The human pancreatic cancer cell lines PaTu-S, Colo357, PaTu-T, PANC-1 and MIAPaCa-2 were cultivated in Dulbecco's Modified Eagle Medium (DMEM; cat. no. D5796), supplemented with 10% FBS (cat. no. F9665) and 1% penicillin/streptomycin (PS; cat. no. P0781; all Sigma-Aldrich; Merck KGaA). The cell lines Capan-1 and Capan-2 were cultured in RPMI-1640 medium (cat. no. 21875091; Thermo Fisher Scientific, Inc.) with 20% FBS and 1% PS. All cell lines were obtained from the cell bank of the Department of Medicine II at LMU University Hospital and verified using STR profiling. All cell lines were incubated at 37°C in a humidified chamber with saturated atmosphere containing 5% CO₂. Cell lines were passaged at a confluency of 80-90% with 1X Trypsin/EDTA (cat. no. T3924; Sigma-Aldrich; Merck KGaA) in Dulbecco's PBS. Cell lines were mycoplasma-negative, as tested by PCR Mycoplasma Detection Kit (cat. no. ABM-G238, Biozol). Light microscopy images (magnification, ×100) were taken under normal cell culture conditions.

Cell proliferation and viability was measured with colorimetric conversion of MTT (cat. no. M2128; Sigma-Aldrich; Merck KGaA). To determine cell proliferation PaTu-S, Colo357, PaTu-T, PANC-1 and MIAPaCa-2 cells were seeded at a density of 2,000 cells/well and Capan-1 and Capan-2 cells were seeded at a density of 5,000 cells/well in 96-well plates. At 0, 24, 48, 72 and 96 h, 20 μl MTT (2.5 mg/ml) was added to the cell culture media and incubated for 3 h at 37°C with 5% CO₂. Media was aspirated and replaced with 100 μl DMSO. Following 30 min incubation, optical density was measured at 570/690 nm using a microplate reader. Gemcitabine cytotoxicity was determined by seeding 8,000-10,000 cells/well in 96-well plates. After 24 h, cell lines were treated with 1×10⁻¹⁰–1×10⁻⁵ M gemcitabine (cat. no. G6423; Sigma-Aldrich; Merck KGaA) for 72 h before adding MTT.

A selective p300/CREB-binding protein (CBP) HAT domain inhibitor A485 (cat. no. 6387; Tocris Bioscience) and a pan-HDACi, SAHA (cat. no. SML0061; Sigma-Aldrich; Merck KGaA) inhibiting class I and class II HDACs were used for short-(24 or 96 h) and long-term treatment (4-30 weeks) of PDAC cell lines, both dissolved in DMSO. To determine the half-maximal inhibitory concentration (IC50) for A485 and SAHA, 6,000 cells/well were seeded in 96-well plates. After 24 h, cell lines were treated with A485 (1×10⁻¹⁰–1×10⁻⁵ M) or SAHA (1×10⁻⁸–1×10⁻⁵ M) for 72 h before MTT assay. For short-term treatment, all cell lines were treated with 10 nM or 1 μM A485 or 10 nM or 0.5 μM SAHA. For long-term treatment all cell lines were treated with 1 μM A485 or 0.5 μM SAHA, except MIAPaCa-2, which was treated with 0.05 μM SAHA because of cell toxicity. Controls were treated with DMSO. Cell lines for long-term treatment were treated Monday, Wednesday and Friday each week.

A total of 500 PaTu-T and MIAPaCa2 and 2,000 PaTu-S, Capan-1, Capan-2, Colo357 and PANC-1 cells/well were seeded on six-well plates. Capan-2, PaTu-T and MIAPaCa2 cell were cultured for 10 days; PaTu-S, Capan-1, Colo357 and PANC-1 cells were cultured for 14 days as stated above. Cell colonies were washed with PBS and fixed with 100% methanol. Colonies were stained with 1% crystal violet (cat. no. C3886; Sigma-Aldrich; Merck KgaA) in 25% methanol, 10 min each step at room temperature. Pictures were taken with a camera. Colony density was measured with ImageJ (1.53t) software (NIH) and normalized to controls.

Tumor cells were seeded on six-well plates and grown to 100% confluence. PANC-1 cells were treated with 3 μg/ml Mitomycin C (cat. no. 4150.1; Carl Roth); PaTu-S, Capan-2, Colo357, PaTu-T and MIAPaCa-2 cells were treated with 15 μg/ml Mitomycin C for 3 h to inhibit proliferation. Capan-1 was not used due to cell death under low concentration treatment with Mitomycin C. A wound was made with a 200-μl sterile pipette-tip. Cells were washed with PBS and incubated at 37°C with 5% CO₂. Light microscopy pictures (magnification, ×50) of wound closure were captured at 0, 6, 15 and 24 h. Size of the area was measured by using wound healing size tool plugin (16) for ImageJ (1.53t) software and calculated as cell migration in percent over time.

5,000-10,000 tumor cells were seeded on 8-well chamber slides (cat. no. 154534; Thermo Fisher Scientific, Inc.) and staining was performed according to the manufacturer's protocol using 5% donkey serum for blocking (cat. no. D9663¸ Sigma-Aldrich; Merck KgaA) and the following antibodies: Primary epithelial cell adhesion molecule (EpCAM, 1:200; cat. no. 2929; Cell Signaling Technology, Inc.) and vimentin (VIM; 1:200; cat. no. 5741; Cell Signaling Technology, Inc.) and secondary fluorescence-conjugated Cy™3 AffiniPure Donkey Anti-Mouse IgG (H+L; 1:500; cat. no. 715-165-150; Jackson ImmunoResearch) and fluorescein AffiniPure Donkey Anti-Rabbit IgG (H+L; 1:500; cat. no. 711-095-152; Jackson ImmunoResearch). Nuclei were stained using mounting medium with DAPI (cat. no. ab104139; Abcam). Fluorescence images (magnification, ×200) were taken with Axio Imager M2 microscope using ZEN 3.4.91.0 software (Carl Zeiss Microscopy GmbH).

For protein extraction, tumor cells were lysed in 50 mM Tris-HCl (pH, 7.4) containing 2% SDS. Protein concentration was measured with Pierce™ BCA Protein assay kit (cat. no. 23225; Thermo Fisher Scientific, Inc.). A total of 5-20 μg total protein extract/lane (10% gel) was subjected to western blotting according to manufacturer's instructions with the following antibodies: Primary H3K27ac (cat. no. 8173), H3K9ac (cat. no. 9649), HDAC1 (cat. no. 5356), HDAC2 (cat. no. 5113) and β-actin (cat. no. 4967; all 1:1,000; all Cell Signaling Technology, Inc.) and GAPDH (cat. no. H86504M; 1:5,000; Meridian Bioscience) and secondary Amersham ECL Rabbit (cat. no. NA934) and Mouse IgG, HRP-linked whole Ab (cat. no. NA931; both 1:5,000; both GE Healthcare). Bands were visualized in a Fusion Fx Instrument (Vilber) using Clarity or Clarity Max Western ECL Substrate (cat. nos. 1705060 and 1705062, respectively; Bio-Rad Laboratories, Inc.) and quantified by using ImageJ (1.53t) software.

From cell lines, total RNA was extracted using the RNeasy^® Plus Mini kit (cat. no. 74134; Qiagen GmbH) according to manufacturer's protocol. A total of 2 μg RNA was reverse-transcribed to cDNA with the RevertAid First Strand cDNA Synthesis kit (cat. no. K1621; Thermo Fisher Scientific, Inc.). qPCR was carried out using 20 ng cDNA, 500 nM forward and reverse primer and FastStart Essential DNA Green Master (cat. no. 06924204001; Roche Diagnostics). The reaction (35 cycles consisting of 15 sec denaturation at 95°C, primer annealing for 15 sec at 55°C, and extension for 15 sec at 68°C) was performed on a LightCycler^®96 (Roche Diagnostics). Primer pairs (Invitrogen™ Custom DNA Oligos, Thermo Fisher Scientific, Inc.) are listed in Table SI. The gene encoding TATA box binding protein (TBP) was used as housekeeping gene and relative mRNA expression was calculated with ΔΔCq method (17).

RNAseq was performed at the Laboratory for Functional Genome Analysis LMU Munich, Germany. PolyA-selected RNAseq libraries were prepared using Version 2: CORALL RNA-Seq V2 Library Prep Kits and Bundles (Lexogen GmbH) and quality controlled on 2100 Bioanalyzer (Agilent Technologies, Inc.). Illumina Nextseq 1000 (NextSeq 1000/2000 P2 reagent kit v3 (100 cycles, cat no. 20046811, Illumina, Inc., loading concentration 1 nM) was used to obtain 50 bp single-end reads, with a depth of 10-20 million reads/sample. RNA-seq reads were trimmed using trim galore (version 0.6.5, https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). Ribosomal (r)RNA reads and non-eukaryotic RNA contamination were removed using sort- MeRNA (version 4.3.6) (18). Reads were aligned to the human reference genome (GRCh38, Ensembl annotation release 100) and counted using STAR aligner (version 2.7.3a) (19). Count matrix was normalized using DEseq2 package (20) and rlog transformation was performed for principal component analysis. Differentially expressed genes (DEGs) were analyzed using DEseq2 package. To identify subtype marker genes, a cutoff of P<0.05 was set for the comparison of classical and basal-like cell lines. Candidate genes for treatment comparisons were selected based on a cutoff of adjusted P<0.1. Enriched candidate genes were further subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses using clusterProfiler package (21,22). EMT and gemcitabine score were defined based on published genes signatures (23,24).

Cell lines were cross-linked with 1% PBS-buffered formaldehyde for 10 min at room temperature. The reaction was quenched with 125 mM glycine for 5 min and cells were lysed in cell lysis buffer (5 mM PIPES, 85 mM KCl, 0.5% NP-40). Isolated nuclei (centrifuged at 500 × g, 5 min, 4°C) were resuspended in MNase reaction buffer (50 mM Tris-HCl, 5 mM CaCl₂) and DNA was fragmented with 20 U MNase (cat. no. 88216, Thermo Fisher Scientific, Inc.) for 5 min for Capan-1, Capan-2, PaTu-T and MIAPaCa-2, 6 min for PaTu-S and Colo357 and 3 min for PANC-1. The reaction was inactivated by 20 mM EGTA. Nuclei were pelleted (centrifuged at 500 × g, 10 min, 4°C), lysed in nuclei lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl) on ice for 1 h and sonicated for 10 sec for two cycles at 10% amplitude. Samples were centrifuged (12,000 × g, 10 min, 4°C) and supernatant was brought to a total volume of 2 ml using ChIP Dilution buffer (0.01% SDS, 1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, 167 mM NaCl) and pre-cleared with 100 μl Protein A Agarose/Salmon Sperm DNA beads (cat. no. 16-157; Merck Millipore) for 1 h at 4°C. After removing the pre-clearing agarose beads (centrifuged at 4,000 × g, 1 min, 4°C), 1% Input was collected from each sample. Subsequently, samples were equally divided for ChIP antibody reaction with H3K27ac (0.25 μg; cat. no. 8173; Cell Signaling Technology, Inc.) and IgG control (2.5 μg; cat. no. 3900; Cell Signaling Technology, Inc.) and incubated at 4°C overnight. Protein/DNA complexes were precipitated with agarose beads for 1 h at 4°C. After washing the agarose beads with low (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, 150 mM NaCl) and high salt (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, 500 mM NaCl), LiCl (250 mM LiCl, 1% NP-40, 1% sodiumdeoxycholat, 1 mM EDTA, 10 mM Tris-HCl) and twice with TE (10 mM Tris-HCl, 1 mM EDTA) washing buffer (centrifuged at 4,000 × g, 1 min, 4°C), Chromatin was eluted two times in elution buffer (1% SDS, 100 mM NaHCO₃) for 15 min at room temperature. Chromatin was reverse cross-linked in 75 μM NaCl overnight at 65°C and RNA and proteins were digested in 0.05 μg/μl RNase (cat. no. 19101; Qiagen GmbH) for 30 min at 37°C and 0.05 μg/μl proteinase K (cat. no. EO0492; Thermo Fisher Scientific, Inc.) including 10 μM EDTA and 30 μM Tris-HCl for 2 h at 45°C, respectively. DNA was purified with the QIAquick PCR Purification kit (cat. no. 28104; Qiagen GmbH) and quantified by qPCR, as aforementioned. ChIP-specific primers are listed in Table SI. Sample values were calculated according to the % input method [ΔCt=Ct[ChIP]−(Ct[Input]-Input dilution factor), 2^(−ΔCt [normalized ChIP]) ×100).

GraphPad Prism version 9.0 (Dotmatics) was used for statistical analysis using two-tailed, unpaired Student's t test, one-way ANOVA followed by Fisher's LSD post hoc test or non-parametric Mann-Whitney or Kruskal-Wallis test followed by Dunn's post hoc test. P<0.05 was considered to indicate a statistically significant difference. RNAseq data preprocessing and analyses were performed using R 4.0.3 software and visualized in RStudio 1.3.959 (RStudio). Changes in EMT and gemcitabine scores were tested using Wilcoxon test.

To determine the epigenetic reprogramming ability of pancreatic cancer cells of different molecular subtypes, human PDAC cell lines PaTu-S, Capan-1, Capan-2, Colo357, PaTu-T, PANC-1 and MIAPaCa-2 were used. Cell morphology and expression of specific subtype marker proteins were determined by immunofluorescence staining. PaTu-S, Capan-1, Capan-2 and Colo357 cells showed an epithelial phenotype and expressed epithelial marker EPCAM, which indicated a classical subtype. PaTu-T, PANC-1 and MIAPaCa-2 cells exhibited mesenchymal, spindle-shape morphology and a positive staining for the mesenchymal marker VIM, which is associated with a basal-like subtype (Fig. 1A) (25). Whole-transcriptome analysis was performed by RNAseq and the cell lines clustered into two main groups in the principal component analysis (Fig. 1B). Differential gene expression analysis demonstrated expression of classical subtype genes, such as EPCAM, ELF3 and KLF5 (KLF transcription factor 5), in PaTu-S, Capan-1, Capan-2 and Colo357 cells, whereas PaTu-T, PANC-1 and MIAPaCa-2 cells showed high expression of VIM, ZEB1 and JUN, which are markers of a basal-like gene expression network (Figs. 1C and S1A) (25). Classical cell lines were associated with lower proliferation rates compared with the basal-like cell lines, except for Colo357 (Fig. 1D).

Next, it was determined whether the expression of selected subtype marker genes for classical and basal-like PDAC cell lines was regulated by the activating histone modification H3K27ac. ChIP-qPCR demonstrated high levels of H3K27ac at the promoter of classical marker genes EPCAM and ELF3 and low levels at the basal-like marker gene VIM in classical cell lines. In basal-like cell lines low levels of H3K27ac were detected at the promoter of classical marker genes EPCAM and ELF3 and high levels at the basal-like marker gene VIM (Figs. 2A and S1B). Notably, H3K27ac levels were associated with the gene expression of EPCAM, ELF3 and VIM (Fig. 2A). To assess whether the classical and basal-like PDAC cell lines exhibited differences in the overall H3K27ac levels and associated histone modifying enzymes, immunoblot and qPCR analyses were performed. Basal-like cell lines showed a significantly increased protein expression of HDAC2 and low abundance of H3K27ac in comparison with the classical cell lines (Fig. 2B). On the other hand, HDAC1 was slightly reduced in basal-like cells. The overall abundance of histone marker H3K9ac did not differ between classical and basal-like PDAC cell lines (Fig. 2B). In line with lower H3K27ac levels in basal-like cell lines, the basal-like cells expressed significantly lower levels of the HAT genes EP300 (E1A binding protein p300) and CREBBP (CREB binding protein) compared with the classical cell lines and showed reduced gene expression of BRD4 (bromodomain containing 4) and SMARCA4 (SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4), two epigenetic readers that bind to acetylated histone residues to activate transcription (26) (Fig. 2C). The aforementioned data demonstrated a H3K27ac-driven epigenetic regulation of specific PDAC subtype marker genes. Moreover, the classical cell lines revealed high abundance of H3K27ac and strong expression of HAT genes, whereas basal-like cell lines showed low H3K27ac levels and high HDAC2 expression.

Following identification of distinct histone acetylation levels and differences in the expression of associated histone acetylation key factors, HATi A485 and HDACi SAHA were used to treat classical and basal-like cell lines to determine epigenetic reprogramming abilities. For cell lines Capan-1, Capan-2, PaTu-T and PANC-1, it was not possible to determine IC50 for A485 and SAHA because the cell lines did not respond to A485 or SAHA treatment (Fig. S2). Therefore, the cell lines were treated with two dosages of A485 (10 nM and 1 μM) and SAHA (10 nM and 0.5 μM) for 24 h and overall H3K27ac levels were determined by immunoblot analysis (Fig. 3A and B). The classical and basal-like cell lines responded well to the treatment with 1 μM A485 and showed significantly decreased H3K27ac levels (Fig. 3A). Only the basal-like cell lines exhibited significantly increased H3K27ac levels after treatment with 0.5 μM SAHA. The high abundance of H3K27ac in the classical cell lines was not further increased by treatment with SAHA (Fig. 3B), indicating that SAHA treatment did not affect PDAC cells of the classical subtype. To assess the effect of HATi and HDACi treatment on epigenetic reprogramming of classical and basal-like cell lines, cell lines were treated with 1 μM A485 and 0.5 μM SAHA for 24 h and analyzed by RNAseq to detect changes in gene expression. The classical and basal-like cell lines treated with A485 showed 619 and 120 downregulated genes, respectively (Fig. 3C; Tables SII and SIII). Although the basal-like cell lines exhibited increased H3K27ac levels after SAHA treatment, the number of DEGs was low. SAHA treatment of classical and basal-like cell lines resulted in an upregulation of 49 and 25 genes, respectively (Fig. 3C; Tables SIV and SV). To identify biological effects of HATi and HDACi treatment for each tumor subtype, GO and KEGG pathway analyses were conducted based on DEGs. The comparison of the top 100 GO terms between classical and basal-like cell lines for genes downregulated after A485 treatment revealed 54 subtype-specific GO terms. The downregulated genes of classical cell lines were involved in 'lipid metabolic process' and 'drug metabolisms cytochrome P450', while basal-like cell lines showed decreased expression of genes associated with 'mesenchymal cell differentiation' and 'extracellular matrix organization' (Fig. 3D; Table SVI). The comparison of the top 100 GO terms for genes upregulated after SAHA treatment, showed 87 subtype-specific GO terms. However, only the classical cell lines revealed significantly enriched GO terms, which were related to 'cytoskeleton organization', 'leukocyte mediated immunity' and 'secretion by cell' (Fig. 3D; Table SVII). Based on enriched GO terms and KEGG pathways associated with 'mesenchymal cell differentiation' and drug metabolisms, it was analyzed in more detail whether HATi and HDACi treatment had a subtype-specific impact on EMT and chemosensitivity of classical and basal-like cell lines. An EMT signature published by Gröger et al (23) was assessed to determine changes after HATi and HDACi treatment. Notably, basal-like cell lines generally had a higher EMT score than the classical cell lines, but the EMT score of basal-like cell lines did not change after epigenetic drug treatment. However, when treated with 1 μM HATi A485, classical cell lines showed an increased EMT score, indicating an enrichment of an EMT phenotype (Fig. 3E). In addition, a transcriptomic signature to predict gemcitabine sensitivity generated by Nicolle et al (24) was analyzed. Classical cell lines had an overall higher gemcitabine score, indicating higher gemcitabine sensitivity, than basal-like cells. Upon A485 HATi treatment, classical cell lines showed a decreased gemcitabine score, whereas no effect was detected for basal-like cell lines following HATi or HDACi treatment (Fig. 3E). The data suggested that HATi treatment of classical PDAC cells had a negative effect resulting in an increased EMT phenotype and decreased drug response towards gemcitabine.

To determine the effects of HATi and HDACi treatment on gemcitabine chemosensitivity, drug response to gemcitabine alone was measured first in classical and basal-like cell lines (Fig. 4A). The drug response assay revealed that classical cell lines responded better to gemcitabine treatment with lower relative IC50 values than basal-like cell lines (Fig. 4B), which was consistent with the transcriptomic gemcitabine score (24). Solute carrier family 29 member 1 (SLC29A1, also known as hENT1) is an important transporter for gemcitabine uptake and its expression is associated with gemcitabine sensitivity in patients with pancreatic cancer (27-29). Thus, it was investigated whether SLC29A1 expression was regulated by histone acetylation and altered by epigenetic drug treatment. Although the comparison of SLC29A1 expression and presence of H3K27ac at the gene promoter revealed two clusters of cell lines, in which the abundance of H3K27ac was associated with gene expression, these were independent of the tumor subtypes (Figs. 4C and S3A and B). Treatment with 1 μM HATi A485 resulted in a significant reduction of H3K27ac at the SLC29A1 promoter in all classical cell lines, whereas the SLC29A1 promoter showed an enrichment of H3K27ac in basal-like cell lines after 0.5 μM SAHA treatment (Fig. 4D). In line with this, SLC29A1 expression was decreased in all cell lines after A485 treatment, whereas SAHA treatment did not result in increased SLC29A1 expression (Fig. S3C). Next, the impact of HATi and HDACi treatment on gemcitabine sensitivity was assessed by pre-treating cell lines with 1 μM and 0.5 μM SAHA for 96 h, followed by 72 h gemcitabine treatment. There were no significant changes in gemcitabine response after HATi or HDACi pre-treatment (Figs. 4E and S3D). In summary, the epigenetic reprogramming efficiency after short-term HATi or HDACi treatment was low and did not change chemosensitivity of pancreas tumor cells.

To analyze if longer epigenetic treatment had a more pronounced effect on the transcriptional reprogramming of tumor cells, cell lines were treated for 4-30 weeks with 1 μM and 0.5 μM SAHA. With the exception of MIAPaCa-2, for which SAHA dose was decreased to 0.05 μM, the long-term epigenetic treatment was well-tolerated by all cell lines showing no visible signs of cell death. Long-term HATi treatment with 1 μM A485 significantly decreased H3K27ac levels in classical and basal-like cell lines, whereas long-term SAHA treatment with 0.5 μM (or 0.05 μM for MIAPaCa-2) caused no subtype-dependent H3K27ac enrichment (Fig. 5A). Next, it was investigated whether long-term HATi or HDACi treatment affected the epithelial and mesenchymal phenotype of classical and basal-like cell lines, respectively, by determining H3K27ac levels at promoter regions of the classical marker gene ELF3 and basal-like marker gene VIM. Classical cell lines showed a significant loss of H3K27ac at the ELF3 promoter and a significant enrichment of H3K27ac at the VIM promoter after long-term A485 treatment, whereas basal-like cell lines exhibited no such changes (Fig. 5B). In line with the epigenetic changes, the classical cell lines showed significantly decreased ELF3 expression and increased VIM expression after long-term treatment with A485, indicating a switch from an epithelial to a mesenchymal phenotype (Fig. 5C). Notably, these data were consistent with short-term A485 treatment of the classical cell lines, which showed an increased EMT score (Fig. 3E). To evaluate the effects of HATi and HDACi treatment on the phenotypical behavior of the tumor cells, classical and basal-like cell lines were analyzed for changes in cell proliferation, colony formation and migration upon epigenetic drug treatment. Only the classical cell lines exhibited decreased cell proliferation after long-term SAHA treatment, while long-term A485 treatment did not affect proliferation of classical and basal-like cell lines (Figs. 5D and S4A). Although long-term A485 treatment resulted in significantly lower colony forming ability in classical and basal-like cell lines (Fig. 5E), the migration of classical cell lines was significantly increased under long-term A485 treatment (Figs. 5F and S4B). In summary, prolonged treatment with HATi induced changes in overall H3K27ac abundance in classical and basal-like cells. Classical cell lines revealed decreased epithelial and increased EMT marker gene expression after HATi treatment, as well as increased cell migration.

As mentioned above, long-term epigenetic treatment of PDAC cell lines caused a significant transcriptional and phenotypical reprogramming, particularly of classical cells. Thus, it was investigated if long-term epigenetic drug treatment also had an influence on tumor cell chemosensitivity to gemcitabine. After long-term HATi treatment with 1 μM A485, there was a significant reduction of H3K27ac at the SLC29A1 promoter only in classical cells (Fig. 6A). The loss of H3K27ac following HATi treatment resulted in a significant decreased expression of SLC29A1 in classical cell lines (Fig. 6B). In contrast, the SLC29A1 expression was not affected by HATi in basal-like cell lines. Likewise, long-term HDACi treatment had no effect on epigenetic regulation of SLC29A1 in all tumor cell lines (Fig. 6A and B). Next, it was tested whether long-term epigenetic drug treatment affected gemcitabine drug response of classical and basal-like cell lines. The classical cell lines exhibited impaired gemcitabine drug response after long-term treatment with SAHA, whereas no alterations were detected following A485 treatment. Notably, neither long-term A485 nor SAHA treatment altered the gemcitabine drug response of basal-like cell lines (Figs. 6C and S5). Overall, the data suggested that epigenetic and transcriptional changes of SLC29A1 induced by long-term A485 or SAHA treatment did not affect the gemcitabine response of the tumor cells. However, pancreatic tumor cells respond with decreased chemosensitivity to epigenetic drug treatment depending on the molecular subtype.