Abstract

   Arsenic (As^3+) is a well-established environmental carcinogen known to
   induce malignant transformation and cancer stem-like cell (CSC)
   properties in somatic cells, with Nrf2 functioning as a central
   regulator. However, the impact of chronic As^3+ exposure on pluripotent
   stem cells, particularly through Nrf2-mediated epigenetic and metabolic
   reprogramming, remains largely unexplored. In this study, we
   chronically exposed human induced pluripotent stem cells (iPSCs,
   Nips-B2) to an environmentally relevant concentration of trivalent
   arsenic (0.25 μM, As^3+) for three months. The tumorigenic potential of
   exposed iPSCs was evaluated using anchorage-independent growth assays
   and xenograft models, while mechanistic insights were gained via
   chromatin immunoprecipitation sequencing (ChIP-seq) for Nrf2 and key
   histone modifications (H3K4me3, H3K9me3, H3K27me3, H3K36me3, and
   H4K20me3), alongside transcriptomic profiling by RNA sequencing
   (RNA-seq). Prolonged exposure markedly enhanced tumor sphere formation
   in vitro and accelerated tumor growth in vivo, indicating the
   acquisition of CSC-like traits. Integrated ChIP-seq and RNA-seq
   analyses revealed widespread Nrf2 chromatin binding and global
   epigenetic remodeling, characterized by increased levels of H3K27me3,
   H3K36me3, and H4K20me3, a modest rise in H3K9me3, and reduced H3K4me3.
   Notably, As^3+ exposure enhanced Nrf2 binding at loci regulating
   glycolysis, cholesterol biosynthesis, self-renewal, and oncogenesis.
   Functional analyses confirmed that transcriptional and metabolic
   changes were Nrf2-driven and closely linked to H3K36me3 and H3K27me3
   dynamics. Collectively, our findings demonstrate that chronic As^3+
   exposure reprograms iPSCs through Nrf2 activation and coordinated
   epigenetic remodeling, revealing a novel mechanism by which
   environmental carcinogens exploit stem cell plasticity to initiate
   CSC-like transformation.

   Keywords: Arsenic (As^3+), Nrf2, iPSCs, CSCs, Histone methylation

Graphical abstract

   [43]Image 1
   [44]Open in a new tab

Highlights

     * •
       Chronic arsenic exposure induces CSC-like transformation in human
       iPSCs.
     * •
       Nrf2 activation drives metabolic and transcriptional reprogramming
       in iPSCs.
     * •
       Arsenic alters histone marks, enhancing H3K27me3 and H3K36me3
       deposition.
     * •
       Upregulated genes are enriched in glycolysis, cholesterol
       biosynthesis, and stemness.
     * •
       Nrf2 and histone remodeling coordinately regulate CSC-associated
       gene expression.

1. Introduction

   Environmental exposure to arsenic, particularly in regions with
   contaminated drinking water, poses a major public health risk
   worldwide. Inorganic arsenic exists primarily as trivalent (As^3+) and
   pentavalent (As^5+) species, with human exposure typically involving a
   mixture of both. Among these, As^3+ is recognized as the more toxic and
   biologically active form and is the predominant intracellular species
   following cellular uptake and reduction [[45]1]. Extensive
   epidemiological and case-control studies have established a strong link
   between arsenic exposure and elevated risks of cancers in the lung,
   liver, prostate, skin, and other organs [[46]2]. Although As^3+ has
   limited mutagenic potential, experimental evidence from in vitro and in
   vivo models indicates that it is highly effective in promoting
   malignant transformation and carcinogenesis.

   Mechanistically, As^3+ activates a board range of stress-responsive and
   oncogenic kinases through both oxidative stress-dependent and
   -independent mechanisms. Using prolonged low-dose exposure to mimic
   environmental relevant conditions, our previous studies and that of
   others have demonstrated that As^3+-transformed non-cancerous cells
   acquire features of cancer stem-like cells (CSCs). These include
   enhanced tumor sphere formation, resistance to apoptosis, elevated
   expression of stemness transcription factors, and metabolic
   reprogramming—specifically, reduced mitochondrial activity and a shift
   toward glycolysis [[47][3], [48][4], [49][5]]. We previously showed
   that this metabolic switch from mitochondrial tricarboxylic acid (TCA)
   cycle activity to cytosolic glycolysis is largely driven by
   Nrf2-dependent transcriptional activation of glycolytic enzymes and
   antioxidant genes, thereby conferring a survival advantage under
   As^3+-induced stress [[50]5]. In parallel, As^3+-induced CSCs display a
   distinct epigenetic landscape, with increased levels of the
   transcriptionally active histone mark H3K4me3 and reduced levels of the
   repressive mark H3K27me3 [[51]6].

   Nrf2 (nuclear factor erythroid 2–related factor 2) is traditionally
   known for orchestrating the cellular antioxidant response
   [[52]7,[53]8]. However, emerging evidence positions Nrf2 as a key
   modulator of redox-sensitive transcriptional networks, metabolic
   plasticity, carcinogenesis, and CSC maintenance [[54][9], [55][10],
   [56][11]]. In As^3+-exposed cells, Nrf2 is frequently activated and has
   been implicated in driving oncogenic transformation by coordinating
   both metabolic and epigenetic reprogramming [[57]12,[58]13]. Given this
   dual role in stress adaptation and oncogenesis, we hypothesized that
   Nrf2 play a pivotal role in initiating As^3+-induced transformation. To
   test this hypothesis, we investigated genome-wide Nrf2 binding patterns
   alongside histone methylation dynamics, aiming to delineate the
   transcriptional and epigenetic landscape associated with chronic As^3+
   exposure.

   Most prior studies investigating As^3+ carcinogenicity have relied on
   terminally differentiated or immortalized cancer cell lines, which
   often harbor genetic abnormalities and chromosomal instability that may
   confound mechanistic interpretation. In contrast, human induced
   pluripotent stem cells (iPSCs) offer a genetically stable and renewable
   model capable of differentiating into all three germ layers [[59]14].
   Although As^3+ is predominantly metabolized in the liver into
   monomethylarsonic acid (MMA) and dimethylarsinic acid (DMA), iPSCs lack
   substantial hepatic metabolic capacity and are unlikely to efficiently
   methylate As^3+. Therefore, iPSCs provides a unique opportunity to
   investigate the direct effects of unmetabolized As^3+ on early stem
   cell states and epigenetic regulation.

   Recent studies have highlighted the value of iPSCs in toxicology and
   carcinogenicity testing. For instance, Jahan et al. utilized human
   episomal iPSCs to assess the neuroprotective effects of Nobiletin
   against sodium arsenate-induced toxicity [[60]15]. Similarly, Graham et
   al. reported that As^3+ alters morphology, reduces viability, and
   induces DNA damage in foreskin fibroblast-derived iPSCs [[61]16]. Bao
   et al. further showed that As^3+ disrupts cardiomyocyte differentiation
   by downregulating key mesodermal and cardiac progenitor markers,
   including TBX1, EOMES, GATA4, MESP1, TBX5, and ACTN4 [[62]17].
   Additionally, Wang et al. demonstrated that chronic As^3+ exposure
   drives transformation in a mouse intestinal organoid model [[63]18].
   Importantly, emerging attention has turned to non-classical oncogenic
   drivers. Many cancer hallmarks, such as metastasis, therapy resistance,
   and CSC persistence, are sustained by pathways not traditionally
   classified as oncogenic. This concept of “non-oncogenic addiction”
   highlights molecular dependencies, like sustained Nrf2 activation, that
   may represent actionable vulnerabilities in tumors lacking canonical
   mutations. Understanding these alternative mechanisms is essential for
   identifying novel therapeutic targets and elucidating early
   carcinogenic processes.

   In the present study, we employed the human iPSCs (Nips-B2), derived
   from nasal epithelial cells of an atopic asthmatic patient using Sendai
   virus-mediated delivery of OCT3/4, SOX2, KLF4, and MYC [[64]19]. We
   exposed the cells to a chronic low dose of As^3+ (0.25 μM) for three
   months to model environmental exposure and investigated the resulting
   alterations in genome-wide Nrf2 occupancy and histone methylation
   patterns. Consistent with our previous findings in human bronchial
   epithelial cells (BEAS-2B cells), we observed that chronic As^3+
   exposure induces profound metabolic and epigenetic reprogramming in
   iPSCs. These data provide new insights into the early molecular events
   underpinning As^3+-induced transformation and support a central role
   for Nrf2 in mediating the stem cell malignancy associated with chronic
   As^3+ exposure.

2. Materials and methods

2.1. Cell culture

   The human iPSC line Nips-B2 (HPS0223) [[65]19] was obtained from the
   RIKEN Bioresource Center (Tsukuba, Japan). Cells were maintained on
   iMatrix-511-coated 60-mm dishes (TaKaRa, cat#T304) in StemFit® Basic04
   Complete Stem Cell Culture Medium (amsbio, cat#SF041-001), supplemented
   with 80 ng/mL basic fibroblast growth factor (bFGF) and 1 %
   penicillin/streptomycin (Gibco, cat#15140122). The culture medium was
   refreshed daily. For routine passaging, iPSCs were dissociated either
   as cell clumps using a pipette or cell scraper, or as single cells
   using TrypLE™ Select (Gibco, cat#12563011). For cryopreservation, cells
   were enzymatically dissociated with TrypLE™ Select and frozen at −80 °C
   or in liquid nitrogen using STEM-CELLBANKER® freezing medium (amsbio,
   cat#11924).

2.2. Establishment of As^3+-transformed iPSCs

   As^3+-transformed iPSCs were generated by continuously exposing Nips-B2
   cells to 0.25 μM arsenite (As^3+) for 105 days. Culture medium was
   replenished daily, and cells were passaged twice weekly. Untreated
   Nips-B2 cells were cultured in parallel as parental controls. To assess
   the malignant transformation potential, anchorage-independent growth
   assays were performed. iPSCs were dissociated into single cells, and
   500 cells were seeded in 3 mL StemFit® Basic04 medium onto
   PolyHEMA-coated (5 mg/mL) dishes. After two weeks of incubation,
   bright-field microscopy was used to quantify sphere formation. The
   sphere size was measured by ImageJ
   ([66]https://imagej.net/ij/index.html).

2.3. In vivo tumorigenicity assay

   All animal procedures were conducted in compliance with the guidelines
   of the Division of Laboratory Animal Resources (DLAR) at Stony Brook
   University (IACUC protocol #IACUC202100104). Mice were housed in a
   maximum isolation facility with unrestricted access to food and water
   and maintained on a 12-h light/dark cycle. Eight-week-old female NSG
   mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ; Jackson Laboratories) were
   randomly assigned to two groups (n = 5 per group): control and
   As^3+-transformed iPSCs. Cells were harvested using TrypLE™ Select and
   resuspended on ice in phosphate-buffered saline (PBS) mixed with phenol
   red-free Cultrex basement membrane extract (R&D Systems,
   cat#3432-005-01) at a final BME concentration of 10 mg/mL. A total of 1
   × 10^6 cells in 200 μL of the cell/BME mixture was injected
   subcutaneously into each mouse. Tumor growth was monitored every other
   day, and tumor volumes were measured starting two weeks post-injection.
   Mice were euthanized if the subcutaneous tumor reached 2 cm in diameter
   or if no tumor developed within 30 days. Harvested tumor tissues were
   fixed in 10 % formalin and processed for histological analysis using
   Hematoxylin and Eosin (H&E) staining.

2.4. Western blotting

   Proteins were extracted using 1 × RIPA buffer (CST, cat#9806)
   supplemented with protease and phosphatase inhibitor cocktail (Thermo
   Fisher Scientific, cat#A32957) and PMSF (Thermo Fisher Scientific,
   cat#36978). Protein concentrations were determined using the BCA
   Protein Assay Kit (Thermo Fisher Scientific, cat#23225). Samples were
   mixed with 4 × LDS sample buffer (Thermo Fisher Scientific, cat#NP0008)
   and dithiothreitol (Thermo Fisher Scientific, cat#R0861), then
   denatured at 95 °C for 10 min. Equal amounts of protein were separated
   on 10 % SDS-PAGE gels and transferred to PVDF membranes (Thermo Fisher
   Scientific, cat#88518). Membranes were blocked with 5 % non-fat milk
   for 1 h at room temperature, then incubated overnight at 4 °C with
   primary antibodies. The following day, membranes were washed three
   times (10 min each) and incubated with species-appropriate
   HRP-conjugated secondary antibodies for 1 h at room temperature.
   Protein bands were visualized using the ChemiDoc MP Imaging System
   (Bio-Rad). The following antibodies were used: anti-NANOG (CST,
   cat#3580, 1:1000), anti-KCTD10 (Sigma, Cat#HPA014273, 1:1000),
   anti-CD44 (CST, Cat#37259, 1:1000), anti-GAPDH (CST, cat#97166,
   1:1000), anti-rabbit (Thermofisher, cat#31460, 1:3000), and anti-mouse
   (Thermofisher, cat#31430, 1:3000).

2.5. Chromatin immunoprecipitation sequencing (ChIP-seq)

   A total of 5×10^6 iPSCs, treated with or without As^3+ for 3 months,
   were collected for ChIP-seq analysis. The experiment was performed
   following protocols previously described [[67]20,[68]21]. All reagents
   and ChIP-grade antibodies were obtained from Active Motif (Carlsbad,
   CA). Antibodies used in this study were pre-validated by the
   manufacturer to ensure high specificity and a satisfactory
   signal-to-noise ratio. Genome-wide chromatin profiling focused on
   histone 3 or 4 trimethylation marks, H3K9me3, H3K27me3, H3K36me3, and
   H4K20me3, which were used to segment and annotate distinct chromatin
   states across the genome. ChIP-seq data quality was assessed through
   multiple metrics, including mapping ratio, read depth, normalized
   strand coefficient (NSC), background uniformity, GC content bias, and
   signal enrichment at summits. Data visualization and inspection of
   signal distribution were carried out using the UCSC Genome Browser.

2.6. Bulk RNA sequencing (RNA-seq)

   A total of 2×10^6 cells were collected for RNA sequencing services
   following the Active Motif (Carlsbad, CA) protocol. Total RNA was
   extracted and assessed for quality and integrity using an Agilent
   Bioanalyzer. Library preparation and sequencing were performed by
   Active Motif using the Illumina platform. Raw reads were aligned to the
   human reference genome (hg19) using the STAR aligner (v2.5.2b).
   Transcript assembly and quantification were conducted with Cufflinks,
   and gene expression levels were normalized using geometric FPKM
   (Fragments Per Kilobase of transcript per Million mapped reads) values.
   For differential expression analysis, read counts were derived from
   exon-mapped fragments, quantified at the annotated 3′ ends of genes.
   Differentially expressed genes (DEGs) were identified using DESeq2
   package (v1.14.1) in R 4.3.1 following standard preprocessing steps,
   including data normalization and filtering.

2.7. Bioinformatics analysis

   To explore the biological relevance of DEGs identified from both
   ChIP-seq and RNA-seq datasets, functional enrichment analysis was
   conducted using Enrichr ([69]https://maayanlab.cloud/Enrichr). DEGs
   were uploaded and analyzed across multiple curated gene set libraries,
   including Gene Ontology (GO) categories—Biological Process, Molecular
   Function, and Cellular Component—as well as KEGG pathways, Reactome
   pathways, Gene Set Enrichment Analysis (GSEA) gene sets, and the
   Molecular Signatures Database (MSigDB). Additionally, transcription
   factor regulatory networks were explored using ChEA and ENCODE
   databases. Enrichment results were ranked by the combined score, which
   the p-value from Fisher's exact test with a z-score reflecting
   deviation from the expected rank. Terms with an adjusted p-value
   (padj) < 0.05 and strong biological relevance were considered
   significantly enriched. The top-ranking pathways and gene sets were
   selected for further interpretation and visualization.

2.8. Kaplan-meier survival analysis

   To assess the prognostic significance of specific genes in lung cancer,
   Kaplan–Meier survival analysis was conducted using the Kaplan–Meier
   Plotter online tool ([70]https://kmplot.com/analysis/index.php?p=home).
   This platform integrates gene expression and clinical outcome data from
   publicly available datasets, including The Gene Expression Omnibus
   (GEO), European Genome-phenome Archive (EGA), and The Cancer Genome
   Atlas (TCGA). For the current analysis, the lung cancer gene chip
   dataset (Affymetrix microarray) was selected. Gene symbols of interest
   were queried individually, and patients were stratified into high- and
   low-expression groups based on the median expression level. Overall
   survival (OS) was used as the primary endpoint, and the “Auto select
   best cutoff” option was applied unless otherwise specified. The hazard
   ratio (HR) with 95 % confidence interval (CI) and the log-rank p-value
   were calculated and displayed on the corresponding Kaplan–Meier plots,
   which were generated for all selected genes.

2.9. Statistical analysis

   Unless otherwise specified, all experiments were independently repeated
   at least three times, each with a minimum of three biological
   replicates. Statistical comparisons between groups were conducted using
   either Student's t-test or one-way analysis of variance (ANOVA),
   depending on the experimental design. A two-tailed p-value less than
   0.05 was considered statistically significant. All data were analyzed
   using GraphPad Prism 8.0, and results are expressed as mean ± standard
   deviation (SD).

3. Results

3.1. As^3+ enhances sphere formation and tumorigenesis of iPSCs

   To assess the cytotoxicity of As^3+ in iPSCs (Nips-B2), we initially
   treated the cells with a concentration range of 0–2.5 μM As^3+ in
   StemFit Basic04 medium for 10 days. No viable iPSCs were observed after
   4 days of exposure to 2.5 μM As^3+ or after 10 days of exposure to
   1.0 μM As^3+. Based on these findings, we selected lower, sub-lethal
   concentrations (0.1–0.25 μM As^3+) for subsequent experiments. As
   depicted in [71]Fig. 1A, iPSCs tolerated 0.25 μM As^3+ well over a
   period of 1, 2, and 3 months of continuous treatment. To examine
   whether As^3+ affects the expression of stemness-related proteins, we
   performed Western blot analysis of CD44 and NANOG in iPSCs at passage
   24 (day 91) and passage 26 (day 105) following 0.25 μM As^3+ exposure.
   While As^3+ only slightly reduced the level of full-length CD44, it
   markedly increased the cleaved form of CD44 in both passages
   ([72]Figure 1B), suggesting enhanced CD44 processing. The effect of
   As^3+ on NANOG expression was inconclusive. Despite the absence of
   significant morphological differences between control and treated
   cells, As^3+ treatment led to a notable increase in sphere size in an
   in vitro sphere formation assay ([73]Figure 1C), indicating enhanced
   self-renewal potential. To determine whether As^3+ exposure promotes
   tumorigenicity, we subcutaneously injected 1 × 10^6 cells from both
   untreated and 3-month As^3+-treated iPSCs into NSG mice and monitored
   tumor development. Tumors were detectable 20–25 days post-injection in
   both groups. However, by days 32–39, tumors derived from As^3+-treated
   iPSCs were significantly larger than those from control cells
   ([74]Figure 1D). Histopathological analysis of excised tumors revealed
   that those from control iPSCs displayed typical teratoma-like
   structures containing tissues representative of all three germ layers
   (ectoderm, mesoderm, and endoderm), along with inflammatory
   infiltration, ductal hyperplasia, and some differentiated papillary
   structures. In contrast, tumors from 3-month As^3+-treated iPSCs
   exhibited additional malignant features, including keratinized
   epithelium, mucin-filled cysts, pigment deposition, and nests of
   neoplastic squamous cells ([75]Figure 1E). Together, these findings
   demonstrate that chronic low-dose As^3+ exposure enhances the
   tumorigenic potential of iPSCs.

Fig. 1.

   [76]Fig. 1
   [77]Open in a new tab

   Chronic low-dose As^3+ exposure enhances sphere formation and
   tumorigenicity of iPSCs. A Bright-field images of iPSCs (Nips-B2) after
   continuous exposure to 0.25 μM As^3+ for 1, 2, and 3 months. Cells
   maintained normal morphology and confluency, indicating tolerance to
   prolonged low-dose As^3+ treatment. B Western blot analysis of stemness
   markers CD44 and NANOG in iPSCs at passage 24 (day 91) and passage 26
   (day 105) following 0.25 μM As^3+ exposure. As^3+ treatment reduced
   full-length CD44 while markedly increased its cleaved form; effects on
   NANOG expression were variable. CIn vitro sphere formation assay
   showing increased sphere size in iPSCs treated with As^3+ for 3 months,
   indicating enhanced self-renewal capacity under anchorage-independent
   conditions. Data are presented as mean ± SD (N = 10 per group). D Tumor
   growth curves in NSG mice subcutaneously injected with control or
   3-month As^3+-treated iPSCs (1 × 10^6 cells/mouse, N = 5 per group).
   Tumors derived from As^3+-transformed iPSCs grew significantly larger
   between days 32 and 39 post-injection. Data are presented as mean ± SD.
   E Representative H&E-stained tumor sections from control and
   As^3+-transformed iPSC-derived xenografts. Control tumors exhibited
   typical teratoma architecture with ectodermal, mesodermal, and
   endodermal elements. In contrast, tumors from As^3+-transformed iPSCs
   showed malignant features, including keratinized epithelium,
   mucin-filled cysts, pigment deposition, and nests of neoplastic
   squamous cells.

3.2. As^3+ treatment alters histone methylation landscapes in iPSCs

   To investigate the molecular basis of the enhanced tumorigenicity
   observed in 3-month As^3+-treated iPSCs, we performed ChIP-seq to
   profile histone trimethylation patterns. Specifically, we analyzed five
   key histone modifications: H3K4me3, H3K9me3, H3K27me3, H3K36me3, and
   H4K20me3, using antibodies specific to each mark. In our previous
   studies involving As^3+-induced cancer stem-like cells (CSCs) derived
   from human bronchial epithelial cells (BEAS-2B), we observed a global
   increase in H3K4me3 enrichment [[78]2,[79]6]. Interestingly, the
   pattern in iPSCs diverged significantly: prolonged exposure to As^3+
   led to a marked decrease in H3K4me3 levels across gene-enriched
   regions. Conversely, the levels of H3K27me3, H3K36me3, and H4K20me3
   were substantially elevated in As^3+-transformed iPSCs ([80]Figure 2A).
   Given the established role of Nrf2 as a central regulator of redox
   homeostasis, metabolic reprogramming, and stemness—particularly in the
   context of As^3+-induced transformation [[81]5,[82]13,[83]22]—we
   additionally profiled genome-wide Nrf2 binding to determine its
   contribution to the observed chromatin alterations. Consistent with its
   proposed function in driving As^3+-mediated CSC phenotypes, we detected
   a marked increase in Nrf2 occupancy across the genome following As^3+
   exposure ([84]Figure 2A). To further characterize the effects of As^3+
   on the chromatin landscape, we visualized the distribution of
   methylation peaks across the genome using the UCSC Genome Browser. As
   expected, over 90 % of H3K4me3 peaks localized sharply to promoter
   regions or the transcription start site (TSS), while H3K36me3 exhibited
   broad enrichment across gene bodies—as exampled by the TBCK locus on
   chromosome 4 ([85]Figure 2B, left panels). TBCK encodes a putative
   kinase implicated in mTOR signaling and cell proliferation [[86]23]. It
   has not been previously reported as a canonical Nrf2 target. In the
   region of As^3+-enhanced Nrf2 enrichment peak, there is a conserved
   Nrf2 binding element TGAGTGA, suggesting TBCK is an undocumented Nrf2
   target gene. We selected the TBCK locus as a representative site to
   illustrate the spatial distribution of active and repressive histone
   marks due to its well-annotated structure, detectable expression in
   iPSCs, and dynamic epigenetic response to As^3+ exposure observed in
   our dataset. On the entire chromosome level, H3K27me3 peaks were more
   variably distributed, appearing at promoters, gene bodies, and flanking
   regulatory regions. In contrast, both H3K9me3 and H4K20me3—histone
   marks classically associated with transcriptional repression and
   heterochromatin—formed large clusters in gene-poor regions lacking
   active marks like H3K4me3, H3K27me3, and H3K36me3. These clusters were
   predominantly localized to pericentromeric regions and the terminal
   ends of the q arms on several chromosomes, including chromosome 22, a
   representative small autosome often used to visualize compact
   heterochromatin domains [[87]24]—as highlighted by red filled triangles
   in the right panels of [88]Figure 2B. While As^3+ had minimal impact on
   H3K9me3 distribution, it significantly enhanced H4K20me3 enrichment,
   with new or enlarged clusters aligning with existing H3K9me3 domains.
   These results not only highlight the mutually exclusive nature of
   active (H3K4me3, H3K27me3, H3K36me3) and repressive (H3K9me3, H4K20me3)
   chromatin marks but also underscore the profound remodeling induced by
   chronic As^3+ exposure in iPSCs. Notably, due to the female origin of
   the Nips-B2 iPSC line, ChIP-seq analysis revealed no significant
   methylation peaks on the Y chromosome, except for a few non-specific
   background signals (data not shown).

Fig. 2.

   [89]Fig. 2
   [90]Open in a new tab

   Chronic As^3+ exposure alters histone methylation landscapes and Nrf2
   chromatin binding in iPSCs. A ChIP-seq analysis showing average
   enrichments of H3K4me3, H3K9me3, H3K27me3, H3K36me3, H4K20me3, and Nrf2
   in control and 0.25 μM As^3+-treated iPSCs after 3 months. Chronic
   As^3+ treatment led to a global reduction in H3K4me3, accompanied by
   increases in H3K27me3, H3K36me3, H4K20me3, and Nrf2 occupancy. B
   Representative genome browser views of the TBCK locus (left panel)
   showing gene-specific enrichment of H3K4me3 and Nrf2 at the promoter
   and H3K36me3 across the gene body in control and As^3+-transformed
   iPSCs. A conserved Nrf2 binding element in the Nrf2 peak region,
   TGAGTGA, inserted in the panel. The right panel displays genome-wide
   histone methylation profiles—H3K4me3, Nrf2, H3K27me3, H3K36me3,
   H3K9me3, and H4K20me3—across chromosome 22. Red filled triangles
   indicate regions of broad H3K9me3 and H4K20me3 enrichment predominantly
   localized to gene-poor heterochromatic domains, including
   pericentromeric regions and the distal ends of chromosome arms. These
   repressive clusters become further intensified following As^3+
   exposure.

3.3. H3K4me3 is enriched on genes involved in differentiation and stemness
regulation

   H3K4me3 is widely recognized as an active histone modification
   associated with gene transcription. However, its functional impact is
   often cell context-dependent and influenced by the surrounding
   chromatin environment, including DNA methylation, additional histone
   modifications (e.g., H3K27me3, H3K79me3), and transcription factor
   binding dynamics [[91]25]. In line with the global reduction observed
   in H3K4me3 enrichment across gene bodies ([92]Figure 2A), we also
   observed a marked decrease in H3K4me3 signals at gene promoters and
   transcription start sites (TSSs) following chronic As^3+ exposure in
   iPSCs ([93]Figure 3A, left panel). To illustrate these changes, we
   selected representative loci on chromosomes 10 and 3 for genome browser
   visualization. These chromosomes harbor several biologically relevant
   genes—including MYOD1 and DPPA4—that showed prominent and readily
   interpretable alterations in H3K4me3 enrichment in our dataset.
   Moreover, many of the genes with the most significant changes in
   H3K4me3 following As^3+ exposure were located on these two chromosomes.
   Chromatin peak analysis further confirmed that approximately 90 % of
   H3K4me3 peaks showed reduced intensity after As^3+ treatment
   ([94]Figure 3A, right panel). However, a small fraction of peaks
   displayed increased intensity post-treatment, highlighted by red filled
   triangles. To further explore the functional relevance of H3K4me3 loss,
   we performed pathway enrichment analysis (WikiPathways) on 9325 genes
   that showed a log[2] fold change < -0.1 in H3K4me3 levels. The top
   enriched pathways included dopaminergic neurogenesis, epinephrine
   (EPI)/non-EPI signaling, and various cell differentiation programs
   ([95]Figure 3B, left panel). Notably, many of these genes are
   well-established tumor suppressors—such as Myogenic Differentiation 1
   (MYOD1), Interferon Regulatory Factor 4 (IRF4), Inhibitor Of DNA
   Binding 2 (ID2), Cyclin Dependent Kinase Inhibitor 2A (CDKN2A), CDKN2B,
   and BRCA1 Associated Deubiquitinase 1 (BAP1)—which exhibited
   substantial H3K4me3 loss ([96]Figure 3B, right panels). These findings
   suggest that epigenetic repression of differentiation-related tumor
   suppressor genes may contribute to the acquisition of malignant
   potential in As^3+-exposed iPSCs.

Fig. 3.

   [97]Fig. 3
   [98]Open in a new tab

   Chronic As^3+ exposure reshapes H3K4me3 enrichment at genes regulating
   cell differentiation and stemness. A ChIP-seq analysis showing the
   average profile of merged H3K4me3 peaks across regions spanning -5 kb
   to +5 kb relative to gene loci in control and As^3+-transformed iPSCs.
   The right panel illustrates a global reduction in H3K4me3 peak
   intensities across chromosomes 3 and 10 in iPSCs following 3 months of
   chronic 0.25 μM As^3+ exposure. A subset of peaks displayed increased
   intensity post-treatment, highlighted by red triangles. B WikiPathways
   enrichment analysis of 9325 genes with decreased H3K4me3 levels (log[2]
   fold change < -0.1) revealed top pathways related to dopaminergic
   neurogenesis, norepinephrine signaling, and cell differentiation. The
   right panel shows genome browser tracks with reduced H3K4me3 enrichment
   at several tumor suppressor genes, including MYOD1, IRF4, ID2, CDKN2A,
   CDKN2B, and BAP1, in As^3+-transformed iPSCs. C WikiPathways analysis
   of 1050 genes with increased H3K4me3 levels (log[2] fold change >0.1)
   highlighted enrichment in RNA processing and metabolic pathways (left).
   The right panel (highlighted by red triangles) displays representative
   genome browser views showing elevated H3K4me3 enrichment at core
   stemness genes such as NANOG, SLC2A3, DPPA4, NR6A1, and TDGF1 in
   As^3+-transformed iPSCs.

   Interestingly, although As^3+ broadly diminished H3K4me3, a subset of
   1050 genes exhibited notable elevation of H3K4me3 enrichment. This
   small group represents a focal upregulation module that stands out
   against the backdrop of global loss. Pathway analysis revealed that
   these genes were enriched in RNA metabolism, pre-mRNA splicing, and
   RNA/rRNA processing pathways ([99]Figure 3C, left panel). Importantly,
   several core pluripotency regulators, including NANOG, Developmental
   Pluripotency Associated 4 (DPPA4), Nuclear Receptor Subfamily 6 Group A
   Member 1 (NR6A1), Teratocarcinoma-Derived Growth Factor 1 (TDGF1), and
   Solute Carrier Family 2 Member 3 (SLC2A3), were among the top genes
   showing increased H3K4me3 levels in response to As^3+ ([100]Figure 3C,
   right panels). While enhanced H3K4me3 does not necessarily lead to
   increased transcription, this modification likely poises these genes
   for rapid activation, thereby supporting the maintenance of a stem-like
   or CSC-like state in As^3+-transformed iPSCs. This dual effect—global
   repression of differentiation programs and selective activation of
   stemness-associated genes—highlights a reprogramming mechanism that may
   underlie As^3+-induced malignant transformation.

3.4. Overlapping of As^3+-induced Nrf2 activation with H3K36me3 and H3K27me3

   At the chromosome-wide level, we observed co-localization of H3K4me3,
   H3K27me3 and H3K36me3 clusters in both control and As^3+-transformed
   iPSCs ([101]Figure 2B). This overlap, particularly between H3K36me3 and
   H3K27me3, was further validated through the pathway enrichment analysis
   of genes targeted by H3K36me3 (5943 genes) and H3K27me3 (6511 genes) in
   the As^3+-transformed iPSCs. The top ten pathways enriched in genes
   with As^3+-induced H3K36me3 included thromboxane A2 receptor (TXA2R),
   lysophosphatidic acid receptor (LPAR), C-X-C Motif Chemokine Receptor 3
   (CXCR3), Protein Tyrosine Phosphatase Non-Receptor Type 1 (PTP1B),
   Hedgehog, endothelin, RhoA, ephrin B (EPHB), EPHA, and α−Synuclein
   (SNCA) ([102]Figure 4A). Notably, eight of these pathways overlapped
   with those enriched in H3K27me3-targeted genes ([103]Figure 4B,
   highlighted by pink dots in left panel). Approximately 54 % (3208/5943)
   of the genes enriched with H3K36me3 and 49 % (3208/6511) of those
   enriched with H3K27me3 also showed co-enrichment for both marks upon
   As^3+ treatment ([104]Figure 4B, right panel). Interestingly,
   transcription factor perturbation assays revealed that six of the top
   ten pathways in the H3K27me3-targeted genes were enriched in Nrf2
   signaling and Nrf2-regulated genes ([105]Figure 4C). We had previously
   shown that As^3+ activates Nrf2, contributing to metabolic
   reprogramming in As^3+-induced CSCs from human bronchial epithelial
   cells BEAS-2B [[106]5]. In iPSCs, As^3+ also strongly activated Nrf2,
   as indicated by enhanced Nrf2 peaks in the ChIP-seq data, which were
   detected in both gene bodies ([107]Figure 2A, bottom right) and
   promoters ([108]Figure 4D, left panel). While As^3+ treatment amplified
   the existing Nrf2 peaks in control iPSCs, approximately 10–15 % of Nrf2
   peaks in As^3+-transformed iPSCs were newly formed and absent in
   control cells ([109]Figure 4D, peaks marked with red triangle in the
   right panels). These de novo Nrf2 peaks may represent a shift in Nrf2
   distribution on the genome in As^3+-transformed cells. Specifically,
   As^3+ reduced Nrf2 binding in the proximal promoter and 5′ UTR regions,
   while increasing its distribution in the exon and intron regions
   ([110]Figure 4E). Consistent with these findings, WikiPathways analysis
   of 1905 genes with enhanced Nrf2 enrichment following As^3+ treatment
   revealed a strong association with Nrf2 signaling, glycolysis, aryl
   hydrocarbon receptor (AHR) pathways, oxidative stress, and amino acid
   metabolism ([111]Figure 4F).

Fig. 4.

   [112]Fig. 4
   [113]Open in a new tab

   Overlap of As^3+-induced Nrf2 activation with H3K36me3 and H3K27me3
   enrichment in iPSCs. A Pathway enrichment analysis of 5943 genes marked
   by H3K36me3 following chronic 0.25 μM As^3+ exposure revealed
   activation of signaling pathways including TXA2R, LPAR, CXCR3, PTP1B,
   Hedgehog, endothelin, RhoA, EPHB, EPHA, and SNCA. B Comparative pathway
   analysis showing that 8 of the top 10 pathways enriched in
   H3K36me3-marked genes also overlapped with those enriched in
   H3K27me3-marked genes (6511 genes), suggesting epigenetic crosstalk
   (left). Right: Venn diagrams showing that ∼54 % of H3K36me3-marked
   genes and ∼49 % of H3K27me3-marked genes were co-enriched for both
   modifications in As^3+-transformed iPSCs. C Transcription factor
   perturbation analysis revealed that 6 of the top 10 pathways enriched
   in H3K27me3-targeted genes are regulated by Nrf2 signaling, supporting
   a potential link between Nrf2 activity and repressive chromatin
   remodeling. D ChIP-seq analysis showing average Nrf2 peak distribution
   across ±5 kb of gene loci in control and As^3+-treated iPSCs. Right:
   Genome browser views highlighting amplified Nrf2 peaks and de novo
   As^3+-specific Nrf2 peaks (red triangles), not present in control. E
   Genomic distribution of Nrf2 binding sites before and after 3-month
   As^3+ exposure. Chronic As^3+ treatment reduced Nrf2 occupancy at
   promoter and 5′ UTR regions while increasing its binding within exons
   and introns, indicating a redistribution of Nrf2 across the genome. F
   WikiPathways enrichment analysis of 1905 genes with increased Nrf2
   binding following As^3+ treatment revealed significant associations
   with Nrf2 signaling, glycolysis, aryl hydrocarbon receptor (AHR)
   signaling, oxidative stress response, and amino acid
   metabolism—highlighting widespread transcriptional reprogramming
   induced by As^3+.

3.5. As^3+ induces limited cross-talk between Nrf2 and H3K4me3 in iPSCs

   Nrf2 is widely recognized as a master regulator of oxidative stress
   response genes. More recently, its oncogenic potential has emerged in
   various cellular and animal models [[114]8,[115]26]. At the human
   γ-globin gene locus, Nrf2 binding has been shown to be a prerequisite
   for the enrichment of the active transcription marker H3K4me3 and for
   facilitating long-range chromatin interactions between the locus
   control region and the γ-globin gene promoter [[116]27]. Genome-wide
   analysis revealed that Nrf2 binding sites largely overlapped with
   clusters of H3K4me3, H3K27me3, and H3K36me3 ([117]Fig. 2, [118]Fig.
   5C). To explore the interplay between Nrf2 and H3K4me3 at the gene
   level, we compared the gene sets with elevated Nrf2, H3K4me3, and
   H3K36me3 following As^3+ exposure in iPSCs. Notably, only 6 % of Nrf2
   target genes showed concurrent increases in H3K4me3, suggesting a
   limited but potentially functionally significant correlation. Despite
   this modest overlap, ChIP-seq Enrichment Analysis (ChEA) of the genes
   with concurrent Nrf2 and H3K4me3 enrichment identified pathways
   associated with key transcriptional regulators, including FOXP3, SOX2,
   EST1, and MYC ([119]Figure 5A). FOXP3 had been implicated in promoting
   metastasis in human lung cancer and maintaining stemness in
   hematopoietic stem cells [[120]28,[121]29]. Similarly, SOX2, MYC, and
   EST1 are well-established factors involved in stem cell self-renewal
   and differentiation. These findings suggest that the limited
   cooperation between Nrf2 and H3K4me3 under As^3+ exposure may
   preferentially enhance tumorigenic stemness in iPSCs. H3K27me3, a
   well-characterized repressive histone modification, also displayed
   large-scale co-localization with H3K36me3 and H3K4me3 across the genome
   ([122]Fig. 2, [123]Fig. 5C). However, at the gene-specific level, less
   than 1 % of genes enriched for H3K27me3 also exhibited increased
   H3K4me3 ([124]Figure 5B). Among the 47 genes that were marked by both
   modifications—indicative of a bivalent chromatin state—many were
   involved in pluripotency-associated pathways such as those governed by
   SOX2 and NANOG ([125]Figure 5B, right panel). This supports the notion
   that key developmental and pluripotent genes remain transcriptionally
   poised via bivalent histone configurations. Furthermore, Nrf2 binding
   clusters generally paralleled those of H3K9me3, H3K27me3, and H3K36me3
   ([126]Figure 5C, green-filled circles), while being largely excluded
   from regions enriched with H3K9me3 and H4K20me3, which are typically
   associated with heterochromatin and transcriptional silencing
   ([127]Figure 5C, yellow box). Across all autosomes and the X
   chromosome, most Nrf2 clusters were not co-located with H3K9me3 and
   H4K20me3 domains ([128]Figure 5C, yellow box in left panel), except
   those in gene-poor regions or loci expressing long non-coding RNAs
   ([129]Figure 5C, right panels).

Fig. 5.

   [130]Fig. 5
   [131]Open in a new tab

   Enrichment analysis of genes co-targeted by Nrf2 and H3K4me3 following
   As^3+ exposure. A ChEA pathway enrichment analysis of genes with
   concurrent Nrf2 binding and H3K4me3 enrichment in iPSCs after chronic
   As^3+ treatment. Identified transcriptional regulators include FOXP3,
   SOX2, EST1, and MYC, implicating their potential roles in promoting
   stemness and tumorigenic features. B ChEA analysis of bivalent
   chromatin states in As^3+-transformed iPSCs. Left: Venn diagram showing
   minimal overlap (<1 %) between genes marked by both H3K4me3 and
   H3K27me3. Right: Enrichment analysis of 47 genes exhibiting a bivalent
   chromatin state (co-enrichment of H3K4me3 and H3K27me3), revealing
   strong association with pluripotency-related factors such as SOX2 and
   NANOG, suggesting these genes are poised for developmental regulation.
   C Genome browser snapshot showing the alignment of Nrf2 binding peaks
   with clusters of H3K27me3 and H3K36me3, as well as selected overlap
   with H3K9me3 (indicated by green-filled circles). These Nrf2-enriched
   regions are largely excluded from domains marked by H3K9me3 and
   H4K20me3, as highlighted in the yellow box. Right panels illustrate
   rare instances of Nrf2 occupancy within H3K9me3-enriched regions, all
   of which are located in gene-poor areas or near long noncoding RNA
   (lncRNA) loci.

3.6. Upregulated oncogenic gene expression in As^3+-transformed iPSCs

   Above data demonstrated that As^3+ exposure induces both Nrf2
   activation and broad epigenetic reprogramming in iPSCs. However, these
   chromatin-level changes do not necessarily reflect direct
   transcriptional outcomes related to oncogenesis or CSC formation. To
   investigate this, we performed RNA-seq to profile gene expression
   changes in control and As^3+-transformed iPSCs. Using a shrunken log[2]
   fold change cutoff of >0.1 for upregulation and < -0.1 for
   downregulation, we identified 2133 upregulated and 2369 downregulated
   genes following As^3+ exposure. The top 20 most significantly altered
   genes are presented in [132]Figure 6A. Among the most upregulated,
   ALPPL2 (ALPG), a folate metabolism-associated alkaline phosphatase, is
   a known surface marker for the naïve pluripotent state and plays a
   critical role in both establishing and maintaining pluripotency
   [[133]30]. XYLT1 and XXYLT1, which encode xylosyltransferases involved
   in glycosaminoglycan biosynthesis, are essential for extracellular
   matrix function; mutations in XYLT1 cause Desbuquois dysplasia
   [[134]31]. Other upregulated genes include TRNP1, a neural stem cell
   DNA-binding protein, and PLA2G2A, a niche factor involved in intestinal
   stem cell regulation [[135]32]. Gene Ontology analysis using EnrichNet
   showed that the top five enriched biological processes among
   upregulated genes were associated with cholesterol metabolism and
   vesicle trafficking—including COPI and COPII vesicle coating,
   intra-Golgi transport, isoprenoid biosynthesis, and cholesterol
   biosynthesis ([136]Figure 6B). Notably, 20 out of 34 genes in the
   EnrichNet cholesterol biosynthesis pathway were significantly
   upregulated, suggesting a potential role of cholesterol homeostasis in
   the tumorigenicity of As^3+-transformed iPSCs (data not shown).
   Consistent with our earlier studies [[137]26,[138]33], Molecular
   Signatures Database (MSigDB) analysis further revealed enrichment of
   upregulated genes in pathways associated with Myc signaling,
   glycolysis, hypoxia, and cholesterol metabolism ([139]Figure 6C),
   highlighting an oncogenic transcriptional program linked to stemness
   and metabolic reprogramming. In contrast, over 2000 genes were
   downregulated by As^3+. Transcription factor perturbation analysis
   revealed enrichment of repressed genes downstream of regulators such as
   GATA6, SPDEF, ATF6, and FOXQ1 ([140]Figure 6D). GATA6 is a key
   transcription factor for early cardiac differentiation [[141]34] and
   endodermal lineage commitment in human iPSCs [[142]35]. Similarly,
   SPDEF, ATF6, ETV2, and FOXQ1 have been implicated in stem cell or
   cancer cell differentiation [[143][36], [144][37], [145][38],
   [146][39]]. Thus, repression of these factors may diminish
   differentiation potential while reinforcing the pluripotent and
   self-renewing phenotype of As^3+-transformed iPSCs. MSigDB analysis of
   downregulated genes further supported this interpretation, showing
   significant enrichment in pathways related to p53 signaling, apoptosis,
   and myogenesis ([147]Figure 6E), all of which are commonly associated
   with cell differentiation and tumor suppression.

Fig. 6.

   [148]Fig. 6
   [149]Open in a new tab

   Oncogenic transcriptional reprogramming in As^3+-transformed iPSCs. A
   Heatmap showing the top 20 differentially expressed genes in iPSCs
   chronically exposed to 0.25 μM As^3+ for 3 months, as identified by
   RNA-seq. Significantly upregulated genes include ALPPL2, XYLT1, TRNP1,
   XXYLT1, and PLA2G2A. B Gene Ontology (GO) analysis using EnrichNet
   revealed significant enrichment of upregulated genes in pathways
   related to cholesterol biosynthesis, isoprenoid metabolism, and vesicle
   trafficking (COPI/COPII coating and intra-Golgi transport). C MSigDB
   enrichment analysis showing upregulated genes are associated with
   oncogenic and stemness-related pathways, including Myc signaling,
   glycolysis, hypoxia response, and cholesterol metabolism. D
   Transcription factor perturbation analysis of downregulated genes
   revealed enrichment for targets of GATA6, SPDEF, ATF6, ETV2, and FOXQ1,
   indicating suppression of key regulators involved in lineage
   specification and differentiation. E MSigDB analysis of downregulated
   genes showing significant association with tumor suppressive and
   differentiation-related pathways, including p53 signaling, apoptosis,
   and myogenesis, suggesting impaired differentiation potential and
   enhanced self-renewal in As^3+-transformed iPSCs.

3.7. As^3+-induced Nrf2 amplifies glycolytic metabolism via the pentose
phosphate pathway (PPP) and hexosamine biosynthetic pathway (HBP) in iPSCs

   We previously demonstrated that As^3+-induced Nrf2 drives a metabolic
   reprogramming from mitochondrial TCA cycle activity toward cytosolic
   glycolysis in CSCs derived from bronchial epithelial cells. This
   metabolic shift diverts glycolytic intermediates into ancillary
   biosynthetic routes, including the HBP and the serine/glycine pathway
   [[150]5]. Consistent with this, we observed a marked enhancement of
   Nrf2 signaling in As^3+-transformed iPSCs. ChIP-seq data confirmed
   widespread enrichment of Nrf2 binding across the genome ([151]Fig. 2,
   [152]Fig. 4D), while RNA-seq revealed increased expression of key Nrf2
   regulators and target genes ([153]Figure 7A). These include the
   upstream activators VCP, SQSTM1 (p62), and KLF2, as well as canonical
   Nrf2 downstream genes involved in antioxidant defense such as PRDX1,
   FTL, HMOX1, and NQO1. Interestingly, ChIP-seq also showed strong Nrf2
   occupancy at the promoter regions of SQSTM1, VCP and PRDX1, highlighted
   by red arrows in [154]Figure 7B, suggesting the presence of a positive
   feedback loop that amplifies Nrf2 signaling in response to As^3+. This
   autoregulatory circuit likely reinforces a broader oncogenic and
   metabolic program encompassing redox homeostasis, glycolysis, stemness,
   and transcriptional plasticity. Interestingly, TXNRD2, which was
   downregulated following As^3+ exposure, lacked Nrf2 binding, further
   supporting a selective transcriptional program. RNA-seq also revealed
   marked upregulation of genes encoding rate-limiting enzymes in
   glycolysis, especially those involved in early steps, such as HK2 and
   TKT, and in the glycolytic shunt pathways—PPP and HBP ([155]Figure 7C).
   In contrast to our previous findings in CSCs derived from
   As^3+-transformed BEAS-2B cells [[156]5], several genes critical for
   the serine/glycine pathway were downregulated in As^3+-transformed
   iPSCs, suggesting cell-type-specific metabolic rewiring. The oxidative
   branch of the PPP, crucial for the generation of nucleotides and NADPH
   in rapidly proliferating cells, was strongly activated. All three
   rate-limiting enzymes—G6PD, PGLS, and PGD—were significantly
   upregulated, accompanied by robust Nrf2 and H3K36me3 enrichment at the
   gene bodies and/or first exon regions ([157]Figure 7D, top panels),
   indicating transcriptional activation through coordinated Nrf2
   signaling and chromatin remodeling. Similarly, most key enzymes in the
   HBP—GFPT, PGM3, UAP1, and UAP1L1—were also upregulated in
   As^3+-transformed iPSCs, with corresponding Nrf2 and H3K36me3
   co-enrichment at their loci ([158]Fig. 7C and bottom panel of
   [159]Figure 7D). GNPNAT1 was the only HBP gene that was not induced. As
   the HBP is crucial for protein glycosylation, including modification of
   stemness regulators and SCAP (SREBP cleavage-activating protein), its
   activation supports cholesterol and fatty acid biosynthesis [[160]40].
   Collectively, these findings suggest that Nrf2-mediated activation of
   the PPP and HBP contributes to the acquisition of stem-like and
   tumorigenic properties in As^3+-transformed iPSCs through metabolic
   reprogramming and epigenetic regulation.

Fig. 7.

   [161]Fig. 7
   [162]Open in a new tab

   Nrf2-mediated metabolic reprogramming in As^3+-transformed iPSCs. A
   RNA-seq analysis reveals significant upregulation of key Nrf2 target
   genes in As^3+-treated iPSCs, including Nrf2 regulators (VCP, SQSTM1,
   KLF2) and antioxidant defense genes (PRDX1, FTL, HMOX1, NQO1),
   confirming robust activation of Nrf2 signaling. B ChIP-seq analysis
   demonstrates strong Nrf2 binding at the promoter regions of SQSTM1, VCP
   and PRDX1 in As^3+-transformed iPSCs, highlighted by red arrows,
   suggesting a positive feedback loop that amplifies Nrf2 signaling. No
   Nrf2 binding was detected at the promoter region of TXNRD2, indicating
   selective regulation. C Bar graph showing significant upregulation of
   key enzymes involved in glycolysis and glycolytic shunt pathways,
   including the pentose phosphate pathway (PPP) and hexosamine
   biosynthetic pathway (HBP). Key enzymes such as HK2 and TKT
   (glycolysis) and G6PD, PGLS, PGD, GFPT, PGM3, and UAP1 (PPP/HBP) are
   markedly upregulated in As^3+-transformed iPSCs. D ChIP-seq analysis
   shows significant Nrf2 binding and H3K36me3 histone modification at the
   gene bodies and first exons of key metabolic genes in the PPP (G6PD,
   PGLS, PGD) and HBP (GFPT, PGM3, UAP1), supporting transcriptional
   activation through both Nrf2 signaling and histone modification.
   GNPNAT1 in the HBP pathway was not induced.

3.8. Nrf2 promotes cholesterol metabolism and oncogenic stemness in
As^3+-transformed iPSCs

   The pathological role of elevated blood cholesterol in metabolic
   diseases such as atherosclerosis, myocardial infarction, and stroke is
   well-established. More recently, aberrant cholesterol homeostasis has
   also been implicated in cancer progression, with studies linking
   dysregulated cholesterol metabolism to breast cancer metastasis and
   poor prognosis across various malignancies [[163]41]. Intriguingly,
   increased cholesterol efflux associated with hypercholesterolemia has
   been shown to expand hematopoietic stem and progenitor cells, as well
   as leukemia-initiating cells [[164]42]. Our previous study revealed
   that chronic As^3+ exposure reprograms non-cancerous epithelial cells
   into CSCs, characterized by persistently upregulated cholesterol
   metabolism [[165]33]. Consistent with these findings, cholesterol
   metabolism emerged as one of the top 10 enriched pathways in
   As^3+-transformed iPSCs ([166]Figure 6C). RNA-seq data indicated that
   genes involved in every step of cholesterol biosynthesis were
   upregulated following As^3+ treatment, including key regulators of
   sterol regulatory element-binding protein (SREBP) signaling ([167]Fig.
   8A and B, highlighted in red). Moreover, ChIP-seq revealed pronounced
   Nrf2 enrichment at the promoter or gene body regions of several
   cholesterol biosynthetic genes such as ACAT2, HMGCS1, MMAB-MVK, and
   FDPS, supporting a direct transcriptional role of Nrf2 in driving
   cholesterol metabolism in iPSCs ([168]Figure 8C). We previously
   reported that BEAS-2B cells transformed by chronic As^3+ exposure
   exhibit suppressed ROS production, reduced oxidative phosphorylation,
   enhanced glycolysis, and stem-like properties [[169]2,[170]4,[171]43].
   Given that a major function of Nrf2 is to promote antioxidant defense
   and reduce intracellular ROS—a hallmark of CSCs—and that accumulated
   cholesterol can further shift cellular metabolism from oxidative
   phosphorylation to glycolysis [[172]44], it is plausible that
   Nrf2-driven cholesterol accumulation enhances the CSC-like potential of
   iPSCs in response to As^3+. Gene expression profiles of
   As^3+-transformed iPSCs support this hypothesis. Several transcription
   factors essential for oncogenic stemness were markedly upregulated,
   including PRDM14, KLF4, NR6A1, STAT5A, NANOG, and SOX2 ([173]Figure
   8D). In contrast, the expression of MYC, VEGFA, and PDGFB was
   significantly suppressed by As^3+, though the underlying mechanism
   remains unclear. For the upregulated genes, increased Nrf2 occupancy
   may contribute to their transcriptional activation. For example, within
   the STAT5 gene cluster on chromosome 17, As^3+ induced prominent Nrf2
   peaks specifically at STAT5A (highlighted by red arrows), which showed
   elevated expression, whereas neighboring genes STAT5B and STAT3 showed
   no Nrf2 binding and no change in expression ([174]Figure 8E). Another
   notable feature of As^3+-transformed iPSCs is the reprogramming of
   histone trimethylation marks ([175]Figure 2A), likely resulting from
   altered expression of genes encoding chromatin-modifying enzymes. As^3+
   treatment increased the expression of DNA methyltransferases DNMT3B and
   DNMT3L, as well as demethylation-associated enzymes TET and GADD45G,
   while reducing expression of GADD45A and GADD45B ([176]Figure 8F).
   These changes may shift the balance of DNA methylation. It is known
   that DNA hypermethylation is positively associated with histone marks
   H3K9me3, H3K27me3, and H4K20me3, and negatively associated with
   H3K4me3, due to methylation-dependent repulsion of H3K4
   methyltransferases [[177]45]. Thus, the observed reduction of H3K4me3
   and increase in H3K27me3 and H4K20me3 may result from elevated DNMT3B
   and DNMT3L activity in response to As^3+. As^3+ also modulated
   expression of several histone methyltransferases and demethylases,
   including EZH2, SETD4, and the KDM4 family ([178]Figure 8F). However,
   the extent to which these transcriptional changes directly contribute
   to altered histone methylation remains to be elucidated. Notably,
   changes in expression of these epigenetic regulators did not
   consistently correlate with Nrf2 binding, suggesting involvement of
   additional regulatory mechanisms beyond Nrf2.

Fig. 8.

   [179]Fig. 8
   [180]Open in a new tab

   Nrf2 directly regulates cholesterol biosynthesis and oncogenic stemness
   transcriptional programs in As^3+-transformed iPSCs. A Schematic
   representation of the cholesterol biosynthetic pathway, with genes
   upregulated in As^3+-transformed iPSCs highlighted in red. B Bar graph
   showing RNA-seq-based differential expression of key cholesterol
   biosynthetic genes in control and As^3+-transformed iPSCs. Gene
   expression is presented as average log[2] fold change. C ChIP-seq
   tracks illustrating Nrf2 occupancy at promoter or gene body regions of
   representative cholesterol biosynthetic genes—including ACAT2, HMGCS1,
   MMAB-MVK, and FDPS—in control and As^3+-transformed iPSCs. D RNA-seq
   analysis showing upregulation of oncogenic stemness-associated
   transcription factors, including PRDM14, KLF4, NR6A1, STAT5A, NANOG,
   and SOX2, in As^3+-transformed iPSCs. Gene expression is shown as
   average log[2] fold change. E ChIP-seq tracks of the STAT5 gene cluster
   showing specific Nrf2 enrichment at the STAT5A locus (indicated by red
   arrows), which correlates with increased expression. No Nrf2 binding or
   expression changes were observed at neighboring genes STAT5B or STAT3.
   F Pie chart depicting the altered expression profiles of genes involved
   in DNA methylation and histone modification in As^3+-transformed iPSCs
   compared to controls, including upregulated genes such as DNMT3B,
   DNMT3L, TET, GADD45G, EZH2, SETD4, and members of the KDM4 family.

3.9. High expression of cholesterol biosynthesis genes correlates with poorer
survival in lung cancer patients

   While scattered studies have linked genetic polymorphisms in
   cholesterol biosynthesis genes to cancer risk [[181]46,[182]47], the
   prognostic value of these genes remains largely underexplored. To
   evaluate the clinical relevance of our findings, we assessed the
   association between cholesterol biosynthesis gene expression and first
   progression (FP) survival in a cohort of 982 lung cancer patients using
   the Kaplan-Meier Plotter database. Among 15 cholesterol biosynthesis
   genes examined, high expression of nine genes—including SREBF2, ACAT2,
   HMGCS1, MVK, MVD, FDFT1, SQLE, LSS, and DHCR7—was significantly
   associated with shorter FP survival ([183]Figure 9A). Conversely,
   elevated expressions of HMGCR, IDI1, FDPS, and DHCR24 correlated with
   improved FP survival. No survival correlation was observed for SC5DL.
   These findings suggest a general trend wherein upregulation of specific
   cholesterol biosynthesis genes may predispose to more aggressive tumor
   behavior, supporting the pro-oncogenic role of hypercholesterolemia. To
   further underscore this relationship, we evaluated the expression of
   three cytochrome P450 family members involved in cholesterol metabolism
   or catabolism: CYP51A1 that catalyzes conversion of lanosterol to
   cholesterol; CYP11A1, which catalyzes cholesterol side-chain cleavage
   potentially linked to the generation of methylmalonic acid (MMA) for
   stemness factor activation [[184]48]; and CYP27A1, which promotes
   cholesterol excretion via bile acid synthesis ([185]Figure 9B). High
   expressions of CYP51A1 and CYP11A1 were linked to worse FP survival,
   while elevated CYP27A1 expression predicted better survival outcomes
   ([186]Figure 9C). These results reinforce the notion that disrupted
   cholesterol homeostasis—especially increased synthesis and impaired
   clearance—may contribute to tumor progression and poor clinical
   prognosis.

Fig. 9.

   [187]Fig. 9
   [188]Open in a new tab

   Increased expression of cholesterol biosynthesis genes is associated
   with poorer prognosis of lung cancer. A First Progression (Progression
   Free, FP) survival of lung cancer patients (n = 2167; data from the
   Kaplan-Meier Plotter) stratified by expression levels of key
   cholesterol biosynthesis genes. The optimal cutoff was used to define
   high and low expression groups for survival comparison. B Schematic
   diagram showing contribution of the indicated CYP family members to
   cholesterol metabolism. C FP survival curves of lung cancer patients
   stratified by expression levels of the indicated CYP genes, based on
   Kaplan–Meier analysis.

4. Discussion

   The association between environmental As^3+ exposure and human cancers
   is well-established. A key mechanism underlying As^3+ carcinogenicity
   is its ability to induce cancer stem-like cell (CSC) formation, either
   by reprogramming normal stem cells or by generating CSCs de novo
   [[189]2,[190]49]. In this study, we used human iPSCs to investigate the
   impact of chronic As^3+ exposure on metabolic and epigenetic programs.
   Our findings suggest that the transcription factor Nrf2 may contribute
   to these reprogramming events, particularly through the modulation of
   glycolytic and cholesterol biosynthesis pathways. Building on previous
   work demonstrating that As^3+ activates Nrf2 and induces a metabolic
   shift from mitochondrial oxidative phosphorylation to glycolysis
   [[191]5], we observed robust upregulation of glycolytic genes in iPSCs
   following long-term As^3+ exposure. Genome-wide Nrf2 binding was
   enriched at the loci of these genes. Interestingly, whereas
   As^3+-transformed bronchial epithelial cells redirected glycolytic
   intermediates toward the hexosamine biosynthetic and serine/glycine
   synthesis pathways [[192]5], transformed iPSCs preferentially funneled
   these metabolites into the pentose phosphate pathway (PPP) and the
   hexosamine biosynthetic pathway (HBP). We also observed upregulation of
   genes involved in cholesterol biosynthesis, which aligns with our
   recent metabolomic data and further implicates cholesterol metabolism
   as a conserved response to chronic As^3+ stress [[193]33].

   Epigenetic remodeling—particularly alterations in histone
   methylation—represents a hallmark of both stem cells and CSCs. While
   the exact influence of As^3+ on epigenetic regulation remains
   incompletely understood, our data reveal a surprising global reduction
   in H3K4me3 levels and concurrent increases in H3K27me3, H3K36me3, and
   H4K20me3 in iPSCs following As^3+ exposure. H3K4me3, typically
   associated with transcriptional activation, was still enriched at
   transcription start sites (TSS), albeit at lower levels. This reduction
   may be attributed to increased expression of DNA methyltransferases
   such as DNMT3B and DNMT3L, which were upregulated by As^3+.
   Interestingly, colocalization patterns revealed enrichment of H3K4me3,
   H3K27me3, H3K36me3, and Nrf2 within active chromatin regions for a
   number of genes, suggesting coordinated regulation of gene expression.
   These epigenetic markers showed minimal overlap with repressive marks
   such as H3K9me3 and H4K20me3, which were largely confined to
   heterochromatin or gene-poor regions.

   One of the most striking findings of this study is the robust
   activation of cholesterol biosynthesis in iPSCs following As^3+
   exposure. This mirrors our previous observations in As^3+-transformed
   bronchial epithelial cells and aligns with emerging evidence linking
   environmental As^3+ exposure to elevated serum cholesterol levels in
   humans and cardiovascular pathology in animal models
   [[194]33,[195]50,[196]51]. While the direct contribution of cholesterol
   to As^3+-induced carcinogenesis remains uncertain, several studies
   suggest a role for cholesterol as a pro-oncogenic metabolite. For
   example, hypercholesterolemia has been correlated with increased cancer
   risk and progression [[197]52]. Statins, cholesterol-lowering drugs,
   have been shown to reduce recurrence and mortality in breast cancer
   patients [[198]53,[199]54]. In triple-negative breast cancer (TNBC),
   metastatic tumors exploit SREBP2-driven cholesterol synthesis to
   support angiogenesis [[200]55]. Estrogen receptor-positive (ER^+)
   breast cancers overexpress CYP27A1, converting cholesterol into
   27-hydroxycholesterol (27HC), which mimics estrogen to promote tumor
   growth and metastasis [[201]56,[202]57]. 27HC also modulates immune
   evasion by recruiting immunosuppressive myeloid populations and
   depleting CD8^+ T cells [[203]57]. Notably, CYP27A1 was the most highly
   expressed cholesterol metabolism gene in As^3+-transformed iPSCs
   ([204]Figure 8B). Cholesterol also promotes survival in glioblastoma,
   and its depletion leads to apoptosis via liver X receptor (LXR)
   signaling [[205]58]. In diffuse large B-cell lymphoma (DLBCL),
   cholesterol enhances B-cell receptor (BCR) clustering and PI3K/AKT
   signaling [[206]59]. Moreover, cholesterol derived from tumor cells can
   induce CD8^+ T-cell exhaustion, facilitating immune evasion
   [[207]60,[208]61].

   Whether iPSCs and CSCs actively produce cholesterol and how this
   impacts their self-renewal and differentiation remain open questions.
   Previous studies have highlighted a role for cholesterol in stem cell
   biology. For instance, Gu et al. demonstrated that hypercholesterolemia
   drives hematopoietic stem and progenitor cell (HSPC) expansion via
   SREBP2 [[209]62]. In zebrafish, cholesterol efflux activates
   endothelial SREBP2 and Notch signaling to promote HSPC emergence during
   development. Similarly in mammalian systems, cholesterol boosts
   intestinal stem cell (ISC) proliferation, as shown in mouse models
   where high-cholesterol diets increase ISC numbers and promote
   tumorigenesis [[210]63]. Meta-analyses have further linked
   hypercholesterolemia with clonal hematopoiesis and an increased risk of
   hematologic malignancy [[211]64]. In our study, both As^3+-induced CSCs
   and iPSCs exhibited enhanced Nrf2 binding to key cholesterol metabolic
   genes [[212]5], along with a marked upregulation of cholesterol
   metabolism, suggesting that cholesterol serves as a critical metabolic
   node supporting stemness. While the full extent of Nrf2's role in
   regulating cholesterol metabolism remains to be fully recognized
   currently, previous studies in LDLR^-/- mice implicate Nrf2 in
   cholesterol-driven pathologies. In the context of atherosclerosis, a
   disease characterized by excessive cholesterol accumulation in arterial
   walls, Nrf2 deficiency led to a reduction in atherosclerotic lesions
   [[213]65].

   Overall, our findings identify a novel mechanism by which As^3+
   enhances oncogenic stemness through Nrf2-mediated metabolic
   reprogramming. This rewiring features heightened PPP and HBP activity,
   supplying key metabolites for nucleotide biosynthesis, glycosylation,
   and methylation. For example, S-adenosylmethionine (SAM), derived from
   the methionine cycle linked to PPP, serves as the universal methyl
   donor for DNA and histone methylation. UDP-GlcNAc, produced via HBP,
   supports glycosylation of stemness-related transcription factors such
   as OCT4, SOX2, and MYC, as well as epigenetic regulators like TET
   enzymes and H3K4 methyltransferase complexes [[214][66], [215][67],
   [216][68]]. SCAP glycosylation is also essential for SREBP activation,
   enabling Nrf2-SREBP synergy in promoting cholesterol biosynthesis
   [[217]40]. In the nucleus, SREBP acts in concert with Nrf2 and other
   transcription factors to drive the transcription of genes involved in
   cholesterol biosynthesis. Once synthesized, cholesterol integrates into
   both the plasma membrane and endoplasmic reticulum (ER) membrane, where
   it facilitates the assembly and activation of membrane-associated
   growth signaling pathways—including B-cell receptor (BCR), Notch, and
   Hedgehog (Hh)—which are critical for promoting metabolic and epigenetic
   reprogramming, as well as the proliferation of stem cells and CSCs
   [[218]69]. Notably, cholesterol metabolites can directly modulate
   epigenetic machinery. For instance, the nuclear metabolite
   25-hydroxycholesterol (25HC) has been shown to act as a direct
   activator of DNMT1, enhancing DNA methylation [[219]70]. In contrast,
   sulfated cholesterol derivatives such as 25HC3-sulfate (25HC3S) and
   27HC3S function as potent inhibitors of DNMT1, DNMT3A, and DNMT3B
   [[220]71]. Moreover, an earlier study demonstrated that cholesterol
   could upregulate DNMT1 expression and promote the DNA-binding activity
   of methyl-CpG binding protein 2 (MeCP2) and the histone
   methyltransferase EZH2 [[221]72].

   While our data support a potential role for Nrf2 in orchestrating these
   interconnected pathways, we cannot rule out the contribution of other
   regulatory mechanisms beyond Nrf2 for the As^3+-induced metabolic
   reprogramming. Moreover, neither Nrf2 nor As^3+ can be equalized to an
   independent reprogramming factor capable of inducing pluripotency in
   somatic cells. Instead, our findings reveal a previously unrecognized
   mechanism involved in As^3+-induced carcinogenesis, consistent with
   prior reports that Nrf2-dependent metabolic shift improves the
   efficiency of iPSC reprogramming from human dermal fibroblast cells
   [[222]73]. Indeed, the supportive role of Nrf2 to the self-renewal,
   pluripotency, and survival has been widely reported across various stem
   cells and progenitor cell types, including human embryonic stem cells,
   mouse hematopoietic stem cells, human mesenchymal stem cells, human
   endothelial progenitor cells, and mouse neural stem cells, among others
   [[223]74]. Although complete disruption of the Nrf2 gene NFE2L2 through
   CRISPR-Cas9 gene editing in the iPSCs used in this study may pose
   challenges due to severe cell death, which precludes further analysis
   of As^3+- or other carcinogen-induced metabolic reprogramming and
   malignant transformation, it is imperative to pursue alternative
   loss-of-function approaches to clarify the specific contributions of
   Nrf2 and its downstream effectors in this process. Nonetheless, our
   findings suggest that As^3+ exposure, at least in part through
   Nrf2-associated metabolic rewiring, engages a complex network of
   interrelated pathways, including glycolysis, nucleotide biosynthesis,
   protein glycosylation, cholesterol metabolism, and epigenetic
   regulation, that collectively promote cancer development and the
   emergence of CSCs ([224]Figure 10).

Fig. 10.

   [225]Fig. 10
   [226]Open in a new tab

   Schematic representation of potential mechanisms underlying
   Nrf2-mediated metabolic rewiring in response to As^3+exposurein iPSCs.
   As^3+ induces Nrf2-dependent metabolic reprogramming that coordinates a
   network of interconnected pathways—including enhanced glycolysis,
   nucleotide biosynthesis, protein glycosylation, cholesterol metabolism,
   epigenetic modifications, and weakened mitochondrial citric acid cycle
   (TCA or Krebs cycle). Together, these alterations contribute to cancer
   development and promote the acquisition of CSC properties.

CRediT authorship contribution statement

   Akimasa Seno: Methodology, Investigation, Data curation. Zhuoyue Bi:
   Methodology, Investigation, Formal analysis. Lisa Polin: Methodology,
   Investigation. Ziqi Liu: Validation, Software, Resources. Yiran Qiu:
   Methodology, Investigation, Data curation. Wenxuan Zhang: Validation,
   Methodology, Data curation. Aashna Pawar: Software, Investigation.
   Chitra Thakur: Resources, Investigation, Data curation. Masaharu Seno:
   Resources. Ziwei Wang: Writing – original draft, Visualization,
   Validation, Methodology, Investigation. Fei Chen: Writing – review &
   editing, Supervision, Funding acquisition, Conceptualization.

Declaration of competing interest

   The authors declare no competing financial interests or personal
   relationships that could have appeared to influence the work reported
   in this paper.

Acknowledgements