Initial epigenome-wide association studies (EWASs) in keloids have relied on microarray platforms such as the Illumina 450K array, which identified thousands of differentially-methylated CpG sites (DMCs) associated with collagen genes and tumor suppressors (15). The subsequent adoption of the Illumina EPIC 850K array, which exhibits higher coverage than the Illumina 450K array, supported and expanded these signatures, revealing additional DMCs in enhancer regions and demonstrating notable cross-platform concordance. Specifically, >90% of the 450K array's CpG content is preserved on the EPIC 850K array, with overlapping probes exhibiting high correlation (R² >0.98), ensuring comparability between legacy and newer datasets (16). Although these bulk-tissue analyses have been shown to provide a comprehensive overview of epigenetic changes, they can also mask cellular heterogeneity. The emergence of single-cell methyl-CpG binding domain sequencing (scMBD-seq) and single-nucleus DNA-methylation sequencing has begun to deconvolute this complexity, having revealed that a notable proportion of scar-specific DMCs are concentrated in α-SMA-positive fibroblasts (16). The cell-type-specific resolution of these techniques is important for pinpointing the primary drivers of fibrosis, such as upregulation of the de novo methyltransferase DNMT3B within the fibroblast population (16). Furthermore, a number of techniques, such as whole-genome bisulfite sequencing, offer base-pair resolution, uncovering hypermethylation events at key regulatory loci, for example the forkhead box F2 (FOXF2) promoter in keloid fibroblasts (17). The functional consequences of these methylation changes have been further supported by studies using CRISPR-dCas9 systems fused to the catalytic domains of DNMTs, such as DNMT3A, or demethylases, for example TET1; this has allowed for locus-specific epigenetic editing to elucidate the relationships between specific modifications and scar formation (18).

A rapidly expanding area of the epigenetic toolbox has focused on sequencing RNA modifications, particularly m6A, which has emerged as an important regulator of mRNA stability, splicing and translation in fibrosis (13,22). Key components of the m6A machinery, including writers, such as METTL3 and METTL14, erasers, for example α-ketoglutarate-dependent dioxygenase FTO (FTO) and α-ketoglutarate-dependent dioxygenase alkB homolog 5 (ALKBH5), and readers, such as YTH domain-containing family protein 1/2/3 and YTH domain-containing protein 1, have been implicated across various fibrotic conditions, including cutaneous scarring (13,22). Multiple studies have consistently reported an upregulation of METTL3 in fibrotic conditions. For example, METTL3-mediated m6A modifications have been shown to: i) Promote the progression of HTSs by stabilizing primary-miR-31 transcripts (19); i) contribute to renal fibrosis by enhancing Ena/vasodilator-stimulated phosphoprotein-like protein and SPARC-related modular calcium-binding protein 2 mRNA stability (20,21); and iii) drive cardiac fibrosis post-myocardial infarction (21). Conversely, upregulation of the demethylase FTO has also been frequently reported, and has been shown to promote keloid formation by increasing COL1A1 expression (22) and aggravate renal fibrosis via runt-related transcription factor 1 upregulation (23). This apparent paradox, in which both the METTL3-mediated addition and FTO-mediated removal of m6A can exert pro-fibrotic effects, highlights the notable importance of elucidating the specific mRNA targets and cellular context of such epigenetic modifications.

The m6A methylation of lncRNAs, such as MALAT1, has been shown to create feed-forward loops that recruit DNMTs, for example DNMT3A, to silence anti-fibrotic miRNAs, such as miR-29b. This establishes a direct link between RNA and DNA methylation layers (19). Technologies such as methylated-RNA immunoprecipitation sequencing or m6A sequencing are instrumental in generating transcriptome-wide m6A maps (24–26), whereas techniques such as methyl-RNA immunoprecipitation followed by DNA pull-down are capable of physically connecting m6A-modified RNAs with DNA loci (19). The functional role of m6A has been validated using pharmacological inhibitors, such as STM2457 for METTL3 (19,27), and CRISPR-dCas13b for site-specific m6A erasure (19).

The integration of single-cell RNA sequencing (scRNA-seq) and epigenetic assays represents a transformative advancement in epigenomics. scRNA-seq of human fibrotic skin has unveiled immune-cell heterogeneity and identified distinct fibroblast subpopulations driving collagen production (28,29). This cellular resolution is now being coupled with epigenetic readouts. scATAC-seq has revealed cell-type-specific chromatin accessibility changes, demonstrating that although fibroblasts exhibit the majority of epigenetic alterations in scar tissues, endothelial and immune cells also contribute notably (30). The most recent advancements in the epigenetic toolbox comprise spatial transcriptomics and spatial-assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq), which preserve the architectural context of tissues (8). For example, spatial enhanced resolution omics-sequencing (Stereo-seq; also referred to as spatio-temporal enhanced resolution omics-sequencing in some literature) can associate open chromatin regions at fibrotic-gene enhancers with mechanical-force vectors within the scar tissue microenvironment, providing a direct spatial link between physical cues and epigenetic states (30). Collectively, the integration of single-cell multi-omics data (including scRNA-seq and scATAC-seq) with spatial transcriptomic maps has emerged as a powerful strategy to reconstruct the gene regulatory networks that sustain fibrosis across different cell types and spatial niches (28–30). Furthermore, the simultaneous spatial co-profiling of DNA methylation and the transcriptome from the same tissue section has recently become feasible at near single-cell resolution, opening new avenues for dissecting the spatial epigenetic landscape of cutaneous fibrosis (1).

The collective evidence obtained using these advanced epigenomic sequencing technologies, including bulk, single cell and spatial platforms, highlights the complex but converging nature of epigenetic dysregulation in fibrosis. There has been notable consistency across studies regarding the hypermethylation of anti-fibrotic gene promoters, such as RASAL1 and PTEN, and the central role of DNMT3B in fibroblasts (15,16). Similarly, the involvement of the m6A machinery, particularly METTL3 and FTO, has proven a reproducible finding, albeit with context-dependent effects on different target genes (19,20,22,23). However, it should be noted that discrepancies remain, such as variations in the reported expression and function of the lncRNA MALAT1, which may be influenced by biopsy site, hypoxia and cellular heterogeneity (31,32). A notable limitation of numerous studies is the reliance on bulk-tissue analysis, which averages signals across diverse cell types and may obscure key cell-specific events. The transition to single-cell and spatial multi-omics directly addresses this limitation, revealing that epigenetic changes are not uniformly distributed across cell types but are concentrated in specific pathogenic subpopulations (28–30). Another challenge lies in the functional validation of specific epigenetic modifications; although CRISPR-based epigenome editing has proven effective, off target effects persist as a concern, as the efficiency and specificity of epigenetic modification are highly dependent on the native chromatin environment at the target locus (5). Furthermore, the difficulty of fully recapitulating this endogenous chromatin context, referring to the native, three dimensional chromatin microenvironment including nucleosome positioning, histone marks, chromatin looping and nuclear architecture in which regulatory elements naturally reside, in simplified experimental systems remains an obstacle to accurately determining the specificity of effects (11).

A critical appraisal of the evidence level underscores the translational gap. The evidence level for epigenetic modifications varies substantially by mechanism and by the rigor of functional validation. As summarized in Table SI, functional studies derived from in vitro fibroblast cultures or small-scale animal models (rabbit ear, murine) represent Level III supporting pre-clinical evidence, whereas human tissue correlative studies (EWAS, RNA-seq) constitute Level II translational evidence. First-in-human interventional data remain scarce and represent Level IIa evidence (early-phase clinical trials). A number of mechanistic insights into the epigenetic regulation of scar formation, especially those involving functional validation via miRNA mimics, lncRNA knockdown or epigenetic editors, have been derived from in vitro fibroblast cultures or small-scale animal models, such as rabbit ear and murine models, which represent level III supporting pre-clinical evidence according to the Oxford Centre for Evidence-Based Medicine (OCEBM) Levels of Evidence (March 2009), where Level III encompasses evidence obtained from well-designed, quasi-experimental studies including non-randomized controlled single-group, pre-post, cohort, time series or matched case-control series (33). Landmark correlative studies in human tissue, such as EWAS and studies employing RNA-sequencing (RNA-seq), provide robust level II translational evidence, but the functional implications of the results of such studies often require validation in more physiologically relevant systems. First-in-human interventional data, such as the aforementioned trial using topical 5-aza-2′-deoxycytidine, remains scarce and constitutes level IIa evidence, representing early-phase clinical trials; this highlights the necessity for larger, controlled studies in patient cohorts (11). The hierarchy of evidence highlights a notable translational gap, in which robust mechanistic data derived from level III pre-clinical models currently outweighs level I/II clinical evidence, underscoring the necessity of high-quality, randomized controlled trials to validate the roles of epigenetic interventions in humans. This gap encompasses the entire pathway from target identification through functional validation, delivery system development, and regulatory approval, rather than merely the interpretation of sequencing results (8,11,28,34,35). Table SI (8,11,12,18,28–30,34–40) summarizes the level of evidence and key limitations for major study types discussed in the present review, providing a structured framework for assessing the current state of the field. As these technologies and studies evolve, they will enable a more precise, dynamic and integrated understanding of the epigenetic regulation of scar formation, paving the way for novel, mechanism-based therapeutic interventions.

The first EWAS performed in keloid tissues was reported by Jones et al (34), which identified 1,819 DMCs across 28 keloid lesions using the Illumina 450K array. Hypomethylation was found to be enriched in collagen-related genes, for example COL1A1 and COL3A1, while hypermethylation was observed in anti-fibrotic loci, such as RASAL1 and PTEN. A more recent study reported by Alghamdi et al (35) expanded the cohort of sequenced pathological scar samples to 48 keloids using the higher-resolution EPIC 850K array and detected 3,214 DMCs, 62% of which overlapped with those identified in the study reported by Jones et al (34). This cross-platform concordance strengthened the validity of the identified keloid-specific methylation signatures. However, the EPIC 850K array also captured an additional 1,395 DMCs, a number of which resided in enhancer regions not covered by the 450K array. Notably, both studies reported consistent Δβ values (i.e., the difference in methylation beta-values between scar tissue and normal skin, ranging from −1 to 1, where positive values indicate hypermethylation and negative values indicate hypomethylation) (±0.20) at key fibrotic loci, suggesting that technical differences did not obscure biological signals. However, the absence of longitudinal sampling in these studies limited insight into whether identified DMCs were causal or consequential to fibrosis, an issue that has been increasingly addressed by single-cell and functional studies.

Although bulk-tissue analyses have provided population-level snapshots of epigenetic modifications in pathological scars, scMBD-seq has revealed that 64% of keloid-DMCs reside within α-SMA-positive fibroblasts (35). Within this cellular subset, DNMT3B expression was revealed to be 2.3-fold higher than in matched normal dermal fibroblasts, associating with de novo methylation at the RASAL1 and PTEN promoters. CRISPR-Cas9 knockout of DNMT3B has been shown to restore RASAL1 expression and reduce collagen-I secretion by 38%, indicating that the DMC at this locus played a causal role in fibrosis (35). Notably, DNMT1 protein levels remained unchanged following DNMT3B knockout compared with control fibroblasts, which aligned with the established maintenance function of DNMT1 (i.e., copying existing methylation patterns during DNA replication rather than creating new ones). This observation reinforced the concept that de novo methylation, driven by DNMT3B, rather than propagative methylation mediated by DNMT1, drove early scar establishment. A study reported by Stevenson et al (42) demonstrated that FOXF2, a key transcriptional regulator of fibroblast identity, was hypermethylated in keloid fibroblasts compared with normal dermal fibroblasts, resulting in FOXF2 silencing and subsequent collagen upregulation. Collectively, these findings have positioned DNMT3B as a prime therapeutic target for reducing pathological scar formation, with treatment strategies involving siRNA-mediated DNMT3B knockdown having already shown efficacy in pre-clinical models (43).

In addition to selective methylation, the active demethylation pathway, which is governed by TET enzymes, has emerged as an important regulator of fibrotic-gene expression. A study reported by Liu et al (44) showed that hypoxia reduced TET2 activity in fibroblasts, leading to a 40% reduction in 5-hydroxymethylcytosine levels at the TGF-β1 promoter, resulting in its subsequent transcriptional de-repression. This observation was corroborated by a study reported by Niu and Tan (45), which demonstrated that TET2 upregulation in keloid explants increased 5-hydroxymethylcytosine levels at the type I collagen α2 chain enhancer and therefore decreased collagen deposition by 31%. Notably, both studies demonstrated that treatment with ascorbic acid, a cofactor for TET enzymes, restored 5-hydroxymethylcytosine levels and attenuated fibrotic-gene expression, suggesting a potential nutraceutical route to modulating scar methylation. These findings aligned with the broader observation that keloid tissue exhibits globally reduced 5-hydroxymethylcytosine content compared with normal skin tissue (44), reinforcing the therapeutic potential of TET activators in reducing scar formation.

Although gene-centric analyses have dominated the literature, emerging evidence has implicated repetitive-element methylation in scar pathogenesis. A study by Meevassana et al (46) reported that Alu (a short interspersed nuclear element, SINE) elements and long interspersed nuclear element-1 (LINE-1) repeats exhibited an 8–12% increase in methylation in burn scar tissues compared with uninjured skin. Another study reported by Prabsattru et al (36) extended this observation to keloids, demonstrating that LINE-1 methylation positively correlated with collagen density (r=0.62; P<0.001) and predicted recurrence after surgical excision. These findings implied that repetitive-element methylation serves as both a biomarker and a potential therapeutic target in fibrosis, although the mechanistic link between repetitive-element methylation and fibrotic-gene expression remains speculative. These findings implied that repetitive-element methylation serves as both a biomarker and a potential therapeutic target in fibrosis, although the mechanistic link between repetitive-element methylation and fibrotic-gene expression remains speculative. The present review proposes a hypothesis that repetitive-element methylation influences chromatin architecture, thereby modulating the accessibility of nearby collagen promoters, a concept that warrants further investigation using chromosome-conformation capture technologies.

The translation of observational data into therapeutic strategies has been demonstrated in several pre-clinical models. In a porcine excisional model of hypertrophic scars (HTSs), intra-lesional delivery of DNMT3B-targeting siRNA encapsulated in polylactic-co-glycolic acid (PLGA) microspheres achieved a 60% knockdown of DNMT3B levels and decreased scar height by 1.2 mm compared with scrambled control siRNA (43). Similarly, a study by Sharma et al (11) demonstrated that subconjunctival administration of 5-aza-2′-deoxycytidine prevented excessive scarring after glaucoma-filtration surgery in rabbits, as evidenced by a 45% reduction in collagen deposition compared with vehicle-treated controls. These intervention studies not only validated DNA methylation as a driver of fibrosis but also provided dose-response benchmarks for future clinical studies. Specifically, 5-aza-2′-deoxycytidine is a well-established DNA methyltransferase (DNMT) inhibitor that induces passive DNA demethylation by incorporating into DNA and trapping DNMT enzymes, thereby reactivating silenced anti-fibrotic genes. In the referenced studies, topical or subconjunctival administration of 5-aza-2′-deoxycytidine led to reduced collagen deposition and scar volume, supporting a causal role of aberrant DNA hypermethylation in maintaining fibrotic phenotypes (11,43). Notably, the absence of systemic toxicity in these models supports the feasibility of localized epigenetic modulation as a clinically viable strategy targeting scar formation.

In summary, DNA methylation acts as a central regulator of cutaneous fibrosis by silencing anti-fibrotic genes and amplifying collagen production. Genome-wide scans have delineated robust keloid-specific signatures and single-cell analyses have pinpointed DNMT3B as a key effector of collagen deposition in fibroblasts. Loss of TET-mediated hydroxymethylation and hypermethylation of repetitive elements have been shown to exacerbate fibrotic-gene expression. Notably, pre-clinical interventions targeting these methylation drivers have demonstrated notable anti-scar therapeutic efficacy, laying the groundwork for precision epigenetic therapies.

In contrast to miR-29, miR-21-5p is the most consistently upregulated miRNA in fibrotic skin, with an average log₂-fold increase of 1.38 across six datasets (38,49–53). The upregulation of miR-21 has been shown to enhance collagen secretion by ~55%, whereas treatment with 25 nM antagomiR-21 (a chemically modified, single-stranded oligonucleotide that specifically binds to and inhibits the function of miR-21-5p) has been shown to reverse this pro-fibrotic phenotype (38,49). Mechanistically, miR-21 targets PTEN and Smad7, thereby amplifying both the PI3K/AKT and TGF-β signaling pathways (50,52,53). However, not all cohorts behave identically. Two pediatric studies on HTSs in white individuals reported only a marginal upregulation of miR-21 expression (0.2-fold), highlighting the influence of ancestry, age and anatomical site on miR-21 expression levels, which may in turn affect scar pathogenesis (49,50). Similarly, miR-155 has been found to exhibit a 1.29 log₂-fold upregulation and promote connective tissue growth factor-mediated α-SMA expression (54,55). Studies have demonstrated that a dual antagomiR-21/155 cocktail achieved 58% collagen reduction ex vivo, outperforming single reagents by ~30% and logically supporting combined miRNA targeting as a next-generation strategy (49,55).

Beyond the aforementioned groupings, several miRNAs have displayed reproducible context-specific alterations in epigenetic programming in scar tissue. For instance, miR-145-5p and miR-31-5p are frequently upregulated in scar tissue and have been found to enhance myofibroblast conversion by targeting Krüppel-like factor 4 and Ras homolog family member A, respectively (56,57). Conversely, miR-152-5p, miR-194-5p and miR-203 are downregulated in scar tissue; the experimental overexpression of these miRNAs (achieved via transfection of miRNA mimics) has been shown to suppress fibroblast proliferation and migration, both of which are key processes in scar formation and pathological fibroblast activation, by targeting Smad3, nuclear receptor subfamily 2 group F member 2 and early growth response 1 (58–61). Single-cell quantitative PCR (qPCR) has further revealed that miR-200b/c loss is restricted to a COL1A1-high fibroblast subset, explaining why bulk-tissue analysis occasionally misses this signal (coefficient of variation, CV, 28%) (49,72).

While miRNAs act as fine-tuners in epigenetic reprogramming, lncRNAs serve as scaffolding molecules that recruit chromatin modifiers or as competitive endogenous RNAs that sponge miRNAs. This dual functionality makes them attractive but complex therapeutic targets.

H19 is the most well-characterized lncRNA in keloid biology. Chromatin isolation by RNA purification with high-throughput sequencing (ChIRP-seq) data has demonstrated H19 occupancy at the miR-29b promoter, where it recruits DNMT3B, thereby coupling RNA-guided targeting with DNA methylation (39). ASO-mediated H19 knockdown has been shown to de-repress miR-29b by ~2.5-fold and reduce collagen-I expression by ~30% in keloid-derived fibroblasts (39). However, it has been noted that the expression and functional impact of H19 in pathological scars may exhibit heterogeneity and that this could potentially be influenced by a number of factors, such as biopsy site (62).

MALAT1 displays divergent trends in expression; for example, a 2.4-fold increase in normoxic cultures of hypertrophic scar fibroblasts vs. a 0.1-fold decrease in dexamethasone-treated hypertrophic scar fibroblasts under standard oxygen conditions, depending on biopsy site and variable oxygen tension (ranging from normoxia to hypoxia) (63,64). siRNA-mediated MALAT1 silencing has been shown to reduce fibroblast migration by 22%, whereas 5-aza-2′-deoxycytidine-induced demethylation of MALAT1 has been found to upregulate MALAT1 expression while still reducing collagen deposition (64). This apparent contradiction suggests that MALAT1 is not consistently a pro-fibrotic driver across all contexts. Instead, its upregulation following DNMT inhibitor treatment may reflect a pharmacodynamic by-product of on-target demethylation, positioning MALAT1 as a potential pharmacodynamic biomarker rather than a direct effector of fibrosis. Notably, this interpretation does not exclude the possibility that MALAT1 exerts pro-fibrotic effects in other settings, as reported in certain scar subtypes or under hypoxic conditions (64). This paradox positions MALAT1 as a pharmacodynamic biomarker of on-target demethylation rather than a driver, logically supporting a ‘dual-hit’ strategy to prevent rebound, for example co-treatment with a DNMT inhibitor and MALAT1 ASO (64,65).

HOXA11-AS, TUG1 and COL1A2-AS1 have been repeatedly found to be upregulated in scar tissue; these lncRNAs have been demonstrated to promote proliferation, collagen synthesis and epithelial-mesenchymal transition via miR-124-3p sponging and subsequent activation of Smad5 signaling, miR-27b-3p sponging and phosphorylated-Smad3 sponging, respectively (39,66,67). Individual knockdown of these lncRNA has been shown to decrease collagen-I expression by 20–30%, and the pro-fibrotic role of HOXA11-AS has been functionally validated, providing supportive evidence that targeting this lncRNA may serve as a potential therapeutic strategy for reducing scar formation (39).

High-depth rRNA-depleted sequencing has uncovered 91 differentially expressed circRNAs in keloid tissue (40,68). CircPTK2 and circFNDC3B act as miR-19a-3p and miR-29b sponges, respectively; their suppression has been shown to increase the availability of these miRNA and therefore decrease collagen-I expression by ~20% (40). Similarly, circ_0057452 and circSLC8A1 have been demonstrated to sponge miR-7-5p and miR-27b-3p, relieving vascular endothelial growth factor A (VEGFA) and C-X-C motif chemokine 2 repression (69,70). However, the sample sizes in these high-depth rRNA-depleted sequencing studies remain small, typically involving no more than six biological replicates (n≤6) (40,69), and to the best of our knowledge, in vivo delivery methods for therapeutic modulation of circRNAs in cutaneous fibrosis remain unexplored (8).

The RNA-m6A writer METTL3 has been found to be upregulated 2.1-fold in scar fibroblasts compared with normal dermal fibroblasts, and its pharmacological inhibition by STM2457 (5 µM) has been shown to reduce collagen secretion by 28% (13). Mechanistically, m6A-marked MALAT1 transcripts physically tether DNMT3A to the miR-29b promoter, creating a feed-forward loop of RNA-methylation to DNA-methylation (13). Site-specific m6A erasure on MALAT1 transcripts via CRISPR-dCas13b has been shown to decrease DNMT3A occupancy of the miR-29b promoter by 27% and elevate miR-29b expression 1.7-fold, underscoring the therapeutic value of targeting RNA modifications upstream of DNA methylation (13). Additionally, ALKBH5-mediated demethylation of circGLIS3 has been demonstrated to ameliorate ECM deposition, which is a key pathological event in scar formation, as excessive accumulation of ECM components, particularly collagen, leads to dermal thickening, tissue stiffness and loss of normal skin architecture (71). This finding reveals a broader m6A-based circuitry that intersects with both lncRNA and circRNA networks.

Although antagomiRs and miRNA mimics have demonstrated efficacy in reducing collagen deposition and suppressing fibrotic gene expression in vitro (37,38,48,49), the RNase-rich dermis and negatively charged cell membranes pose notable barriers to therapeutic delivery. Specifically, transfection of miR-29b mimics into keloid-derived fibroblasts reduced COL1A1 mRNA expression by >40%, while antagomiR-21 treatment reversed the pro-fibrotic phenotype by downregulating collagen synthesis and promoting matrix degradation (38,49). These effects are mediated through post-transcriptional regulation of target mRNAs involved in ECM production and turnover, rather than through direct modification of DNA methylation. Cationic DOPC liposomes have been shown to achieve 600-µm penetration in human scar explants and maintain miR-29b activity for 72 h (73). Furthermore, PLGA-polyethylene glycol microspheres engineered to release H19 ASOs have been shown to exhibit zero-order kinetics over 14 days of administration, resulting in a 38% reduction in scar height in rabbit ears (73). Additionally, dissolvable hyaluronic acid (HA)-microneedle (MN) arrays loaded with miR-141-3p-functionalized exosomes have produced a 1.2 mm reduction in HTS thickness (i.e., flattening) in a porcine model, with no detectable systemic leakage of exosomes or their miRNA cargo into the circulation (73). These studies have collectively provided level IIa evidence that localized ncRNA delivery represents a feasible therapeutic strategy targeting scar formation and logically support a ‘needle-free’ patient-friendly formulation for future clinical studies.

Histone acetylation, which is generally associated with open chromatin and gene activation, is tightly regulated by histone acetyltransferases and HDACs. In fibrotic conditions, a pattern of HDAC upregulation and hyperactivity emerges, leading to the repression of anti-fibrotic genes. For example, HDAC4 has been identified as an important mediator of transcriptional activity that deacetylates the transcription factor myocyte-specific enhancer factor 2A (MEF2A), which in turn silences the expression of Smad7, a potent inhibitor of TGF-β signaling. This HDAC4/MEF2A/Smad7 axis is important for fibroblast activation and HTS formation (74). Similarly, HDAC5 contributes to fibrosis by deacetylating MEF2A and, by extension, repressing Smad7 (74), indicating a shared mechanism of anti-fibrotic gene repression among class IIa HDACs (74) (Fig. 2).

The roles of specific HDACs in skin fibrosis can be context-dependent. HDAC6, for example, has been reported to promote wound healing in skin tissues by regulating the migration and differentiation of fibroblasts in aged mice (75), suggesting that this enzyme plays a beneficial role in acute wound repair. Conversely, HDAC6 inhibition has been found to suppress inflammatory responses and invasiveness in fibroblast-like synoviocytes (76), and its upregulation has also been implicated in additional fibrotic pathways (77). This functional duality underscores the importance of elucidating the specific cellular and pathological context-such as age, inflammatory milieu, and injury type - for understanding the role of HDAC6, as its pro-repair vs. pro-fibrotic effects are critically dependent on these environmental and disease-specific factors. Furthermore, sirtuin 1 (SIRT1), an NAD⁺-dependent deacetylase, exerts anti-fibrotic effects by deacetylating CCAAT/enhancer-binding protein β and histone H3, which has been shown to suppress YAP transcription and ameliorate HTS formation (78). The therapeutic potential of targeting the dynamic balance between histone acetylation and deacetylation, a balance that determines whether HDAC mediated deacetylation exerts pro-fibrotic or anti-fibrotic effects, has been highlighted by studies that have shown that downregulation of HDAC8 (79) or inhibition of HDAC9 (80) can alleviate fibrosis in oral submucous-fibrosis models.

EZH2 represents the catalytic subunit of the polycomb repressive complex 2 and has been shown to mediate H3K27me3, a canonical repressive epigenetic mark. In keloid fibroblasts, EZH2 is frequently upregulated and facilitates the silencing of tumor-suppressor genes and anti-fibrotic mediators (81). This activity can be regulated by other epigenetic players; for example, the deacetylase SIRT1 has been shown to deacetylate EZH2, enhancing its stability and exerting pro-fibrotic effects (81). The notable role of EZH2 in fibrosis is further demonstrated by its ability to be recruited by lncRNAs, such as FEZF1-AS1; this lncRNA has been shown to promote pulmonary fibrosis by upregulating EZH2, subsequently repressing miR-200c-3p and leading to activation of the zinc finger E-box-binding homeobox (ZEB)1 pathway (82). This establishes a direct link between the lncRNA and histone-methylation layers of fibrotic regulation. Beyond EZH2, other methyltransferases, such as SET domain containing 2, which deposits the histone 3 lysine 36 trimethylation epigenetic mark, also play notable roles in cell proliferation and migration; for example, downregulation of SET domain containing 2 was shown to inhibit these processes in hepatocellular carcinoma (83). This suggests that, beyond repressive histone methylation, activating methylation marks (e.g., H3K36me3) may play a broader yet underexplored role in promoting or regulating the fibrotic process itself, rather than merely representing a potential therapeutic target.

BET family proteins, such as bromodomain-containing protein (BRD)2, BRD3 and BRD4, recognize acetylated lysine residues on histones and recruit transcriptional co-activators to promote the expression of pro-fibrotic genes (84). As such, BET protein inhibition has emerged as a promising anti-fibrotic strategy. Pharmacological inhibition of BRD4 has been shown to alleviate cutaneous fibrosis in scleroderma models by suppressing the transcription of key fibrotic genes (84). Similarly, in hepatic fibrosis, BRD4 has been shown to promote hepatic stellate cell activation via the P300 (a histone acetyltransferase)/histone H3 lysine 27 acetylation (an activating chromatin mark)/Polo-like kinase 1 axis (85). The therapeutic efficacy of BET protein inhibition has also been shown to extend to other organs, as BRD4 inhibition has been shown to reduce fibrous scarring in the brain following an ischemic stroke by inhibiting Smad2/3 phosphorylation (86). A recent innovative approach targeting BET family proteins utilized a silk-based core-shell MN system to program the degradation of BRD9, a member of the same chromatin remodeling complex as BRD4, which effectively activated sonic hedgehog signaling and promoted diabetic wound healing (87). This highlights the potential of targeting BET family proteins with advanced delivery systems for anti-scar therapies.

The functional outcome of histone modifications is a change in chromatin accessibility, which can be comprehensively mapped using ATAC-seq. A comprehensive analysis of histone modifications in keloid tissue identified distinct chromatin accessibility patterns, with repression of breast cancer type 1 susceptibility protein highlighted as a potential pathological factor of fibrous scarring (88). This suggests that the silencing of key DNA repair and tumor-suppressor genes via chromatin closing contributes to the keloid phenotype. Furthermore, scATAC-seq has begun to deconvolute the cellular heterogeneity within scars. A landmark study on burnt-skin healing in rats obtained single-cell chromatin landscapes, revealing dynamic and cell-type-specific changes in chromatin accessibility throughout the repair process (89). These sequencing methods have also uncovered how master transcription factors, such as Twist-related protein 1, promote TGF-β receptor 1 expression in keloids by regulating the stability of MEF2A (90), and how factors such as forkhead box protein C1 activates Notch3 signaling to promote the inflammatory phenotype of keloid fibroblasts (91).

The collective evidence have strongly and consistently implicated the increased activity of HDACs, particularly class IIa HDACs, and EZH2 in promoting a pro-fibrotic fibroblast phenotype via repression of key negative regulators of TGF-β and other fibrotic signaling pathways. Similarly, BET proteins have been consistently identified as pro-fibrotic transcriptional co-activators across different organs and models. However, some apparent discrepancies remain, which are exemplified by the context-dependent role of HDAC6. This context-dependence is supported by evidence showing that HDAC6 promotes acute wound healing in aged mouse skin, whereas its upregulation or inhibition is associated with pro-fibrotic pathways in other models of chronic inflammation and fibrosis (89). These divergent roles may be explained by differences in model systems (e.g., acute wound healing vs. chronic fibrosis) and cell types. The emergence of single-cell and spatial multi-omics technologies have begun resolving these complexities by revealing that epigenetic changes are not uniform but are concentrated in specific pathogenic fibroblast subpopulations (89).

Therapeutically, this knowledge is being rapidly translated. Inhibition of HDACs, EZH2 and BET proteins has shown reproducible efficacy in pre-clinical models. The development of targeted delivery systems, such as MNs for BRD9 degraders (87), represents the next frontier in minimizing off-target effects and achieving localized epigenetic reprogramming. Future work should focus on defining the temporal dynamics of these changes during scar progression and on identifying synergistic combinations of epigenetic drugs that can permanently reverse the pro-fibrotic cellular memory.

Genome-wide methyl-CpG capture followed by qPCR has repeatedly identified the promoter region of miR-29b-2 as one of the most consistently hypermethylated loci in both keloid and HTS fibroblasts (Δβ, 0.18–0.24) (where Δβ represents the difference in methylation beta-values between scar tissue and normal skin, ranging from −1 to 1; positive values indicate hypermethylation and negative values indicate hypomethylation) (92,93). A study reported by Cui et al (92) first demonstrated in gastric cancer that DNMT3A and DNMT3B physically occupy this promoter and that forced expression of miR-29b in turn reduced reporter activity at the DNMT3A 3′-UTR by 35%. Another study extended this feedback loop to cutaneous fibrosis: Transfection of 50 nM miR-29b mimic into keloid fibroblasts decreased DNMT3B mRNA and protein expression by 42 and 38%, respectively, whereas transfection with antagomiR-29b (a chemically modified antisense oligonucleotide that specifically inhibits miR-29b) increased DNMT3B expression (93). Notably, CRISPR-dCas9-TET1-mediated demethylation of the miR-29b promoter not only restored its expression to physiological levels but also downregulated DNMT3B, indicating that the double-negative loop between miR-29b and DNMT3B is DNA-methylation-dependent rather than a simple targeting event (93). Collectively, these findings position miR-29b as both a downstream sensor and an upstream brake of DNMT3B activity, resulting in a circuit that tips the balance toward sustained collagen expression upon disruption.

In addition to the canonical miRNA/DNMT axis, lncRNAs provide an additional layer of interaction specificity by bridging RNA-chromatin contacts. This specificity refers to the ability of lncRNAs to recruit DNA methyltransferases to defined genomic loci via sequence complementarity or structural motifs, as validated by ChIRP-seq showing H19 occupancy at the miR-29b promoter (94). As discussed in Section 4, H19 is the best-characterized lncRNA in keloid biology, functioning as a scaffold that recruits DNMT3B to the miR-29b promoter (39). Extending this observation, recent ChIRP-seq analyses have quantified that H19 transcripts exhibit 42–46% enrichment at the miR-29b promoter, and ASO-mediated H19 knockdown reduces DNMT3B occupancy at this locus by 31%, elevates miR-29b expression 2.5-fold and decreases COL1A1 mRNA by 34% (94,95). Thus, H19 exemplifies a direct mechanistic link between lncRNA scaffolding activity and DNA-methylation-driven gene silencing in cutaneous fibrosis. Beyond H19, the m6A-modified lncRNA MALAT1 provides another example of RNA-guided DNA methylation. Unlike the METTL3/pri-miR-31 axis described earlier (which focuses on m6A writer function in miRNA maturation), MALAT1 acts as a physical scaffold that directly recruits DNMT3A to the miR-29b promoter region, thereby establishing a feed-forward loop between m6A RNA methylation and DNA methylation (96). Site-specific erasure of m6A modification on MALAT1 transcripts via CRISPR-dCas13b systems has been shown to reduce DNMT3A occupancy at the miR-29b promoter by 27% and restore miR-29b expression, further confirming the direct role of RNA methylation in dictating local DNA methylation patterns (96). Thus, lncRNAs do not only associate with but actively orchestrate the deposition of repressive CpG marks.

Although circRNAs are not well-established in fibrosis research, emerging evidence indicates that they indirectly govern DNA methylation by sponging miRNAs that target DNMTs. A study reported by Zhang et al (41) identified circPTK2 and circFNDC3B as circRNAs that were abundant in keloid tissue; both circRNAs were found to harbor seed matches for miR-19a-3p and miR-29b. Silencing these circRNAs elevated the availability of their cognate miRNAs, leading to a 25% reduction in DNMT3A protein expression and a 20% decrease in global 5-methylcytosine content (41). Similarly, circ_0057452 has been shown to sponge miR-7-5p, an miRNA that directly targets DNMT1 (97). Furthermore, overexpression of circ_0057452 has been shown to attenuate the miRNA-mediated inhibition of DNMT1, enhance hypermethylation of the promoter of the anti-fibrotic gene bone morphogenetic protein 4 and increase collagen-I secretion by 28% (97). Despite smaller sample sizes (n=4-6), the findings of the aforementioned studies have converged on a common mechanism in which circRNAs act as competitive endogenous RNAs that buffer miRNA-dependent repression of DNMTs, thereby reinforcing CpG hypermethylation.

Not every interaction between ncRNAs and DNMT promoters follow the paradigm that ncRNA downregulation results in DNMT upregulation. miR-21, a well-characterized pro-fibrotic miRNA, exhibits enhanced expression in HTSs yet paradoxically has been reported to target DNMT3A for repression in certain cancer types (98,99). By contrast, two independent keloid cohorts have shown that following a 1.4-fold increase in miR-21 levels, DNMT3A expression remained unchanged in one cohort and exhibited a modest increase in the other (99,100). Detailed 3′-UTR mapping has revealed the presence of a single-nucleotide polymorphism, known as rs10891883, within the DNMT3A seed match that disrupts miR-21 binding (100). Consequently, the expected downregulation of DNMT3A by miR-21 is overridden, highlighting how germline variants can reshape the ncRNA-DNMT interactome. Such population-specific findings caution against universal extrapolation of mechanism-driven data and underscore the necessity for ancestry-stratified analyses.

The reciprocity between DNA methylation and ncRNA activity offers immediate therapeutic implications. As detailed in the Introduction (Section 1), the clinical efficacy of topical 5-aza-2′-deoxycytidine has been demonstrated in a phase IIa trial; post-hoc RNA-seq analysis of this trial revealed a 2.3-fold induction of miR-29b and a 40% reduction in DNMT3B expression, corroborating the in vitro feedback loop (93). Pre-clinical studies have further shown that co-delivery of miR-29b mimics with low-dose 5-aza-2′-deoxycytidine achieves additive anti-fibrotic effects without detectable systemic toxicity, resulting in a 55% reduction in scar-height vs. 34% with monotherapy (93,101). Taken together, the convergence of mechanistic and translational data supports a dual-hit strategy in which DNMT inhibitors reactivate silenced anti-fibrotic ncRNAs and exogenous ncRNA supplementation reinforces DNMT downregulation, thereby locking fibroblasts into a quiescent epigenetic state.

Dissolving MNs remain the most advanced platform for epigenome-targeting therapeutic delivery. In a seminal keloid-relevant study reported by Yeo et al (102), 600-µm HA needles loaded with 5-fluorouracil (0.3 mg per patch) were manufactured. A single 30-sec application reduced rabbit-ear scar height by 28% after 28 days, which was equivalent to daily intra-lesional 5-fluorouracil administration but without the epidermal atrophy commonly associated with injection therapy. Pharmacokinetic analysis showed undetectable plasma levels of 5-fluorouracil, supporting the localization of administered therapeutics. The therapeutic benefit, namely equivalent scar height reduction compared with daily intralesional 5-fluorouracil administration while avoiding the epidermal atrophy commonly seen with injections, was attributed to burst release of 5-fluorouracil (~80% within 15 min) directly into the avascular dermis, bypassing efflux transporters that limit the efficacy of topical gels.

To prolong drug release, a study by Yang et al (103) reported the development of a bilayer dissolving microneedle in which triamcinolone was loaded into the rapid-release layer and 5-fluorouracil was loaded into the sustained-release layer, enabling biphasic drug release from a single administration (rapid release of triamcinolone primarily within the first day, followed by sustained release of 5-fluorouracil over a 7-day period). Scar height was reduced 38% compared with the 22% reduction observed for either monotherapy. Additionally, intra-lesional TGF-β1 protein expression dropped 34%, a mechanistic endpoint rarely captured in earlier MN-based studies. Notably, the coefficient of variation (CV) across six independent rabbit and porcine studies reached only 14%, underscoring the reproducibility of the findings of these studies. Furthermore, mechanical signaling can also confer therapeutic effects. A study reported by Zhang et al (104) detailed the use of manufactured blank polycaprolactone MNs that downregulated YAP/TAZ nuclear translocation by altering the mechanical microenvironment of scar tissue, which were administered every 48 h for 2 weeks. This downregulated YAP/tafazzin nuclear translocation and resulted in a 41% reduction in collagen-I expression in pig HTS explants. Notably, mechanical off-loading via silicone sheeting abolished 30% of the therapeutic benefits resulting from MN use, implying that tension and MN-induced tissue trauma should be optimized, as complete mechanical off-loading (i.e., maximized unloading) reduces therapeutic efficacy.

Burst-release kinetics (defined as the rapid release of ≥70% of the therapeutic payload within the first 2 h of application) remain a shared limitation of HA-based MNs. Two studies have therefore introduced ‘separating’ MNs that have tips capable of detaching only under conditions typical of keloid tissue, including pH<6.5 and reactive oxygen species (ROS) >500 µM (104). This stimulus-responsive geometry has achieved a 48% reduction in scar height while sparing adjacent unwounded skin, providing level IIa evidence that smart materials widen the therapeutic index (104).

Nanomicelle-generating MNs and MOF systems have been engineered for the zero-order release and light-triggered potentiation of epigenome-targeting therapeutics. A study reported by Chien et al (105) described the use of manufactured tranilast-loaded MNs that self-assembled into 25-nm micelles upon dermal insertion; these MNs penetrated tissues to a depth of 900 µm and exhibited >90% drug stability over 14 days. A single application of tranilast via preloaded MNs reduced rabbit-ear scar height by 42% compared with the 22% reduction observed following daily applications of tranilast gel over a 21-day treatment period (once daily, as described in the original study), whereas plasma levels remained <5 ng/ml, confirming the local confinement of therapeutics.

Photodynamic therapy has been integrated with MOF chemistry to simultaneously inhibit pathological fibroblast proliferation and reduce microvascular density within established scar tissue. A study reported by Chen et al (106) developed MNs containing MOF-armored porphyrin that release ROS under 660-nm illumination. The study demonstrated that treatment with a single 10-min exposure to the specific light wavelength reduced scar height by 46% and downregulated VEGFA by 52% and α-SMA by 38%. A follow-up study introduced 5-aminolevulinic acid-loaded MOF-MNs that metabolized endogenous ROS into cytotoxic singlet oxygen, driving fibroblast ferroptosis yet sparing keratinocytes (107). Scar height was reduced by 52% with no rebound after 8 weeks of treatment, which represented a notable increase in treatment durability compared with burst-release delivery systems. The inter-study CV for the aforementioned studies was 9%, supporting their reproducibility; however, both studies were performed only in albino rabbits. Validation of these therapeutic strategies in larger pigmented animals is therefore obligatory before human translation.

Exosomes derived from mesenchymal or epidermal stem cells are rich in anti-fibrotic miRNAs, such as miR-29a and miR-200s, and pro-resolution proteins; however, topical application of these exosomes remains limited by RNase degradation of miRNAs and poor exosomal penetration into target tissues. Dissolvable MNs have therefore been repurposed as ‘exosome docks’. A study reported by Yuan et al (108) embedded miR-29a-overexpressing adipose-derived mesenchymal stem cell exosomes into gelatin MNs at 1×10⁹ exosomes per patch; the study found that 80% of exosomes were released within 48 h of MN application, reducing scar height by 35% and causing a 44% drop in phosphorylated-Smad3 expression.

Another study reported by Zhen et al (72) advanced this concept by using core-shell MNs: The outer shell released miR-200s-enriched epidermal stem cell exosomes whereas the core provided TGF-β-neutralizing antibodies, achieving dual blockade of ZEB1/2 and TGF-β signaling. In porcine HTSs this combination treatment resulted in 1.2 mm scar flattening vs. 0.6 mm flattening upon treatment with exosome-only MNs. Additionally, collagen-I mRNA expression remained <50% of the baseline value at week 12, representing the longest post-treatment follow-up period (84 days) among microneedle-based scar treatment studies in pre-clinical animal models reported to date. These findings provided level IIa evidence that cell-free biologics targeting scar formation can be deployed via MNs with sustained efficacy.

The translational portfolio now spans small-molecule epigenetic editors, RNA therapeutics and two categories of device-assisted platforms: i) Dissolvable MN arrays used as passive drug-delivery vehicles for DNMT inhibitors, ASOs, miRNA mimics or exosomes; and ii) MOF-armored PDT patches, wherein 660-nm visible light irradiation triggers ROS release from the MOF-photosensitizer complex, directly exerting a physical therapeutic effect on scar tissue. Each category has achieved a 25–55% reduction in scar height across pre-clinical models; the CV across the nine pre-clinical studies listed in Table SIV ranged from 9 to 28%, depending on the platform and the specific intervention [e.g., the 28% CV corresponds to the Chen et al (107) MOF-MN-PDT study from 2025, whereas the 9% CV corresponds to the Chen et al (106) MOF-MN-PDT study from 2023].

Ancestry-stratified mechano-epigenomic biomarkers are already available: rs10891883, a DNMT1 intronic single nucleotide polymorphism, predicts poor response to DNMT1 monotherapy in African populations (Δβ, 0.22; P=2×10⁻¹¹) but has no notable effect on patient outcomes in East-Asian cohorts (36). Integrating methylation quantitative trait locus data, such as ancestry-stratified mechano-epigenomic biomarkers, with real-time mechanical-stress mapping (e.g., via smartphone-based 3D scanning) could yield a composite algorithm that captures a meaningful portion of scar-height variation, as supported by the correlation between repetitive-element methylation and collagen density (35) and by the therapeutic efficacy of microneedle-mediated mechanical unloading (101). Such an algorithm could guide patient stratification in adaptive trial designs.

Regulatory science remains the final hurdle to clinical translation. The US Food and Drug Administration (FDA) requires replication-competent retrovirus testing and insertional oncogenesis risk assessment for products delivered via retroviral or lentiviral vectors prior to Investigational New Drug submission for first-in-human trials, regardless of whether genome-wide, unbiased identification of double-strand breaks enabled by sequencing shows no off-target insertions (109). It is noted that the term ‘advanced therapy medicinal products’ is used by the European Medicines Agency rather than the FDA. Conversely, the European Medicines Agency now requests germ-line transmission data in zebrafish for any topical epigenetic drug, irrespective of systemic exposure. Balancing these divergent requirements, possibly through a ‘topical ncRNA’ subclass with reduced genotoxicity packages (110), will be important for preventing multi-jurisdictional gridlock and accelerating patient access to the next generation of scar-modifying medicines.

The transition of epigenetic scar therapies into late-stage clinical trials necessitates a strategic approach integrating biomarker-guided patient stratification and objective endpoint assessment. Adaptive trial designs are particularly suited to address patient heterogeneity through biomarker-driven stratification approaches, such as the ancestry-linked germline variants discussed in the beginning of this chapter. Integrating these predictive molecular signatures with quantitative scar phenotyping (e.g., via 3D imaging) can define enriched patient populations and enhance trial power (35,36,88). Furthermore, primary endpoints of therapeutic efficacy should evolve beyond subjective scales to include robust, quantifiable measures, such as volumetric reduction observed via 3D imaging and histopathological analysis of collagen architecture, complemented by molecular pharmacodynamic readouts, for example intra-lesional miR-29b induction, to confirm target engagement (11,93).

Safety and pharmacokinetic monitoring plans must be tailored to the local delivery paradigm while vigilantly assessing long-term local tolerance to therapeutic agents. Early-phase trials should include pharmacokinetic sampling to verify the minimal systemic exposure anticipated from advanced localized delivery systems, such as MNs or MOF patches, that were consistently observed in pre-clinical models (14,107). Concurrent long-term follow-up is important to monitor for potential local adverse effects, including changes in pigmentation or dermal tissue texture, ensuring that the local therapeutic benefits of therapeutic strategies are not offset by unintended sequelae.

Finally, proactive navigation of the evolving regulatory landscape is important for the timely development of therapeutic strategies. Defining an acceptable path for RNA-based therapeutics and other epigenetic modulators requires early dialogue with regulatory agencies so that therapeutic strategies align with specific requirements for genotoxicity assessments, pharmacodynamic biomarker validation and safety monitoring tailored to locally acting agents (108). Addressing these translational pillars, precision trial design, objective efficacy and safety assessment, and regulatory strategies (i.e., proactive alignment with divergent agency requirements such as FDA vs. EMA classification of epigenetic therapeutics), will be important for converting relevant epigenetic discoveries into approved scar-sparing therapies.