The Toledo cell line (ATCC-CRL-2631) was purchased from American Type Culture Collection. NU-DUL-1 cells (ACC 579) were purchased from GSMZ-German Collection of Microorganisms and Cell Cultures GmbH. The GNE-587170 cell line was supplied by Dr Mark Minden's laboratory (Ontario Cancer Institute, Toronto, Canada). Non-Hodgkin lymphoma-derived cell lines were cultured in RPMI-1640 medium (Euroclone SpA) with penicillin-streptomycin at 10 U/ml (Euroclone SpA) and 10% FBS (Gibco; Thermo Fisher Scientific, Inc.). Cells were routinely tested for mycoplasma contamination using the MycoAlert™ Mycoplasma Detection Kit (Lonza Group, Ltd.). Cell lines were authenticated by PCR-single-locus-technology (Eurofins Scientific).

Hypomethylating drugs were used at a concentration ranging between 0.05–10 µM, which is specified in each figure. The drugs were dissolved in water and the control is represented by untreated cells. Cells were incubated in a humidified incubator at 37°C with 5% CO₂ for the indicated time each experiment.

Viability assays were performed by using a Cell Counting Kit-8 (CCK-8; Boster Biological Technology). Briefly, 20,000 cells were seeded in a 96-well plate and, after 48 h of incubation, 10 µl/well tetrazolium salt was added. Spectrophotometric reading at 450 nm with a 620 nm filter was performed after 1 h of incubation.

Time-course apoptosis assays following either 5AZA or decitabine treatment were performed using an IncuCyte^® S3 (Essen Bioscience). Cells were seeded at a density of 20,000 cells/well. The reagents used to label the cells were IncuCyte^® Nuclight Rapid Red and Cell Health Reagent Caspase-3/7 (both used at a 1:2,000 final dilution in the wells; Essen Bioscience). Cells were incubated in a humidified incubator at 37°C with 5% CO₂. Each condition was performed in triplicate.

The cytologic specimen was prepared using Cytospin™ 4 (Thermo Fisher Scientific, Inc.) and the glass was fixed with 95% ethanol for 20 min at room temperature. Immunocytochemistry was performed with Monoclonal Mouse Anti-Human Ki-67 Antigen Clone MIB-1 (1:100; cat. no. M724001-2; Dako; Agilent Technologies, Inc.) for 40 min at room temperature. The slides were pre-treated with heat antigen retrieval in EDTA buffer for 64 min. The immunocytochemical staining was developed on Ventana BenchMark ULTRA platform using the pre-diluted UltraView DAB Detection Kit (cat. no. 05269806001; Ventana Medical Systems, Inc.; Roche Diagnostics) for 8 min at 36°C. Slides were counterstained with hematoxylin for 3 min at room temperature. Slides were observed on a Nikon Eclipse 80i light microscope (magnification, ×40; Nikon Corporation).

Quantitative gene-specific methylation was performed via EpiTect Methyl II qPCR (Qiagen S.r.l.) as previously described (4).

Global chromatin methylation was performed using the MethylFlash Methylated DNA Quantification kit (EpiGentek Group, Inc.) with 100 ng DNA/sample, according to the manufacturer's instructions. Methylation-specific PCR (MS-PCR) was performed with 50 ng bisulfite-converted genomic DNA. Primers and PCR conditions are presented in Tables SI and SII, respectively. Bisulfite conversion and cleanup was performed with the EpiTect Bisulfite kit (Qiagen S.r.l.). Genome-wide methylation analysis was performed on an Infinium Human Methylation 450K BeadChip array (Illumina, Inc.). DNA samples underwent bisulfite treatment with the EZ DNA methylation kit (Zymo Research Corp.) and were then hybridized on the Infinium 450K BeadChip arrays, following the manufacturer's protocols. Signal intensities and β-values were exported from Beadstudio 2.0 software (Illumina, Inc.), applying default settings. All bioinformatics processing was performed with R version 3.2.0 (https://cran.r-project.org/bin/windows/base/old/3.2.0/). Raw intensity signals were imported and processed using the minfi package (https://www.bioconductor.org/help/course-materials/2015/BioC2015/methylation450k.html). All samples had an average detection P<0.001, indicating good quality data. For quality control (QC), histograms and boxplots were plotted for signal intensities and β-values. All samples were quantile normalized, β-values and M-values were also calculated. The probes containing SNPs were removed as they can have important consequences on the downstream analysis. For clustering and statistical tests, probes mapped on sex chromosomes were excluded. Unsupervised analysis was performed on all the retained probes using hierarchical clustering with Euclidean distance and Ward linkage on the β-values. Supervised analysis of differential methylation between groups was performed using Limma package (https://bioconductor.org/packages/release/bioc/html/limma.html). The Benjamini-Hochberg multiple test correction was applied to control for false positives. Probes showing an unadjusted FDR <0.05 were considered differentially methylated.

Total RNA was extracted using TRIzol (cat. no. 15596026; Invitrogen; Thermo Fisher Scientific, Inc.) with the RNA isolation protocol (W.M. Keck Foundation Biotechnology microarray Research Laboratory at Yale University) and RT was performed with the iScript™ cDNA Synthesis kit (Bio-Rad Laboratories, Inc.) according to the manufacturer's protocol. Gene expression was performed via qPCR with the SsoFast™ EvaGreen^® Supermix (Bio-Rad Laboratories, Inc.) following the manufacturer's protocols. The Cq value of 37 cycles was chosen as the lower threshold (cutoff) for gene expression using the 2^−∆∆Cq method of quantification (18). Ribosomal protein S7 was used as the reference gene in RT-qPCR. Primers and amplification conditions are reported in Tables SI and SII, respectively.

Immunoblotting was performed as previously described (19). Briefly, 2×10⁶ cells were lysed in RIPA buffer (150 mM NaCl, 5 mM EDTA pH 8.0, 50 mM Tris pH 8.0, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS in water). Total lysates were quantified by spectrophotometric reading at 750 nm [FLUOstar^® Omega (BMG Labtech GmbH)] with the Lowry method (DC™ Protein Assay Kit; Bio-Rad Laboratories, Inc.) and 20 µg protein/sample were separated on Bolt Bis-Tris Plus precast 4–12% polyacrylamide gels (Thermo Fisher Scientific, Inc.) and blotted onto nitrocellulose membranes. Blocking was performed with 4% milk in PBS with 0.1% Tween-20 for 30 min at room temperature. Rabbit primary antibodies against DNMT1 (cat. no. 5032), DNMT3a (clone D23G1; cat. no. 3598), SPG20 (cat. no. 13520S), KLF4 (cat. no. 12173) and poly (ADP-ribose) polymerase (PARP; cat. no. 9532) were purchased from Cell Signaling Technology, Inc. Anti-DAPK1 (cat. no. MA1-24696) was purchased from Thermo Fisher Scientific, Inc. All antibodies were used at a final dilution of 1:2,000. Mouse anti-β-actin antibody (clone AC-15; cat. no. A1978) was purchased from Sigma-Aldrich (Merck KGaA) and used at a final dilution of 1:20,000. The incubation with the primary antibodies was performed overnight at 4°C. The membrane was then incubated for 1 h at room temperature with anti-rabbit (cat. no. NA934V; Amersham; Cytiva) and anti-mouse (cat. no. NXA931V; Amersham; Cytiva) secondary antibodies at dilutions of 1:10,000. For visualization ECL detection reagent was used (Amersham; Cytiva) and the chemiluminescence signal was acquired through ChemiDoc MP (Bio-Rad Laboratories, Inc.) and analyzed by ImageLab version 5.2 (Bio-Rad Laboratories, Inc.).

Gene silencing was performed through 4D-Nucleofector™ Core Unit (Lonza Group, Ltd.). The buffer used was the Amaxa™ SF cell line 4D Nucleofector™ X kit (cat. no. V4XC-2024; Lonza Group, Ltd.). Small interfering (si)RNAs were purchased from Ambion (Thermo Fisher Scientific, Inc.), diluted in molecular biology grade water and used at a final concentration of 100 nM. The cells were either transfected with siRNAs against DNMT1 (cat. nos. 4390824, s4216 and s4217; Ambion; Thermo Fisher Scientific, Inc.), DNMT3a (cat. nos. 4392420 and s200426; Ambion; Thermo Fisher Scientific, Inc.) or both. Negative control #2 siRNA (cat. no. 4390846) was used as the control (scramble). Sequences are listed in Table SI. Cells were incubated in a humidified incubator at 37°C with 5% CO₂ for 48 h before being collected.

PBMCs were purified from the peripheral blood using Histopaque^®−1077 (Sigma-Aldrich; Merck KGaA) according to the manufacturer's instructions. Blood samples from healthy donors were collected at the Azienda Unità Sanitaria Locale - IRCCS di Reggio Emilia (Reggio Emilia, Italy), between October 2017 and June 2018. The age range was 19–60 years (five female patients and 22 male patients). The use of PBMCs upon written informed consent was previously approved by the Local Ethics Committee of Reggio Emilia (Comitato Etico Provinciale; approval no. 2016/0025053).

IncuCyte^® S3 Software 2020B (Essen Bioscience) was used to analyze and graph the time-course fluorescence data. The probes specific to SPG20 locus were sorted from the Infinium Human Methylation 450K BeadChip array (Illumina, Inc.) and the HumanMethylation 27 BeadChip (Illumina, Inc., available through the repository ‘ArrayExpress’ https://www.ebi.ac.uk/arrayexpress/; accession no. E-MTAB-417), and subsequently mapped onto NCBI's genome browser for humans (https://www.ncbi.nlm.nih.gov/genome/gdv/; GRCh38.p13; accession no. GCA_000001405.28). Heatmaps of the β-values were generated through Displayr (https://www.displayr.com/).

Data analysis was performed with GraphPad Prism 5.0 (GraphPad Software, Inc.). One-way ANOVA followed by Bonferroni's post hoc test (multiple comparisons) were performed. Data are represented as the mean ± standard error of the mean. Figures were prepared with Adobe Illustrator CS4 (Adobe Systems, Inc.). P<0.05 was considered to indicate a statistically significant difference.

It was observed that lymphoma cell lines displayed a characteristic profile of promoter methylation on specific targets, including tumor suppressors. A panel of nine target promoters were previously selected and assessed on a series of primary B lymphocytes sorted from healthy and tumor lymph nodes (4). Among the tested target promoters the three targets that were found to be highly methylated in the tumor samples were chosen. The promoters of the KLF4, DAPK1 and SPG20 genes were fully methylated in Toledo and NU-DUL-1 cell lines (germinal center- and activated B cell like derived, respectively; Fig. 1A). This was consistent with the tumor and metastasis suppressor role reported for KLF4 and DAPK1 (20,21). The positions of the CpG islands investigated by EpiTect Methyl II qPCR are reported in Table SIII.

The epigenetic regulation of these targets was initially evaluated through the use of hypomethylating agents. Apoptosis was triggered by 5AZA and decitabine, as demonstrated by caspase-3/7 activation after 24 h (Fig. S1).

DNMT1 was downregulated and PARP was cleaved upon treatment with either 5AZA or decitabine (Fig. 1B and C), which was consistent with the reported mechanism of action for these nucleoside analogs (13). PARP was partially activated by 5AZA and completely activated by decitabine in NU-DUL-1 cells, while Toledo cells only presented a partial cleavage at the highest dose of decitabine. After one week from the day of administration of the hypomethylating drug, DNMT1 was still downregulated and PARP was notably cleaved with both drugs (Fig. 1D and E). Doses of decitabine >1 µM for one week were lethal and therefore data were not analyzable (data not shown).

To assess the specific activity of hypomethylating agents in this study, global chromatin methylation was measured via a fluorescence immunoenzymatic assay with anti-5mC antibody. As expected, chromatin was demethylated upon treatment with 5AZA in both cell lines after 48 h and up to 7 days later (Fig. 2A), which was consistent with the observed DNMT1 downregulation and caspase-3/7 activation (Figs. 1B-E and S1). Next, it was investigated whether the previously selected KLF4, DAPK1 and SPG20 promoters could be modulated by these treatments. These were not affected by 5AZA treatment at any of the tested time points (Fig. 2B, left panel).

The effect of decitabine treatment was also evaluated. Global chromatin methylation decreased after 48 h in the Toledo and NU-DUL-1 cell lines compared with the NT groups, but the levels of treated samples returned to the same levels observed in the NT groups after 7 days (Fig. 2A, right panel), which differed from the results observed for 5AZA. Quantitative methylation of the selected targets demonstrated that KLF4, DAPK1 and SPG20 promoters were unaffected (Fig. 2B, right panel).

The gene expression of the three genes of interest was assessed by RT-qPCR. Only KLF4 was found to be partially upregulated by 5AZA treatment in both cell lines, while DAPK1 and SPG20 were still undetectable upon both 5AZA and decitabine treatments (Fig. 2C; mean Cq of 37 was chosen as a cutoff). The immunoblotting confirmed these observations since these three targets were almost undetectable in all the conditions tested. The cell line GNE-587170 was reported as a positive control for KLF4 and DAPK1 proteins (Fig. 2D).

These data collectively suggested that 5AZA and decitabine selectively affected only some regions of the genome, which was consistent with previous findings on colon cancer cells (14). However, unlike in these previous findings, the drugs did not preferentially affect the highly methylated CpG dinucleotides in our targets, which remained unaffected.

To investigate the role of specific DNMTs in the maintenance of methylation at the level of the selected targets, DNMT1 and DNMT3a were individually silenced and promoter methylation was assessed.

DNMT1 and DNMT3a were consistently and specifically downregulated at the RNA and protein levels 48 h after transfection with siDNMT1 and siDNMT3a compared with the scramble siRNA group (Fig. 3A and B). DNMT3a RNA levels were low in both cell lines (Table SIV) and the corresponding protein was undetectable by immunoblotting. This was consistent with the known role played by this enzyme in embryonic development (22).

Gene expression analysis showed that none of the KLF4, DAPK1 and SPG20 targets were affected by any of the silencing conditions (Fig. 3C). This was consistent with the finding that the percentage of methylation was not affected after DNMT1 silencing, nor after simultaneous DNMT1- and DNMT3a-silencing (Fig. 3D-F).

To assess whether proliferation was affected by the knockdown of DNMT, Ki67 expression was determined by immunocytochemistry. The majority of the nuclei were Ki67-positive in all the conditions tested after 48 h of incubation (Fig. 3G). In addition to this, viability measured by CCK-8 was not affected after either DNMT1- or DNMT3a-silencing in the two cell lines (Fig. 3H). Therefore, the methylation stability observed in the different conditions was the result of an active process maintaining methylation marks during DNA replication even when DNMTs were deficient.

It was demonstrated that genomic DNA was strongly demethylated by 5AZA and decitabine (Fig. 2A). DNMT1-specific silencing did not lead to hypomethylation at the promoter-specific levels of KLF4, DAPK1 and SPG20 (Fig. 3D-F). Thus, these data collectively demonstrated that DNMT1 downregulation alone is not responsible for chromatin demethylation, even when lymphocytes are actively proliferating.

The data collected thus far demonstrated that the SPG20 promoter was hypermethylated in all the tested cell lines by qPCR and was not affected by hypomethylating drugs nor by the silencing of DNMT1. At present, this locus is poorly characterized in B cell lymphomas, and therefore deserves further investigation.

To investigate the whole SPG20 locus, the data from an Infinium Human Methylation 450K BeadChip array was analyzed. The β-values for 28 DLBCL cell lines were filtered for the SPG20 locus and the results were assembled in a heatmap (Fig. 4A). The β-value distribution showed a hypomethylated region (light blue, bottom) and a hypermethylated region (dark blue, top) in all the tested cell lines. A heterogeneous region (mid-upper) is common to 24 out of 28 cell lines. These three regions were quite clearly clustered and corresponded to exact positions in the SPG20 locus. Looking at the precise mapping of the probes spanning the gene locus, the highest and more homogeneous β-values were clustered in the open sea regions (away from the islands), upstream of the SPG20 transcription starting site (TSS; Fig. 4B). The mean β-values for the nine probes located upstream the TSS were in the 0.60–0.85 range, indicating a hypermethylated status (Fig. 4D). These data suggested a transcriptional silencing, mediated by the hypermethylation of the promoter region.

To confirm the methylation status of this specific open sea region, primers were designed for MS-PCR on the probe cg15754752, located upstream of the TSS (Fig. 4B and D). The selected region was methylated regardless of the silencing of DNMT1, thus confirming the results of the Infinium Human Methylation 450K BeadChip array (Fig. 4C).

It is noteworthy to underline that the data obtained with an older type of array (Illumina 27K) in a colon cancer cell line and available through the repository ‘ArrayExpress’ confirm our findings (14). The probes specific for SPG20 displayed high β-values even after treatment with 5AZA or decitabine (Fig. S2).

Once it had been demonstrated that the SPG20 promoter was fully methylated in the tested DLBCL cell lines (Fig. 4), it was next assessed whether PBMCs from healthy donors displayed a different methylation profile. PBMCs from healthy donors displayed the same methylation pattern as the one shown by non-tumor B lymphocytes immunologically sorted from follicular hyperplasia (Fig. 5A) (4). KLF4, DAPK1 and SPG20 were almost completely unmethylated in the tested PBMCs (Fig. 5A and B), which differed from the results observed in the NHL cell lines (Figs. 2B and 3D-F). This led us to assess Spartin protein levels in healthy, circulating mononuclear cells. Spartin was expressed in all the tested PBMCs (representative of a total of 27 samples; Fig. 5C), demonstrating the epigenetic regulation of this gene through methylation of the promoter. DNMT1 expression levels were highly heterogeneous among different subjects, which suggested that SPG20 methylation status did not rely on this DNMT (Fig. 5C). These data were also consistent with the fact that DNMT1 silencing did not affect SPG20 methylation in Toledo and NU-DUL-1 cell lines (Figs. 3 and 4).