Epigenetic and transcriptional control of classical and basal-like cell states in pancreatic ductal adenocarcinoma

Abstract

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies, with a 5-year survival of only ∼13%. Despite incremental advances through combination chemotherapy, most patients relapse rapidly due to profound molecular heterogeneity and intrinsic resistance. Recent genomic and transcriptomic studies have defined distinct PDAC molecular subtypes, classical and basal-like, which differ in differentiation state, prognosis, and therapeutic vulnerability. Classical tumors, marked by GATA6 and hepatocyte nuclear factors, exhibit epithelial identity and relative chemosensitivity, whereas basal-like tumors driven by ΔNp63 and MYC display mesenchymal and inflammatory programs associated with resistance and poor outcome. Importantly, these subtypes are dynamic, with single-cell and spatial analyses revealing frequent coexistence and therapy-induced transitions, highlighting cellular plasticity as a major determinant of treatment response. Subtype identity is governed by lineage-defining transcription factors, chromatin regulators, and stromal cues that integrate to form reversible epigenetic states. Targeting these mechanisms with inhibitors of EZH2, BET proteins, or CDK9 can restore differentiation programs and resensitize tumors to chemotherapy. Integrating molecular subtyping with epigenetic modulation thus offers a rational path toward biomarker-guided therapy. Continued efforts combining spatially resolved profiling, organoid modeling, and liquid-biopsy monitoring will be essential to capture tumor evolution in real time. Understanding and therapeutically exploiting the transcriptional and epigenetic plasticity in PDAC may ultimately enable reprogramming of resistant states and improve clinical outcomes in this intractable disease.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) accounts for approximately 95% of pancreatic cancer and remains one of the deadliest malignancies worldwide. With a 5-year overall survival (OS) rate of only ∼13%, PDAC has the highest mortality rate among major cancers and is projected to become the second leading cause of cancer-related deaths by 2030 [1, 2]. The poor prognosis is primarily due to late diagnosis and propensity of PDAC to metastasize, as PDAC progresses asymptomatically until advanced stages, when surgical resection is rarely possible. Only 15%–20% of patients are eligible for potentially curative surgery, and even in this subset, long-term survival is less than 30% [3–5].

Systemic chemotherapy is the mainstay of treatment for most patients. Three combination regimens have been shown to improve survival: FOLFIRINOX or its modified variant (mFFX), NALIRIFOX, and gemcitabine plus nab-paclitaxel [6, 7]. These combinations modestly improve survival, but PDAC is intrinsically resistant even to first-line chemotherapy, and responses are often short-lived. Moreover, higher efficacy is frequently offset by increased toxicity, which limits tolerability and excludes many patients from treatment intensification [8].

Therapeutic options beyond first-line therapy are limited with only ∼50% of patients being fit enough to receive second-line therapy, and fewer than 25% proceeding to third-line treatment, where no standardized protocols exist [9, 10]. Thus, achieving effective first-line therapy is critical. Clinical decision-making is still primarily guided by patient-related factors such as performance status, age, and comorbidities, rather than tumor biology. As a result, current therapies represent evidence-based averages that inadequately address the molecular and clinical diversity of PDAC.

The recently reported PASS-01 trial directly compared mFFX and gemcitabine/nab-paclitaxel in untreated metastatic PDAC and confirmed similar efficacy overall, with gemcitabine/nab-paclitaxel showing a slight survival advantage and better tolerability, while serious adverse events were more frequent with mFFX [11]. Importantly, PASS-01 embedded comprehensive molecular correlatives, including whole-genome and transcriptome profiling as well as patient-derived organoids, and demonstrated the feasibility of integrating such approaches into prospective trials. Early analyses suggest that molecular subtypes and biomarkers, such as GATA-binding factor 6 (GATA6) expression may influence regimen sensitivity, but no validated predictor is yet available. Strikingly, nearly half of patients did not receive second-line therapy, underscoring the critical importance of optimizing first-line treatment and of developing biomarkers that can reliably guide regimen selection.

A defining feature of PDAC is its profound inter- and intratumoral heterogeneity, which contributes to variable therapeutic responses and rapid relapse [12]. Between patients, tumors show distinct genetic, transcriptional, and stromal landscapes; within individual tumors, heterogeneous subclones with different drug sensitivities coexist and evolve under treatment pressure. This complexity underlies the limited efficacy of standard regimens and the frequent emergence of resistant populations [13–15].

Attempts to overcome this challenge include intensified or sequential chemotherapy strategies that aim to target diverse subclones [16]. While early trials alternating gemcitabine/nab-paclitaxel with FOLFIRINOX showed promising response rates, cumulative toxicity remains a major limitation, and superiority over standard regimens has not been demonstrated [17]. Similarly, efforts to remodel the tumor stroma, which plays a central role in drug accessibility and subtype plasticity, have been largely unsuccessful, reflecting the functional heterogeneity of cancer-associated fibroblasts and the risks of indiscriminate stromal targeting [18].

Despite advances in molecular characterization, no robust predictive biomarkers are available in clinical practice to guide treatment selection in PDAC. Beyond small subsets of patients with homologous recombination repair deficiencies (e.g. BRCA1/2 or PALB2 mutations) or rare actionable fusions, most patients receive unstratified chemotherapy. In contrast to other cancers where biomarker-driven precision therapy has transformed outcomes, PDAC continues to be treated as a uniform disease [19–21].

Over the past decade, transcriptomic and epigenomic profiling has revealed that PDAC represents a spectrum of molecularly distinct subtypes, each characterized by specific biology, clinical behavior, and therapeutic vulnerabilities. A consistent dichotomy has emerged between classical (epithelial-differentiated) and basal-like (squamous, quasi-mesenchymal) tumors, with the former often associated with better prognosis and relative chemosensitivity, and the latter linked to dedifferentiation, resistance, and poor outcomes [12, 22–24].

Importantly, PDAC subtypes are not static. Single-cell analyses and longitudinal studies have demonstrated that tumors frequently harbor mixed subtype features and can undergo dynamic transitions under therapeutic pressure [15, 25, 26]. Patient-derived organoid models have further shown that distinct phenotypic states differ in drug sensitivity and that therapeutic exposure can drive subtype switching [27]. The tumor microenvironment also plays a critical role: for instance, stromal niche signals can sustain classical-like states, while their absence promotes a shift toward basal-like phenotypes [28]. These transitions are regulated by transcription factors, epigenetic mechanisms, and chromatin state changes, as well as microenvironmental cues, together highlighting both the complexity of PDAC identity and the opportunity to therapeutically redirect resistant states [29].

Together, these insights point to molecular subtyping as a promising strategy for precision medicine in PDAC. Reliable stratification could refine patient selection, improve therapeutic matching, and uncover novel vulnerabilities. At the same time, targeting the molecular regulators of subtype identity, particularly transcriptional and epigenetic mechanisms, offers a potential approach to reprogram resistant tumors into more treatable states. This review will summarize current PDAC molecular subtypes and their clinical relevance, discuss the challenges posed by tumor heterogeneity and plasticity, explore the molecular regulation of subtype identity, and highlight therapeutic opportunities, with a particular focus on epigenetic strategies to rewire PDAC cell identity for improved treatment outcomes.

Molecular subtypes of PDAC

Genetic drivers and transcriptomic diversity

Over the past decade, large-scale genomic and transcriptomic profiling studies have defined molecular subtypes of PDAC that differ in therapeutic sensitivity, metastatic behavior, and patient prognosis [12, 22–24]. These efforts reaffirmed the central role of recurrent driver alterations. Notably, activating KRAS mutations occur in >90% of cases and are thought to represent initiating events, while loss-of-function alterations in TP53, CDKN2A, and SMAD4 dominate the later stages of progression [30–32]. Each of these drivers contributes unique features to PDAC biology. For example, SMAD4 loss is associated with poor prognosis and enhanced metastatic spread [33], CDKN2A deletions often co-occur with interferon cluster loss leading to an immune-cold phenotype [34], and TP53 mutations promote genomic instability, metabolic rewiring, and aggressive disease behavior [35, 36]. Although PDAC is nearly universally Ras-driven, not all tumors are strictly dependent on mutant Ras, and experimental Ras extinction models reveal the capacity of resistant clones to survive independently of oncogenic Ras signaling [37, 38].

Beyond these canonical drivers, additional subsets harbor therapeutically relevant alterations. Approximately 5%–7% of patients carry germline or somatic mutations in homologous recombination repair genes (BRCA1/2, PALB2), conferring platinum sensitivity and enabling the use of polyadenosine-diphosphate-ribose polymerase (PARP) inhibitors in selected settings [39, 40]. Chromatin regulators such as KDM6A, ARID1A, and SMARCA4 (also called BRG1) are frequently altered, and their loss accelerates tumorigenesis or promotes transitions toward more aggressive basal-like states [41–44]. Finally, a minority (∼10%) of tumors that are KRAS wild-type harbor alternative drivers, including ALK, RET, NRG1, or BRAF fusions, which activate mitogen-activated protein kinase (MAPK) signaling and are potentially targetable with existing inhibitors [45, 46].

Transcriptional subtypes: classical versus basal-like

While these genomic alterations set the landscape of PDAC tumorigenesis, transcriptional and epigenomic programs provide an additional layer of heterogeneity with direct clinical implications. Despite differences between classification systems, a consensus has emerged around two major identities: a classical/epithelial-differentiated subtype and a basal-like/squamous/quasi-mesenchymal subtype [12, 22–24]. Classical tumors are characterized by epithelial differentiation programs and the expression of transcription factors, such as GATA6, hepatocyte nuclear factor 1 homeobox A (HNF1A), and pancreatic and duodenal homeobox 1 (PDX1), and are associated with improved survival and relative chemosensitivity [47, 48]. In contrast, basal-like tumors display mesenchymal gene signatures, higher grade, increased metastatic potential, and are linked to therapy resistance and poor prognosis, with regulators including MYC proto-oncogene (MYC) and transcription factor TP63 (ΔNp63) [49, 50]. These transcription factors act through enhancers and are further shaped by epigenetic regulators that either control their expression or act as transcriptional co-factors [29]. Importantly, accumulating evidence indicates that these transcriptional identities do not represent fixed tumor subtypes, but rather dynamic cellular states capable of interconversion. As a result, subtype identity is highly plastic and responsive to microenvironmental cues. Given that classical tumors are more chemosensitive, modulating PDAC tumor cell plasticity and redirecting resistant basal-like states toward a classical phenotype (subtype switching) may represent a viable and promising strategy to increase therapeutic efficacy [51]. Given the reversibility of epigenetic regulation, pharmacological targeting of chromatin regulators or other transcriptional processes represents a compelling strategy to reprogram PDAC subtype identity and exploit this vulnerability.

Despite the robust association of classical and basal-like programs with prognosis and therapy response, it remains debated whether these subtypes represent discrete biological entities or extremes along a continuous transcriptional spectrum. The degree to which subtype identity is predetermined by lineage versus dynamically imposed by microenvironmental cues also remains unresolved.

Consistent with this concept of plasticity, single-cell and spatial analyses have provided critical insights into the dynamic regulation of PDAC subtypes. Rather than representing mutually exclusive subtypes, classical and basal-like programs often coexist within the same tumor, and intermediate “co-expresser” populations can bridge these states [12, 15, 26]. Longitudinal analyses further indicate that basal-like features can emerge during progression and after therapy, consistent with state transitions under selective pressure [52]. The tumor microenvironment plays a central role in driving these transitions. Cancer-associated fibroblast (CAF)-derived niche signals can maintain classical differentiation or drive shifts toward mesenchymal/basal-like programs, with corresponding changes in drug response [28]. These findings underscore that PDAC subtypes are dynamic rather than fixed, with important implications for treatment resistance and opportunities for therapeutic reprogramming.

An important unresolved question is whether subtype plasticity primarily represents an adaptive response to therapeutic and microenvironmental stress, or whether pre-existing plastic cell populations actively drive resistance and disease progression. Distinguishing cause from consequence will be critical for designing effective intervention strategies.

Clinical relevance of subtyping

Clinically, subtype assignment carries both prognostic and predictive significance. Classical tumors are generally more responsive to chemotherapy and associated with better outcomes, while basal-like tumors frequently underlie treatment failure [53]. This has been demonstrated in retrospective cohorts and translational studies, such as COMPASS, which provided proof-of-principle that classical tumors respond better to 5-fluorouracil-based regimens than basal-like tumors [4, 54]. These observations are now being tested in larger multi-center trials that also integrate patient-derived organoids to assess drug sensitivities ex vivo. More recently, analyses from PASS-01 further supported the potential predictive role of subtype biomarkers, with GATA6 expression linked to regimen sensitivity [11]. However, while subtype identity is consistently associated with differential treatment response, it remains an open question to what extent current subtyping frameworks are directly predictive rather than correlative, particularly in the context of intratumoral heterogeneity and dynamic state transitions under therapy.

To date, no subtype classifier has yet been validated for routine clinical use, and molecular subtyping is therefore not incorporated into current clinical guidelines for PDAC. Barriers to implementation include challenges in tumor sampling and tissue quality, pronounced intratumoral heterogeneity, and the absence of standardized, clinically validated assays that can be deployed rapidly in real-world settings. In addition, the aggressive clinical course of PDAC and frequent deterioration in patient performance status limit the feasibility of extensive molecular profiling for many patients.

Beyond these methodological and biological constraints, real-world clinical factors further complicate the implementation of molecular subtyping in PDAC. Delays in diagnosis arising from fragmented care pathways, prolonged intervals between initial presentation, imaging, tissue acquisition, and referral to specialized oncology centers often constrain the time window available for molecular profiling and treatment stratification. As a consequence, many patients initiate therapy before comprehensive molecular information can be generated, or experience rapid clinical deterioration that precludes biomarker-guided decision making. Addressing these systemic challenges through streamlined diagnostic workflows, improved integration between primary care and oncology services, and the development of rapid assays applicable to small biopsy specimens will be as critical as biomarker discovery itself for enabling subtype-guided therapy in clinical practice.

An additional and underappreciated barrier to clinical implementation is sampling bias arising from intratumoral spatial heterogeneity. Increasing evidence from spatial transcriptomic and single-cell analyses demonstrates that classical and basal-like transcriptional programs can coexist within distinct regions of the same tumor, such that the molecular identity captured by a biopsy may strongly depend on the sampled location. In this context, sampling a single tumor region may classify a PDAC as predominantly classical or basal-like, while unsampled regions harbor alternative or mixed transcriptional states. This spatial heterogeneity has important clinical implications. Misclassification due to regional sampling bias may lead to inappropriate therapy selection, particularly if subtype identity is used to guide first-line chemotherapy choice or enrollment into biomarker-driven clinical trials. Moreover, therapy may preferentially target one regional state while allowing resistant subpopulations in other regions to persist and expand. These considerations underscore that molecular subtyping based on limited tissue samples must be interpreted with caution and highlight the need for approaches that better capture spatial and temporal tumor diversity, including multi-region sampling, spatially resolved profiling, or longitudinal liquid-biopsy strategies.

In summary, PDAC subtyping has provided important biological and clinical insights, defining patient groups with markedly different prognosis and therapy sensitivity. The refinement of subtype definitions through single-cell and multi-omics approaches highlights both the plasticity of PDAC and the opportunity to integrate molecular classifiers into clinical trials. Ultimately, robust and clinically feasible subtyping tools will be essential for advancing precision-medicine strategies in PDAC. These insights converge on a model in which PDAC subtypes represent dynamic, epigenetically regulated cell states influenced by both intrinsic and extrinsic cues. Figure 1 summarizes the defining features of the basal-like and classical programs, their transcriptional and epigenetic regulators, and the microenvironmental and therapeutic factors that drive plasticity between these states.

Figure 1.

Epigenetic regulation, plasticity, and microenvironmental control of PDAC molecular subtypes. (A) Schematic overview of the transcriptional and epigenetic programs defining basal-like and classical PDAC subtypes. Basal-like tumors are defined by ΔNp63, MYC, ZBED2, and GLI2 activity, enhancer activation upon KDM6A loss, and PRC2/EZH2-mediated repression of GATA6. BET/BRD4 and CDK9 maintain transcriptional elongation at basal-specific enhancers, supporting mesenchymal, glycolytic, and chemoresistant phenotypes. Classical tumors express GATA6, HNF1A/4A, and SPDEF, maintained by p300/CBP-driven histone acetylation, TET2-dependent 5-hydroxymethylcytosine (5hmC) marking, and KDM6A-HNF1A enhancer regulation. These programs promote epithelial differentiation, oxidative metabolism, and chemosensitivity. Subtype plasticity is regulated by chromatin state and can be pharmacologically modulated, for example through EZH2 inhibition (EZH2i) promotes GATA6 re-expression and classical reprogramming, while BET inhibition (BETi) represses MYC/ΔNp63-driven basal programs. (B) Differentiation landscape model illustrating dynamic interconversion between progenitor, basal-like, intermediate co-expressor, and classical transcriptional states. Epigenetic and microenvironmental perturbations such as TNFα, IL6, WNT, AXL, and therapy-induced stress promote basal-like reprogramming, whereas differentiation-supportive cues including glucocorticoids, anti-inflammatory signaling, p300 and TET2 activity, and CAF-derived niche factors stabilize the classical identity. Intermediate co-expressor populations bridge both programs, highlighting the reversible and plastic nature of PDAC cell states. Adapted from Hayashi et al. [55] and Lomberk et al. [56]. Created in BioRender. Parassiadis, C. (2026) https://BioRender.com/nfax8bs.

Regulation of subtype identity

The regulation of PDAC subtype identity arises from an intricate network of lineage-defining transcription factors, chromatin regulators, and microenvironmental signals that collectively establish and maintain the classical and basal-like phenotypes. While subtypes are primarily defined by transcriptional programs, a large body of work has demonstrated that these programs are not static, but plastic, shaped by epigenetic landscapes and influenced by extrinsic cues. This plasticity underpins both the heterogeneity observed in PDAC and its capacity to evolve under therapeutic pressure.

The GATA6/HNF network and classical differentiation

A key regulator of the classical PDAC identity is the transcription factor GATA6, a member of the GATA family that plays an essential role in pancreas development. Human genetic studies have shown that heterozygous inactivating mutations in GATA6 lead to pancreatic agenesis [57], with complete loss of both exocrine and endocrine compartments and resulting diabetes, while knockout mice require combined loss of Gata4 and Gata6 to recapitulate this phenotype [58]. In the adult pancreas, GATA6 is required for acinar differentiation and maintenance of exocrine function [59]. In PDAC, GATA6 is frequently amplified and overexpressed in classical tumors, whereas its locus is frequently hypermethylated in basal-like cancers, reinforcing suppression of the classical program [60–62]. Functionally, GATA6 defines classical identity by sustaining epithelial gene expression, often in concert with hepatocyte nuclear factors HNF1A and HNF4A, which act as co-regulators of pancreatic differentiation programs [62–64]. Recent work demonstrated that GATA6 specifically promotes classical-specific gene expression by recruiting the positive transcription elongation factor-b (P-TEFb) complex containing cyclin-dependent kinase-9 (CDK9) [65]. By doing so it promotes the release of RNA polymerase II (RNAPII) from a promoter-proximal paused state to an elongation competent form. Notably, acute loss of GATA6 or inhibition of CDK9 resulted in pausing at both enhancers and target genes and, interestingly, stabilization of enhancer-gene interactions.

Loss of GATA6 has been reported to result in the upregulation of basal-like gene expression, while overexpression can restore epithelial features, indicating that GATA6 is both necessary and sufficient for maintaining the classical phenotype [47, 62, 65]. Moreover, low-GATA6 tumors that retain HNF expression appear to represent an intermediate state, whereas low-GATA6 basal tumors epigenetically silence HNFs, suggesting that HNFs provide a barrier to complete basal transdifferentiation [62, 64]. A study in 2021 further showed that epigenetic inactivation of GATA6 through reduced tet methylcytosine dioxygenase 2 (TET2)-mediated hydroxymethylation promotes basal identity, while TET2 stabilization via ascorbate or metformin restores GATA6 expression and classical features, highlighting a direct link between epigenetic regulation and subtype identity [66]. Consistent with these observations, a p300/GATA6 regulatory axis has been identified in which p300-driven histone acetylation sustains GATA6 expression and classical subtype features. Loss of p300, on the other hand, promotes a shift toward basal-like identity, especially in tumors harboring RNF43 mutations and wingless-related integration site (WNT) pathway dependency [67, 68]. This highlights how GATA6 and its regulatory network safeguard the classical program, while epigenetic perturbations can destabilize lineage identity.

The ΔNp63/MYC axis and basal-like reprogramming

In contrast, the basal-like subtype is defined by ΔNp63, an isoform of the tumor protein 63 (TP63) transcription factor that lacks the full-length N-terminal transactivation domain but retains DNA-binding capacity. Normally expressed in keratinocytes and required for epidermal development [69], ΔNp63 is aberrantly activated in basal PDAC, where it drives enhancer reprogramming and squamous transdifferentiation [49, 50]. Chromatin run-on and sequencing (ChRO-seq) studies have revealed that ΔNp63-bound enhancers produce basal-specific enhancer RNAs (eRNAs), providing a transcriptional fingerprint of this program [70]. Notably, in contrast to the effects observed on GATA6-dependent enhancers in classical PDAC, acute loss of ΔNp63 or inhibition of CDK9 had a more drastic effect on basal-specific enhancers and resulted in decreased RNAPII at both the enhancers and target genes and decreased enhancer-gene interactions [65]. Thus, the transcriptional programs driving the two major molecular identities show differential sensitivities which may have significant clinical implications as noted below. Among these target genes ΔNp63 induces inflammatory cytokines, such as interleukin-1 alpha (IL1A) and C-X-C motif chemokine ligand 1 (CXCL1), which recruit neutrophils and activate cancer-associated fibroblasts, thereby coupling transcriptional identity to tumor-stroma crosstalk [71]. In addition, ΔNp63 interacts with the mediator kinase module via mediator complex subunit 12 (MED12) to activate basal enhancers, linking lineage specification to core transcriptional machinery [72]. Rare ΔNp63-positive basal progenitor-like cells have also been identified in the normal pancreas, suggesting that this program may hijack latent developmental lineages [73]. Beyond ΔNp63, other transcription factors reinforce basal identity. Zinc finger BED domain-containing protein 2 (ZBED2), a zinc finger transcriptional repressor, directly antagonizes GATA6 by binding to its promoter and displacing interferon regulatory factor 1 (IRF1), thereby repressing the classical program [74]. IRF1 itself promotes metabolic programs of low-grade PDAC, and its balance with ZBED2 may help determine lineage identity [75]. Similarly, GLI family zinc finger 2 (GLI2), a transcription factor downstream of Hedgehog signaling, can drive basal-like features independently of ligand, repressing GATA6 and inducing epithelial-mesenchymal transition (EMT) and squamous markers, while secreted osteopontin (SPP1) acts as a paracrine mediator of basal reprogramming [76]. These findings highlight how basal identity emerges from a coordinated transcriptional network, sustained by enhancer activation and reinforced by autocrine and paracrine signaling loops.

Epigenetic regulators of subtype identity

The dichotomy between GATA6-driven classical and ΔNp63-driven basal programs is shaped and reinforced by transcriptional and epigenetic regulators. Polycomb repressive complex 2 (PRC2), which mediates deposition of the repressive H3K27me3 mark, has emerged as a key suppressor of the classical program. Enhancer of zeste homolog 2 (EZH2), the catalytic subunit of PRC2, represses GATA6, thereby favoring basal identity. Inhibition of EZH2 or knockdown of PRC2 components reactivates GATA6 and restores classical features [77]. Conversely, loss of the H3K27 demethylase lysine demethylase 6A (KDM6A) promotes basal reprogramming through activation of ΔNp63 and MYC super enhancers, in a sex-specific manner, reflecting compensation by the Y-linked paralog ubiquitously transcribed tetratricopeptide repeat containing, Y-linked (UTY) in males [42]. Mechanistically, HNF1A recruits KDM6A to enhancer-promoter contacts to sustain acinar differentiation, and loss of either HNF1A or KDM6A disrupts this network and promotes basal-like programs [78]. These findings establish KDM6A as a central guardian of the classical lineage and help explain why its loss is enriched in aggressive PDAC. More broadly, enhancer regulation of subtype programs also involves SWI/SNF remodeling complexes and bromodomain and extraterminal (BET) proteins, such as bromodomain-containing protein 4 (BRD4). Notably, BET inhibitors have shown preferential activity against KDM6A-deficient, squamous-like tumors, underscoring the therapeutic relevance of enhancer control in PDAC [42].

Integrative epigenomic profiling has provided a more global view of how chromatin landscapes underpin subtype identity. Lomberk et al. combined ChIP-seq, RNA-seq, and methylation profiling to map the chromatin states of PDAC using patient-derived xenografts and identified distinct enhancer and Polycomb landscapes corresponding to classical and basal subtypes [56]. Classical tumors were enriched for developmental transcription factors and metabolic regulators, while basal tumors showed activation of EMT and mesenchymal-epithelial transition (MET)-driven programs. ChromHMM analysis revealed that nearly 30% of the epigenome is occupied by repressive heterochromatin or Polycomb marks, with functional consequences for cell cycle and adhesion pathways. Importantly, these landscapes were shown to be dynamic and reversible, underscoring the therapeutic potential of targeting epigenetic regulators, such as EZH2. These findings align with subsequent evidence that enhancer reprogramming is central to subtype transitions, whether through ΔNp63, KDM6A loss, or TET2-mediated hydroxymethylation changes.

Microenvironmental and metabolic control of subtype identity

Recent studies have highlighted additional mechanisms linking transcriptional identity to metabolic and microenvironmental states. Loss of GATA6 and HNF4A shifts metabolic programs toward glycolysis, aligning with the metabolic profile of basal tumors [63]. TEA domain transcription factor 2 (TEAD2)-driven enhancer activation has been implicated in anchorage-independent growth and basal reprogramming, linking biomechanical stress to enhancer rewiring and angiogenic programs [79]. Similarly, fibroblast-derived Interleukin 6 (IL6) can drive AXL-JAK-STAT3 signaling through roundabout guidance receptor 3 (ROBO3), promoting basal identity and linking inflammatory cues to transcriptional fate [80]. Tumor necrosis factor alpha (TNFα)-producing macrophages have been shown to reprogram activator protein 1 (AP-1) enhancers and promote basal identity, while colony stimulating factor 1 receptor (CSF1R)+ macrophages sustain squamous programs through T-cell suppression [81, 82]. Spatial single-cell studies further reveal that classical and basal programs coexist in defined niches, with stromal heterogeneity likely facilitating subtype co-existence and influencing therapy response [14]. Moreover, axon guidance cues, such as semaphorin 3A (SEMA3A), can reinforce basal aggressiveness by engaging invasive programs [83]. Together, these findings support that metabolic phenotypes in PDAC are closely linked to transcriptional subtype identity and are influenced by the same transcriptional and epigenetic programs that govern enhancer activity, rather than reflecting purely cell-autonomous metabolic adaptation. These observations place the tumor microenvironment not only as a modulator of therapy resistance but also as an active participant in the determination of subtype identity.

Additional layers of complexity arise from transcription factors beyond GATA6 and ΔNp63. The transcription factor SAM pointed domain containing ETS transcription factor (SPDEF) has been implicated in defining a subset of classical tumors enriched for mucus production. Knockout of SPDEF induces a shift toward basal identity, although it remains unclear whether this occurs directly or through effects on GATA6 [76]. Redundancy between GATA4 and GATA6 has also been observed, and combined low expression is associated with poor outcome, further supporting a cooperative model of classical lineage maintenance [64]. These examples emphasize that subtype regulation is not binary but relies on a network of transcriptional regulators that can substitute or reinforce one another.

Integrated model of transcriptional-epigenetic regulation

Taken together, these findings support a model in which PDAC subtype identity emerges from the interplay of lineage-defining transcription factors (GATA6, HNF1A/4A, ΔNp63, MYC, SPDEF, ZBED2, GLI2), transcriptional regulators (CDK9, MED12), epigenetic regulators (PRC2, KDM6A, BRD4, SWI/SNF, TET2, p300), and extrinsic signals from the tumor microenvironment (IL6, TNFα, macrophages, CAFs, axon guidance cues). Classical identity is characterized by epithelial differentiation, developmental transcription factors, and enhancer landscapes maintained by GATA6, HNFs, and supportive epigenetic regulators. Basal identity, in contrast, arises through enhancer activation by ΔNp63 and MYC, repression of GATA6, and reinforcement by inflammatory and metabolic cues.

In addition to transcription factors and chromatin regulators, non-coding RNAs (ncRNAs) have emerged as modulators of transcriptional and epigenetic programs in cancer, including PDAC [84]. Long non-coding RNAs (lncRNAs) and microRNAs (miRNAs) can influence subtype-associated gene expression by interacting with chromatin-modifying complexes, regulating transcription factor stability, or shaping enhancer activity [85].

Several lncRNAs have been implicated in promoting EMT, stem-like features, and chemotherapy resistance in PDAC models, often by acting as competing endogenous RNAs or by interfacing with transcription factor networks. For example, linc-DYNC2H1-4 promotes EMT and cancer stem cell phenotypes by sponging miR-145 and has been linked to gemcitabine resistance in PDAC cells [86]. In addition, lncRNAs, such as maternally expressed gene 8 (MEG8), have been shown to contribute to epigenetically regulated EMT programs through interactions with chromatin-modifying machinery, providing a mechanistic link between ncRNA function and transcriptional state control, including in pancreatic cancer cells [87]. While direct evidence linking specific ncRNAs to stable classical or basal-like subtype identity remains limited, another review highlights ncRNAs as potential modulators of transcriptional plasticity and EMT-associated programs in PDAC, warranting further investigation [88].

In this framework, metabolic heterogeneity is closely linked to transcriptional and epigenetic state, reflecting subtype-specific transcriptional programs and associated enhancer landscapes. Moreover, metabolic stress and nutrient availability may further modulate chromatin regulation and thereby contribute to cell-state stability. Crucially, these states are not fixed. Single-cell and longitudinal studies consistently demonstrate that PDAC tumor cells are highly plastic and can shift between subtypes under therapeutic and environmental pressure, with intermediate populations bridging classical and basal programs. This plasticity provides both a challenge and an opportunity. On the one hand, it underlies treatment resistance and intratumoral heterogeneity. On the other hand, it highlights the potential of therapeutic reprogramming strategies, particularly through pharmacological targeting of chromatin regulators, to redirect resistant basal-like states toward more chemosensitive classical programs. As such, the mechanistic regulation of PDAC subtype identity is not only central to our understanding of tumor biology but also represents a promising axis for therapeutic innovation.

Heterogeneity in PDAC

Spatial and cellular heterogeneity

PDAC is characterized by remarkable heterogeneity that manifests not only between patients but also within individual tumors and metastatic sites. This variability includes genetic, epigenetic, metabolic, and microenvironmental dimensions and represents one of the principal causes of therapy resistance and disease relapse [89, 90]. Although the distinction between classical and basal-like subtypes provides a useful framework to describe dominant transcriptional programs, accumulating evidence demonstrates that PDAC tumors often contain a mixture of cell states that frequently coexist in spatially organized regions and change dynamically in response to therapy and environmental cues [14, 51].

Recent spatial transcriptomic and imaging studies have demonstrated that intratumoral heterogeneity in PDAC is highly organized rather than random [14, 91, 92]. Classical and basal-like transcriptional programs frequently coexist within the same tumor but are segregated into distinct spatial niches. Within these regions, tumor-intrinsic AP-1 activity together with macrophage-derived TNFα signaling appear to determine which state predominates locally. This spatial organization helps stabilize the coexistence of different subtypes and shapes immune infiltration and therapeutic response. Moreover, inhibition of TNFα in combination with chemotherapy has been shown to enhance CD8⁺ T-cell infiltration and tumor regression, suggesting that local immune-tumor interactions actively modulate drug sensitivity [14].

Additional spatial analyses further confirm this concept. Multiplex immunohistochemistry and quantitative image analysis have shown that the spatial proximity of immune populations, such as CD8⁺ T cells adjacent to IL10⁺ myelomonocytes, correlates with patient outcome [91]. Regions where different immune subsets are positioned closely together show distinct survival patterns compared to spatially segregated areas, underscoring that the physical organization of immune cells is a key component of functional heterogeneity. Complementary spatial-omics approaches have mapped structured patterns of cancer cells, fibroblasts, and immune subsets within the same tissue section, providing direct evidence that PDAC heterogeneity is highly organized in space [14, 91].

Clonal evolution and genetic diversity

Beyond spatial diversity, PDAC displays substantial clonal heterogeneity and branched evolution. Multi-region sequencing and phylogenetic reconstruction have shown that subclones diverge early and acquire additional mutations during tumor progression [93]. Under therapy, small pre-existing subclones may expand and dominate, explaining both intrapatient variability and recurrence after initial response. This evolutionary “forest” model reflects how selective pressure drives the emergence of resistant phenotypes while maintaining shared founder mutations [93]. However, how or whether this genetic diversity is related to molecular subtype identity is unclear. An earlier study from the same group suggested that intrapatient molecular subtype identity did not appreciably correlate with genetic differences, thereby suggesting identity is probably driven largely by environmental signals and epigenetic regulation [55].

Metabolic heterogeneity and adaptive flexibility

Distinct metabolic programs add another dimension to PDAC heterogeneity. Early metabolomic and transcriptomic profiling of PDAC cell lines revealed two dominant metabolic subtypes: a glycolytic subtype, marked by enhanced glucose metabolism and hypoxia-related pathways, and a lipogenic subtype, enriched for lipid biosynthesis and oxidative phosphorylation [94]. These metabolic configurations partially overlap with transcriptional subtypes and predict differential sensitivity to metabolic inhibitors. More recent studies have demonstrated that PDAC cells can flexibly shift between these metabolic states depending on nutrient availability and microenvironmental stress. Even genetically identical clones can differ in nutrient preferences, and secreted metabolites such as asparagine can support neighboring cells during starvation, highlighting how metabolic cooperation contributes to tumor resilience [95]. Beyond intrinsic flexibility, the tumor microenvironment plays a decisive role in shaping these metabolic patterns. Poor vascularization and limited nutrient supply impose selective pressure that promotes nutrient scavenging mechanisms, such as macropinocytosis and autophagy, enabling PDAC cells to recycle macromolecules and sustain growth [96]. In parallel, cancer-associated fibroblasts provide metabolic support by releasing lipids that fuel oxidative metabolism in cancer cells, thereby reinforcing local metabolic heterogeneity [97]. At the mitochondrial level, PDAC cells exhibit remarkable plasticity, dynamically rewiring energy production between glycolysis and oxidative phosphorylation in response to environmental changes [98]. Thus, metabolic heterogeneity in PDAC is a dynamic and adaptive feature, shaped by both intrinsic plasticity and microenvironmental constraints, contributing to the overall complexity of tumor behavior.

Heterogeneity in PDAC is also linked to transcriptional plasticity. Tumor cells can transition between classical and basal-like states through enhancer remodeling and changes in chromatin organization. These transitions are strongly influenced by microenvironmental signals. For example, macrophage-derived cytokines and epigenetic modulation of AP-1 targets can stabilize epithelial programs or drive basal reprogramming depending on the niche context [51]. The presence of intermediate “co-expressor” populations expressing both lineage signatures supports the concept of a continuous spectrum rather than discrete subtypes. Such plastic populations may act as a reservoir for adaptation and resistance. It is, however, worth noting that while TNFα treatment induces a certain degree of “basalness,” it is not sufficient for inducing ΔNp63 expression and a full basal-like phenotype as observed in patients [81, 99].

Stromal and immune heterogeneity

The PDAC stroma contributes substantially to overall heterogeneity. CAFs comprise several functionally distinct subtypes, including myofibroblastic CAFs (myCAFs), inflammatory CAFs (iCAFs), and antigen-presenting CAFs (apCAFs), which differ in their cytokine secretion, matrix remodeling, and interactions with immune cells [100]. MyCAFs maintain extracellular-matrix integrity and stiffness, while iCAFs secrete IL6 and other inflammatory mediators that promote EMT and therapy resistance. Macrophage-fibroblast crosstalk further shapes the immune landscape by coordinating T-cell exclusion and cytokine gradients, reinforcing resistant tumor niches [101]. These findings highlight that stromal heterogeneity is functionally connected to tumor cell plasticity and therapeutic outcome.

Modeling and monitoring heterogeneity

To model this diversity, organoid systems have been developed that preserve both intra- and inter-tumoral heterogeneity. Branched-organoid cultures display distinct morphologies, transcriptional programs, and drug sensitivities that mirror the diversity observed in patient tumors [27]. Such models provide a translational platform to study how heterogeneity influences drug response and to test reprogramming or combination strategies in controlled experimental settings.

The high degree of PDAC heterogeneity complicates molecular classification, as single biopsies often fail to capture the full diversity of the tumor. Moreover, subtype composition and molecular features can evolve during treatment. An unresolved challenge is how best to reconcile high-resolution insights from single-cell and spatial profiling with clinically feasible bulk-based classifiers. While bulk transcriptomic signatures are more readily deployable in routine diagnostics, they may obscure rare but functionally relevant subpopulations that drive resistance and disease progression. To overcome this limitation, liquid-biopsy approaches using cell-free DNA (cfDNA) and methylation profiling are increasingly being explored for dynamic disease monitoring. Serial analysis of the tumor-derived cfDNA fraction, known as circulating tumor DNA (ctDNA), can reveal emerging mutations and methylation patterns associated with therapy resistance and prognosis [102]. These minimally invasive approaches offer the potential to track tumor evolution in real time and to guide adaptive treatment decisions based on dynamic molecular changes. It is conceivable that advances in liquid-biopsy approaches may even facilitate the monitoring of molecular subtype identity longitudinally during therapy in order to adjust treatment approaches. However, the necessary biomarkers and methodology first need to be developed.

Integrating heterogeneity into therapeutic strategy

Taken together, PDAC heterogeneity spans genetic, metabolic, epigenetic, and microenvironmental layers. This complexity explains why PDAC often shows mixed therapeutic responses and rapid adaptation to therapy. Future strategies should aim to integrate spatially resolved and longitudinal profiling into clinical decision-making. Approaches that constrain cellular plasticity, such as epigenetic modulation, or that disrupt supportive niches, for example, IL6/JAK-STAT3 or TNFα signaling, may help reduce heterogeneity and sensitize tumors to treatment. Ultimately, dynamic biomarker-guided approaches will be essential to translate our understanding of heterogeneity into improved patient outcomes.

Clinical implications and therapeutic targeting

Subtype-associated therapy response in PDAC

Despite major advances in our molecular understanding, PDAC remains one of the most treatment-refractory cancers. Chemotherapy remains the backbone of therapy for nearly all patients, yet responses are typically transient and limited by rapid emergence of resistance. Translational research has revealed that the therapeutic intractability of PDAC arises not only from its desmoplastic stroma and late diagnosis but also from its profound molecular heterogeneity, transcriptional plasticity, and metabolic adaptability [4, 53, 103]. These insights now guide efforts to align treatment with tumor biology and to identify targetable vulnerabilities across genetic, epigenetic, and microenvironmental layers.

The most robust biological distinction with clinical implications remains the classical versus basal-like subtype dichotomy. Classical tumors, marked by GATA6 and HNF1A/4A expression, exhibit epithelial differentiation, lower metastatic potential, and better response to fluoropyrimidine-based regimens, such as FOLFIRINOX or NALIRIFOX [22–24]. In contrast, basal-like tumors with ΔNp63 or MYC activation display squamous-like features, transcriptional plasticity, and chemoresistance, leading to inferior outcomes [12, 104, 105].

Transcriptomic subtyping as a clinical stratification tool

Prospective translational trials, such as COMPASS and PASS-01, have confirmed that transcriptomic subtyping can stratify treatment response, with classical tumors showing greater sensitivity to 5-FU-based regimens and basal-like tumors deriving limited benefit, but generally responding better to gemcitabine/nab-paclitaxel treatment compared to FOLFIRINOX [11, 54, 106]. To enable implementation in clinical settings, purity-independent subtyping of tumors (PurIST) provides a robust classifier that distinguishes basal-like and classical tumors from individual samples, independent of tumor purity [105]. Validation studies demonstrated that PurIST classification improves regimen selection and identifies patients more likely to benefit from specific chemotherapeutic combinations. It is currently being prospectively evaluated as a biomarker tool in clinical trials (NCT04683315). Complementary approaches, such as the pancreatic adenocarcinoma molecular gradient (PAMG), extend this binary model into a continuous scale reflecting mixed or intermediate phenotypes that better capture the spectrum of PDAC biology and prognosis [104]. Integration of multi-omics data further refines prognostic modeling, as shown by Fraunhoffer et al. [107], who combined transcriptomic, epigenomic, and metabolomic signatures to define regulatory gradients underlying PDAC heterogeneity and patient outcome. However, this approach remains challenging for routine diagnostic purposes.

Collectively, these stratification systems are beginning to bridge the gap between molecular biology and clinical decision-making. Embedding molecular classifiers into ongoing trials, as exemplified by COMPASS and PASS-01, represents a key step toward biomarker-guided first-line therapy and rational treatment personalization.

Epigenetic and transcriptional targeting to exploit subtype plasticity

While transcriptomic stratification improves patient selection, it does not directly address the dynamic and adaptive nature of PDAC cell states. Given that PDAC subtypes and resistance phenotypes are governed by transcriptional and chromatin networks, epigenetic modulation has emerged as a particularly promising therapeutic avenue [29]. Epigenetic therapies can restore differentiation programs, reduce cellular plasticity, and enhance chemosensitivity, without requiring defined mutations.

At the same time, it is important to recognize that the clinical efficacy of epigenetic therapies as single agents has thus far been limited in solid tumors. Early clinical trials of BET inhibitors, histone deacetylase inhibitors, and other chromatin-modifying agents have generally shown modest response rates and dose-limiting toxicities, underscoring the strong context dependence of epigenetic vulnerabilities and the lack of robust predictive biomarkers [108, 109]. Unlike hematologic malignancies, where epigenetic alterations can represent dominant oncogenic drivers, solid tumors, such as PDAC, exhibit pronounced cellular heterogeneity and adaptive capacity, which may constrain durable responses to epigenetic monotherapy [110]. These limitations have prompted a conceptual shift toward the use of epigenetic agents as priming or sensitizing therapies rather than stand-alone treatments.

Among epigenetic strategies, inhibition of BET proteins and CDK9 shows particular promise. BET inhibitors (e.g. JQ1, OTX015) displace BRD4 from super-enhancers that sustain MYC and enhancers controlling the basal/squamous transcriptional programs, thereby repressing ΔNp63- or MYC-driven basal-like phenotypes, with heightened sensitivity observed in KDM6A-deficient or enhancer-hyperactivated tumors [42, 111, 112]. CDK9 inhibitors, which suppress transcriptional elongation, have demonstrated synergy with BET inhibition and enhanced efficacy when combined with cytotoxic chemotherapy in preclinical PDAC models [113, 114].

Epigenetic regulators also critically influence subtype plasticity. Inhibition of the Polycomb Repressive Complex 2 (PRC2) component EZH2 can derepress GATA6 and promote basal-to-classical subtype switching, restoring differentiation programs and enhancing chemosensitivity [77]. This principle aligns with the demonstration that the glucocorticoid receptor suppresses GATA6-mediated RNAPII pause release, thereby stabilizing transcriptional pausing and repressing classical identity genes, highlighting a dynamic axis through which transcriptional pausing and enhancer regulation shape PDAC subtype identity and therapeutic response.

Building on these insights, epigenetic priming (pretreatment with enhancer-targeting drugs such as BET or EZH2 inhibitors) can reprogram transcriptional networks, inhibit or reactivate tumor cell identity programs, and increase subsequent chemotherapy sensitivity in PDAC models [29, 115]. Mechanistically, therapy adaptation converges on interactive enhancer hubs (iHUBs) that are AP-1/JunD-driven, BRD4-enriched enhancer clusters characterized by high enhancer RNA output that coordinate adaptive transcriptional reprogramming and drug resistance [116]. Deletion of a single iHUB or pharmacologic attenuation of iHUB activity using low-dose CDK9 inhibition (enitociclib) or mitogen-activated protein kinase (MEK) inhibition (trametinib) restores chemosensitivity in vitro and in patient-derived xenografts [116]. Together, these data support priming-based combination or sequential strategies to prevent enhancer-mediated adaptation and to integrate epigenetic therapy into precision treatment paradigms for PDAC. Nevertheless, it remains an open question whether epigenetic reprogramming can induce stable and durable subtype transitions in vivo, or whether tumor cells ultimately reacquire resistant transcriptional states through alternative adaptive pathways once selective pressure is relieved. Defining the durability, timing, and reversibility of such reprogramming will be critical for the rational clinical deployment of epigenetic priming strategies.

Translating these mechanistic insights into clinical benefit requires integration of molecular subtyping, dynamic monitoring, and functional profiling. The PASS-01 trial demonstrated that classical PDAC responds markedly better to FOLFIRINOX, whereas basal-like tumors are comparatively resistant, emphasizing the therapeutic relevance of subtype identity. Epigenetic inhibition of EZH2 or BET proteins may promote classical differentiation and restore FOLFIRINOX sensitivity, suggesting that transcriptional reprogramming could modify therapy response rather than merely predict it. Figure 2 illustrates this concept.

Figure 2.

Translational framework linking PDAC subtype biology to therapeutic personalization. (A) Conceptual framework for epigenetically guided precision therapy in PDAC, integrating patient-centered supportive care, molecular and epigenetic profiling (GATA6, ΔNp63, EZH2, BRD4, CDK9, KDM6A, TET2), and patient-derived organoid chemoprofiling to inform therapy selection. Epigenetic priming with EZH2 or BET inhibitors may promote classical differentiation, enabling subtype-matched chemotherapy (FFX/NALIRIFOX for classical, gemcitabine ± epigenetic therapy for basal-like). Longitudinal monitoring through liquid biopsy and transcriptomic profiling allows dynamic adaptation of treatment according to resistance and transcriptional drift. (B) Schematic representation inspired by the PASS-01 clinical trial [11], illustrating differential response of classical and basal-like PDAC to FOLFIRINOX. The conceptual model illustrates how epigenetic reprogramming (EZH2 or BET inhibition) could shift basal-like, chemoresistant tumors toward a classical, more chemosensitive state, thereby improving therapeutic efficacy. Created in BioRender. Parassiadis, C. (2026) https://BioRender.com/6061a0q.

Functional genomic discovery and rational combination strategies

To systematically identify actionable vulnerabilities and rational combinations, functional genomic approaches have become increasingly important. Large-scale clustered regularly interspaced short palindromic repeats-associated protein 9 (CRISPR-Cas9) screens have revolutionized the discovery of genetic dependencies and therapy sensitizers in PDAC. Ramaker et al. [117] performed genome-wide CRISPR activation and knockout studies and showed that increased expression of NCoR/SMRT-HDAC co-repressor complex components promotes EMT and drug resistance, whereas their inhibition can restore sensitivity to gemcitabine and 5-FU. Martinez et al. [118] used an in vivo CRISPR screen in autochthonous PDAC models to identify SR-related CTD associated factor 1 (SCAF1) and ubiquitin specific peptidase 15 (USP15) as context-dependent vulnerabilities whose loss enhanced tumor growth but also increased sensitivity to gemcitabine and PARP inhibition. In a complementary study, Ubhi et al. [119] demonstrated that APOBEC3C/D cytidine deaminases sustain replication-stress tolerance and protect PDAC cells from DNA-damage inducing therapies, thereby linking the APOBEC-DNA damage response axis to drug resistance. Together, such unbiased functional screens provide powerful tools to uncover druggable epigenetic and transcriptional nodes that can be leveraged for synthetic-lethal or combination-therapy approaches in PDAC. Beyond discovery, CRISPR-based perturbation models may enable systematic evaluation of epigenetic-chemotherapy interactions, identifying candidate targets that sensitize basal-like PDAC to cytotoxic or targeted therapies. These functional genomic frameworks thus help translate mechanistic insights into actionable strategies.

Resistance in PDAC arises through multiple, often overlapping mechanisms, including stromal shielding, enhancer remodeling, metabolic adaptation, and stress-response activation [53, 103]. Enhancer reprogramming via iHUBs contributes to adaptive tolerance by reorganizing transcriptional output in response to chemotherapy [116]. Moreover, FOSL1-RELA (AP-1/NF-κB) signaling has been identified as a driver of PDAC metastasis and its inhibition reduces invasion [99].

High-throughput drug-combination screens have revealed reproducible synergistic interactions in PDAC. Jaaks et al. [120] reported strong synergy for BET + MEK, CDK9 + gemcitabine, and HDAC + chemotherapy combinations, underscoring the potential of data-driven approaches to rational combination design. In parallel, efforts targeting Kirsten rat sarcoma viral oncogene homolog (KRAS), the principal oncogenic driver in PDAC, have achieved major breakthroughs with KRAS^G12D-selective inhibitors such as MRTX1133, which show potent preclinical activity and are advancing into clinical trials [121]. Combining KRAS blockade with epigenetic or metabolic inhibitors may help prevent rapid adaptive resistance and represents a promising next wave of targeted strategies in PDAC.

Integrating molecular stratification into future therapeutic frameworks

Recent integrative reviews emphasize that meaningful clinical progress will require a multimodal strategy combining molecular subtyping, functional genomics, and rational therapy design [4, 53, 103]. Epigenetic and transcriptional targeting must be incorporated into precision oncology frameworks, ideally guided by biomarkers, such as GATA6, ΔNp63, KDM6A, or MYC. Patient-derived organoid testing, spatial transcriptomics, and liquid-biopsy-based monitoring can help track evolving tumor states in real time, supporting adaptive therapy selection [11, 107, 115]. In addition, the mentioned studies advocate combining molecular profiling with clinical and metabolic parameters to optimize predictive models. Stromal reprogramming, rather than ablation, remains a parallel objective, with selective targeting of IL6/JAK-STAT3, TNFα/AP-1, or fibroblast subtypes offering opportunities to improve chemotherapy efficacy [100, 103]. Altogether, the convergence of molecular stratification, epigenetic modulation, and rational combination therapy defines a translational roadmap for PDAC treatment. By integrating these layers into biomarker-guided clinical trials, PDAC therapy may evolve from empiric chemotherapy toward personalized, mechanistically informed intervention.

Future perspectives and conclusions

Over the past decade, major progress has been made in deciphering the transcriptional and epigenetic basis of PDAC. These studies have redefined PDAC as a continuum of molecular identities: from classical, epithelial-differentiated to basal-like, mesenchymal, and therapy-resistant states, rather than a static disease entity. This concept of subtype plasticity, governed by transcriptional regulators such as GATA6, HNF1A/4A, and ΔNp63, together with chromatin modifiers including EZH2, KDM6A, and BRD4, has fundamentally reshaped our understanding of PDAC biology. Yet, despite these insights, the translation of this molecular knowledge into durable clinical benefit remains limited.

Epigenetic mechanisms offer both a challenge and an opportunity: they drive the adaptability that enables therapeutic resistance, but they also provide a means to therapeutically reprogram resistant states. Targeting key chromatin regulators, such as EZH2 inhibition to reactivate GATA6 or BET and CDK9 inhibition to suppress enhancer-driven basal programs, has shown potential to restore differentiation and chemosensitivity. These strategies, combined with rational chemotherapy or targeted agents, may transform transient responses into sustained control.

Future efforts should aim to integrate molecular subtyping, dynamic disease monitoring, and epigenetic targeting into a unified clinical framework. Transcriptomic classifiers, patient-derived organoids, and liquid biopsies can collectively inform precision treatment, capturing evolving tumor states in real time. Embedding these approaches within biomarker-driven clinical trials will be essential to validate predictive markers, refine therapeutic timing, and identify effective drug combinations.

Ultimately, the convergence of molecular stratification, epigenetic modulation, and adaptive trial design represents a realistic path forward. By harnessing transcriptional plasticity rather than fighting it, it may become possible to reprogram resistant pancreatic cancer toward more differentiated, therapy-sensitive states, bringing the goal of truly personalized treatment closer to clinical reality.

Authors’ contributions

C.P. conceived the review concept, performed the literature research, and wrote the manuscript. S.A.J. supervised the work, provided conceptual guidance, and critically revised the manuscript. Both authors read and approved the final version.

Funding

This work was supported by the Robert Bosch Stiftung and conducted within the Robert Bosch Center for Tumor Diseases (RBMF/RBCT), Stuttgart, Germany.

Conflicts of interest

The authors declare that there is no conflict of interest in this study.