Epigenetic reprogramming of lineage switching in cancer

Abstract

Lineage plasticity develops as tumor cells overcome terminal differentiation barriers, adapt to environmental stressors, and acquire metastatic, therapy‐resistant phenotypes. These phenotypic transitions are driven by nonmutational epigenetic mechanisms, including dynamic shifts in chromatin accessibility, histone modifications, DNA methylation, and noncoding RNA activity beyond genetic alterations. While epithelial–mesenchymal plasticity (EMP) remains a foundational model, cancer‐cell plasticity extends into a broader pan‐plasticity spectrum encompassing dedifferentiation, transdifferentiation into neuroendocrine or squamous lineages, and stem cell–like reprogramming. In this review, we evaluate how the antagonistic interplay between Polycomb and Trithorax complexes, chromatin modifier activity, and enhancer reprogramming affects lineage identity and enables multi‐lineage plasticity in cancer. We further discuss extrinsic mechanisms reinforced by the tumor microenvironment, particularly hypoxia, chronic inflammation, and metabolic stress, that rewire the epigenetic landscape to stabilize plastic states. Collectively, these data underscore epigenetic deregulation as a primary driver of lineage plasticity and intratumoral heterogeneity, while revealing therapeutic vulnerabilities and the potential to reverse plasticity and overcome therapy resistance by targeting the epigenome.

Cancer cells rarely commit to a single identity. Epigenetic mechanisms and tumor microenvironment cues push epithelial cells toward flexible, hybrid states that can shift into mesenchymal, neuroendocrine, or stem‐like fates, driving metastasis, drug resistance, and tumor heterogeneity. Targeting the epigenetic regulators behind these transitions, using EZH2, BET, HDAC, LSD1, or DNMT1 inhibitors, may restore therapeutic vulnerability and anti‐tumor immune recognition. Figure partially created in BioRender. Boyvatlı, E. (2026) https://BioRender.com/4yerlz4

Abbreviations

5mC, 5‐methylcytosine

BET, bromodomain and extra‐terminal

CAF, cancer‐associated fibroblast

DNMT, DNA methyltransferase

DTP, drug‐tolerant persister

ECM, extracellular matrix

EMP, epithelial–mesenchymal plasticity

EMT, epithelial–mesenchymal transition

EMT‐TF, EMT‐associated transcription factor

E–NE, epithelial‐to‐neuroendocrine

H3K18la, H3K18 lactylation

HDAC, histone deacetylase

HERV, human endogenous retrovirus

ICB, immune checkpoint blockade

lncRNA, long noncoding RNA

miRNA, microRNA

NED, neuroendocrine differentiation

NEPC, neuroendocrine prostate cancer

NK, natural killer

PRC1, Polycomb repressive complex 1

PRC2, Polycomb repressive complex 2

ROS, reactive oxygen species

SAM, S‐adenosylmethionine

SCLC, small‐cell lung cancer

TAM, tumor‐associated macrophage

TE, transposable element

TME, tumor microenvironment

TNBC, triple‐negative breast cancer

TrxG, Trithorax group

Introduction

Cell differentiation and lineage specification were once thought to be one‐way and irreversible, but it is now understood that cell identity is more plastic than previously assumed. This plasticity is essential for restoring homeostasis, but it can also contribute to tumor development [1, 2]. Cancer cells rarely maintain a single identity; instead, they exist along a spectrum of differentiation programs. They can sometimes exhibit multi‐lineage fates or evolve into entirely new histological identities without necessarily acquiring new driver mutations [3, 4]. This characteristic is known as lineage plasticity and has recently been recognized as a hallmark of cancer. This capability plays a central role in metastasis, treatment resistance, and disease recurrence by enabling cells to rapidly adopt a range of states or phenotypes in response to environmental and therapeutic pressures [5]. Cellular plasticity is fundamentally an epigenetically driven phenomenon that tumors exploit across diverse contexts, from endocrine therapy–resistant prostate carcinoma to targeted therapy–resistant lung adenocarcinoma [6, 7].

Epithelial–mesenchymal plasticity (EMP) represents one of the most extensively characterized forms of lineage plasticity in cancer. Classical epithelial–mesenchymal transition (EMT) was initially defined as a binary transition in which cells lose epithelial characteristics and acquire mesenchymal features [8, 9]. However, in vivo lineage tracing and single‐cell profiling now reveal a spectrum of partial or hybrid EMT states, collectively termed EMP [10, 11]. Importantly, EMP represents only one axis of a broader pan‐plasticity spectrum. Here, we use the term pan‐plasticity to describe the broader ability of cancer cells to move across multiple phenotypic and lineage states beyond the epithelial–mesenchymal axis, including neuroendocrine, squamous, endothelial‐like, dedifferentiated, and stem‐like states [12, 13, 14].

Advances in epigenomic profiling and single‐cell technologies have demonstrated that phenotypic switching is frequently driven by chromatin remodeling rather than irreversible genetic alterations. Epigenetic mechanisms allow cancer cells to rapidly rewire transcriptional programs, enabling adaptation on timescales that far exceed those of clonal evolution [15, 16]. This adaptability is further reinforced by the tumor microenvironment (TME) and metabolic perturbations. Factors such as hypoxia, chronic inflammation, and accumulation of oncometabolites can lock or unlock specific epigenetic states. Rather than being a passive consequence of tumor heterogeneity, cellular plasticity is an active survival mechanism regulated by cancer cells to escape selective pressures imposed by treatment and the TME.

In this review, we evaluate how epigenetic reprogramming regulates lineage plasticity in cancer. We first discuss EMP as a model of plasticity and place it within the broader framework of pan‐plasticity. We then examine the major epigenetic mechanisms that govern lineage switching, including histone modifications, DNA methylation, noncoding RNAs, chromatin remodelers, and enhancer reprogramming. Finally, we discuss how tumor microenvironmental signals reinforce plastic cell states and how targeting the epigenome may provide therapeutic opportunities to limit plasticity and overcome therapy resistance.

From EMP to pan‐plasticity

Our understanding of cancer‐cell plasticity has considerably expanded over the past decade. Accumulating evidence indicates that EMP represents only one dimension of a much broader landscape of epigenetically regulated cell‐state plasticity. Beyond movement along the epithelial–mesenchymal continuum, cancer cells can reversibly acquire multiple developmental and lineage‐associated identities in response to genetic alterations, epigenetic remodeling, and selective pressures imposed by the TME or therapeutic interventions. Here, we use the term pan‐plasticity to describe this wider landscape of reversible or semi‐stable phenotypic transitions, including EMP, lineage switching, transdifferentiation, dedifferentiation, blocked differentiation, metaplasia, and transdetermination [12, 17].

The recognition that EMP represents only one axis of pan‐plasticity has shifted attention from individual transcription factors toward the epigenetic mechanisms that govern cellular identity. Rather than functioning as simple binary switches, EMT‐associated transcription factors (EMT‐TFs), including SNAIL, ZEB, and TWIST, act as coordinators of chromatin remodeling. By interacting with histone deacetylases (HDACs), Polycomb repressive complexes (PRC2), and DNA methylation machinery, these factors help establish chromatin states that simultaneously repress epithelial programs while permitting the activation of alternative transcriptional identities [18, 19, 20]. Consequently, plasticity is increasingly recognized as an epigenetically coordinated process that reshapes cellular identity and expands phenotypic competence, rather than as a transcription factor‐driven event alone [9, 10, 17].

The diversity of pan‐plasticity is reflected in the wide range of lineage transitions observed across human cancers. Epithelial‐to‐neuroendocrine (E–NE) transitions are well characterized in prostate and lung cancers, where they correlate strongly with therapy resistance and aggressive disease progression. These switches are frequently driven by combined genetic alterations, such as loss of TP53 and RB1 [21, 22]. Similarly, epithelial‐to‐squamous transitions represent a recurrent mechanism of lineage switching across multiple tumor types, often associated with large‐scale chromatin remodeling and transcriptional reprogramming [23]. In parallel, epithelial‐to‐endothelial–like transitions facilitate tumor vascularization through vasculogenic mimicry, enabling cancer cells to sustain perfusion independently of canonical angiogenesis pathways [24]. Collectively, these examples illustrate that cancer cells are capable of navigating multiple lineage trajectories that extend well beyond the classical epithelial–mesenchymal continuum.

Despite their phenotypic diversity, these lineage transitions converge on common epigenetic regulatory networks. Large‐scale pan‐cancer studies have identified conserved transcriptional signatures associated with these plastic tumor states [25]. Among these, dysregulation of HOX gene expression (often referred to as altered ‘HOX codes’) is linked to developmental reprogramming and lineage infidelity [26]. RUNX1T1, a transcriptional corepressor belonging to the MTG family, recruits histone deacetylase complexes to suppress downstream target genes [27]. Integrative transcriptomic analyses have further identified RUNX1T1 as a recurrent regulator within these networks, which interacts with histone deacetylase complexes to modulate HOX‐driven transcriptional programs and maintain lineage flexibility [28].

Not all tumor cells are equally plastic. Instead, pan‐plasticity appears to arise from subsets of cells that are epigenetically primed for phenotypic switching. Single‐cell multi‐omics analyses reveal that these cells exhibit open chromatin configurations and transcriptional signatures associated with stemness and developmental potency. Importantly, these epigenetically primed states may emerge prior to overt phenotypic transitions, allowing for rapid activation of alternative lineage programs under selective pressures, such as targeted therapies [3, 29]. Rather than representing terminal differentiation states, these highly plastic populations constitute dynamic reservoirs from which diverse adaptive phenotypes can emerge, thereby promoting tumor evolution, metastatic dissemination, and therapeutic resistance.

Epigenetic mechanisms underlying lineage plasticity

One key feature of cellular transitions is that they typically occur without new genetic changes. This highlights epigenetic reprogramming as the underlying regulatory mechanism of lineage plasticity [15, 22]. Growth factors and inflammatory signals in the TME (e.g., TGF‐β, EGF, and WNT) activate signaling pathways through transcription regulators, initiating plastic responses [30, 31, 32]. The transcription factor binding to regulatory DNA elements is dynamic and binding‐dependent. This results in many binding events being short‐lived, and transcription factors alone are insufficient to establish stable transcriptional states. On the contrary, epigenetic mechanisms such as DNA methylation and histone modifications make transcriptional states long but reversible gene expression programs [33, 34]. Here, we describe the major classes of epigenetic mechanisms that govern gene expression and regulate lineage plasticity in cancer: Histone modifications and bivalent chromatin, DNA methylation, RNA interference, chromatin remodelers, and enhancer reprogramming (Fig. 1).

Fig. 1.

Epigenetic mechanisms underlying lineage plasticity in cancer. Cancer cells exploit a broad repertoire of epigenetic mechanisms to remodel their transcriptional landscape and enable lineage plasticity. DNA methylation at CpG sites, histone modifications including acetylation and tri‐methylation, and noncoding RNA‐mediated gene silencing collectively regulate chromatin accessibility and gene expression programs. ATP‐dependent chromatin‐remodeling complexes, such as SWI/SNF reposition nucleosomes to expose regulatory elements, while transcription factor‐driven enhancer reprogramming, including the formation of super‐enhancer domains, establishes lineage‐specific gene expression networks. Together, these epigenetic alterations destabilize epithelial identity and drive transitions through a hybrid state, ultimately enabling cancer cells to acquire mesenchymal, cancer stem cell, or neuroendocrine identities. Figure partially created in BioRender. Boyvatlı, E. (2026) https://BioRender.com/4qds54m.

Histone modifications and bivalent chromatin: A balance between Polycomb and Trithorax

Histone modifications represent an important epigenetic mechanism that regulates gene expression by altering chromatin structure and accessibility without changing the DNA sequence. Post‐translational modifications of histone tails, including methylation, acetylation, phosphorylation, and ubiquitination, modulate the interaction between DNA and histones, thereby influencing transcriptional activity. Histone modifications are highly dynamic and rapidly and reversibly regulate gene expression programs that play a role in cell identity and cell plasticity [35, 36, 37]. In general, histone acetylation marks, such as H3K27ac at active promoters and enhancers, and activating methylation marks, including H3K4me3 at promoters and H3K36me3 within actively transcribed gene bodies, promote gene expression, whereas repressive methylation marks, including H3K27me3 and H3K9me3, contribute to transcriptional silencing [38, 39, 40]. These modifications are installed, interpreted, and removed by three classes of enzymes: writers, which deposit histone marks; erasers, which remove them; and readers, which recognize specific modifications and recruit downstream regulatory complexes [41].

One of the best‐characterized mechanisms linking histone modifications to lineage plasticity is chromatin bivalency. Bivalent chromatin refers to the coexistence of the activating H3K4me3 and repressive H3K27me3 marks at promoters of developmental and lineage‐regulatory genes. These genes are typically transcriptionally silent or expressed at low levels but remain poised for rapid activation or stable repression in response to developmental, environmental, or oncogenic cues [42]. In normal development, chromatin bivalency is enriched in elements that regulate the expression of lineage‐specific TFs and other developmental factors, enabling dynamic and rapid control of cell state. During lineage plasticity, changes in bivalent chromatin help activate new alternative lineage programs while repressing genes that maintain the original cell identity [42, 43].

At the molecular level, this balance is governed by antagonistic activities of the Polycomb group (PRC1 and PRC2) and Trithorax group (TrxG) proteins. TrxG complexes catalyze activating marks, such as H3K4me3. Within Polycomb‐mediated repression, PRC2 catalyzes H3K27me3 through EZH2/EZH1 methyltransferase activity, whereas PRC1 recognizes Polycomb‐marked chromatin and catalyzes H2AK119ub, contributing to chromatin compaction and transcriptional repression [40, 44, 45].

In cancer, this poised chromatin architecture is frequently exploited to enable EMP, a dynamic process underlying metastasis and therapy resistance. Bivalent chromatin domains at key regulators of epithelial and mesenchymal identity, including EMT‐associated transcription factors, such as ZEB1, TWIST, and SNAI, allow cells to reversibly transition between cellular states without genetic alterations [3, 46]. Shifts in this balance between Polycomb‐mediated repression and Trithorax‐regulated activation can bias transcriptional programs toward epithelial or mesenchymal identities, enabling hybrid EMT states [18]. These intermediate phenotypes are now recognized as critical determinants of tumor heterogeneity, invasiveness, and therapeutic adaptation [14, 47, 48, 49]. Accordingly, the Polycomb–Trithorax axis emerges as a core epigenetic framework linking bivalent chromatin to EMP [42].

The study of SW480 colon cancer cells using a TGF‐β‐induced EMT model provides a clear example of how plasticity is shaped by histone modifications. Following TGF‐β induction, EZH2, the catalytic subunit of PRC2, binds to the WNT5A promoter, increasing H3K27me3 accumulation, thereby leading to the transcriptional silencing of WNT5A, loss of E‐cadherin, and reduced adhesion [18]. An EZH2 inhibitor reduced H3K27me3 levels at the WNT5A locus, partially restoring WNT5A/E‐cadherin expression and cell adhesion. These results demonstrate that H3K27me3 acts as a reversible repressive mark supporting EMT‐associated phenotypic plasticity [18]. This finding supports the concept that PRC2‐mediated H3K27me3 deposition may function as a dynamic epigenetic barrier that regulates the stability of the epithelial lineage during plastic state transitions. Similarly, histone modification switching has been shown to promote EMP in gastric cancer. In this context, the long noncoding RNA HOTAIR acts as a key epigenetic regulator promoting EMT. HOTAIR controls cellular plasticity through a ‘histone mark switch’ mechanism that mediates the transition from H3K27ac to H3K27me3 at the E‐cadherin promoter [50]. By recruiting PRC2 (EZH2/SUZ12/EED) to chromatin, HOTAIR suppresses H3K27 acetylation, shifting the promoter from a transcriptionally active state to a repressive state, thereby silencing E‐cadherin expression and enhancing EMT and metastatic behavior [51].

Beyond epithelial plasticity, histone modification‐driven chromatin reprogramming also contributes to lineage switching events involving neuroendocrine differentiation (NED). In genetically engineered mouse models and human prostate cancer transcriptomes, overexpression of N‐Myc was observed following AR‐targeted therapy in prostate adenocarcinoma. Following its overexpression, N‐Myc has been identified as a key driver that both initiates and sustains the transition from adenocarcinoma to NEPC by epigenetically silencing the AR signal (via EZH2/PRC2 and H3K27me3) and activating neuroendocrine, EMT, and stem cell–like programs [21].

DNA methylation

DNA methylation is an important mechanism for stabilizing transcriptional states; however, in cancer, it is controlled in a dynamic and context‐dependent manner that supports cellular plasticity and remains reversible. DNA methylation occurs through the transfer of a methyl group from S‐adenosylmethionine (SAM) to the fifth carbon of a cytosine residue, catalyzed by a family of DNA methyltransferases (DNMTs), resulting in the formation of 5‐methylcytosine (5mC) [52, 53]. Rather than permanent gene silencing, restricted genes are frequently regulated through dynamic promoter or enhancer methylation, allowing reversible transitions between transcriptional states [54].

In the context of EMP, studies have shown that coordinated methylation remodeling occurs during EMT‐like transitions. One of the best‐characterized examples linking DNA methylation to lineage plasticity is the hypermethylation of the promoter region of the CDH1 gene during EMP. Hypermethylation of the CDH1 promoter is associated with loss of epithelial identity and repression of E‐cadherin expression, a key event in EMT‐associated state transitions. Mechanistically, EMT‐associated transcription factors actively participate in this process: ZEB1 has been shown to recruit DNMT1 to the CDH1 promoter, thereby maintaining promoter hypermethylation and reinforcing transcriptional repression beyond transient transcription factor binding. Through this coupling of transcriptional repression with epigenetic modification, EMT‐TFs translate short‐lived signaling inputs into more stable, yet reversible, epigenetic states [55, 56, 57]. In another study, it was determined that DNA hypermethylation in some genes in breast epithelial cells contributes to breast cancer progression by suppressing these genes, thereby promoting EMT, invasion, and tumor cell plasticity [58]. In addition to breast cancer, it has been shown that the suppression of differentiation‐related genes via DNA methylation in colorectal cancer contributes to EMP and metastatic progression. For example, hypermethylation of the ANGPTL4 promoter leads to transcriptional silencing of this epithelial‐associated regulator, thereby facilitating activation of EMT programs and promoting invasive behavior in tumor cells [59]. In another study, DNA methylation remodeling in hepatocellular carcinoma was associated with activation of EMT programs and the gain of invasive phenotypes. While demonstrating that methylation‐mediated suppression of epithelial differentiation genes contributes to sorafenib resistance and promotes metastatic behavior, the study revealed that pharmacologically or genetically modulating methylation could partially restore drug sensitivity [60]. These studies demonstrate that DNA methylation‐dependent suppression of epithelial differentiation programs is a common mechanism underlying EMP across various tumor types.

Beyond EMP, reconfiguration of DNA methylome plays a role in lineage‐reprogramming events, such as NED. For instance, DNA methylation mediated by DNA methyltransferases (particularly DNMT1/3A) sustains lineage plasticity and an aggressive phenotype in castration‐resistant prostate adenocarcinoma and neuroendocrine prostate cancer, thereby promoting therapy resistance and aggressive disease evolution [61]. Similarly, studies on small‐cell lung cancer have shown that DNMT1‐mediated DNA methylation keeps tumor cells undifferentiated and proliferating. Removing this epigenetic constraint through DNMT1 inhibition induces terminal NED. It suppresses tumor growth via a noncytotoxic mechanism, underscoring the role of methylation as a key regulator of cell lineage stability and plasticity [62, 63, 64].

Collectively, these findings highlight that, rather than serving solely as a stable silencing mechanism, DNA methylation acts as a dynamic regulator of cell identity, enabling reversible reprogramming during tumor progression.

RNA interference

Noncoding RNAs constitute a critical post‐transcriptional and epigenetic regulatory layer that shapes lineage identity and plasticity without altering the underlying DNA sequence. Unlike histone modifications and DNA methylation, which directly remodel chromatin architecture, RNA‐based mechanisms primarily operate by modulating the stability, translation, and chromatin recruitment of key lineage regulators, enabling rapid and reversible control of cell‐state transitions [65, 66, 67].

MicroRNAs (miRNAs) are short (~22 nucleotides) noncoding RNAs that are primarily responsible for silencing mRNA translation by directly interacting with transcription factors or their associated cofactors. miRNAs influence many cellular processes, including cell proliferation, adhesion, cell death, and differentiation. They are particularly important in cancer treatment because they directly silence genes involved in plasticity [68]. The miR‐200 family provides the best‐characterized example: its members strengthen epithelial identity by directly suppressing ZEB1 and ZEB2 mRNA, while ZEB1 reciprocally suppresses miR‐200 gene transcription, creating a double‐negative feedback loop that confers bistability to the epithelial–mesenchymal spectrum [69, 70]. A parallel feedback loop is observed in the binding of miR‐34a/b/c to SNAIL1. While miR‐34 directly targets SNAIL1 mRNA, SNAIL1 transcriptionally represses miR‐34 promoters; together, these two determine the threshold at which cells decide to undergo mesenchymal transition [71]. These interconnected loops reveal that miRNA‐mediated regulation not only responds to lineage signals but also actively encodes and maintains plastic cell‐state boundaries.

Long noncoding RNAs (lncRNAs) are noncoding RNAs that are longer than 200 nucleotides and regulate gene expression at epigenetic, transcriptional, and post‐transcriptional levels through interactions with DNA, proteins, mRNAs, and miRNAs [66, 72]. A central mechanism through which lncRNAs exert epigenetic control is the recruitment of chromatin‐modifying complexes to target loci [73]. HOTAIR exemplifies this principle: acting as a molecular scaffold, it recruits PRC2 to EMT gene promoters to enforce H3K27me3‐mediated repression, while simultaneously engaging the LSD1/REST/CoREST complex to catalyze H3K4me2 demethylation, thereby establishing a coordinated chromatin‐repressive mechanism that reinforces the mesenchymal state [74].

In the context of NED, lncRNAs similarly function as epigenetic scaffolds that redirect chromatin‐remodeling activity toward lineage‐specific loci. lncRNA H19 is significantly overexpressed in NEPC, drives NE gene expression in vitro, and its knockdown reverses the NE phenotype and restores enzalutamide sensitivity; mechanistically, H19 modulates EZH2‐dependent H3K27me3 deposition at lineage‐regulatory loci through direct PRC2 interaction, coupling lncRNA activity to the Polycomb‐mediated mechanisms [75]. Similarly, lncRNA‐p21, upregulated by enzalutamide treatment, promotes NED by redirecting EZH2 from its canonical histone methyltransferase function toward nonhistone substrate activity, thereby reprogramming AR‐target loci [76]. Together, these findings position ncRNA‐mediated chromatin control not as a peripheral regulatory layer, but as mechanistically integrated with the epigenetic state transitions that enforce or dissolve lineage boundaries in cancer.

Chromatin remodelers in lineage switching

Chromatin remodelers are ATP‐dependent molecular machineries which can reposition, slide and evict nucleosomes, playing critical roles in gene regulation and genome maintenance. Chromatin‐remodeler complexes are broadly categorized into four classes as SWI/SNF, CHD, INO80, and ISWI families. Although each of these remodelers might have slightly different actions on nucleosomes, their proper action is required for the maintenance of cell type‐specific gene expression programs [77].

The role of chromatin remodelers in maintaining developmental plasticity is well established, providing a conceptual framework to understanding their roles in tumor plasticity. It has been identified that loss of Arid1a in mice results in increased regeneration and repair capacity in the liver [78]. In mouse models, it was also determined that loss of Arid2 disrupts the differentiation capability of hematopoietic stem cells [79]. The proper function of Chd1 has been shown to be required for maintenance of the pluripotent state of mouse embryonic stem cells [80]. Additionally, it was demonstrated that the INO80 complex acts together with pluripotency factors to keep the promoters of pluripotency gene promoters in an active configuration [81]. The esBAF complex, a specialized SWI/SNF chromatin‐remodeling complex in embryonic stem cells, has been identified as a critical regulator of pluripotency through its functional relationship with Polycomb complexes [82].

Deregulation or mutation of chromatin remodelers has detrimental effects on cancer‐cell plasticity as well. It was shown that loss of SMARCB1, a member of the SWI/SNF remodeling complex, results in global epigenetic reprogramming, disruption of kidney‐specific programs and gives rise to molecular features associated with NED [83]. Disruption of SWI/SNF complex members in lung cancer gave rise to re‐sensitization of the cells to EGFR inhibitor osimertinib via reprogramming chromatin accessibility at genes involved in epithelial differentiation and EMT [84]. It has been further identified that the SWI/SNF remodeler complex regulates transdifferentiation of prostate adenocarcinoma to a neuroendocrine phenotype via switching subunit composition and associating with different lineage‐specific transcription factors [85]. Downregulation of CHD4 and also MBD3, which are both components of the NuRD complex, in ovarian cancer cells resulted in reversible, chemoresistant quiescent state [86]. Mutations in CHD4 have also been identified to activate the TGF‐β signaling pathway and upregulate the expression of cancer stemness genes in endometrial cancer [87]. All these examples illustrate the deregulation of chromatin‐remodeling machinery in cancer lineage transitions and reveal the potential therapeutic vulnerabilities in this direction.

Enhancer reprogramming in lineage plasticity

Enhancers are distal regulatory elements which bind specific transcription factors and function in establishing and maintaining lineage identities. A subset of enhancers, termed super‐enhancers, contain binding sites for lineage‐specific and pioneering transcription factors, critical for stem cell plasticity, and thus specific lineage choices [88]. Given this critical role of enhancers in gene regulation and maintenance of cellular identities, enhancer reprogramming has emerged as a factor that substantially contributes to tumorigenesis [89]. Enhancer reprogramming can occur via several mechanisms, including mutations, structural variations and changes in enhancer accessibility. Rewired enhancer landscapes may provide unlimited proliferation potential to cells, affect metabolic pathways, activate metastasis, and drive drug resistance [89, 90].

Enhancer reprogramming enables lineage switching by redistributing enhancer activity to adopt alternative lineage‐specific transcriptional programs. It has been identified that FOXA1 behaves as a master transcription factor for the expression of neuroendocrine‐lineage‐specific genes via differential binding at epithelial versus neuroendocrine‐linked regulatory elements [91]. Within the frame of lineage identity in prostate cancer, it was recently identified that the transcription factor NKX2‐1 could recruit chromatin modifiers involved in active chromatin organization and rewire the enhancer landscape resulting in luminal to NED [92]. Regarding the regulatory landscape of neuroblastoma, Banerjee et al. identified that ATRA treatment results in shifts in super‐enhancer landscapes: while stem cell specific SEs were downregulated, neuronal specific SEs were upregulated, reinforcing the differentiation program [93]. In breast cancer, it was shown that differential function of the two transcription factors GATA3 and AP1 could reprogram the enhancer landscape, resulting in increased phenotypic plasticity and ultimately giving rise to therapy resistance [94]. Furthermore, in the mouse breast cancer model it was identified that loss of the genes involved in active chromatin organization resulted in the creation of alveolar specific enhancers, causing conversion of basal cells to alveolar‐like lineage [95]. Among the enhancer reprogramming mechanisms, enhancer hijacking, a process whereby a gene is regulated by an enhancer that is not normally involved in its regulation, plays an important role. Enhancer hijacking generally occurs via genomic rearrangements, such as duplications and translocations [96]. One of the earliest findings of enhancer hijacking was activation of the GFI1B oncogene in medulloblastoma as a result of repositioning of GFI1 gene proximal to a super‐enhancer [97]. Similar enhancer hijacking mechanisms were later described for many other cancer types as well; such events cause misregulation of lineage‐specific enhancers, altering cellular plasticity [98, 99, 100, 101, 102, 103].

While enhancer reprogramming and hijacking mechanisms establish the epigenetic framework needed for shifts in phenotypic identity, the dynamic regulation of these processes is achieved through transcription factors that recruit chromatin‐modifying complexes to distal regulatory regions [104].

Transcription factors driving pan‐plasticity networks

The most well‐documented drivers of EMP are a core network of transcription factors collectively known as EMT‐TFs [9, 105]. While initially characterized by their ability to repress epithelial identity and induce mesenchymal features, it is now clear that they function as dynamic regulators of cellular plasticity rather than simple binary switches of EMT [106].

Among the canonical EMT‐TFs, the SNAIL family (SNAI1 and SNAI2) plays a central role in initiating epithelial repression. SNAIL proteins directly suppress epithelial markers such as E‐cadherin while recruiting chromatin‐modifying complexes, including HDACs and PRC2, to establish stable and durable transcriptional silencing [107]. This coupling of transcriptional repression with epigenetic remodeling facilitates transitions in cellular identity.

The TWIST family (TWIST1 and TWIST2) contributes to plasticity by promoting mesenchymal traits, metastatic competence, and stem cell–like properties [108, 109]. TWIST factors are particularly important in facilitating reversible transitions, allowing tumor cells to dynamically switch between epithelial and mesenchymal states in response to microenvironmental cues.

Similarly, ZEB1 and ZEB2 function as potent regulators of both EMT and stemness. ZEB proteins repress epithelial gene expression while simultaneously activating transcriptional programs associated with cellular plasticity [107, 108]. Notably, ZEB1 has been shown to interact extensively with chromatin‐remodeling machinery, reinforcing its role as a bridge between transcriptional control and epigenetic regulation during lineage transitions [20]. Although ZEB1 is best known as a transcriptional repressor of epithelial genes, such as CDH1, recent genome‐wide studies in triple‐negative breast cancer (TNBC) have revealed that it can also act as a transcriptional activator [110]. Mechanistically, this ZEB1/YAP/AP‐1 complex is reported as a hallmark of pan‐plasticity due to its activatory roles for a tumor‐promoting gene set independently of canonical ZEB1‐binding motifs. Indirect recruitment of ZEB1 to the DNA through its physical interaction with YAP and AP‐1 proteins bypasses typical lineage constraints to activate programs associated with migration, invasion, and therapy resistance, turning ZEB1 into a main regulator of aggressive cellular states. The translational relevance of this interaction is evidenced by its enrichment in the claudin‐low subtype of breast cancer. Patients with high combined expression of ZEB1 and FOSL1 exhibit a significantly higher risk of distant metastasis and therapy relapses [111].

While the SNAIL, TWIST, and ZEB families constitute the canonical EMT‐TFs, accumulating evidence indicates that their functional roles extend far beyond EMTs. Rather than acting as binary EMT switches, these factors operate within broader regulatory networks that enable multidirectional cell‐state transitions, including dedifferentiation, transdifferentiation, and lineage switching [12, 17]. Consistent with this expanded view, a growing number of transcription factors beyond the canonical EMT‐TFs have been implicated in pan‐plasticity programs (Fig. 2). Members of the SOX family, particularly SOX2, SOX4, and SOX9, are key regulators of lineage plasticity across multiple cancer types [22, 29]. SOX2, for instance, is strongly associated with therapy‐induced lineage switching and the emergence of stem‐like tumor states, whereas SOX9 has been linked to EMP and enhanced tumor‐initiating capacity [112]. Pan‐plasticity factors function primarily by establishing permissive chromatin landscapes that facilitate transcriptional reprogramming. Chromatin‐remodeling complexes and epigenetic modifiers, including the SWI/SNF complex, EZH2, LSD1, and TET family DNA demethylases, are critical in regulating chromatin accessibility and enhancer activity during these transitions [12, 29, 34]. By controlling the accessibility of lineage‐specific regulatory elements, these factors enable tumor cells to activate alternative transcriptional programs in response to environmental stress or therapeutic pressure.

Fig. 2.

Pan‐plasticity framework in cancer. Primary tumor cells undergo dynamic lineage transitions driven by epithelial–mesenchymal transition (EMT)‐associated transcription factors (EMT‐TFs) in coordination with widespread epigenetic reprogramming. Beyond the canonical EMT‐TFs (SNAIL, ZEB, and TWIST), additional regulators, such as SOX2, FOXA1/2, MYC/N‐MYC, and pluripotency factors (OCT4, NANOG, and KLF4) cooperate with epigenetic modifiers (e.g., HDACs, PRC2, SWI/SNF, EZH2, LSD1, and TET) to establish permissive chromatin landscapes that enable transcriptional reprogramming. Within this framework, tumor cells can transition into multiple lineage states, including neuroendocrine, squamous, endothelial‐like, or dedifferentiated phenotypes through EMP and transdifferentiation, ultimately giving rise to metastatic and therapy‐resistant tumor cell populations. Figure partially created in BioRender. Boyvatlı, E. (2026) https://BioRender.com/u5w0m09.

Pluripotency‐associated transcription factors, including OCT4 (POU5F1), NANOG, and KLF4, further contribute to plasticity by reactivating embryonic gene expression programs [113, 114]. These factors promote dedifferentiation and increase developmental potential, thereby enabling tumor cells to escape lineage constraints and adopt alternative cellular identities. Their activity often converges with epigenetic regulators, reinforcing permissive chromatin states that facilitate transcriptional reprogramming [115].

Furthermore, pioneer transcription factors such as FOXA1 and FOXA2 play a critical role in enabling plasticity through chromatin remodeling. By binding to compacted chromatin and initiating enhancer activation, FOXA proteins allow the engagement of previously inaccessible regulatory regions, thereby supporting large‐scale transcriptional reprogramming during lineage transitions [116, 117].

Oncogenic drivers such as MYC and N‐MYC further amplify plasticity by globally enhancing transcriptional output and promoting stem cell–like phenotypes [21, 99]. In particular, N‐MYC has been implicated in NED through cooperation with epigenetic modifiers such as EZH2, highlighting the interplay between transcriptional and epigenetic regulation in lineage plasticity.

Ultimately, although many of these factors are historically categorized as EMT‐TFs or EMT‐associated regulators, their functions are not restricted to EMTs. Instead, they act as central components of a broader pan‐plasticity network that enables tumor cells to traverse multiple lineage states [17]. Within this framework, EMT‐TFs and associated transcriptional regulators function as versatile modulators of cell identity, integrating environmental signals, genetic alterations, and epigenetic states to drive phenotypic diversification.

The recognition of EMT‐TFs as pan‐plasticity regulators has significant conceptual and clinical implications. By facilitating multidirectional lineage transitions, these transcriptional networks contribute to tumor heterogeneity, metastatic dissemination, and resistance to therapy [12]. Targeting the regulatory circuits that sustain these plastic states, particularly in combination with inhibition of oncogenic drivers, may therefore represent a promising strategy to limit tumor adaptability and improve therapeutic outcomes.

Tumor microenvironment and chromatin plasticity

The TME functions as a dynamic and heterogeneous ecosystem comprising stromal cells, diverse immune populations, extracellular matrix (ECM) components, and fluctuating physicochemical conditions. Rather than serving as a passive backdrop, the TME actively shapes cancer‐cell identity by driving epigenetic remodeling and transcriptional plasticity. Through the integration of metabolic, hypoxic, and inflammatory signals, TME facilitates the acquisition of reversible cellular states that support EMP, stemness, and therapy resistance (Fig. 3).

Fig. 3.

Tumor microenvironmental signals converge on epigenetic reprogramming to promote cellular plasticity. The tumor microenvironment (TME) provides critical extrinsic cues that rewire the epigenetic landscape to stabilize plastic cell states. Metabolic alterations directly modulate the activity of chromatin‐modifying enzymes to alter histone and DNA methylation patterns. These include lactate accumulation, oncometabolite production (2‐HG, succinate), and shifts in SAM/Acetyl‐CoA availability. Hypoxia‐driven HIF‐1α activation and reactive oxygen species (ROS) further reshape chromatin accessibility, while TAM‐ and CAF‐derived cytokines, such as IL‐6 and TGF‐β engage NF‐κB and SNAIL‐mediated transcriptional programs to reinforce mesenchymal and stem‐like states. Collectively, these signals converge on chromatin regulation to drive epithelial cells through a hybrid EMP state toward mesenchymal, neuroendocrine, or cancer stem cell identities. Figure partially created in BioRender. Boyvatlı, E. (2026) https://BioRender.com/20kt3ej.

Metabolism‐driven chromatin plasticity

Cellular metabolism is tightly coupled to epigenetic regulation through the availability of metabolic intermediates that function as essential substrates or cofactors for chromatin‐modifying enzymes. These ‘metabolic fuels’ directly modulate the chromatin states and transcriptional programs associated with plasticity [118]. Key metabolites such as SAM serve as methyl donors for histone and DNA methylation reactions, including those mediated by EZH2 and DNMTs [119, 120]. Similarly, the availability of acetyl‐CoA dictates histone acetylation levels, effectively linking the cell's nutrient status to global chromatin accessibility.

Cancer‐associated metabolic rewiring further influences the epigenome through the accumulation of oncometabolites, such as 2‐hydroxyglutarate, succinate, and fumarate. These molecules act as competitive inhibitors of α‐ketoglutarate–dependent dioxygenases, including TET DNA demethylases and histone demethylases. The inhibition leads to global hypermethylation and transcriptional repression of differentiation‐related programs, effectively locking cells into plastic, undifferentiated states. More recently, histone lysine lactylation (Kla) driven by lactate accumulation in glycolytic tumors has emerged as an additional novel mechanism linking metabolic rewiring to chromatin plasticity [121]. In response to both endogenous and exogenous lactate, enrichment in histone Kla has been observed in different cell types [122, 123]. This modification enhances chromatin accessibility to activate gene expression programs associated with stemness and lineage reprogramming. Notably, lactate enhances histone acetylation, further reinforcing this permissive chromatin state to drive cancer stem cell identity and hinder cell fate commitment [124]. Furthermore, lactate accumulation specifically stabilizes certain plasticity states, such as the hybrid E/M phenotype, through H3K18 lactylation (H3K18la). This modification is important to modify nucleosome structure for enhancing the accessibility of regulatory elements that drive stemness and lineage reprogramming. Thus, glycolytic shift in the TME provides the essential biochemical substrates for H3K18la to sustain lineage infidelity, therapeutic resistance and metastasis [125].

Hypoxia and HIF‐1α‐mediated epigenetic reprogramming

Hypoxia is a hallmark of the TME and a major driver of cancer‐cell plasticity. Under restricted oxygen conditions, stabilization of HIF‐1α initiates comprehensive transcriptional and epigenetic reprogramming.

HIF‐1α directly activates genes involved in glycolysis, angiogenesis, and survival, while also cooperating with other transcription factors, such as NF‐κB, c‐Jun, and Nrf2 to amplify stress‐response and inflammatory signaling. Hypoxia exerts profound and direct effects on the epigenome by limiting the oxygen required for α‐ketoglutarate–dependent enzymes. The impairment of TET DNA demethylases and histone demethylases under hypoxic conditions leads to global DNA hypermethylation and reinforcement of transcriptionally repressive chromatin architecture. This epigenetic‐locking mechanism stabilizes plastic and therapy‐resistant phenotypes, including those associated with EMT and neuroendocrine transdifferentiation. Furthermore, hypoxia promotes the activation of core plasticity factors, such as TWIST, SNAIL, and ZEB1 [126, 127], while simultaneously remodeling the ECM to reinforce spatial and phenotypic heterogeneity within the tumor.

Inflammatory tumor microenvironment

Inflammatory signaling within the TME represents another major extrinsic driver of chromatin plasticity. Many pro‐inflammatory factors originate from the surrounding stromal and immune compartments rather than cancer cells. Tumor‐associated immune cells and stromal components secrete a wide range of cytokines and chemokines that modulate cancer‐cell states [128, 129].

An increased number of tumor‐associated macrophages (TAMs) facilitates EMT characteristics by secreting cytokines, such as IL6, IL10, TGF‐β, and CCL2, which correlate with downregulation of E‐cadherin expression and upregulation of the expression levels of mesenchymal markers like VIM [130]. In addition to soluble factors, TAM‐derived exosomes play a role in intercellular communication. Notably, exosomal miR‐95 derived from TAMs has been demonstrated to induce prostate cancer progression inducing proliferation and EMT [131]. Similarly, cancer‐associated fibroblasts (CAFs) contribute to lineage plasticity through ECM remodeling and establishment of pro‐inflammatory niches that sustain plastic cellular states [132].

Beyond EMT, growth factors and inflammatory cytokines can also drive lineage plasticity. For instance, exogenous IL‐6 and EGF treatment have been implicated in promoting NED in prostate cancer‐cell lines. IL‐6‐induced activation of STAT3 and EGF‐activated MAPK/ERK signaling pathways reprograms transcriptional network and epigenetic state, which suppresses luminal lineage identity while activating neuroendocrine‐associated gene expression. These findings highlight how microenvironmental states can shape cell fate transitions beyond epithelial–mesenchymal states [133, 134, 135, 136, 137].

Bidirectional interactions and feedback loops

The interactions between the TME and cancer‐cell plasticity are fundamentally bidirectional. While microenvironmental cues drive epigenetic changes, EMT‐TFs actively remodel their surroundings by regulating the secretion of cytokines, ECM components, and immune‐modulatory signals.

Epigenetic regulators play a central role in mediating this crosstalk. For instance, DNA methyltransferases (DNMTs) have been shown to influence the methylation status of immune‐related genes, which can affect the infiltration of immune cells, such as neutrophils and dendritic cells. These findings suggest that DNMT‐mediated epigenetic regulation in colorectal carcinoma may contribute to the modulation of immune cell recruitment within the TME [138]. Consistent with this, it has been reported that DNMT inhibition upregulates expression of MHC class I in breast cancer subtypes via multiple mechanisms. These include direct demethylation in promoters of MHC‐I genes, in addition to activation of innate immune mechanisms, such as NF‐κB signaling [139]. Through these mechanisms, epigenetic reprogramming not only alters cancer‐cell intrinsic state but also enhances the recruitment of immunosuppressive cells and promotes fibroblast activation, creating a self‐sustaining feedback loop that sustains mesenchymal and stem‐like states. This reciprocal interaction is crucial for stabilizing plasticity and a central determinant of tumor progression and therapy resistance. Notably, EMT has been associated with the establishment of an immunosuppressive TME, contributing to tumor progression and to resistance to immune checkpoint inhibitors [140, 141].

Impact of epigenetic reprogramming in cancer therapy resistance

In addition to genetic mutations, epigenetic reprogramming has emerged as a fundamental driver of therapy resistance in cancer, facilitating both transient adaptive responses and stable resistance phenotypes. This resistance develops through trajectories of cell‐state transitions that progressively enhance cellular fitness and survival under drug pressure.

A key feature of this process is the emergence of intratumoral heterogeneity. Specific subpopulations, such as drug‐tolerant persister (DTP) cells, utilize reversible chromatin dynamics to survive initial treatment. These shifts are orchestrated by coordinated alterations in histone modifications, DNA methylation, chromatin accessibility, and enhancer reprogramming. Importantly, plastic cell states can become stabilized after prolonged exposure to microenvironmental or therapeutic signals. For example, chronic TGF‐β exposure converts a reversible EMT state into a stable mesenchymal phenotype with persistent stem cell–like properties and increased drug resistance, while also creating new therapeutic vulnerabilities [142].

Specific chromatin regulators act as key mediators of these transitions; for instance, NSD2‐mediated H3K36 demethylation promotes epigenetic plasticity, while the loss of KMT2C induces a switch toward a double‐negative prostate cancer phenotype resistant to androgen receptor targeting therapies [143, 144]. Similarly, the loss of the SWI/SNF subunit of ARID1A facilitates a luminal to basal lineage switch in breast cancer, driving endocrine therapy resistance [145].

Beyond intrinsic tumor evolution, therapeutic interventions can act as instructive signals, whereby treatment‐induced chromatin remodeling reshapes enhancer landscapes and transcription factor networks to force the cell into a drug‐tolerant state. Furthermore, activation of normally silenced genomic elements, such as transposable elements (TE), adds an additional layer in which TE‐derived transcripts modulate innate immune signaling and influence the overall fitness of the tumor under therapeutic pressure.

Targeting the epigenome to limit plasticity

Lineage plasticity is maintained through chromatin remodeling, enhancer reprogramming, and disruption of lineage‐defining programs. Thus, targeting the epigenome represents a direct therapeutic vulnerability (Fig. 4). Each of the inhibitor classes discussed below targets different aspects of this chromatin landscape and collectively aims to disrupt the plastic states that fuel both EMP and NED.

Fig. 4.

Epigenetic targeting of the hybrid state to reduce plasticity and restore immunogenicity. Hybrid E/M cells occupy a central position along the epithelial–mesenchymal transition (EMT) spectrum, characterized by maximal transcriptional plasticity, therapy resistance, and immune evasion through endogenous retrovirus (ERV) silencing. Two complementary epigenetic therapeutic strategies are shown to dismantle this state. Chromatin regulator inhibitors, including HDACi, BETi, LSD1i, and EZH2i, induce differentiation by collapsing the hybrid chromatin state, thereby reducing lineage switching and restoring drug sensitivity. In parallel, DNMTi alone or in combination with HDACi derepresses endogenous retroviruses to generate immunostimulatory dsRNA, triggering interferon signaling and innate immune activation. Together, these approaches converge to reduce plasticity, suppress lineage switching, enhance tumor immunogenicity, and improve therapeutic response. Figure partially created in BioRender. Boyvatlı, E. (2026) https://BioRender.com/w61pigm.

Differentiation‐oriented epigenetic therapies

Differentiation‐oriented epigenetic therapies target chromatin regulators that maintain adaptive epigenetic plasticity programs. The underlying idea behind these therapies is not to directly promote terminal differentiation, but rather to destabilize the hybrid, intermediate cell states that enable EMP and NED, thereby limiting the phenotypic diversity available to cancer cells under therapeutic pressure.

HDAC inhibitors

HDAC activity maintains adaptive transcriptional states and cell lineage plasticity by reinforcing repression at differentiation loci, preventing the reactivation of luminal or epithelial gene programs in transitioned cancer cells [146]. Among HDAC family members, HDAC2 is consistently upregulated in neuroendocrine prostate cancer (NEPC) compared with non‐neuroendocrine CRPC across multiple patient cohorts, where it promotes NED and angiogenesis partly through suppression of lineage‐restricting transcriptional regulators, such as REST and TSP1 [147]. Dual inhibition of HDAC1/2 and PI3K/AKT using fimepinostat has been reported to suppress lineage‐defining transcription factors, such as AR in luminal CRPC and ASCL1 in NEPC, and to provide potent tumor growth inhibition in both CRPC subtypes in cell line and patient‐derived xenograft models [148].

In the context of EMP, pan‐HDAC inhibition with panobinostat suppressed HMGA2, which is a chromatin‐remodeling protein driving the maintenance of the mesenchymal state, and increased H3K27 acetylation at luminal gene loci in a genetically defined murine CRPC model. When combined with castration, this strategy restores sensitivity of mesenchymal‐like castration‐resistant tumor cells to androgen deprivation therapy and significantly prolongs survival [149]. Consistent with these findings, class I HDAC inhibition in small‐cell lung cancer (SCLC) preferentially targets tumors with high neuroendocrine transcriptional states. In these subtypes, entinostat disrupts super‐enhancer–driven programs controlled by ASCL1 and NEUROD1, leading to increased therapeutic sensitivity compared with neuroendocrine‐low tumors [150].

BET inhibitors

BET family proteins are key readers of acetylated chromatin, and act as scaffold identity‐defining transcriptional programs at super‐enhancers. Proteins of the bromodomain and extra‐terminal (BET) family, BRD2, BRD3, BRD4, and BRDT, act as chromatin readers that sustain identity‐defining transcriptional programs at super‐enhancers [151]. BET inhibition disrupts enhancer‐dependent transcription and, rather than broadly suppressing gene expression, preferentially destabilizes identity‐defining transcriptional circuits [152].

In EMP, this principle translates directly to suppression of mesenchymal state maintenance. BRD4 sustains Snail protein stability and transcription via Gli1‐dependent mechanisms in breast cancer; its inhibition attenuates cell migration, invasion, and lung colonization in patient‐derived xenograft models, directly linking BRD4 activity to mesenchymal state maintenance [153, 154].

The connection between BET activity and NED is particularly striking in the context of therapy‐induced lineage switching. In CRPC, anti‐AR treatment reprograms BRD4 genome occupancy from AR‐target loci to the enhancers of neurogenesis, EMT, and stem cell programs, directly activating NE drivers, including BRN2, ASCL1/2, NEUROD1, and DLL3, thereby positioning BRD4 as an active driver rather than a bystander of the adenocarcinoma‐to‐NEPC transition. Accordingly, BRD4 inhibitors (JQ1 and AZD5153) blocked these lineage plasticity programs and suppressed growth of both treatment‐induced and de novo NEPC patient‐derived xenografts in vivo [153]. These findings collectively demonstrate that BRD4 is a context‐dependent master regulator of cell lineage transitions; pharmacological targeting of this protein not only inhibits NED programs but also provides a rationale for preventing chromatin reprogramming, which enables cells to shed their initial identity.

LSD1 inhibitors

LSD1 (KDM1A) is a histone demethylase that functions within the CoREST repressive complex; it silences epithelial and differentiation‐related gene loci by removing mono‐ and di‐methyl marks from H3K4. Its overexpression in many cancer types, including breast, prostate, and lung cancer, is associated with poorly differentiated, aggressive phenotypes, indicating that LSD1 plays a direct role in the epigenetic regulation of plastic cell states [155].

LSD1 functions as the essential enzymatic mediator of Snail‐mediated gene repression: it is physically recruited to epithelial gene promoters, where it demethylates H3K4me2; without this, Snail cannot silence E‐cadherin and other epithelial markers [19]. Consistent with this dependency, pharmacological inhibition of LSD1 in breast cancer increases E‐cadherin expression, reduces Snail and vimentin levels, and reverses the mesenchymal phenotype; specifically, LSD1 inhibition also attenuates chemotherapy‐induced EMT and reduces the cancer stem cell burden in circulating tumor cells from metastatic breast cancer patients [156]. The relationship between LSD1 and neuroendocrine cell lineage differentiation is similar. ASCL1, the primary transcription factor governing the neuroendocrine program in both NEPC and SCLC, is directly supported by LSD1 through its interaction with the SNAG‐domain factor INSM1. LSD1 expression is high in NEPC and is associated with poor clinical outcomes; the selective inhibitor bomedemstat suppresses the ASCL1‐dependent NE transcription program by disrupting LSD1–INSM1 interactions, while simultaneously reversing the silencing of YAP1, leading to tumor regression in xenograft models derived from NEPC patients [157, 158].

EZH2 inhibitors

EZH2 inhibitors restrict cellular lineage plasticity by blocking the methyltransferase activity of PRC2, thereby reducing H3K27me3‐mediated transcriptional repression and destabilizing chromatin states that maintain adaptive phenotypes. Rather than acting like traditional cytotoxic agents, these compounds primarily function as state‐modulating therapies and reduce the epigenetic support for EMP, cell lineage transition, and treatment‐resistant states.

In the context of EMP, the selective EZH2 inhibitor GSK126 provides a direct mechanistic example. As detailed in the section ‘Histone modifications and bivalent chromatin’ above, EZH2‐mediated H3K27me3 deposition at epithelial gene loci, exemplified by WNT5A silencing in TGF‐β‐induced EMT, functions as a reversible chromatin switch that stabilizes the mesenchymal state; pharmacological inhibition with GSK126 partially restores epithelial properties, demonstrating that this switch is therapeutically accessible [18].

Beyond EMT‐associated plasticity, EZH2 inhibition is also relevant to NED and broader lineage plasticity. As discussed in the section ‘Histone modifications and bivalent chromatin’, N‐Myc‐driven neuroendocrine reprogramming in prostate cancer operates through an EZH2/PRC2‐mediated program that suppresses AR signaling and activates neuroendocrine lineage identity [21]; pharmacological targeting of this axis with GSK126, particularly when combined with enzalutamide, can suppress growth of enzalutamide‐resistant prostate cancer, restore AR‐associated transcription, and restrain progression toward more plastic, treatment‐adapted states [159].

The most advanced EZH2 inhibitor in clinical development is Tazemetostat. Tazemetostat has demonstrated clinical efficacy in EZH2‐mutated follicular lymphoma and epithelial sarcoma [160, 161], supporting the broader concept that EZH2‐dependent chromatin states are amenable to drug treatment. Although the strongest direct evidence for EMP and NED remains in the preclinical stage, these clinical results strengthen the rationale for extending EZH2 inhibition to plasticity‐driven solid tumors [162].

Viral mimicry

A second mechanism by which epigenetic therapies may restrict cancer plasticity and improve treatment response involves triggering viral mimicry, a chromatin‐dependent innate immune activation program. Unlike the differentiation‐focused strategies discussed above, which destabilize adaptive transcriptional states, viral mimicry functions primarily by increasing the tumor's immunogenicity through the reactivation of genomic elements that are normally silenced. Human endogenous retroviruses (HERVs) constitute approximately 8% of the human genome and represent residues of ancestral retroviral infections integrated into the germline [163]. In somatic cells, ERVs are silenced by repressive epigenetic marks, primarily DNA methylation and H3K9me3 via the KRAB‐ZNF/KAP1/SETDB1 axis [164]. When epigenetic inhibitors remove the suppression of these loci, the resulting cytoplasmic dsRNA is interpreted as foreign viral RNA and triggers an antiviral response from within the tumor cell itself. This process, termed viral mimicry, operates through cytosolic dsRNA sensing pathways that activate a type I interferon response, upregulate antigen presentation machinery, and prime the TME for immune‐mediated clearance [165].

DNMT inhibitors

DNMTi therapy induces the expression of thousands of transposons, including ERVs and latent cancer‐testicular antigens; this leads to viral mimicry. In ovarian cancer, inhibition of the dsRNA sensors TLR3 and MAVS significantly attenuated this response, confirming that these pathways are mechanistically required [166]. Beyond innate sensing, DNMTi, either alone or in combination with HDACi, induces ERV‐derived t‐neopeptides presented on HLA class I molecules that can trigger tumor‐specific T‐cell responses [167], providing a second adaptive immune arm. DNMTi thus operates on two immunological axes simultaneously: (i) innate sensing through ERV‐derived dsRNA and (ii) adaptive recognition through de novo neopeptide presentation. This positions it as a uniquely multi‐mechanistic primer for immunotherapy sensitization.

HDAC inhibitors

While DNA methylation serves as the primary regulatory mechanism at ERV loci, active ERV promoters are marked by H3K27ac and H3K56ac, direct targets of HDACs, and these marks must be removed by class I HDACs to maintain ERV silencing [168] Consequently, HDACi removes the suppression of ERVs via a complementary chromatin axis, triggering dsRNA accumulation and downstream IFN signaling independently of DNA methylation status. Both DNMTi and HDACi promote cytolytic T‐cell activity against renal cell carcinomas and myeloid malignancies by increasing HERV‐derived ORF expression [169]. By removing the suppression of ERV loci and reshaping the immunosuppressive microenvironment through MHC upregulation and MDSC depletion, HDACi transforms the two most common barriers to immune checkpoint blockade (ICB) efficacy, low immunogenicity and immune exclusion, into actionable therapeutic vulnerabilities [165].

Overall, differentiation‐focused strategies and viral mimicry strategies are not mechanistically independent of one another; rather, they overlap significantly. This has been most strikingly demonstrated in muscle‐invasive bladder cancer through EZH2 inhibition where EZH2 inhibition activates NK cell‐associated signaling, upregulating MIP‐1α, ICAM1, ICAM2, and CD86 alongside IFN‐γ, while simultaneously reducing pluripotency markers and increasing tumor cell death [170]. EZH2 inhibitors further enhance viral mimicry caused by DNMTi via the NFAT: AP‐1 signaling pathway in colorectal cancer [171], demonstrating that the combined targeting of epigenetic nodes enhances both transcriptional reprogramming and immunogenic ERV derepression beyond monotherapy. This convergence defines the emerging rationale for next‐generation epigenetic combinations: by synergistically using inhibitor classes, and simultaneously constraining lineage plasticity, promoting differentiation, and making the tumor immunologically visible, one can transform the epigenome from a driver of therapeutic resistance into a coordinated target for its reversal.

Limitations and future challenges in clinical translation

The clinical translation of epigenetic therapies to address cancer‐cell lineage plasticity remains a challenge, despite a compelling molecular rationale and evidence of success in preclinical settings. A primary translational bottleneck is that the efficacy of epigenetic reprogramming is often dependent upon the presence of reversible cellular states; however, once terminal transdifferentiation is complete, the resulting locked epigenetic state effectively closes the window for therapeutic reversal back to a treatment‐sensitive phenotype [22, 172]. This difficulty is boosted by the discovery that many chromatin modifiers sustain plasticity through noncatalytic or structural functions, such as forming complex assemblies with N‐Myc or the androgen receptor, which standard enzymatic inhibitors are simply unable to address [173, 174]. Furthermore, the pleiotropic nature of epigenetic regulators presents a significant clinical risk, as these enzymes govern essential biological processes across diverse cell populations. Systemic inhibition often leads to severe off‐target toxicity or unintended immunological consequences, such as the paradoxical suppression of natural killer (NK) cell activity even as researchers stimulated T‐cell function [175]. Moreover, the field still lacks reliable biomarkers necessary for lineage‐aware patient stratification, and the serial biopsies required to monitor these dynamic cellular shifts are frequently infeasible in a real‐world clinical environment [176]. Beyond these pharmacological hurdles, cancer cells often utilize multiple redundant adaptation mechanisms to bypass single‐node inhibition, and current in vitro models often fail to replicate the complex, hypoxic, or fibrotic niches of the human TME that impair drug efficacy [177]. There is also a distinct risk that targeting specific lineage programs, such as those controlled by ASCL1, could inadvertently eliminate subpopulations that are naturally sensitive to chemotherapy [178]. Finally, conceptual uncertainties persist regarding the causality of epigenetic shifts and the potential to revert fully transformed states back to a sensitive phenotype [172].

Conclusion and perspectives

Lineage plasticity has emerged as a central hallmark of cancer, reflecting the dynamic and reversible nature of cellular identities during tumor progression. This concept highlights that clonal genetic diversity alone cannot fully explain the phenotypic transitions observed in cancer. Instead, nonmutational epigenetic reprogramming plays a fundamental role in driving transitions across a broad plasticity spectrum, including EMP, neuroendocrine transdifferentiation, and stem cell–like reprogramming. These processes enable tumor cells to bypass differentiation barriers and adapt to selective pressures within the TME.

Despite recent advances, several key challenges remain. It is still difficult to distinguish whether epigenetic changes associated with plasticity act as drivers or consequences of phenotypic transitions. Moreover, it remains unclear whether similar epigenetic alterations produce comparable biological outcomes across different tumor types, to what extent plastic states are reversible, and which cellular states are most critical for therapy resistance. Further studies are also needed to define the bidirectional interactions between EMT‐like programs, metabolic rewiring, and immune evasion.

The integration of single‐cell omics technologies will be essential for monitoring and quantifying cellular transitions over time. These approaches will help define transient and hybrid states associated with plasticity and clarify their clinical relevance. In parallel, therapeutic strategies targeting chromatin‐level regulatory networks may prove more effective than approaches directed at individual epigenetic factors. A deeper understanding of the epigenetic mechanisms underlying lineage plasticity will therefore not only advance cancer biology but also support the development of more precise therapies for aggressive, therapy‐resistant tumors with reprogrammable epigenetic states.

Author contributions

BA contributed to the conceptualization, project administration, writing—original draft, writing—review and editing, and visualization. EB and EB contributed to the writing—original draft and visualization. NA contributed to the supervision, conceptualization, and writing—review and editing. SE‐O contributed to the project administration, supervision, writing—original draft, writing—review and editing, and correspondence.

Conflict of interest

The authors declare no conflict of interest.