue. We then validated and verified the methylation status using additional clinical samples. Furthermore, additional studies on other tumor tissues were performed using data available at the The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. Finally, we investigated hypermethylation marks in cell-free DNA (cfDNA) in plasma samples from pancreatic cancer patients and relevant controls (for a graphical overview of the work, see Fig. S1).

Tissue samples and patient cohort
Human pancreatic tissue samples were collected during surgery. In all cases, written informed consent had been obtained from the patients. The study was approved by the Ethics Committee Heidelberg, Germany, and performed in compliance with the provisions of the Declaration of Helsinki. Demographic and clinical characteristics of the patient study population are shown in Table S1. The frozen tissues were sectioned with a cryotome; three sections from each tumor sample were used for histopathology. Various clinical parameters of the patients including the age, survival time, metastasis, and tumor stage were taken into account (Bauer et al., 2018).

DNA/RNA isolation
Genomic DNA and total RNA were isolated from fresh-frozen tissue samples with the AllPrep Isolation Kit (Qiagen, Hilden, Germany), following the protocol suggested by the manufacturer. RNA integrity was evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). DNA quantities and qualities were determined using a NanoDrop 1000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA).

In-house gene expression data
Expression profiling data are available from extensive previous studies (Bauer et al., 2018). In summary, there are data on 195 cases of PDAC, 30 cases of peritumoral tissue from next to PDAC (N.PDAC), 24 CT, 22 peritumoral tissues from next to cystic tumors, 59 samples of CP, and 41 healthy pancreatic tissues from noncancer patients (N). Total RNA from individual samples was analyzed on the Sentrix Human-6v3 Whole Genome Expression BeadChips (Sentrix Human WG-6; Illumina, San Diego, CA, USA). Differential expression analysis was performed using the LIMMA package by pairwise comparisons of the groups (Ritchie et al., 2015). The resulting Pvalues were adjusted for multiple testing using Benjamini-Hochberg's false discovery rate (FDR) method; features with a FDR < 0.01 and an absolute log 2 -fold change (log 2 FC) > 0.5 were considered significant.

DNA methylation profiling
Genomewide DNA methylation analysis was performed using the Illumina Infinium 450k DNA methylation platform (Illumina) on 26 PDAC tissues, 24 normal pancreases, 12 CPs, 12 NETs, and 12 peritumoral samples close to PDAC. The analysis procedure followed the manufacturer's standard workflow. The resulting raw data were preprocessed using the RnBeads routine (Assenov et al., 2014) at default settings. Quality control, probe filtering, background correction, batch effect correction, and selection of differentially methylated regions (DMRs) were performed as recommended. DMRs were selected from the list of Infinium probes that passed quality control assessments (FDR-adjusted P ≤ 0.01, absolute methylation difference ≥ 0.15). The data are available at Table S2.

Pathway analysis and candidate selection
The top genes corresponding to the significantly hypermethylated DMR candidates were analyzed with ingenuity pathway analysis (IPA) software (Qiagen, Redwood City, CA, USA) and ConsensusPathDB (Herwig et al., 2016). For the selection of final candidates, we focused on anticorrelated (hypermethylated at DNA level and downregulated at RNA level) genes in PDAC vs. normal pancreas samples. Of these anticorrelated genes, protein encoding genes were selected for further analyses that are involved in cancer-related pathways.

Gene-specific methylation analysis
For verification and validation of the DMRs obtained from the discovery phase, combined bisulfite restriction analysis (COBRA) was performed. (Xiong and Laird, 1997). In short, 1 µg of DNA was treated with sodium bisulfite (EpiTect Bisulfite Kit; Qiagen, Germany) following the manufacturer's protocol. Amplification was performed using specific primer pairs (Table S3) and HotStarTaq DNA Polymerase (Qiagen) as described (Moskalev et al., 2011). Bisulfite sequencing was performed to confirm the methylation in the CGI region of the SST gene promoter. The sequencing data were analyzed using the BIQANALYZER software (Agilent Technologies, Palo Alto, CA, USA) (Bock et al., 2005) .

Bisulfite pyrosequencing
Quantitative validation of DNA methylation was performed using bisulfite pyrosequencing according to a protocol described previously (Moskalev et al., 2015). The sequences of all primers are listed in Table S3. The software PYROMARK Q24 v.2.0.6 (Qiagen) was used for quantification of CpG methylation percentages. The amplification bias toward unmethylated alleles was corrected using the calibration data derived from a set of EpiTect control DNA samples of 0%, 25%, 50%, 75%, 100% methylation (Qiagen) and cubic polynomial regression as described (Moskalev et al., 2011).

Real-time qPCR
The relative quantification of SST gene expression was performed by real-time qPCR on a Light Cycler 480 instrument (Roche Diagnostics, Mannheim, Germany) using SYBR Green (QuantiFast SYBR Green PCR Kit; Qiagen). Gene expression was analyzed in various pancreas samples; GAPDH was used as an internal reference control.

DNA methylation and expression in pancreatic cell lines
To define the SST methylation and expression in pancreatic cell lines, four tumor cell lines-AspC1, Mia-PaCa2, T3M4, and QGP1-were used. The mycoplasma-free cells were grown in Dulbecco's modified Eagle's medium (Gibco, Paisley, UK) supplemented with 10% FBS and penicillin/streptomycin at 37°C and 5% CO 2 in a humidified incubator. Direct bisulfite sequencing and qPCR were performed to define the pattern of SST promoter methylation and gene expression levels, respectively.

DNA demethylation using 5-aza-2´deoxycytidine
MiaPaCa2 cells were grown and treated with 10 µM 5aza-dC (Sigma-Aldrich, Heidelberg, Germany) for 72 h. The medium was changed every 24 h in order to supply fresh drug to the cells. Treated and untreated control cells were harvested and total RNA was extracted. RT-qPCR was performed to document SST expression levels in both treated and untreated cells.

Validation of results using external datasets
For external validation, different DNA methylation and expression datasets from GEO and TCGA were explored. First, for validation of SST downregulation in PDAC, we applied three microarray expression datasets from GEO (Zhang et al., 2012: accession number GSE62452;Yang et al., 2016: GSE28735;and Pei et al., 2009: GSE16515) containing data on 150 PDACs and 122 normal samples. For validation of SST DNA hypermethylation, we analyzed 184 tumor and 10 normal samples from the TCGA-PAAD project and 167 tumor and 29 normal samples from GEO (Nones et al., 2014: GSE49149). Since the Genotype-Tissue Expression (GTEx) project data showed that SST was highly expressed in different parts of brain and gastrointestinal tissues (Mele et al., 2015), we also investigated the expression and methylation of the SST gene across other gastrointestinal cancer entities in the TCGA database: tumors and normal esophagus (TCGA-ESCA), stomach (TCGA-STAD), colon (TCGA-COAD) and rectum (TCGA-READ) tissues. Subsequently, also data from various other tumor entities were looked at lung adenocarcinoma, breast invasive carcinoma, prostate adenocarcinoma, head neck squamous cell carcinoma, liver hepatocellular carcinoma, bladder carcinoma, and kidney renal clear cell carcinoma. These datasets are available for download at the UCSC Xena database (http://xena.ucsc.edu/).

Analysis of cell proliferation and migration
Cell viability and proliferation upon treatment with different concentrations of octreotide were assessed by resazurin assay (Sigma-Aldrich) on the PDAC cell lines MiaPaCa-2 and AsPC-1. In summary, 4000 cells were seeded in each well of a 96-well plate and grown for 24 h. Then, the cells were treated with 10, 20, and 40 lM of octreotide, respectively. After 96-h drug treatment, the relative fluorescence was measured. Cell migration was assayed using an in vitro wound-healing assay (Ibidi, Gr€ afelfing, Germany) according to the manufacturer's protocol. Briefly, 30 000 MiaPaCa-2 cells were seeded in each compartment of 2-well inserts. After 24 h, the inserts were removed and the cells were cultured with 1% FBS medium in presence or absence of 5 lM octreotide. Cell migration was observed under the microscope after 24-, 48-, and 72-h periods, and images were acquired. The NIH (Bethesda, MD, USA) image processing software IM-AGEJ was used to quantify the cf area from different locations of three replicate experiments each.

Cell-free DNA isolation and bisulfite conversion
Circulating cfDNA was extracted from 1 mL of plasma of PDAC patients and healthy individuals using the QIAamp MinElute ccfDNA Kit (Qiagen). In total, cfDNA was isolated from 30 PDAC samples [15 nonmetastatic (M0) and 15 metastatic (M1)] and 18 control samples from healthy donors. In order to avoid large DNA fragments, which originate from white blood cells and represent the main part of ctDNA contamination (Moss et al., 2018;Mouliere et al., 2018), size selection was performed using Agencourt Ampure XP beads (Beckman Coulter, Krefeld, Germany) according to the protocol recommended by the manufacturer. Bisulfite conversion was carried out on total cfDNA samples (EpiTect bisulfite kit; Qiagen) using the manufacture's protocol for low concentration DNA samples.

MethyLight droplet digital PCR
The droplet digital PCR (ddPCR) method was employed for sensitive detection of aberrant methylation at target CpG sites of SST in cfDNA isolated from PDAC and normal plasma samples. The list of primer and Taqman probe sequences is provided in Table S3. All steps including droplet generation, thermal cycling, and droplet reading were performed according to the manufacturer's protocols (Bio-Rad).
To evaluate the sensitivity of the primer/probe assay for detection of methylated alleles, initial experiments were performed using calibration DNA samples of 100%, 10%, 1%, 0.1%, 0.01%, and 0% methylation that were prepared by mixing control DNA of 100% and 0% methylation (EpiTect PCR Control DNA Set; Qiagen).

Statistical analysis
Statistical analyses were performed using GRAPHPAD PRISM 6 software (GraphPad Software Inc., San Diego, CA, USA). Results are presented as the mean and SD unless mentioned otherwise. Normal distribution of variables was computed using D'Agostino-Pearson normality test. Comparison of a continuous variable in two or more than two groups with normal distribution was performed using parametric test (t-test or ANOVA). If the variable was not normally distributed, a nonparametric test (Mann-Whitney or Kruskal-Wallis) was applied. All P-values were two-sided, and P < 0.05 was considered statistically significant. Survival analysis was performed on expression and methylation data by KM plotter (http://kmplot.com) and MethSurv (https://biit.cs.ut.ee/methsurv/) tools (Modhukur et al., 2018;Nagy et al., 2018). A Cox proportional hazards regression model was used to determine hazard ratios.

Genomewide DNA methylation profiling identified multiple aberrantly methylated loci in PDAC tissue DNA
After performing a genomewide analysis of DNA methylation by assaying more than 450 000 CpG sites across the genome with the 450k Illumina microarray, principal component analysis (PCA) of the data was performed to compare different samples based on the b-values (degree of DNA methylation 0 < b < 1). In the analysis, a distinct cluster is formed by the normal tissue samples (Fig. 1A). The PDAC and CP samples were rather close with respect to methylation, which is different to a clustering according to transcriptional data, in which the CP samples fall in between normal and PDAC samples. This intermediate kind of pattern was actually exhibited by the methylation of the phenotypically normal-looking peritumoral tissues (N.PDAC). The NET samples defined the second most principal component, indicating a particular methylation pattern that is not shared with any of the other sample groups. A Venn diagram shows which DMRs compared to normal tissues differ between or are shared among PDAC, NET, CP, and N.PDAC samples (Fig. 1B). Unsupervised hierarchical clustering was conducted using DMRs that were significantly differentially methylated between PDAC and normal pancreas tissues (Fig. 1C). The location of the CpGs that were differentially methylated in pancreatic tumors compared to normal tissues was annotated according to the UCSC classification of CGI (Dreszer et al., 2012) (Fig. S2).

Selection of SST hypermethylation as candidate for further analysis
The genes associated with the top hypermethylated DMR candidates (622 genes, provided in Table S4) were used for functional enrichment analysis with Ingenuity IPA and ConsensusPathDB-human predicting their relevance to cancer and gastrointestinal disease, besides other diseases and biofunctions (Fig. S3). Data from transcriptional profiling (Bauer et al., 2018) showed that 2341 genes exhibit significant downregulation at the transcript level in PDAC compared to normal pancreas. The overlap between the two lists consists of 70 genes of both a higher degree of methylation and a concomitantly lower level of the respective transcript (Table S5). Based on the functional enrichment analysis, we focused on genes involved in important cancer processes such as proliferation, invasion, migration, cell death, and apoptosis signaling, thereby prioritizing the genes further. Furthermore, we looked for genes encoding proteins with known function in the gastrointestinal tract (Fig. S4).
Among the 70 genes, there were several with functions in either cell proliferation (GPAM, KLB, NUPR1, PLA2G1B, REG1A, VIPR2, and SST), cell migration and invasion (FAM107A and SST), apoptosis or cell death (NUPR1, EEF1D, GPAM and SST) as well as gastrointestinal function (SLC9A4, AMY2A, CEL, CLPS, CTRB1, CTRB2, PNLIP, PNLIPRP2, PRSS1, and SST). However, SST was the only gene that was involved in all categories; therefore, we selected it for further analysis. SST is a highly expressed gene in normal pancreas. It encodes for the hormone somatostatin, which was initially known as a regulator of growth hormone released from the anterior pituitary and has inhibitory effects on all known hormones of the gastrointestinal tract (Modlin et al., 2010). In our analysis, the SST methylation and expression data showed a significant inverse difference in PDAC compared to the levels in normal samples ( Fig. 2A,B). DNA methylation is higher while simultaneously SST expression is lower. Interestingly, in tissues from CP patients, expression of SST was not different compared to that in healthy tissues. To confirm that the SST gene is regulated through DNA methylation, we also looked at the gene's methylation and expression in cell lines of PDAC and endocrine tumor. As expected, SST expression was significantly higher in the endocrine somatostatinoma cell line QGP1 compared to the three PDAC cell lines AspC1, MiaPaCa2, and T3M4 (Fig. 2C). DNA methylation analysis of the respective SST promoter revealed that all CpG sites were completely unmethylated in QGP1, which highly expresses SST, but is methylated in the other cell lines with very low SST expression (Fig. 2C). For one of the PDAC cell line-MiaPaCa2 -we looked further at the change in SST expression upon treatment with 5-aza-dC, a compound that results in DNA demethylation. As expected, the SST expression increased (Fig. 2D) as a consequence of the treatment.

Validation of SST gene hypermethylation and downregulation in PDAC
The SST promoter contains a CGI, which covers the region upstream of the transcription starting site (TSS) and reaches into exon 1 (Fig. 3A). A 464-bp region containing 26 CpG dinucleotides (location À222 to +242 relative to TSS) was studied by COBRA with samples already used on the 450k array. Normal control samples were not cut indicating a lack of DNA methylation; however, there was cleavage in the DNA isolated from tumor samples (Fig. 3B). Quantification by direct bisulfite sequencing of the amplified SST 464-bp region showed that an average 37.2% of CpGs in tumor samples were methylated while only 2.2% methylation was observed in normal pancreas tissues (Fig. 3C). In addition, quantitative methylation results obtained by pyrosequencing were produced to validate further SST hypermethylation. Four CpG sites at positions +12, +23, +30, and +39 of the coding sequence of exon 1 were assessed. The four CpGs were studied in 49 PDAC, 17 NET, and 33 normal pancreas tissues. Of these samples, 26, 12, and 23, respectively, had already been used in the genomewide analysis (technical validation) while 23, 5, and 10 samples, respectively, were additional and independent ones (biological validation). Tumors exhibited a significant hypermethylation (Fig. 3D). Also, real-time qPCR was performed for quantifying SST gene expression in normal and tumor tissue samples. There was a significant decrease in SST expression levels in PDACs and NETs compared to normal tissue (Fig. 3E). A Pearson's correlation analysis of the association of SST methylation and the gene's expression level yielded a significant (P = 0.001) negative correlation (r = À0.6641).
For further confirmation, we also looked at methylation data available from the pancreatic cancer study (PAAD) within TCGA data repository and the GSE49149 dataset available in the GEO. Put together, they contain information about 351 pancreatic tumor and 39 normal tissues. For the analysis, the b-values of the 10 Illumina probes were used that target the SST gene. There was significant hypermethylation in the tumors compared to normal tissues, confirming our findings (Fig. 4A). Moreover, for validation of SST downregulation, we used expression data from other, independent cohorts of pancreatic cancer available in GEO. This dataset contains expression profiles of 150 PDAC and 122 normal tissue samples. Congruent with our results, significant downregulation of the SST gene could be observed in PDAC tumors. Moreover, a Pearson's correlation of the SST methylation profile with expression levels using methylation and expression data of pancreatic cancer in TCGA-PAAD found a significant inverse correlation (r = À0.4161, P < 0.0001).

SST methylation as a pan-cancer biomarker
Somatostatin exhibits inhibitory effects on all known hormones of the gastrointestinal tract (Modlin et al., 2010). In order to learn how specific the variations in SST methylation and expression are to pancreatic cancer, we took advantage of TCGA expression and methylation data for an analysis of other gastrointestinal cancers. The corresponding data for tumors and normal ESCA, TCGA-STAD, colon, and TCGA-READ tissues were studied. As true for PDAC, SST methylation and expression in other gastrointestinal cancers showed highly significant hypermethylation and downregulation of expression (P < 0.0001; Fig. 4B). Subsequently, we extended the investigation to tumor entities beyond the gastrointestinal tract, for which appropriate data were available: TCGA methylation data from breast, lung, prostate, and head and neck cancer as well as liver, bladder and kidney tumors were analyzed. For all cancers, a comparison of tumor and healthy tissue samples revealed significant SST hypermethylation in tumor (Fig. 4C), implying that SST hypermethylation is a general and common feature in a broad range of tumors and could be utilized as a pan-cancer biomarker.

Pharmacologic agonist of somatostatin reduced cell proliferation and migration
We determined the effect of the somatostatin agonist octreotide on cell proliferation and migration. Treatment with different octreotide concentrations resulted in an inhibitory effect on cell proliferation in the PDAC cell lines MiaPaCa-2 and AsPC-1 compared to untreated controls (Fig. 5A). Afterward, we studied whether octreotide had an effect on PDAC cell migration by assessing the motility of MiaPaCa-2 cells in a wound-healing assay in the presence of 5 lM octreotide. Octreotide treatment significantly suppressed the migration of MiaPaCa-2 cells (Fig. 5B,C). These data confirmed that SST regulation modulates cancer-related pathways involved in cell proliferation and migration.

Disease prognosis and diagnostic accuracy
To evaluate whether there would be any prognostic value of SST promoter methylation, a univariate survival analysis regarding overall survival of PDAC patients was performed on our dataset using the median value as cutoff. The analysis revealed no (D) Effect of treatment with 5-aza-dC, which causes complete DNA demethylation. Cell line MiaPaCa2 was grown in absence (ctrl) or presence of 5-aza-dC. The resulting variation in gene expression is shown as bands after gel electrophoresis of RT-PCR products as well as a quantification of the band intensities. Results are shown as mean values AE SD of the experiments. A two-tailed unpaired Student's t-test was applied. ns: P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. statistically significant association between SST methylation and expression with the survival rate of PDAC patients. However, analyzing public datasets, which contain results from more samples (in total 184), did yield a prognostic value of SST methylation (P = 0.016 for cg02164046 from TCGA-PAAD) and expression (P = 0.008 for TCGA-PAAD) (Fig. 6). The Kaplan-Meier plots of SST methylation and expression exhibit very good congruence, although resulting from different datasets. Taking into account that low For an evaluation of the diagnostic performance of SST methylation, a receiver operating characteristic (ROC) curve analysis was conducted using the average methylation degree of the four CpGs analyzed in the pyrosequencing (validation) experiment. Applying a threshold of 20% methylation, all PDAC tissue samples could be accurately discriminated from control tissue samples based solely on the degree of SST methylation. The area under the curve (AUC) value was 100% (P < 0.0001). The SST gene methylation could also differentiate NETs from healthy pancreas, although with a substantially lower AUC value of 86.3% (P < 0.001).

Sensitive detection of SST methylation from cfDNA by ddPCR
For clinically applicable diagnosis, it would be advantageous, if a noninvasive process could be implemented. Therefore, we wondered, if tumor-specific methylation may be detectable in circulating free (cf) DNA in the plasma of PDAC patients. ddPCR was applied for detection and quantification; the primer and probe sequences are provided in Table S3. The assay was found to be capable of detecting methylated SST from 100% down to 0.01% (methylated/ unmethylated SST allele) in a bulk of unmethylated DNA. For the actual analysis, cfDNA was isolated from plasma samples. ddPCR detected SST methylation in nearly all cfDNA samples ranging from 0.09 to 4.79 copies per lL. However, the number of copies per lL of SST methylation was significantly higher in PDAC (M1 and M0) than control samples (Fig. 7A). Also, there was a trend of more copies per µL in M1 than M0, although not at a level of statistical significance. With a threshold of 0.825 copies per lL, our SST methylation biomarker assay by ddPCR exhibited a sensitivity of 93.3% and a specificity of 88.9%. ROC analyses of the data from plasma samples yielded an AUC value of 89% (P < 0.0001; Fig. 7B).

Discussion
Using genomewide DNA methylation experiments, we showed that the SST gene is hypermethylated in pancreatic tumors. In addition, a significant inverse correlation between DNA methylation and gene expression was revealed. A most interesting result was the fact the SST hypermethylation and concomitant downregulation occurred in tumor DNA but not in material isolated from CP patients, which showed similar methylation variations but a different response at the transcript level. The combined effect is clearly tumorassociated and not related to inflammation processes.
The differences of PDAC and NET, most apparent in the PCA (Fig. 1), indicate that different CpGs are affected by methylation changes, which overall have the same consequence, however, with respect to downregulation of the SST gene expression.
While earlier data indicated that SST methylation may be a marker for PDAC, our results imply that it does not exhibit specificity for PDAC but could act as a marker useful for detecting many different tumors. SST gene expression may simultaneously discriminate tumor from inflammation (AUC = 0.74). Expression data on normal tissue samples from several projects, such as GTEx (Mele et al., 2015), showed that somatostatin is expressed throughout the body, but at higher levels in the brain, pancreas, and across the various tissues of the gastrointestinal tract. Therefore, we explored the SST gene methylation and expression in other cancer entities available in the TCGA database (Tang et al., 2017). Interestingly, the results showed that SST was significantly downregulated in the  (left  panel) and expression (right panel) of SST. HR stands for hazard ratio; also, the respective P-value is given. Fig. 7. Results from analyzing cfDNA from plasma samples. ddPCR was used to study the abundance of the methylated SST allele in plasma samples of PDAC patients (M0: nonmetastatic; M1: metastatic) and healthy controls. ns: P > 0.05; ***P ≤ 0.001; ****P ≤ 0.0001. gastrointestinal tract malignancies of esophageal, stomach, colon, and rectal cancers. The methylation levels in tumor samples of the digestive system from TCGA were concomitantly significantly higher compared to normal tissue suggesting SST hypermethylation as a general mechanism of somatostatin downregulation in these types of tumors. In agreement with our results, there are reports on SST hypermethylation in gastrointestinal tract malignancies including esophageal (Jin et al., 2008), stomach (Jackson et al., 2011), and colorectal cancer (Mori et al., 2006). It is interesting to note that gastrointestinal tumors exhibit a high expression of SST receptors, a fact that might represent a feedback mechanism following the decreased expression of somatostatin (Reubi et al., 1994;Tang et al., 1998;Zhao et al., 2014).
To learn whether SST hypermethylation is specific to gastrointestinal tumors, we explored TCGA data of five additional major tumor entities beyond the gastrointestinal tract: lung, breast, prostate, head and neck, as well as liver cancer. In all the eventually 10 tumor types looked at, there were strong SST hypermethylation and downregulation of SST expression. This inverse correlation is not necessarily expected; transcription factors have been found that activate gene expression despite or rather because of a hypermethylated promoter (Wan et al., 2017). Obviously, SST hypermethylation is a general and common process in a broad range of tumors, suggesting that somatostatin may function as a tumor suppressor.
The somatostatin receptors belong to the G proteincoupled receptor superfamily, which recruits several downstream transduction signals such as adenylyl cyclase and calcium channels upon somatostatin binding. Through binding to high-affinity G protein-coupled somatostatin receptors (SSTR1 to 5) on the cell membrane, somatostatin plays different roles. It is a negative regulator of cell proliferation and migration, has cytostatic effects on tumor cells, and induces apoptosis through different pathways (Modlin et al., 2010). Furthermore, somatostatin may also have antiangiogenic properties by inhibiting the production and release of angiogenic factors (Bocci et al., 2007). Synthetic analogs of somatostatin have made possible advances in the treatment of NETs (Modlin et al., 2010). Somatostatin and analogs thereof bind to different receptors and arrest cell growth using direct and indirect mechanisms (Susini and Buscail, 2006). It has also been reported that somatostatin analogs reduce hormone-related morbidity and increase the time to progression in NET of the small intestine, which is the most common small intestinal malignancy (Stalberg et al., 2016). Somatostatin and dopamine have some similar structural and functional characteristics, and heterodimerization of SSTRs with dopamine receptors create a novel receptor with enhanced functional activity (Rocheville et al., 2000). Interestingly, in a previous study, deregulation of dopamine receptor D2 (DRD2) was shown in PDAC tissues compared with nontumor tissues and DRD2 inhibition strongly affected PDAC metastasis (Jandaghi et al., 2016). We also found that the somatostatin agonist octreotide has antiproliferative and antimigratory effects on PDAC cell lines, providing more evidence on its tumor suppressive roles in pancreatic cancer. In agreement with our results, a study showed that PDAC cells upregulate SST expression following demethylation by 5-aza-dC treatment (Gailhouste et al., 2018), and the somatostatin analog octreotide decreases the PDAC cell growth, suggesting a possible therapeutic application of somatostatin analogs for combination therapy of PDAC.
Aberrant SST hypermethylation and expression was associated with the survival rate of PDAC patients based on TCGA data, implying its prognostic value for stratifying patients into high-risk and low-risk groups (HR = 1.67 and HR = 0.57, respectively). The result showed a low-risk group of patients with higher SST mRNA levels or lower SST methylation score, had significantly longer survival time. So far, carbohydrate antigen (CA) 19-9 remains the only approved biomarker also for response assessment but is limited by low sensitivity and specificity (Chang and Kundranda, 2017); therefore, further investigations are needed for the discovery of new markers for pancreatic cancer outcome prediction and management. Our result is in agreement with a previous study, which showed that PDAC patients survival was associated with hypermethylation of several individual genes including SST (HR = 1.63, P = 0.03) (Henriksen et al., 2016). In this regard, Bydoun et al. (2018) reported a S100A10 mRNA and methylation status as predictive of pancreatic cancer survival where hypermethylation and lower expression were correlated with better survival. In another study, DNA methylation and altered expression of MET and ITGA2 was associated with patient survival and patients with coordinated hypomethylation and high expression of MET and ITGA2 strongly correlated with poor outcome (Pvalue = 0.007 and 0.040, respectively) (Nones et al., 2014).
With respect to diagnostics, the AUC value of SST methylation in tissue and plasma samples were 100% and 89%, respectively, demonstrating its high potential as a biomarker for cancer diagnosis. In comparison with the only available and approved serum biomarker for PDAC, CA 19-9, our result shows a higher diagnostic sensitivity and specificity with about 93% vs. 80% and 89% vs. 86%, respectively (Hasan et al., 2019;Zhang et al., 2015). In addition, 10% of Caucasians lack the ability to produce CA-19-9. Although the difference between the PDAC M0 and M1 groups is not statistically significant (P-value = 0.08), the mean level of methylated SST allele was higher in M1 compared to M0 (2.409 vs. 1.577 copies per µL). The increase supports the notion of the relevance of methylation for tumor progression.
The result of the gene's methylation status in plasmaderived cfDNA is in agreement with a previous study by Henriksen et al. (2017) which showed DNA methylation in cfDNA to be diagnostically relevant for pancreatic adenocarcinoma. The overall diagnostic performance of the selected 28-gene promoter panel was demonstrated by an AUC value of 0.86. In a study by Kisiel et al. on tumor tissue and pancreatic juice using several DNA methylation biomarkers, CD1D was reported as the best marker that highly discriminated pancreatic cases from controls. Detection of CD1D methylation in pancreatic juice provided a better AUC compared to our blood biomarker (92% vs. 89%); however, collecting pancreatic juice is a more invasive procedure compared to blood collection. Another study identified BNC1 and ADAMTS1 hypermethylation as potential biomarker to detect early-stage pancreatic cancer, with sensitivities of 79% and 48% and specificities of 89% and 92%, respectively. In our study, SST hypermethylation shows a higher sensitivity (93%) and similar specificity (89%).
From our tissue data, one would expect that SST methylation will not be usable as a marker of pancreatic cancer in particular. Instead, it is likely that SST hypermethylation has the potential as a blood-based pan-cancer biomarker for a broad range of tumors for initial screening in order to stratify individuals into high-risk and low-risk groups. The high-risk group could then be further evaluated by other means, such as CT imaging and/or endoscopic screening. In combination with existing biomarkers such as CA.19-9 or KRAS mutations, the SST methylation should improve the diagnostic accuracy. Apart from its potential for noninvasive diagnosis, SST hypermethylation and its concomitant downregulation may lead to increased cell proliferation and migration. This supports a possible therapeutic application of somatostatin analogs in combination therapies.

Conclusion
Our results show that SST gene hypermethylation and simultaneous downregulation of expression occurs in pancreatic cancer and a broad range of other tumor entities, acting as a pan-cancer molecular biomarker. Furthermore, SST hypermethylation and expression have prognostic value and are associated with the survival rate of PDAC patients. In addition, the methylation changes could also be found in liquid biopsy samples of PDAC patients, holding promise for a noninvasive diagnostic use.
FH and EAM designed and executed the pyrosequencing experiments. TH, MM, ASB, and JDH interpreted the results and wrote the manuscript. All authors read and revised the manuscript critically and approved the final version.

Ethics approval
The study was approved by the local ethics committee at Heidelberg University.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Graphical overview of the analysis processes and sample numbers used in the study. Fig. S2. Pie charts are illustrating the proportions of probes with significant methylation differences in PDAC vs. normal tissue according to the UCSC classification of functional regions. Fig. S3. Result of a functional enrichment analysis of genes associated with genomic regions that were found to be significantly hypermethylated. Fig. S4. Schematic flow chart on the gene selection process. Table S1. Demographic and clinical characteristics of the patients in this study. Table S3. Sequences of the primers used in this study. Table S4. Hypermethylated genes. Table S5. Hypermethylated and down-regulated genes. Table S2. List of differentially methylated regions (DMRs).