Aspirin Mediates Its Antitumoral Effect Through Inhibiting PTTG1 in Pituitary Adenoma

Abstract

Context

DNA demethylation and inhibitory effects of aspirin on pituitary cell proliferation have been demonstrated.

Objective

Our aim was to clarify the molecular mechanisms behind the aspirin-related effects in pituitary cells.

Methods

DNA methylome and whole transcriptome profile were investigated in RC-4B/C and GH3 pituitary cell lines upon aspirin treatment. Effects of aspirin and a demethylation agent, decitabine, were further tested in vitro. PTTG1 expression in 41 human PitNET samples and whole genome gene and protein expression data of 76 PitNET and 34 control samples (available in Gene Expression Omnibus) were evaluated.

Results

Aspirin induced global DNA demethylation and consequential transcriptome changes. Overexpression of Tet enzymes and their cofactor Uhrf2 were identified behind the increase of 5-hydroxymethylcytosine (5hmC). Besides cell cycle, proliferation, and migration effects that were validated by functional experiments, aspirin increased Tp53 activity through p53 acetylation and decreased E2f1 activity. Among the p53 controlled genes, Pttg1 and its interacting partners were downregulated upon aspirin treatment by inhibiting Pttg1 promoter activity. 5hmC positively correlated with Tet1-3 and Tp53 expression, and negatively correlated with Pttg1 expression, which was reinforced by the effect of decitabine. Additionally, high overlap (20.15%) was found between aspirin-regulated genes and dysregulated genes in PitNET tissue samples.

Conclusion

A novel regulatory network has been revealed, in which aspirin regulated global demethylation, Tp53 activity, and Pttg1 expression along with decreased cell proliferation and migration. 5hmC, a novel tissue biomarker in PitNET, indicated aspirin antitumoral effect in vitro as well. Our findings suggest the potential beneficial effect of aspirin in PitNET.

In previous work we demonstrated the interdependence between DNA demethylation and proliferation behavior and differentiation stage of pituitary neuroendocrine tumors (PitNET) (1). Indeed, the global DNA demethylation level in tumor tissues negatively correlated with Ki-67 proliferation rate and the ratio of 5-hydroxymethylcytosine (5hmC) to 5-methylcytosine (5mC) was higher in less differentiated adenomas (1). Additionally, increased expression of the DNA demethylating enzymes TET2 and TET3 was identified as a potential cause behind increased 5hmC level. TET2-3 exhibited significantly higher expression in adenomas with higher proliferation rate (1). Furthermore, significant positive correlation was detected between TET-cofactor UHRF2 expression and 5hmC level.

Aspirin (acetylsalicylic acid, ASA) is a commonly used nonsteroidal anti-inflammatory drug with antipyretic, analgesic, anti-inflammatory, and antithrombotic effect (2). Its antitumoral effect was demonstrated in several cancer types (3), and epidemiological and clinical studies indicated that aspirin reduced the cancer risk in several tumor types (3–7). While these data are encouraging, the mechanisms behind aspirin antitumoral effect are not entirely clarified. It is partially attributed to the inhibition of cyclooxygenase-2 (COX-2) which is upregulated in various cancer cells (8, 9, 10). Recently, more COX-independent effects of aspirin have been revealed (11, 12). ASA regulates multiple signaling pathways, biological functions, and molecules, including cell cycle, apoptosis, cell differentiation, proteasome and redox-mediated signaling, NF-κΒ, and VEGFs (13–15). Published data indicated that aspirin also had diverse and complex epigenetic influence, which were implicated in its antitumoral effect (9, 13, 16). Indeed, it was suggested that ASA was able to reverse tumor suppressor gene methylation in cancer tissues. In addition, ASA was shown to modulate histone structure through histone acetylase and deacetylase enzymes (13).

Regarding pituitary, it was also shown that ASA exerted significant inhibitory effect on pituitary adenoma cells directly by targeting cell cycle regulator cyclin A, cyclin dependent kinase 2, and indirectly through decreasing expression of survivin on mRNA and protein levels (17).

Based on these previous findings, in this study our aim was to investigate the genomic and epigenomic effects of ASA in pituitary adenomas using high-throughput profiling approaches (global methylation-demethylation analysis by high-performance liquid chromatography–tandem mass spectrometry [HPLC-MS/MS], whole transcriptome sequencing) and validate our findings by targeted methodologies, using in vitro functional assays and by investigating human pituitary tissue samples (mRNA microarray).

Methods

Cell Culture and Treatment

RC-4B/C (CRL-1903) and GH3 (CCL-82.1) pituitary cell lines were obtained from LGC Standards GmbH (Wesel, Germany) in the frame of LGC-ATCC partnership with a corresponding authentication certificate. Cells were cultured as previously reported (1) following ATCC recommendations, and they were used for experiments between 5 and 20 passages.

Aspirin (acetylsalicylic acid, ASA) from Sigma Chemical Co. (#A5376, St. Louis, MO, USA) and decitabine from Adooq (#A10292) were purchased and used similarly to our previous published experiments (ASA in 5 mM, decitabine in 10 μM final concentration, as formerly reported) (1, 17).

Patient Cohorts

As a discovery cohort, we correlated the methylation/demethylation status of 41 human pituitary adenoma samples that we previously reported (1) with currently measured gene expression data on the same samples (Table 1). Tissue samples were collected after surgical removal (National Institute of Clinical Neuroscience, Budapest, Hungary, between 2007 and 2017). Histological evaluation was performed routinely at the First Department of Pathology and Experimental Cancer Research, Semmelweis University, Budapest, Hungary. The histological diagnosis was based on immunohistochemistry for anterior pituitary hormones and transcription factors following the 2017 World Health Organization instructions. Altogether, we analyzed 12 growth hormone (GH)-producing; 25 gonadotropic (18 follicle-stimulating hormone [FSH]/luteinizing hormone [LH]+; 7 hormone-negative [HN] steroidogenic factor-1 [SF-1]+); 3 HN, T-pit+ corticotropic, and one null cell tissues (1). Patients were informed and gave consent in writing; this study was approved by the Scientific and Research Committee of the Medical Research Council of Hungary (0618/15).

Table 1.

Discovery sample cohort

Sex	Age	Clinical diagnosis	Cell lineage	Immunohistochemistry	Ki 67 (%)	Ellipsoid volume (mm3)	Primary/ recurrent
F	76	NFPA	gonadotroph	FSH	2-3%	1767	primary
M	44	NFPA	gonadotroph	FSH	2-3%	n.a.	primary
M	51	NFPA	gonadotroph	FSH	2-3%	10	recurrent
F	39	NFPA	gonadotroph	FSH	1-2%	n.a.	primary
M	38	NFPA	gonadotroph	FSH	3-4%	n.a.	primary
F	49	NFPA	gonadotroph	FSH, LH	< 1%	1950	recurrent
F	74	NFPA	gonadotroph	FSH, LH	< 3%	n.a.	primary
M	62	NFPA	gonadotroph	FSH, LH	< 3%	n.a.	primary
M	64	NFPA	gonadotroph	FSH, LH	< 3%	87114	primary
M	63	NFPA	gonadotroph	FSH, LH	< 1%	21501	primary
F	68	NFPA	gonadotroph	FSH, LH	3-4%	n.a.	recurrent
M	43	NFPA	gonadotroph	FSH, LH	7-10%	n.a.	primary
F	37	NFPA	gonadotroph	FSH, LH	7-8%	1150	recurrent
F	80	NFPA	gonadotroph	FSH, LH	2-3%	16366	primary
M	73	NFPA	gonadotroph	FSH, LH	2-3%	n.a.	primary
M	38	NFPA	gonadotroph	FSH, LH	3-4%	n.a.	primary
M	73	NFPA	gonadotroph	LH	2%	5964	primary
M	72	NFPA	gonadotroph	LH	< 2%	2145	primary
M	50	NFPA	gonadotroph	T-Pit-; PIT1-; SF-1+	< 2%	n.a.	recurrent
F	64	NFPA	gonadotroph	T-Pit-; PIT1-; SF-1+	5%	n.a.	primary
F	60	NFPA	gonadotroph	T-Pit -; PIT1-; SF-1+	5%	n.a.	recurrent
M	58	NFPA	gonadotroph	T-Pit-; PIT1-; SF-1+	5-7%	14137	primary
F	64	NFPA	gonadotroph	T-Pit-; PIT1-; SF-1+	< 1%	n.a.	primary
F	50	NFPA	gonadotroph	T-Pit-; PIT1-; SF-1+	2%	n.a.	primary
F	43	NFPA	gonadotroph	T-Pit-; PIT1-; SF-1+	1%	n.a.	primary
F	58	NFPA	corticotroph	T-Pit+; PIT1-; SF-1−	4%	1950	primary
F	65	NFPA	corticotroph	T-Pit+; PIT1−; SF-1−	3-4%	905	primary
F	49	NFPA	corticotroph	T-Pit+; PIT1−; SF-1−	3-4%	n.a.	primary
M	73	NFPA	null cell	T-Pit−; PIT1−; SF-1−	3-4%	6371	recurrent
M	35	GH-producing	somatotroph, lactotroph	GH, PRL	6%	n.a.	primary
M	30	GH-producing	somatotroph, lactotroph	GH, PRL	8%	2185	primary
M	51	GH-producing	somatotroph, lactotroph	GH, PRL	< 3%	n.a.	primary
F	22	GH-producing	somatotroph, lactotroph	GH, PRL	4-5%	n.a.	recurrent
F	35	GH-producing	somatotroph, lactotroph	GH, PRL	5-6%	n.a.	recurrent
F	48	GH-producing	somatotroph, lactotroph	GH, PRL	10%	n.a.	primary
M	22	GH-producing	somatotroph, lactotroph	GH, PRL	3%	8181	primary
F	49	GH-producing	somatotroph, lactotroph	GH, PRL	3-4%	1023	primary
M	33	GH-producing	somatotroph, lactotroph	GH, PRL	3-4%	n.a.	primary
F	60	GH-producing	somatotroph, lactotroph	GH, PRL	< 3%	n.a.	primary
M	49	GH-producing	somatotroph	GH	< 1%	n.a.	recurrent
F	43	GH-producing	somatotroph	GH	1-3%	n.a.	primary

Abbreviations: FSH, follicle-stimulating hormone; GH, growth hormone; LH, luteinizing hormone; NFPA, nonfunctioning pituitary adenoma; PRL, prolactin; SF-1, steroidogenic factor-1.

For cross-validation, an independent cohort of human cases was used: altogether pituitary tissue specimens of 76 nonfunctioning adenoma and 34 normal samples (Table 2). Data for gene and protein expression of these samples were obtained from Gene Expression Omnibus and literature mining (Table 2).

Table 2.

Validation sample cohort

	# Pituitary adenoma samples (NFPA)	# Normal control samples (NP)	Platform (used data)
Gene expression studies
Morris et al, 2005 (18)	5	5	Affymetrix HG-U133 Plus 2.0 Array (GEO Acc. Number: GSE2175)
Michaelis et al, 2011 (19)	14	9	Affymetrix HG-U133A Array (GEO Acc. Number: GSE26966)
Moreno et al, 2005 (20)	11	3	Affymetrix HG-U95A V.2 Array (significant gene list published)
Elston et al, 2008 (21)	13	3	Affymetrix HG-U133 Plus 2.0 Array (significant gene list published)
Feng et al, 2015 (22)	7	3	Agilent-014850 Whole Human Genome Microarray (GEO Acc. Number: GSE51618)
Protein array studies
Moreno et al, 2005 (20)	11	3	2DGE-MS (significant protein list published)
Zhan et al, 2010 (23)	15	8	2DGE-MS (significant protein list published)

Nucleic Acid Isolation

Total RNA was extracted with Qiagen MiRNeasy Mini kit (217004, Qiagen, Hilden, Germany) or Macherey Nagel NucleoSpin miRNA kit (740971.50, Duren, Germany). Genomic DNA was extracted with QIAamp DNA Mini Kit (51104, Qiagen, Hilden, Germany). The isolation protocol was performed strictly following the manufacturer's instructions. Nucleic acid purity and quantity were analyzed with a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

DNA Methylation-Demethylation Analysis by HPLC-MS/MS

Chemicals and reagents

Acetonitrile and formic acid were purchased from VWR International Ltd. (Debrecen, Hungary). Water was prepared by MilliQ Purification System from Millipore (Bedford, MA, USA). The 5-methylcytosine and 5-hydroxymethylcytosine DNA Standard Set was purchased from Zymo Research Corporation (Biocenter Ltd., Szeged, Hungary).

Sample preparation

DNA solution containing 1.0 μg DNA was mixed with 50 µL of 98% formic acid and then hydrolyzed at 140 °C for 90 minutes. The sample was cooled at room temperature and evaporated under nitrogen. The residue was reconstituted in 70 µL acetonitrile-water (50:50, v/v) containing 0.1% formic acid.

The HPLC-MS/MS method was performed as we previously published (18) on an Agilent 1100 Series HPLC coupled with a Sciex 6500 QTRAP tandem mass spectrometer equipped with TurboV ion source instrument.

Transcriptome Sequencing and Data Analysis

PolyA NGS library preparation was performed using NEBNext Ultra II Directional RNA Library Prep Kit for Illumina with Purification Beads (NEB #E7760S/L) strictly following the manufacturer's instructions. Sequencing was run on Illumina NovaSeq platform (NovaSeq 6000 SP 300 cycles [2 × 150 bp]) with data output 100 M PE reads/sample. Fastq file processing was done using R package. For bioinformatic analysis, the Bioconductor package edgeR was applied for investigating differential expression. False discovery rate (FDR) was used for multiple testing correction. Unsupervised cluster analysis was done using Genesis 1.8.1 (https://genome.tugraz.at/genesisclient/genesisclient_news.shtml), t-distributed stochastic neighbor embedding (t-SNE) analysis was performed using iDEP.91 (http://bioinformatics.sdstate.edu/idep/). For pathway analysis, gene set enrichment analysis was performed using Rno Reactome pathway gene sets. Gene ontology analysis was performed using Generic GO Term Finder. For network construction, the String database (https://string-db.org/) was applied. For analyzing transcription factor regulatory relationships, TRRUST was used (https://www.grnpedia.org/trrust/). Raw data was uploaded to NCBI GEO database (GSE202934).

Gene and Protein Array Expression Profile Analysis

Briefly, where raw data were available, we reanalyzed it using Genespring GX 12 Software (Agilent Tech Inc, Santa Clara, CA, USA), avoiding biases originating from different analysis settings and usage of different software packages. Data analysis details were previously published (19). Fold change filter was set to 2-fold, and then unpaired t test was used to identify significant (P < 0.05) gene expression changes with multiple testing correction (Benjamini-Hochberg). When raw data were not available, significant gene and protein lists extracted from manuscripts and supplementary materials were used (Table 2).

In Vitro Functional Assays (Viability, Proliferation, Dead Cell Ratio, Migration, Cell Cycle Analysis, and Ki-67 Staining by Flow Cytometry)

For determining cell viability, Alamar Blue assay (DAL1025, Invitrogen, Thermo Fisher Scientific, Grand Island, NY, USA) was used on 96 well plates, that is a commonly used method for detection of cell viability and in vitro cytotoxicity. Fluorescent signals with excitation at 560 nm and emission at 590 nm were detected using a flash spectral scanning multimode reader (5250040, Varioskan, Thermo Fisher Scientific, Waltham, MA, USA) with SkanIt Software 2.4.5 RE. To investigate cell proliferation and dead cell ratio, cell numbers on 6-well plates were determined using 0.4% Trypan Blue staining (15250061, Gibco, Thermo Fisher Scientific, Waltham, MA, USA).

To assess the effect of ASA on migration, a wound healing assay was performed on 24-well plates. Following ASA treatment, the cell monolayer was wounded using a 200-μL pipette tip and floating cells were washed with phosphate-buffered saline (PBS) (21-040-CV, Corning, Corning, NY, USA). Photos were taken after 0, 24, and 48 hours and analyzed with ImageJ Software (https://imagej.nih.gov/ij/, Bethesda, MD, USA) to calculate cell-free area (CFA %: [(cell-free area 24 or 48 hours/CFA 0 hours) × 100]), and migration rate (cell-covered area: (100-CFA %) (20).

Cell Cycle Analysis by Flow Cytometry

500 000 cells were incubated with 0.5 mL Cycloscope Reagent DNA labeling solution (cat: CYT-CS-R-50; Cytognos, S.L., Salamanca, Spain) for 10 minutes at room temperature in the dark and then measured with a BD FACS Lyric flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). At least 10 000 events were acquired. Flow cytometry data were analyzed using Kaluza 2.1.1 software (Beckman Coulter, Brea, CA, United States).

Ki-67 Staining by Flow Cytometry

500 000 cells were fixed for 15 minutes at RT with 100 µL IntraStain reagent A (Dako-Agilent, CA, USA, cat: K2311). After washing (5 minutes, 400g, room temperature) with PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH = 7.4), the pellet was resuspended in 100 µL IntraStain reagent B containing 0.4% Triton X-100 and 5 µL Ki67 antibody (Sony Biotechnology, Weybridge, UK, Cat# 2352515, RRID:AB_2920575). After 15 minutes of incubation at room temperature in the dark, the cells were washed again (5 minutes, 400g, room temperature) and the pellet was resuspended in 0.5 mL PBS containing 0.01 mg/mL 2-(4-amidinophenyl)-6-indolecarbamidine (DAPI) and incubated for 30 minutes at room temperature in the dark. The samples were measured with a BD FACS Lyric flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). At least 50 000 events were acquired. Flow cytometry data were analyzed using Kaluza 2.1.1 software (Beckman Coulter, Brea, CA, United States).

Vector Construction and Luciferase Reporter Gene Assay

The vector construction and luciferase reporter gene assay were performed as we previously reported (21). Briefly, 2090 bp of the Pttg1 5′-UTR 201 ENSRNOT00000005070.5 region was cloned (5′→3′) into pGL3 promoter vector (Promega, Madison, WI) at the 5′ end of the firefly luciferase gene. Activity of Pttg1-luc construction was controlled by direct Sanger sequencing and basic luciferase activity detection. RC-4B/C cells were plated at 10⁴ cells per well in 96-well plates on the day before transfection. Cells were cotransfected with 150 ng Pttg1-luc or control pGL3-promoter plasmid and 150 ng renilla luciferase vector (pRL-TK; Promega, Madison, WI) using Lipofectamine 3000 (Thermo Fisher Scientific, Grand Island, NY, USA) according to the manufacturer's instructions. Luciferase assay was performed 24 hours later using Dual-Glo luciferase assay system (#E2920; Promega, Madison, WI), as previously reported (21).

Targeted Gene Expression Measurements

The measurements were performed as previously described (1) using TaqMan Gene Expression Assays (Thermo Fisher Scientific, Waltham, MA, USA). Dnmt1: Rn007009664_m1; Tet1: Rn01428192_m1; Tet2: Rn01522037_m1; Tet3: Rn01425643_m1; Uhrf1: Rn02346366_m1; Uhrf2: Rn01502134_m1; Pttg1: Rn00574373_m1; Tp53: Rn00755717_m1 and PTTG1: Hs00864094_g1. Glyceraldehyde 3-phosphate dehydrogenase (Rn01775763_g1) or β-actin (Hs01060665_g1) were applied as endogenous control. To calculate the relative gene expression changes we used the fold change formula (fold change [FC] = 2-ΔΔCt).

Protein Extraction and Western Blot

Protein extraction and Western blotting were done as we previously reported (17, 22) using the following antibodies: p53 (1C12) mouse mAb (Cell Signaling Technology Cat# 2524, RRID:AB_331743) (1:1000, Cell Signaling Technology, ZA, Leiden, The Netherlands) and acetyl-p53 (K382) rabbit mAB, (1:500, Bioss Cat# bs-0905R, RRID:AB_10855372) (Bioss Antibodies, Woburn, MA, USA); anti-PTTG antibody (DCS-280) (1:2000, Santa Cruz Biotechnology Cat# sc-56207, RRID:AB_785382) Santa Cruz Biotechnology (Santa Cruz, CA, USA); anti-β-actin (1:1000, Cell Signaling Technology Cat# 4967, RRID:AB_330288) (Cell Signaling Technology, ZA, Leiden, The Netherlands) and secondary antibodies as goat anti-mouse HRP-conjugated (1:2000, Agilent Cat# P0447, RRID:AB_2617137) and goat anti-rabbit HRP-conjugated (1:2000, Agilent Cat# P0448, RRID:AB_2617138) (Agilent, Santa Clara, CA, USA).

Statistical Analysis

Statistical analyses were performed using GraphPad Prism 6.01 software. Results were presented by mean ± SD. For assessing the differences in DNS methylation/demethylation, cell viability, proliferation and dead cell ratio, and mRNA and protein expression, unpaired t test with Welch's correction was used. Investigating correlation between gene expression and methylation/demethylation status Pearson correlation was used. P values < 0.05 were considered significant.

Results

Methylation-Demethylation Alteration Upon Aspirin Treatment

Global methylation-demethylation levels were determined by HPLC-MS/MS following ASA treatment in RC-4B/C and GH3 cells. In RC-4B/C cells, a significant increase of 5hmC (fold change: 1.95; P = 0.0003) and 5hmC/5mC ratio (fold change: 2.49; P = 0.0012) were observed (Fig. 1A). In GH3 cells, similar results were detected: both 5hmC (fold change: 3.5; P = 0.0005) and 5hmC/5mC ratio were elevated (fold change: 3.28; P = 0.0078) (Fig. 1B). Therefore, expression of enzymes and their cofactors regulating the methylation-demethylation cycle was measured by quantitative reverse transcription–polymerase chain reaction (RT-qPCR). In line with the mass spectrometry results, Dnmt1 responsible for cytosine methylation and its cofactor Uhrf1 showed no change in any of the cell lines (Fig. 1C-1D). However, in RC-4B/C cells, among Tet enzymes (converting 5mC to 5hmC) Tet2, Tet3, and their cofactor Uhrf2 were significantly overexpressed in line with increased 5hmC level (Fig. 1C). GH3 cells showed very similar expressional alteration, including increased Tet1 level also (Fig. 1D). Accordingly, Tet1, 2, and 3 showed significant correlation with 5hmC upon ASA treatment in both cell types (Fig. 1E).

Figure 1.

ASA effect on DNA methylation-demethylation. A-B, Change of global methylation-demethylation in RC-4B/C (A) and in GH3 (B) cells. C-D, Gene expression change behind global methylation/demethylation in RC-4B/C (C) and in GH3 (D) cells. E, Correlation between 5hmC and demethylation enzymes.

Global Transcriptome Alteration Upon Aspirin Treatment Indicates Inhibition of Cell Cycle and Proliferation

As global DNA demethylation showed marked alteration after ASA treatment, we investigated its consequence, a change in global transcriptome of pituitary cells following 5 mM ASA treatment.

By next-generation sequencing, 15 655 exons belonging to 3278 genes were identified as significant differentially expressed (< or > fold change 2 and FDR P < 0.05) after 5 mM ASA treatment. Unsupervised hierarchical clustering and t-distributed stochastic neighbor embedding analysis showed that ASA had a significant effect on overall gene expression profile (data not shown).

Pathway and gene ontology analysis indicated that the most affected pathway affected was the cell cycle (Table 3). This was reinforced by network analysis of gene interactions that revealed the top 10 most important genes governing ASA antitumoral effect on transcriptional level (Mad2l1, Pcna, Ube2c, Mcm4, Chek2, Brca1, Smc4, Cdc45, Cdk2, Rbx1), which also influenced proliferation, cell division, DNA damage response, and ubiquitination.

Table 3.

Pathway analysis of transcriptional changes upon ASA treatment (only the first 10 significant pathways are presented)

Reactome pathway ID	Reactome pathway Name	Observed gene count	Background gene count	False discovery rate
RNO-69278	Cell Cycle, Mitotic	71	387	4.82e-06
RNO-1640170	Cell Cycle	77	449	7.19e-06
RNO-1430728	Metabolism	162	1330	0.00035
RNO-194315	Signaling by Rho GTPases	56	324	0.00035
RNO-69306	DNA Replication	27	102	0.00035
RNO-194840	Rho GTPase cycle	28	117	0.00058
RNO-453279	Mitotic G1-G1/S phases	26	103	0.00058
RNO-69002	DNA Replication Pre-Initiation	20	65	0.00058
RNO-69620	Cell Cycle Checkpoints	41	218	0.00058
RNO-68886	M Phase	48	283	0.00097

Several cyclins and cyclin dependent kinase were downregulated, while Cdkn1a was upregulated (Fig. 2A). The net outcome of the ASA effect on cell cycle was represented by the marked decrease of Ki-67 (log2FC: −2.66; P = 2.71E-04). In addition, expressional alteration of multiple genes implicated in proliferation, DNA repair, and genome stability were identified (Fig. 2B)

Figure 2.

A, Gene expression change behind cell cycle and proliferation alteration upon aspirin administration. Expression of genes implicated in cell cycle phase G1, S, G2, and M are indicated by blue, green, yellow, and red, respectively. Expressions of genes that inhibit cell cycle are illustrated by gray color. B, ASA genomic effect in pituitary adenoma. Expression changes: Tp53 (log2CPM:-0.403135161, FDR:0.475431307); E2f1 (log2CPM:-1.699322161, FDR:0.017677985); Birc5 (log2CPM:-1.669861341, FDR:0.002634309); Aurka (log2CPM:-2.693827176, FDR:0.000112511); Aurkb (log2CPM:-1.994711397, FDR:0.009565016); Pttg1 (log2CPM:-2.359268022, FDR:3.27562E-05); Ha2 × (log2CPM:-1.59, FDR:5.86E-03); Ezh2 (log2CPM:-1.187080264, FDR:0.067119053); Ccnb1 (log2CPM:-2.995668611, FDR:1.09241E-06); Ccne1 (log2CPM:-1.533242072, FDR:0.015933026); Ccna2 (log2CPM:-2.642928343, FDR:0.050601632); Cdk1 (log2CPM:-2.137090443, FDR:0.000341413); Cdkn1a (log2CPM:1.591983374, FDR:0.003852956); Gadd45a (log2CPM:1.907358455, FDR:0.000777632). Genes indicated by gray represent links between the transcriptional and epigenomic level. See further details in the discussion.

In Vitro Functional Validation of Effects of Aspirin on Viability, Proliferation, and Cell Death

Validation of transcriptional findings using metabolic viability, cell proliferation and cell death of RC-4B/C and GH3 cells by functional in vitro assays were carried out. Treatment with 5 mM ASA led to significant viability decrease of RC-4B/C cells. Inhibition of cell proliferation was also observed, as cell number was decreased by 46.5% (P < 0.0001) (Fig. 3A). Since GH3 cells are characterized by loosely adherent appearance with floating clusters we performed functional assays without media change during ASA treatment. We found slight but significant decrease in viability and cell proliferation by 19.3% (P = 0.0028) upon treatment compared with DMSO control. Upon cell cycle analysis by flow cytometry, reduced number of cells entered to the cell cycle and reduced number of cells in S phase were identified in the ASA-treated group compared with control (4.38% vs 10.25% and 3.88% vs 14.94% in RC-4B/C and GH3 cells, respectively.) This was verified by decreased Ki-67 staining following ASA treatment compared with control: 0.09% vs 0.34% in RC-4B/C cells and 0.34% vs 1.48% in GH3 cells. ASA resulted in only a minor increase (1.66% and 1.11%) of dead cell ratio in RC-4B/C and GH3 cells, respectively (Fig. 3A). Additionally, we observed decreased cellular migration in the scratch assay upon ASA treatment (Fig. 3A).

Figure 3.

In vitro validation. A, microscopic images of RC-4B/C and GH3 cells upon ASA treatment for morphological assessment: significant morphologic changes cannot be observed upon DMSO or ASA treatment, only reduced cell numbers are suggested in ASA group. B, Decreased optical density following ASA treatment was observed compared to controls, due to decreased proliferation. No cytotoxic effect of DMSO was observed vs nontreated (NT) group. C, ASA effect on cell proliferation was further investigated by living cell numbers. Cell proliferation was 46% (P < 0.0001) and 80% (P = 0.028) in ASA group vs DMSO control in RC-4B/C and GH3 cells, respectively. Significant cell death was not observed in either cell line upon treatment. D, Cell cycle analysis by flow cytometry indicated reduced number of cells entering to the cell cycle and reduced number of cells in S phase upon ASA treatment vs control (4.38% vs 10.25% and 3.88% vs 14.94% in RC-4B/C and GH3 cells, respectively). E, ASA decreased RC-4B/C cell migration to 34% (P = 0.01) and 47% (P = 0.02) for 24 and 48 hours, respectively.

Aspirin Effect on Transcription Factor Activity

Transcription factor regulatory relationships were assessed using RNA sequencing data. Following ASA treatment, genomic effect of 12 transcription factors was revealed (Table 4). Among these, Tp53 activity was the most significant. Numerous (25) transcripts regulated by Tp53 showed significant expressional change (Table 5). Among Tp53 controlled genes, Pttg1 was also identified as downregulated (log2FC: −2.36; P = 3.275E-05). The E2F1 transcription factor activity was also modulated by ASA, which was also supported by Birc5 downregulation (log2FC: −1.72; FDR P = 0.052). Furthermore, the expression of Tp53 downstream target Gadd45α was also upregulated (log2FC: 1.90; FDR P = 0.0007) upon ASA treatment (Table 5).

Table 4.

Activity of transcription factors changed following aspirin treatment

Transcription factor	P value	Adjusted P value	Genes regulated by the transcription factor
TP53	9.244E-9	5.28E-06	MCM7; MKI67; FOXM1; AQP3; EGFR; RECQL4; CCNB1; PTTG1; GPNMB; CHEK2; CASP3; E2F7; PDGFRB; PLK3; SMAD3; GADD45A; GDF15; PLK1; CCNA2; PRC1; CDK1; CRYAB; ATF3; MAD1L1; EZH2
E2F1	8.646E-8	2.47E-05	TOP2A; RRM1; RRM2; HSPA5; PLK1; FANCA; CDC6; FOXM1; AURKB; AURKA; DHFR; CCNB1; RAD51; CHEK2; DDIT3; CDK1; MYBL2; KIF2C; MCM5; ECT2; ASF1B
E2F4	1.851E-6	3.52E-04	PLK4; CCNB1; RAD51; PLK1; PCLAF; MCM10; TTK; AURKB
E2F3	5.498E-6	7.85E-04	CCNA2; CCNB1; PLK1; CDK1; CDC6; AURKA
TRP53	1.265E-5	0.001445	PDGFRB; DUSP4; PLK3; SMAD3; MCM7; GADD45A; GDF15; INSR; FOXM1; CKS1B; CCNA2; PER2; CCNB1; GAP43; KRT19; COL2A1; CASP3; PCBP2; PTGDS; ATF3
E2F4	2.181E-5	0.002076	PLK4; MAP1LC3B; RAD51; E2F2; MYBL2
TFAP2A	1.144E-4	0.009333	ACHE; CCNB1; CRABP2; GALNT3; TH; CYP11A1; MCAM; CGA; ADRA1A; CRYAB; EGFR
ATF1	2.953E-4	0.021081	TOP2A; LDHA; MAP1LC3B; TH; SLC20A1; CGA
ATF4	4.086E-4	0.025929	MAP1LC3B; HSPA5; DDIT3; TRIB3; NDC80; ATF3; DDR2
TFDP1	4.440E-4	0.025353	DHFR; RRM1; CDK1; MYBL2
XBP1	7.446E-4	0.038652	ERN1; HSPA5; GAD1; DDIT3; ERP29
ARNT	9.602E-4	0.045694	JUN; GAD1; BHLHE40; AHR; MFSD2A

Table 5.

Expression changes of genes from Tp53 regulatory network

GeneName	EnsGeneID	log2FC	P value	FDR
Aqp3	ENSRNOG00000009797	-2.30422	5E-05	0.001157
Atf3	ENSRNOG00000003745	3.484486	1.43E-08	1.8E-06
Casp3	ENSRNOG00000010475	-2.04182	1.61E-06	6.51E-05
Ccna2	ENSRNOG00000015423	-2.64293	0.02121	0.050602
Ccnb1	ENSRNOG00000058539	-2.99567	9.47E-09	1.09E-06
Cdk1	ENSRNOG00000000632	-2.13709	1.25E-05	0.000341
Chek2	ENSRNOG00000037509	-1.37363	0.0324	0.083205
Cryab	ENSRNOG00000010524	2.61213	0.000107	0.002052
E2f7	ENSRNOG00000026252	-3.02489	0.000156	0.002484
Egfr	ENSRNOG00000004332	1.951245	0.000199	0.003023
Ezh2	ENSRNOG00000006048	-1.18708	0.01809	0.067119
Foxm1	ENSRNOG00000005936	-2.53223	1.17E-06	5.55E-05
Gadd45a	ENSRNOG00000005615	1.907358	3.82E-05	0.000778
Gdf15	ENSRNOG00000019661	2.696207	2.61E-06	0.000124
Gpnmb	ENSRNOG00000008816	2.678566	3.31E-08	2.54E-06
Mad1l1	ENSRNOG00000001265	-1.90029	0.000841	0.009182
Mcm7	ENSRNOG00000001349	-2.1472	2.97E-05	0.000535
Mki67	ENSRNOG00000028137	-2.66258	1.47E-05	0.000271
Pdgfrb	ENSRNOG00000018461	2.073018	0.002211	0.014145
Plk1	ENSRNOG00000018815	-2.84073	7.04E-09	8.9E-07
Plk3	ENSRNOG00000018484	1.98987	8.11E-06	0.000296
Pprc1	ENSRNOG00000018561	-0.15577	0.587733	0.749968
Pttg1	ENSRNOG00000003802	-2.35927	5.87E-07	3.28E-05
Recql4	ENSRNOG00000032446	-1.78421	0.000247	0.00399
Smad3	ENSRNOG00000008620	1.545729	0.001247	0.009008
Tp53	ENSRNOG00000010756	-0.40314	0.300957	0.475431

Although Tp53 expression itself did not change (P = 0.475) either in transcriptome data or in validation, ASA led to acetylation at the K382 position and demonstrated significant positive correlation with 5hmC level (Fig 4A-4C).

Figure 4.

A, Tp53 expression in transcriptome and RT-qPCR validation. B, Tp53 significantly correlated with 5hmC upon ASA administration. C, Western blot indicated no total p53 protein change following ASA treatment; however, treatment induced p53 acetylation at K382 residue. Densitometry showing the ratio acetyl-p53/total p53. Due to the lack of total p53 band in RC-4B/C cell, and the lack of acetylated p53 in the untreated cells, the ratio acetyl-p53/total p53 can be given only for MDA-MB231 and GH3 cells and it should be interpreted with caution.

Pttg1 Is Regulated by Aspirin Treatment and Demethylation in Human Pituitary Adenoma

Since Pttg1 is a member in the Tp53 regulatory network and it is a main oncogenic factor in pituitary tumorigenesis, we investigated the expression of Pttg1 further. We found that genes regulated by Pttg1 and Pttg1 interacting partners were downregulated (Fig. 5A) after ASA treatment. Besides Pttg1 interaction partners, Pttg1 expression was significantly decreased on mRNA and protein level as well (Pttg1 protein level decreased to 49% ± 2% in RC-4B/C and to 31% ± 22% in GH3 cells, respectively) (Fig. 5A).

Figure 5.

A, Transcriptional change of Pttg1 and its interaction partners upon ASA effect. Decreased levels of Pttg1 was confirmed during RT-qPCR and Western blot experiments as well. B, In vitro luciferase experiment demonstrated decreased Pttg1 promoter activity after ASA and decitabine treatment. Although in human pituitary adenomas, global demethylation level did not show significant association with PTTG1 expression (C), upon ASA administration significant negative correlation was observed between Pttg1 expression and 5hmC level (B).

By constructing a Pttg1 promoter reporter system, we observed marked decrease in Pttg1 promoter activity upon ASA treatment. Additionally, since upon ASA administration, significant negative correlation was observed between Pttg1 expression and 5hmC level (Fig. 5B), we also investigated the effect of the global demethylation agent decitabine on the Pttg1 promoter. We found that decitabine decreased Pttg1 promoter activity in vitro. However, in human pituitary adenomas global methylation and demethylation level did not show significant association with PTTG1 expression (Fig. 5C)

Aspirin-Regulated Gene Signature Can Be Detected in Human Pituitary Adenoma Samples

To cross-validate the ASA effect, transcriptome and protein expressions of independent human pituitary adenoma samples (76 adenoma and 34 normal control specimen) were used (Table 2). Using the differentially expressed genes between human adenomas and normal specimens (3755 genes), we filtered out those that were also regulated by ASA. A high proportion, 20.15%, of the genes (757/3755) dysregulated in human PitNET samples were common with ASA-regulated genes. The 757 common genes were implicated mostly in cell proliferation and cell cycle, including p53 activity and function, while cellular migration and genome stability were also detected (Table 6).

Table 6.

Pathway analysis of commonly regulated genes between human pituitary adenoma vs normal pituitary and aspirin-regulated genes

Category	Name	Source*	q-value FDR B&H	# Affected gene nr/# of genes of pathway	Ratio of affected genes in the pathway
p53-related pathways	p53 Signaling Pathway	1	3.507E-2	5/16	0.3125
	Direct p53 effectors	2	4.360E-2	15/132	0.113636
Proliferation and cell division related pathways	Retinoblastoma Gene in Cancer	1	6.484E-6	20/88	0.227273
	DNA strand elongation	4	9.949E-5	11/32	0.34375
	DNA replication	3; 1	1.188E-3	10/36	0.277778
	Regulation of nuclear SMAD2/3 signaling	2; 1	2.005E-2	12/77	0.155844
	G1 to S cell cycle control	1	3.507E-2	10/64	0.15625
Genome stability related pathways	Activation of ATR in response to replication stress	4	2.193E-2	8/37	0.216216

Abbreviation: FDR B&H, Benjamini-Hochberg false discovery rate adjustment.

Sources*: 1: MSigDB C2 BIOCARTA (v7.3); 2: BioSystems: Pathway Interaction Database; 3: BioSystems: KEGG; 4: BioSystems: REACTOME.

Discussion

Based on our previous findings that in PitNET DNA demethylation was altered and it showed correlation with proliferative behavior and differentiation (1) and the epigenetic effects of aspirin in other malignancies (13), we investigated the complex genomic and epigenomic consequences of aspirin treatment in PitNET.

Upon aspirin treatment, increased global DNA demethylation was observed. In line with this, the overexpression of Tet1, 2, and 3 (enzymes responsible for DNA demethylation), along with their cofactor Uhrf2, was confirmed. In addition, DNA methylation machinery did not show significant change (neither the global level of 5mC or Dnmt1 nor the cofactor Uhrf1 exhibited alteration). Since accurate measurement and detection of demethylation (5hmC) is challenging and requires special instrumentation, only scarce information have been available regarding demethylation change upon aspirin treatment. In a study using methylation-specific polymerase chain reaction, Guo et al demonstrated that aspirin induced demethylation of the Forkhead Box D3 (FOXD3) promoter leading to increased FOXD3 expression (23). They also proved that regular use of aspirin dramatically decreased the number of metastatic nodules of cancer cells in immunodeficient mouse lungs through the FOXD3-OLA1P2-STAT3 axis (23). Similarly, we demonstrated that Pttg1 promoter activity and expression were decreased by treatment with aspirin and by the demethylation agent decitabine, respectively (Fig. 6). In addition, a negative correlation was observed between Pttg1 expression and 5hmC level, indicating the link between aspirin, demethylation, and Pttg1 expression change.

Figure 6.

ASA-demethylation-Pttg1-Tp53 regulatory network. In pituitary adenoma ASA increased global demethylation by elevating TET1-3 enzyme level. Global DNA methylation induces global transcriptional alteration including Pttg1 underexpression that was validated by decitabine treatment. Additionally, ASA inhibits Pttg1 promoter activity. Independently, ASA acetylated p53 leading to increased p53 stability and enhanced transcriptional activity (27). Pttg1 inhibits p53 activity as it reduces p53 protein stability and represses its transcriptional activity (30). As a summary, upon ASA treatment global transcription alteration, and enhanced p53 activity results in inhibition of cell cycle, proliferation, viability. See details in the discussion.

As a consequence of global demethylation, significant transcriptome alteration was observed using whole RNA sequencing. Investigating the biological function of the altered transcriptome profile, mainly cell cycle and proliferation related transcriptional changes were observed. Changes in viability, proliferation, and cellular migration of pituitary cells were also validated by multiple in vitro functional assays. All these are in line with our previous finding that the global increase of demethylation correlated with Ki-67 proliferation index measured in human PitNET tissue samples (1). In addition, cell division and pituitary cell migration were also found to be inhibited by aspirin. However, the exact mechanism needs to be further investigated. It could be hypothesized that this occurred through Pttg1 regulated Aurka and Aurkb downregulation (see details below), because Aurka regulates cilia disassembly, neurite extension, and cell motility, while Aurkb has also been described to reduce the cellular migration of tumor cells (24, 25).

When analyzing transcription factor regulatory relationships, Tp53 and E2f1 showed the most significant alteration in activity. Indeed, the transcription factor E2f1, is a key cell cycle regulator and it targets genes encoding proteins that regulate cell cycle progression through the G1/S transition (26). Therefore, its reduced activity upon aspirin treatment was further confirmed by our in vitro functional results. While the expression of Tp53 did not change significantly upon aspirin administration, the change in its activity was experienced by the transcriptional outcome, as genes regulated by Tp53 showed significant enrichment upon aspirin treatment. In the background we detected increased Tp53 acetylation at the lysine residue at 382 position, which was previously shown to stabilize Tp53 protein and increase its transcriptional activity (27) (Fig. 6). Furthermore, Tp53 expression showed positive correlation with the 5hmC level. Since 5hmC is an indicator of decreased proliferation, our findings recapitulated previous data that aspirin acetylated and thereby activated Tp53 and induced Cdkn1a (p21) in colon tumors (28). Additionally, in PitNET cells, ASA also led to elevated Cdkn1a expression, reflecting the increased Tp53 activity.

Since Pttg1 is a well-known oncogene in pituitary tumorigenesis, overexpressed in approximately 90% of pituitary adenomas (29), and it is a member of Tp53 interaction partners, we further investigated its expression and activity following aspirin treatment. Aspirin effectively decreased Pttg1 expression in pituitary adenoma on both the RNA and protein levels, demonstrated by next-generation sequencing, qPCR, and Western blot analysis. Additionally, members of the Pttg1 regulatory network also showed decreased expression indicating decreased Pttg1 activity. Further, upon aspirin treatment of pituitary adenoma cells, a significant negative correlation was observed between Pttg1 and 5hmC, indicating the link between them. Moreover, a high proportion (20.15%) of genes being common between aspirin-regulated genes and dysregulated genes in human pituitary adenomas further indicates that aspirin may have a beneficial effect in this tumor type. These results are in line with previously reported findings, demonstrating that PTTG1 interacted with p53, and in cooperation with PTTG1 binding factor (PBF), it reduced p53 protein stability (30). Altogether, p53 and PTTG1 showed negative correlation, and head and neck squamous cell carcinoma patients with high tumoral PTTG1 had the poorest overall survival, which reflected a marked impairment of p53-dependent signaling (30).

In summary, aspirin has antitumoral effects on multiple levels in pituitary adenoma, summarized in Fig. 6. However, additional links can also be revealed from transcription alterations. For instance, the Tp53 downstream target Gadd45α, which is upregulated upon ASA treatment, not only inhibits proliferation through its interaction with Pcna and regulating DNA repair, but also induces global DNA hypomethylation (31). Aurkb, the mitosis regulator, is induced by Pttg1 and E2f1. Our findings on rat pituitary tumor cells (negative correlation of 5hmC with Ki-67) were in line with previous results described in human PitNET specimens (1); therefore, we can hypothesize that the modulation of 5hmC by ASA or decitabine and associated inhibition of Pttg1 may have antiproliferative effect on human PitNET as well. Nevertheless, our results should be validated in human PitNET samples as well, in order to avoid potential interspecies differences.

Conclusion

Aspirin exerted its antitumoral effects, among others, on genomic and epigenomic levels in pituitary cells. An aspirin-demethylation-Pttg1 regulatory network was described as an antitumoral mechanism in pituitary. The previously identified biomarker 5hmC, correlating with Ki-67 and differentiation of PitNET, also indicated the antitumoral effect of aspirin. In addition, the extensive overlap between the human pituitary transcriptome changes and aspirin-regulated gene signature further suggests the potential beneficial effect of aspirin in PitNET; however, further studies are necessary to strengthen this finding.

Abbreviations

5hmC

5-hydroxymethylcytosine

5mC

5-methylcytosine

ASA

aspirin (acetylsalicylic acid)

COX

cyclooxygenase

FC

fold change

FDR

false discovery rate

FSH

follicle-stimulating hormone

GH

growth hormone

HPLC-MS/MS

high-performance liquid chromatography–tandem mass spectrometry

LH

luteinizing hormone

PBS

phosphate-buffered saline

PitNET

pituitary neuroendocrine tumors

SF-1

steroidogenic factor-1

Funding

This work was supported by a Hungarian Scientific Research Grant of the National Research, Development and Innovation Office NKFI FK 135065 to H.B. H.B. is a recipient of Bolyai Research Fellowship of the Hungarian Academy of Sciences. H.B. and A.P. acknowledge financial support from the National Laboratories Excellence program (under the National Tumor Biology Laboratory proiect (NLP17)) and the Hungarian Thematic Excellence Programme (TKP2021-EGA-44).

The authors would like to express their acknowledge to the National Program of Bionics (the subtheme medical bioinics led by A.P.). The research was supported by the Higher Education Institutional Excellence Programme of the Ministry of Human Capacities in Hungary, within the framework of the Molecular Biology thematic programme of the Semmelweis University to A.P.

Disclosures

The authors have nothing to disclose.

Data Availability

Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in References.