A time-series analysis of altered histone H3 acetylation and gene expression during the course of MMA^III-induced malignant transformation of urinary bladder cells

Summary

UROtsa cells’ exposure to MMA^III alters histone acetylation levels and changes their binding patterns with gene-regulatory regions across the genome and, hence, deregulates the expression of key tumor control genes and their regulated gene networks. This probably plays an important role in MMA^III-induced urothelial cell malignant transformation.

Abstract

Our previous studies have shown that chronic exposure to low doses of monomethylarsonous acid (MMA^III) causes global histone acetylation dysregulation in urothelial cells (UROtsa cells) during the course of malignant transformation. To reveal the relationship between altered histone acetylation patterns and aberrant gene expression, more specifically, the carcinogenic relevance of these alterations, we performed a time-course analysis of the binding patterns of histone 3 lysine 18 acetylation (H3K18ac) across the genome and generated global gene-expression profiles from this UROtsa cell malignant transformation model. We showed that H3K18ac, one of the most significantly upregulated histone acetylation sites following MMA^III exposure, was enriched at gene promoter-specific regions across the genome and that MMA^III-induced upregulation of H3K18ac led to an altered binding pattern in a large number of genes that was most significant during the critical window for MMA^III-induced UROtsa cells’ malignant transformation. Some genes identified as having a differential binding pattern with H3K18ac, acted as upstream regulators of critical gene networks with known functions in tumor development and progression. The altered H3K18ac binding patterns not only led to changes in expression of these directly affected upstream regulators but also resulted in gene-expression changes in their regulated networks. Collectively, our data suggest that MMA^III-induced alteration of histone acetylation patterns in UROtsa cells led to a time- and malignant stage-dependent aberrant gene-expression pattern, and that some gene regulatory networks were altered in accordance with their roles in carcinogenesis, probably contributing to MMA^III-induced urothelial cell malignant transformation and carcinogenesis.

Introduction

Consumption of arsenic-contaminated drinking water remains one of the most concerning public health issues throughout the world (1). Arsenic is a potent carcinogen and the bladder is one of arsenic’s known cancer target sites. A study suggested that consuming drinking water with arsenic levels even below the WHO recommend level, 10 μg/l, may result in a 40% increased risk of bladder cancer (2). The increased bladder cancer risk has been linked to the historical consumption of water from private wells built in an era when pesticides containing arsenic were widely used in the New England region of the USA (3). Additionally, studies reported that high arsenic exposure in early life is associated with a significant high incidence of and an even higher mortality rate from urinary bladder cancer (4,5). Arsenic is inefficient at inducing point mutations or initiating and promoting the development of tumors in experimental animals (6–8); however, arsenic exposure is known to be associated with large-scale aberrant gene expression (9–11), probably resulting from arsenic-induced aberrant epigenetic modifications, which are thought to play a central role in arsenic-induced carcinogenesis (12).

Epigenomic deregulations caused by arsenic have been observed in various tissues and cells (8,12–14). Previously, we showed that the exposure of human urothelial bladder cells (UROtsa cells) to monomethylarsonous acid (MMA^III), one of the most toxic arsenic metabolites, produced a global effect in acetylation levels of histone H3 and H4 (15–17). Epidemiologic studies also suggest that exposure to high levels of arsenic is linked to aberrant histone modifications, including histone acetylation deregulation (18,19). Acetylation on lysine residues of histone tails is a key regulation mechanism of gene expression. The basic charges of histone tails become neutralized upon acetylation, leading to an open status of chromatin and causing increased accessibility for binding factors and transcriptional machineries to gene regulatory regions across genomic DNA (20–22). Therefore, it is reasonable to speculate that the altered histone acetylation patterns are responsible, at least partially, for the large arsenic-induced scale of aberrant gene expressions, which may result in deregulating cell growth, cell cycle, proliferation and differentiation, and ultimately malignant cell transformation and carcinogenesis (9,23).

In the current study, we examined histone acetylation patterns, the accessibility of gene regulatory regions across the genome due to altered histone acetylation patterns, and global gene expression over the course of MMA^III-induced UROtsa cell malignant transformation. Our integrated analysis provided evidence suggesting that alterations in histone acetylation patterns induced by chronic MMA^III exposure led to changes in the interaction between histone acetylation and gene-regulatory regions across the genome, which resulted in the deregulation of the expression of key tumor control genes and regulated gene networks during the development and progression of malignant transformation of urinary bladder cells.

Materials and methods

Cell culture and treatment

UROtsa cells, an immortal, non-tumorigenic human bladder cell line, were generously provided by Dr D.Zuzana (University of North Carolina, Chapel Hill, NC) in 2012. These UROtsa cells were derived from the left ureter of a 12-year-old girl and immortalized by transfection with a SV40 large T antigen gene (24). The karyotype of this cell line is normal in both quantity and appearance (25). UROtsa cells used in this study were at low passage in our laboratory and were tested for viability, morphology and growth curve analysis on a regular basis. They tested negative for mycoplasma in the latest test, conducted in December 2015. Cells were plated at 2 × 10⁴/ml on 100 mm tissue-culture flasks and maintained in Dulbecco’s modified Eagle’s medium containing 5% vol/vol fetal bovine serum and 1% antibiotic-antimycotic. On the second day, cells were treated and continuously cultured in a medium enriched with 50 nM MMA^III and refreshed every three days with freshly prepared MMA^III solution, for up to 16 weeks. Parallel cultures of UROtsa cells were maintained in an MMA^III-free medium and served as passage-matched controls, as described previously (17). After 8 or 12 weeks of the initial MMA^III treatment, an aliquot of MMA^III-treated cells was cultivated under standard cell culture conditions for an additional 2 weeks after removal of the MMA^III.

Colony formation in soft agar

Anchorage-independent growth was performed by colony formation in soft agar, using cells treated with MMA^III for 4, 8, 12, 14 and 16 weeks and passage-matched control cells (17). Briefly, at each termination treatment, cells were resuspended in culture medium supplemented with 0.3% agar, which was then overlaid onto 0.6% agar medium in a 12-well plate with a density of 2 × 10⁴ cells per well. After 14 days of incubation, colonies were manually counted with an AMG EVOS XL microscope (Advanced Microscopy Group, Bothell, WA). Images were obtained with a Zeiss Axio Observer Z1 microscope (Carl Zeiss Microscopy LLC, Thornwood, NY).

RNA-seq analysis

Samples of untreated (0W) and UROtsa cells treated with 50nM MMA^III for 4 weeks (4W), 8 weeks (8W), 12 weeks (12W), and 14 weeks (14W) plus 14N (12 weeks treatment with an additional 2 weeks’ culture after MMA^III removal) were chosen for the RNA-seq analysis. The extraction of total cellular RNA from cells was carried out using TRIzol Reagent (T9424; Sigma) and purified by the RNeasy Kit (217004; Qiagen). Subsequent RNA preparation steps were carried out at the Genomics and Bioinformatics Core Facility at the University at Buffalo. RNA quality was assessed by agarose gel electrophoresis, spectrophotometry, and a BioAnalyzer (Agilent Technologies, Santa Clara, CA). The samples were then used to generate sequencing libraries with a TruSeq RNA Sample Prep Kit (Illumina) and sequenced on an Illumina HiSeq 2500 sequencer following the manufacturer’s instructions. STAR (version 2.4.2a) was used to align raw sequencing reads to the human reference genome (hg19) and GENCODE annotation (version 19). We used RSEM (version 1.2.23) to measure gene and transcript abundances in transcript per million, then the transcripts per million were upper-quartile normalized across all samples. The genes with transcripts per million larger than one in at least one sample were kept for the downstream analysis. The mapping statistics were summarized in Supplementary Table S1, available at Carcinogenesis Online.

Gene-based genome-wide association analysis and network pathway analysis

The RNA-seq gene lists of differentially expressed genes obtained at each time point after MMA^III treatment with a threshold fold change > 1.5 were used to conduct the functional analysis. Gene Ontology analysis on differential expressed genes are done by PANTHER (http://patherdb.org) online tools. The statistical significance for overrepresentation test is conducted by the binomial test, and the P-value cutoff is 0.05. Moreover, the Ingenuity® Pathway Analysis (IPA) tool from the Ingenuity Systems Inc. was used to perform the functional analysis on the differential expressed genes, including the pathway analysis, network analysis, biological function and upstream activity, as described previously (26,27). Briefly, a right-tailed Fisher’s exact test was applied to identify canonical signaling pathways that were modulated significantly across cell malignant transformation. A corrected P-value < 0.05 was used to define the significant pathways. We further performed network enrichment analysis by overlapped genes with differential expression in the lists onto the molecular networks from IPA database. The connected subnetworks were then constructed based on the algorithm developed by IPA. R package was used to generate the heatmaps for differentially expressed genes in the time course of MMA^III treatment.

ChIP-seq analysis

Following MMA^III treatment, UROtsa cells were cross-linked with 0.75% formaldehyde for 5 min. Chromatin was collected and sonicated using a Misonix Sonicator 3000 (QSonica, Newtown, CT). Antibodies against acetyl-Histone H3 Lys18 (AB1191; Abcam) were used for ChIP. The chromatin immunoprecipitated DNA fragments were blunt-ended and ligated to the Illumina indexed DNA adaptors using a NEBNext ChIP-seq library prep master mix set (E6240; New England BioLabs). The prepared library was sequenced using the Illumina HiSeq 2500. Reads were aligned to the human reference genome (hg19) using BWA with default parameters. The histone data were called peak by MACS2 with matched input data, and the pileup profiles were generated by extending 200 nt at the 5′-end of the mapped reads. Biological replicate experiments from cell cultures at each time point were performed to assess the reproducibility. The mapping statistics of ChIP-seq duplicate experiments were summarized in Supplementary Table S2, available at Carcinogenesis Online. The irreproducible discovery rate analysis methodology was applied to assess the reproducibility of ChIP-seq data sets (Supplementary Figure S4, available at Carcinogenesis Online) (28).

Quantitative PCR analysis

The quality and quantity of collected total RNA for the quantitative real-time PCR (qRT-PCR) samples was evaluated using a Nanodrop 2000 Spectrophotometer (Thermo Scientific, Waltham, MA). The qRT-PCR method has been described previously (17). Briefly, mRNAs were reverse-transcribed using a SuperScript III cDNA Synthesis Kit (Life Tech). Gene expression was measured using primers listed in Supplementary Table S3, available at Carcinogenesis Online. Real-time PCR was performed by Bio-Rad CFX96 Touch™ Detection System and a SYBR Green Supermix Kit (Bio-Rad). The GAPDH gene was used as internal control. Relative expression was calculated using the comparative CT (2^−ΔΔCT) method. A Student’s t-test was performed to evaluate the fold differences between groups.

Histone acetylation analysis

Treated and untreated cells were pelleted, and histones were isolated with EpiQuik Histone Extraction Kit (OP-0006; EpiGentek) following the manufacturer’s protocol. Histone acetylation was evaluated by western blot and/or by a calorimetrically based ELISA kit. Western blot analysis was performed as described previously (17). Total cell lysates were prepared using RIPA lysis buffer (Sigma, St Louis, MO). Protein concentrations of cell lysate and histone were determined by the BCA assay (23225; Thermo) and Bio-Rad Protein Assay Dye Reagent (500-0006, Bio-Rad), separately. Equal protein amounts were subjected to immunoblot analysis with antibodies to acetyl-pan Histone H3 (06-599, Millipore), acetyl-Histone H3 Lys18 (AB1191; Abcam), Histone H4 (07-108, Millipore) and β-actin (A3854, Sigma). The EpiQuik H3/H4 Modification Multiplex Assay (EpiGentek) kit was used to quantify multiple histone H3/H4 acetylation simultaneously in an ELISA-like format. The detailed protocol can be found in the manufacture website (P-3100/P-3102).

Animal experiments

The animal study was approved by the Institutional Animal Care and Use Committee (MED19014). Tumorigenicity assay was performed at the AAALAC-accredited vivarium at the University at Buffalo. Female BALB/c nude mice were purchased from Charles River. Mice were housed in laminar-flow cabinets under specific pathogen-free conditions at room temperature with a 24-h night–day cycle and fed with pellets and water ad libitum. Log growth-phase of MMA^III-treated or untreated UROtsa cells (2 × 10⁶ cells in 0.1 ml phosphate-buffered saline) were injected into the intrascapular region of nude mice using a tuberculin syringe with a 26-gauge needle. Tumor growth was observed every 3 days by measuring its diameter with Vernier calipers. Upon termination of the study, mice were killed by CO₂, followed by cervical dislocation of dead animals. Tumor tissues were removed and portions were fixed in 10% formalin and frozen in liquid nitrogen. Portions of each tumor were sectioned and stained for histopathological analysis using hematoxylin and eosin.

Statistical analysis

Western blot data represent a total of three experiments, and one-way analysis of variance was used to compare the difference. Data represent the mean ± SD of the averages of at least three independent experiments. Significance was noted in corresponding figure legends (*P < 0.05; **P < 0.01; ***P < 0.001).

Results

Malignant transformation of UROtsa cells induced by MMA^III

Exposure of UROtsa cells to a physiologically relevant level of 50 nM MMA^III for ~12 weeks resulted in malignant transformation as indicated by in vitro colony and in vivo nude mice xenograft assays (Figure 1). The number of colonies produced was increased with the extended duration of MMA^III exposure, with 62 colonies per square centimeter (colonies/cm²) at 8 weeks (8W) and 113 colonies/cm² at 12 weeks (12W) (Figure 1A). The colony numbers stabilized and did not significantly increase with the continued MMA^III exposure post-12 weeks (14 and 16 weeks). Removing MMA^III from the culture at 8W and 12W had a dramatically different impact on colony formation for these cells. With continuous culture for an additional 2 weeks of 8W and 12W cells without MMA^III (samples termed 10N and 14N, respectively), the number of colonies formed in 10N cells decreased from 62 colonies/cm² to an average of 38 colonies/cm² in comparison with 8W treated cells, whereas no significant changes of colony numbers were observed between 12W and 14N cells (Figure 1A). These different tumorigenic characteristics between 8W and 12W cells were observed in animal studies too, with tumors formed from 12W cells in the xenograft nude mice assay but not for the UROtsa control (0W) and 8W cells, as also evidenced by pathological analysis (Figure 1B and Supplementary Table S4, available at Carcinogenesis Online).

Figure 1.

MMA^III exposure in UROtsa cells induced cell malignant transformation. (A) A marked increase in the colony number of UROtsa cells were observed by soft agar assay following exposure to 50 nM MMA^III at 4W, 8W, 12W, 14W and 16W. A significant decrease in the colony number resulting from MMA^III withdrawal was observed at 8W but not at 12W following an additional 2 weeks’ cell culture (10N and 14N). (B) Histopathological analysis of mammary gland in nude mice after cells treatment and representative hematoxylin and eosin staining images were shown (size bars denote 1mm). A total of 2 × 10⁶ cells of untreated, 8W and 12W cell samples were subcutaneously injected into a flank of nude mice (n = 6). Tumor was induced in mice post injection of 12W cells but not in mice of untreated or 8W cells.

Global transcriptional alterations during the MMA^III-induced UROtsa cell malignant transformation

RNA-Seq was used to generate global transcription profiles of UROtsa cells over a time course of MMA^III exposure. The number of genes with a significantly altered expression was increased following the MMA^III exposure (Supplementary Figures S2 and S3, available at Carcinogenesis Online). Unsupervised hierarchical clustering analysis of the global transcriptional profiles clustered MMA^III-treated UROtsa cells into two main clusters based on the duration of exposure, in which 0W and 4W samples were clustered together and all other samples were in another cluster (Figure 2A). Within the cluster group of longer exposure duration samples, the 8-week sample was distanced with other samples, including 12W, 14W and 14N. Among the differentially expressed genes, 1222 were identified and changed significantly across samples of 8W, 12W, 14W and 14N relative to the control 0W sample: 576 transcripts upregulated and 646 downregulated (Supplementary Table S6, available at Carcinogenesis Online). Based on the expression patterns and roles in tumorigenicity, we selected eight genes (TIE1, ALDH1A3, PTGS2, TM4SF19, PLEKHA7, BRCA1, SUV39H1 and E2F1) and verified their expression by qRT-PCR. The data showed a high degree of consistency between the RNA-Seq and the qRT-PCR results (Figure 2B). GO enrichment analysis of these 1222 gene sets revealed that the downregulated genes mainly impacted genomic stability, such as the ‘DNA metabolic process’, ‘cell cycle’, ‘DNA replication and repair’ and ‘regulation of chromosome replication’. The upregulated genes were related to ‘extracellular region’, ‘cell communication’ and ‘cell-matrix adhesion’ etc. (Figure 2C).

Figure 2.

MMA^III exposure in UROtsa cells induced the change of global transcriptional profiles. (A) Hierarchical clustering across six RNA-seq tested samples based on differentiation of gene expression shows malignant grades deteriorate in a time-dependent manner. Results are expressed as the correlation ratio of gene expression of MMA^III-treated versus untreated cells at different time points. Global changes in transcript levels of differentially expressed genes identified from the 8W, 12W, 14W and 14N samples, with their expression levels consistently upregulated or downregulated compared with the untreated control. (B) Validation of differentially expressed genes at different stages of malignant transformation via real-time PCR. Genes selected are involved in cell proliferation, apoptotic and genome stability. TIE1, tyrosine kinase with immunoglobulin-like and EGF-like domains 1; ALDH1A3, aldehyde dehydrogenase 1 family member A3; PTGS2 (COX-2), prostaglandin-endoperoxide synthase 2; TM4SF19, transmembrane 4 L six family member 19; PLEKHA7: PH domain-containing family A member 7; BRCA1, breast and ovarian cancer susceptibility protein 1; SUV39H1, Histone H3-K9 methyltransferase; E2F1, retinoblastoma-associated protein 1. (C) GO analysis on genes whose mRNA levels were consistently and significantly upregulated (left) or downregulated (right) from 8W, 12W, 14W and 14N samples.

An IPA package was used to further understand the molecular events altered during the MMA^III-induced cell malignant transformation. Top gene networks were shown in Figure 3A, including those regulated by E2F1/2, nuclear factor-kappaB (NF-κB) complex, Myc and BRCA1 etc. Heat map analysis of top signaling pathways identified by IPA showed that cell signaling networks were largely unchanged post-4 weeks’ MMA^III exposure, but altered significantly with continuous MMA^III exposure, with 8 weeks apparently more significant (Figure 3B). Similar results were observed for the identified top upstream regulators of gene networks (Figure 3C). A more detailed analysis of genes involved in top signaling pathways, including DNA repair, cyclins and cell cycle regulation as well as cell movement, is shown in Supplementary Figures S5–S7, available at Carcinogenesis Online.

Figure 3.

IPA shows transcriptional changes are time- and malignant-status dependent. (A) A representative merged network of differentially dysregulated genes centered around Myc, Histone 3, BRCA1, NF-ĸB and E2F1. Network enrichment analysis was performed by Qiagen IPA software. The top score networks were displayed graphically as nodes and edges. Node symbols were described in the bottom part of the figure. In the network, green indicates that gene expression was downregulated by MMA^III and red indicates that gene expression was upregulated by MMA^III. The node color intensity indicates the expression of genes. The darker the shade of red or green, the greater the fold change. Lines indicate the interactions between proteins, in which solid lines show the direct interaction and dashed lines show indirect interaction. Comparison of enriched pathways (B) and upstream regulators (C) across all stages of malignant transformation in UROtsa cells induced by MMA^III treatment. (D) Molecular signaling pathways enriched in UROtsa cells treated by MMA^III for 8 weeks.

Identification of early alterations of gene regulatory networks that facilitate malignant transformation of UROsta cells

Data presented above repeatedly showed that continual MMA^III exposure for ~8 weeks is probably the most critical time period when cells went from a reversible to a non-reversible malignant transformation. We thus further examined molecular events occurring in UROtsa cells during this critical time period. Utilizing the gene regulatory network knowledge base in the IPA package, we were able to identify the upstream transcriptional regulators whose regulated gene networks were significantly altered at 8 weeks (Supplementary Table S7, available at Carcinogenesis Online). Gene network analyses of significantly altered genes set at 8 weeks showed the interrelated regulation of gene expression by these upstream gene regulators (Supplementary Figure S8, available at Carcinogenesis Online). These activating or inhibiting networks exhibited function changes associated with inflammatory response, cell death and survival, organismal injury and abnormalities, DNA replication, recombination and repair etc. (Supplementary Table S5, available at Carcinogenesis Online). Among these upstream transcriptional regulators, 30 had their expression significantly changed at least 1.5-fold, as summarized in Table 1.

Table 1.

Selected upstream regulators are differentially expressed in response to MMA^III exposure at 8 weeks (P < 0.0001)

Upstream regulator	Fold change (log 2) of expression	Fold change (log 2) of H3K18ac binding	Molecular type	P-value of overlap network
FOXM1	−3.00	−0.30	Transcription regulator	3.64E−14
E2F1	−2.41	−0.58	Transcription regulator	6.92E−36
NGFR	−1.61	−0.10	Transmembrane receptor	8.81E−04
TLR4	−1.19	−0.19	Transmembrane receptor	8.13E−04
TP73	−1.19	−0.20	Transcription regulator	5.59E−05
PI3K	−0.71 (PIK3CA)	0.04	PI3K kinase family	6.13E−04
CBX3	−0.66	0.11	Transcription regulator	9.40E−04
TFDP1	−0.64	−0.15	Transcription regulator	8.81E−04
FAS	−0.63	−0.27	Transmembrane receptor	5.42E−04
IL6	2.55	0.81	Cytokine	2.92E−04
ATF3	1.67	0.82	Transcription regulator	4.61E−05
JUNB	1.56	0.13	Transcription regulator	7.84E−04
IRF9	1.32	0.08	Transcription regulator	4.32E−04
ATF4	1.30	−0.30	Transcription regulator	1.59E−10
IL1B	1.28	0.32	Cytokine	2.07E−10
MAP2K2	1.12	0.35	Group	6.13E−05
CSF2	1.07	−0.02	Cytokine	1.61E−04
CDKN2A	1.10	0.17	Transcription regulator	1.42E−04
NFIL3	1.00	0.71	Transcription regulator	4.61E−04
TGFB1	1.02	−0.03	Growth factor	1.18E−09
NF-ĸB	0.86 (NFKB2)	0.01	NF-ĸB complex	1.29E−11
RUNX2	0.87	0.44	Transcription regulator	5.42E−04
STAT2	0.76	0.13	Transcription regulator	7.03E−07
HSF1	0.79	0.03	Transcription regulator	1.54E−04
FOS	0.64	0.34	Transcription regulator	1.49E−08
FOXO1	0.66	0.38	Transcription regulator	8.84E−11
IL1A	0.67	0.35	Cytokine	3.29E−13
FOXO4	0.71	0.29	Transcription regulator	3.36E−05
TNF	0.56	0.09	Cytokine	6.76E−28
RELA	0.58	0.04	Transcription regulator	1.86E−07

The changes of upstream gene regulators and their regulated gene networks in turn led to changes in cell signaling pathways. IPA revealed that a large number of signaling pathways were significantly altered in UROtsa cells with 8-week-long MMA^III exposure. The top 20 over-represented upregulated and downregulated canonical pathways are shown in Figure 3D. One of the upregulated signaling pathways is the ‘NRF2-mediated Oxidative Stress Response’ pathway, which has been identified previously and is known to play a role in arsenic carcinogenesis (29). In addition, in line with the downregulation of E2F1, a regulator of the cell cycle, several cell cycle-related pathways were becoming inactive (Figure 3D).

Differential gene expression during MMA^III-induced transformation is coupled to altered binding patterns of H3 lysine 18 acetylation at these gene regulatory regions across the genome

We previously showed that MMA^III-mediated transformation results are accompanied by changes in histone H4 acetylation (17). Here, we verified the result by ELISA assay and showed that the total acetylated histone H4 was decreased based on the acetylation levels on histone H4 lysine K5, K8, K12 and K16 at the N-terminal of tails over the time course of MMA^III treatment (Figure 4A). However, the total acetylated histone H3 according to the acetylation levels in histone H3 lysine K9, K14 and K18 was altered in the opposite direction, in which it was increased in a time-dependent manner following chronic MMA^III exposure. The increase of total acetylated histone H3 was verified by Western blots of cell protein extracts (Figure 4B). Histone H3 lysine 18 acetylation (H3K18ac) was one of those sites in histone H3 in which acetylation increased significantly. Notably, at 10N cells, which showed a reduced colony formation ability (Figure 1A), H3K18ac protein level decreased compared with cells without MMA^III removal (10W). In comparison, although the level of H3K18ac was reduced slightly at 14N samples (no significant changes of colony formation abilities, Figure 1A), there was no significant change of H3K18ac between 14W and 14N samples (Supplementary Figure S1, available at Carcinogenesis Online).

Figure 4.

MMA^III-induced increase of H3K18 acetylation led to altered H3K18ac levels at gene regulatory regions across genome. (A) The average acetylation level of lysine sites on H3 or H4, detected by EpiQuik Histone H3/H4 Modification Multiplex Assay Kit, respectively. H3 acetyl-modification data present the average acetylation level on lysine sites of K9, K14 and K18 at N-terminal of Histone 3 tails, whereas H4 acetyl-modification data present the average acetylation level on lysine sites of K5, K8, K12 and K16 at the N-terminal of Histone 4 tails. (B) Activation of H3 pan acetylation and H3K18 acetylation post-chronic MMA^III exposure. Analysis with antibodies against total-H4 and β-actin is shown as a loading control. (C) Level of H3K18 acetylation correlates with magnitude of gene expression. Footprints in 2 kb windows of TSS (Transcriptional Start Sites) for distribution of ChIP-seq reads of H3K18ac signal. Expressed genes were divided into four groups based on their expression levels, from the top 75–100% (blue, group 1) to the lowest 0–25% (black, group 4).

We performed a ChIP-Seq experiment in UROtsa cells over the time course of MMA^III exposure. By mapping the binding pattern of H3K18ac across the genome, we found that H3K18ac signals were highly enriched in the gene Transcriptional Start Sites (TSS) regions, exhibiting a typical priority peak distribution pattern of H3K18ac. This enrichment of H3K18ac in the TSS regions was positively correlated with transcription activities (Figure 4C). Higher H3K18ac peaks were translated to higher gene expressions across all time points examined. As shown in Figure 4C, the top 25% of genes with the highest expressions had the highest binding of H3K18ac in their TSS regions. We were specifically interested in the relationship between the changes of H3K18ac levels and the expressions of those identified upstream regulators of gene networks. Using 8W samples for the comparison, Table 1 shows the expression of the top upstream regulators of gene networks and the average levels of H3K18ac in the TSS regions of these regulators. As expected, although the H3K18ac level was not the only determining factor in the regulation of gene expression—the fold change of H3K18ac signals obtained by ChIP-seq was not always consistent to the changes of RNA-seq (Figure 5A)—our data showed that the altered level of H3K18ac at the regulatory region played a very significant role in regulating the expression of the majority of these upstream regulators in this MMA^III-induced carcinogenic model. Among the top 30 upstream regulators (fold change > 1.5-folds and P-value of the regulated network < 0.0001), 20 showed a positive association between gene expression and H3K18ac level at the 5′ gene regulatory regions of these genes (Table 1 and Figure 5).

Figure 5.

Altered gene upstream regulators occurred early and probably played a driven role in MMA^III-induced malignant transformation of UROtsa cells. (A) Scatter plot of differentially expressed upstream regulators at 8W with fold change of H3K18ac signals. (B) Normalized ChIP-seq enrichment profiles (RPM) for the histone modification H3K18ac at the promoter region of selected upstream regulators: E2F1, FOXM1, IL6 and ATF3. Footprints in 2 kb windows of TSS for the histone marks H3K18ac. (C–F) Gene network depiction of upstream regulators in UROtsa cells exposed to MMA^III versus untreated cells, determined by IPA software. Specific upregulated (red) and downregulated (green) genes from the experimental data set involved in determining the expression state of the upstream regulators are shown with direct (solid lines) and indirect (dashed lines) relationships to the upstream regulators. The darker the shade of green or red, the greater the fold change. Upstream regulator analysis identifies the cascade of upstream transcriptional regulators that can explain the observed gene expression changes in our data set.

The change of H3K18ac levels in the regulatory regions of these genes was apparently time-dependent following MMA^III exposure (Figure 5B). In addition, for an unknown reason, the MMA^III-induced H3K18ac level changes were also gene specific. For example, the H3K18ac levels were reduced for E2F transcription factor 1 (E2F1) and Forkhead Box M1 (FOXM1) genes, but enhanced for Activating Transcription Factor 3 (ATF3) and Interleukin 6 (IL6) over the time course of MMA^III exposure (Figure 5B). Importantly, MMA^III-induced changes of H3K18ac level at the majority of these upstream regulators post-8 weeks’ exposure were in a direction correlated to their roles in carcinogenesis. For instance, FOXM1, the top identified upstream regulator, functions as a tumor suppressor, in which the binding of H3K18ac at its regulatory regions was significantly reduced, similar to its expression, which was significantly downregulated. Furthermore, FOXM1 regulated gene networks were also largely downregulated (Figure 5F). Other examples included tumor suppressor E2F1 and tumor-promoting factors ATF3 and IL6. As shown in Figure 5C–E, the change in H3K18ac binding levels led to the changes of their gene expressions and their regulated gene networks. Moreover, as suggested by Figure 5B, MMA^III had the most significant impact on the interaction of H3K18ac with the regulatory regions of these upstream gene regulators around the 8 weeks’ treatment period, again supporting that this time is critical for malignant transformation.

Discussion

Altered epigenetic modifications have been proposed as the underlying mechanism for arsenic-induced carcinogenesis. Here, we reported that the global acetylation level of histone H3 was significantly increased during the development and malignant transformation of urothelial bladder cells exposure to MMA^III. This elevated level, represented by H3K18ac, resulted in changed binding patterns between H3K18ac and gene regulatory regions across the genome, particularly during the critical stage of MMA^III-induced malignant transformation. These changes in H3K18ac binding patterns at the gene regulatory regions are largely in accord with changes in gene expression. More significantly, among the top identified altered upstream regulators of gene networks during malignant transformation, the altered H3K18ac levels seems play a critical role in the regulation of their expressions. In addition, the change of the majority of upstream gene regulators and their associated molecular events were in line with their roles in tumorigenesis. Together, our data suggests that MMA^III-induced histone modifications, e.g. histone H3 acetylation, probably play a critical role in the MMA^III-induced aberrant gene expressions and oncogenic transformation of human bladder cells.

It is known that arsenic exposure induces large-scale aberrant gene expression (9–11,30,31), which has been associated with deregulating cell growth, proliferation, differentiation and malignant transformation (23,31–37) and has been suggested to play a central role in arsenic-induced carcinogenesis. This agrees with our observation of this MMA^III-induced malignant transformation model, in which a global switch was found in gene expression profiles following chronic MMA^III exposure. It has been proposed that the mechanism of arsenic-induced aberrant gene expression could be through arsenic-induced epigenetic modifications, including alterations in histone acetylation (12). Previously, we and other groups have shown that, both in vitro and in vivo, exposure to arsenic could lead to alterations in histone acetylation globally and focally (15,38). We revealed here that MMA^III-induced bladder cell malignant transformation was accompanied by an increase in global histone H3 acetylation levels, including H3K18ac, in a time-dependent manner. An elevated H3K18ac level was linked to an increased risk of prostate tumor recurrence (39), and associates with a reduced survival in pancreatic cancer patients (40). We showed here that perturbations of the H3K18ac level altered its binding patterns at the promoter-specific regions of a large number of genes, which in turn altered the expression of these genes. However, this altered binding pattern is not universal across the genome; rather, it appears gene specific. More critically, MMA^III-induced changes in H3K18ac binding in the regulatory regions of genes were largely correlated with their functions in tumorigenesis, i.e. downregulation for tumor suppressors (e.g. E2F1 and FOXM1) and upregulation for tumor promoters (e.g. IL6 and ATF3).

The E2F family of transcription factors are key regulators of cell-cycle progression, targeting cyclins, CDKs, checkpoint regulators, DNA repair and replication proteins (41). E2F1 and E2F2 act as transcriptional activators and are important for progression through the G1/S transition and S-phase. Despite their contrasting roles in different contexts (42), numerous studies support a suppressive function for E2F1 and E2F2 (43,44). FOXM1 is another tumor suppressor that was downregulated by MMA^III. FOXM1 has a critical function in the maintenance of genomic stability (45). Inhibition of FOXM1 activity results in aberrancies during mitosis, causing frequent chromosome mis-segregation, defects in cytokinesis and overt aneuploidy (46). ATF3, a stress response transcription factor, is kept at low levels in normal cells. The ATF3 gene regulates cell adhesion and invasion in carcinogenesis (47) enhances cell motility and epithelial-to-mesenchymal transition, and increases tumor-initiating cell features (48). Our results indicate that upregulated ATF3 in turn led to a transcriptional repression of CDK1 (Cyclin-Dependent Kinase 1), AURKA/B (Aurora Kinase A), PLK1 (Polo-Like Kinase 1), and NEK2 (NIMA-Related Kinase 2) and an activation of GSN (Gelsolin) and CHAC1 (ChaC Glutathione-Specific Gamma-Glutamylcyclotransferase 1). These factors also play a critical role in the normal biological process and carcinogenesis (49). IL6 is a multifunctional cytokine that was originally characterized as a regulator of immune and inflammatory responses. There is growing evidence that inflammation may help drive tumor formation, growth, and metastasis. IL6 is implicated in tumorigenesis through numerous downstream mediators to support cancer cell proliferation, survival, and metastatic dissemination (50). Our results revealed the transcriptional network of IL6 comprises lots of cytokines and cyclin-dependent kinase-related genes (Figure 5E), including Transforming Growth Factor Beta-1 (TGFB1), Cyclin-Dependent Kinase Inhibitor 2A (CDKN2A), Chemokine Ligand 2 (CCL2), Chemokine Ligand 10 (CXCL10), Prohibitin (PHB) and Prostaglandin-Endoperoxide Synthase 2 (PTGS2). Notably, it is known that elevated PTGS2 is associated with an increased risk of bladder cancer (51).

Additionally, we identified Myc, Cyclin A, NF-κB complex, BRCA1 and E2F as major affected transcriptional networks in the process of MMA^III-induced malignant transformation (Figure 3A). Myc functions as a transcriptional regulator and it targets genes well known to participate in cell cycle, survival, adhesion and differentiation (52). Most importantly, Myc also serves to regulate global chromatin structure by regulating histone acetylation (53). In the human genome, Myc is believed to regulate expression of 15% of all genes through binding on Enhancer Box sequences (E-boxes) and recruiting histone acetyltransferases. Myc-triggered apoptotic pathways are the essential checkpoints that guard the cell from cancer (33). Alterations in Myc expression have been reported in a variety of human cancers (52). Similar to Myc, the NF-κB complex, as part of a network of interacting factors, controls cytokine production and cell survival and is recognized as a crucial player in many steps of cancer initiation and progression. The activation of NF-κB is associated with resistance to apoptosis due to its fundamental effects on cellular dedifferentiation and proliferation in malignancies (54). Compared with non-treated cells, a higher level of the NF-κB protein complex is positively correlated with progression from normal to malignant UROtsa. Notably, through its ability to upregulate the expression of tumor-promoting cytokines, such as transforming growth factors or tumor necrosis factors, and survival genes, such as Bcl-XL, BCL2A1 or Bax, NF-κB provides a critical link between inflammation and cancer (55). Recently identified nucleotide −1186 in the NF-κB binding promoter region of the PTGS2 (COX-2) gene is associated with an increased risk of bladder cancer (51). Other studies suggest MMA^III stimulates proinflammatory, mitogenic signal transduction pathways in UROtsa cells in humans (56). Therefore, a similar mechanism may play a role in the development of squamous cell carcinoma of the bladder. Interestingly, c-Myc can block the activation of NF-κB (57); thus, downregulated expression level of Myc regulating genes contributes to NF-κB-induced antiapoptotic effects in malignantly transformed cells, which was observed in our study. Increased NF-κB levels repress TP53, thus ensuring cancer cells’ survival and replication, even in the presence of DNA damage (58).

Consistent with a previous report (59), we showed that 8 weeks’ exposure to MMA^III is the most critical window for the MMA^III-induced UROtsa cells transition from a reversible event to an irreversible malignant transformation. Although the level of H3K18ac was continually increased with prolonged MMA^III exposure, the binding patterns of H3K18ac across the genome were altered the most significantly in the 8W samples. As a result, the global gene-expression profile of the 8W samples was changed overwhelmingly. A number of gene network regulators and signaling pathways critical for tumor development and progression were significantly altered at this time and continued throughout the ultimate irreversible malignant transformation. In addition to the four upstream regulators discussed above, many others were identified with different expressions post-8 weeks MMA^III treatment, shown in Table 1 and Figure 5A. Many of them exhibited altered H3K18ac patterns in their promoter regions over the time course of MMA^III exposure and were known to play important roles in tumor development and progression. For example, Toll-like receptor 4 (TLR4) belongs to a group of transmembrane proteins, TLRs, which modulate the release of cytokines and chemokines from cancer cells and induce an anti-tumor immune response to suppress tumor progression (60). TLR-mediated signaling also regulates cell proliferation and survival. Signal transducer and activator of transcription 2 (STAT2), an essential transcription factor in the Type I IFN signal transduction pathway, is central in determining whether immune responses in the tumor microenvironment promote or inhibit cancer (61). Constitutively activated STATs have been found in a variety of human tumors and cancer cell lines. Loss of STAT2 resulted in a decrease in inflammatory factor genes, including IL6 (62). FOS (FBJ Murine Osteosarcoma Viral Oncogene Homolog) has oncogenic activity and is frequently overexpressed in tumor cells (63). The Forkhead transcription factor FOXO1 target genes are involved in cell cycle, proliferation, survival and apoptosis. FOXO1 plays a cooperative role in inflammatory signaling through amplifying NF-ĸB signaling (64). In a chronic low-grade inflammatory environment, FOXO1 activates the C/EBPβ gene transcription (65), thereby increasing proinflammatory genes’ expression, such as IL6, CXCL8 and CCCL20, all of which are also triggered by the NF-ĸB complex in UROtsa cells at 8-weeks’ treatment of MMA^III. The function and regulation of FOXO1 is well documented in many cancers, but the molecular mechanism of its regulation in bladder cancer is largely unknown. A number of gene expressions associated with centrosome separation, spindle checkpoint activity, chromosome positioning and stabilization, and mitotic regulation were suppressed in the downregulated network of FOXO1 at 8 weeks, such as NCAPG, KIF11, CCNB1, and SPC25 (Supplementary Table S6, available at Carcinogenesis Online). Other genes identified with decreased expression include BRIP1 (BRCA1 Interacting Protein C-Terminal Helicase 1), which participates in the repair of DNA double-strand breaks; DLGAP5, the key regulator of adherent junction integrity and differentiation in epithelial cells; and DEPDC1, involved in bladder carcinogenesis via the NF-κB signaling pathway (66).

One finding that needs to be given more thought is the opposite effect of MMA^III on histone H3 and H4 acetylation. Previously, we reported that MMA^III led to a global upregulation of histone H4 acetylation, which was verified again in this study. We also showed previously that the administration of the HDAC inhibitor, suberoylanilide hydroxamic acid, after the initial MMA^III treatment caused an increase in histone acetylation and prevention of malignant transformation (17). In general, it is believed that increased histone acetylation is associated with the open status of chromatin, which translates to an increased activity of gene expression (67). However, emerging evidence suggests that the role of histone H3 and H4 acetylation may not always the same and could be altered in different directions and play a different role in many biological processes. For example, a recent study showed that H3- and H4-acetylation have opposing roles in regulating nucleosome architecture and the distinct aspects of nucleosome dynamics might be independently controlled by individual histone. Moreover, H4-acetylation may act to counteract the effect of acetylated H3 (68). The precise balance of the acetylated and deacetylated states of histones is an important mechanism in regulating gene expression since acetylation/deacetylation of histone tails changes nucleosome organization and thus exposes or hides regions of DNA for transcriptional regulation (69–72). After all, chromatin structure is highly dynamic, and its impact on gene regulation is a complex process as study has revealed that both hyperacetylation and hypoacetylation of histones are correlated with gene activities (73). Acetylation imbalance could be critical and requested for cells undergoing malignant transformation, in which a hyperacetylated histone binding occurred on proto-oncogenes can increase their expression and turn them into functional oncogenes, whereas hypoacetylation on tumor suppressor genes can silence them or reduce their expression levels (74). The data presented here support this hypothesis, but clearly more research is needed to further clarify and verify this observation. We are in the process of generating additional ChIP-Seq data for other upregulated histone H3 acetylation sites and also downregulated histone H4 acetylation sites. We hope this will help us to understand how the opposing changes of acetylation on histone H3 and H4 regulate gene expression and contribute to cells’ malignant transformation.

Taken together, our data suggest that altered histone H3 acetylation is associated with chronic and environmentally relevant exposure to MMA^III, which plays an essential role during MMA^III-induced malignant transformation. However, global gene expression is not mediated by a single histone modification, but rather by the interaction between multiple mechanisms including but not limited to histone modifications and DNA methylation (12,75). Studies are ongoing in our lab to systematically elucidate the integrated effects of histone modifications and DNA methylation in chromatin structure and global gene-expression patterns during MMA^III-induced urothelial bladder cell malignant transformation.

Supplementary material

Supplementary date are available at Carcinogenesis online.

Funding

National Institutes of Health (ES022629 and ES022329 to X.R.).

Abbreviations

ATF3

Activating Transcription Factor 3

H3K18ac

histone 3 lysine 18 acetylation

IL6

Interleukin 6

IPA

Ingenuity® Pathway Analysis

MMAIII

monomethylarsonous acid

NF-κB

nuclear factor-kappaB

qRT-PCR

quantitative real-time PCR

TSS

Transcriptional Start Sites