Melatonin-treated stem cell exosomes deliver USP4 to protect the body clock and reduce cell death, improving flap survival

A retrospective analysis (Figure 1A,B) of flap surgery outcomes in patients with pressure injuries at a hospital in south‐west China from 2014 to 2024 (n = 344 patients) revealed a postoperative flap necrosis incidence of 16.57% based on an electronic medical record system (Table 1). Then, a correlation analysis was conducted between risk factors associated with flap necrosis, revealing a positive correlation between SR and flap necrosis (r = .380, p = .000) (Table 2). SR emerged as a significant risk factor of postoperative flap necrosis (Figure 1C). Compared with other risk factors, SR as a predictor of flap necrosis demonstrated a satisfactory predictive efficacy, as evidenced by an area under the curve (AUC) value of .71 (Table 3). The Kaplan–Meier survival curves further substantiated that the presence of SR in postoperative patients undergoing flap surgery during their hospital stay resulted in a substantial escalation in the incidence of flap necrosis (Figure 1D,E). Finally, a COX proportional hazards regression model was used to adjust for other potential confounding factors associated with flap necrosis, including sedative use, haemodialysis, mechanical ventilation, height, cholesterol, haemoglobin, albumin, FiO₂, Apache II and hospitalisation time (Table 4). This model further substantiated that SR amplifies the risk of flap necrosis by a factor of 4.60 (HR = 4.60, 95% CI = 2.72, 7.80).

In order to further explore the molecular mechanisms existing between flap necrosis and SR, skin tissues from flap necrosis and flap survival were subjected to transcriptome analysis. The volcano plot showed that a total of 3228 differentially differentiated genes (DEGs) were detected. Of these, 1743 were up‐regulated genes and 1485 were down‐regulated genes (Figure 1F). Enrichment analysis of these DEGs revealed that they were enriched in 'wound healing', 'regulation of vasculature development', 'collagen−containing extracellular matrix' and 'extracellular matrix structural constituent'. The KEGG pathway analysis revealed that DEGs were enriched in the 'circadian rhythm', followed by the 'GnRH signalling pathway'. To systematically assess how different forms of cell death—such as apoptosis, necrosis and pyroptosis—affect flap necrosis, we tested corresponding inhibitors in animal models. Inhibition of apoptosis, necrosis, pyroptosis and ferroptosis all led to notable improvements in flap necrosis area, blood flow and skin temperature following SR induction (Figure S1A–F). The most marked improvement, however, was observed after ferroptosis inhibition. A dataset of 419 ferroptosis‐related genes was acquired from the FerrDb database. The intersection between this ferroptosis gene set and our dataset of DEGs was then examined, leading to the identification of 79 ferroptosis‐related DEGs. These were subsequently categorised into 44 up‐regulated and 35 down‐regulated genes (Figure 1H). The cnet‐plot function is used to visualise the association between DEGs and the ferroptosis pathway (Figure 1I). Immunofluorescence (IF) staining further validated the differential expression of the ferroptosis marker gene ACSL4/FTH1 in human flap tissue (Figure 1N–Q). Histological staining demonstrated increased epidermal thickness, decreased collagen deposition during flap necrosis (Figure 1J–M). It is hypothesised, based on the current evidence, that there is an association of flap necrosis with circadian rhythms and ferroptosis.

MT has been identified as a core circadian rhythm‐regulating hormone and possesses antioxidant properties. In addition, the cargo capacity, biocompatibility and resistance of EXOs to degradation render them ideal drug‐delivery vehicles. The present study investigates the characteristics of EXOs secreted by MT‐pretreated BMSCs (MEXOs) and compares them with basic EXOs, in order to assess their therapeutic potential for skin flap necrosis associated with circadian rhythm disorders and ferroptosis. BMSCs were phenotypically characterised using flow cytometry and identified as positive for the surface markers CD90 and CD44, while being negative for haematopoietic markers CD45 and CD34 (Figure 2A,B). According to transmission electron microscopy (TEM) images, EXOs and MEXOs both displayed a characteristic cup‐shaped or spherical morphology, around 100 nm in diameter (Figure 2C). The consistency in exosome preparation was further affirmed by nanoparticle tracking analysis, which indicated analogous size distribution profiles between the two groups (Figure 2D). Additionally, no significant differences were observed in the mean nanoparticle size or concentration (Figure 2E). Western blot (WB) analysis was utilised to assess the protein composition of the exosomes, revealing the presence of typical exosome surface markers such as CD9, CD81 and TSG101. Notably, the negative exosome marker calnexin was absent in both types of exosomes, with no observable differences between EXOs and MEXOs (Figure 2F,G). This study confirmed that the isolated exosomes met the expected standards for morphology, size, concentration and surface markers, establishing the foundation for subsequent functional studies.

With the fundamental properties of both exosome types validated as comparable, the focus shifted to assessing their differential biological activities and underlying molecular profiles. We found that 'ferroptosis', 'circadian rhythm' and 'ubiquitin protein ligase binding' differed between the EXOs and MEXOs in KEGG and GO analyses using transcriptome analysis (Figure 2J). Furthermore, differential expression of deubiquitinating enzymes (DUBs) was observed, with MT treatment resulting in the up‐regulation of 20 DUBs in MEXOs samples. USP4 exhibited the most significant up‐regulation ((Figure 2I). Combined with the finding in an animal model that inhibition of ferroptosis significantly ameliorates SR leading to a decrease in USP4 in vascular endothelial cells (Figure S1G,H). Subsequent experiments focused on the role of USP4 in regulating flap ferroptosis related to SR. Finally, the exosomes were labelled with Dil dye and co‐cultured with human umbilical vein endothelial cells (HUVECs) for 24 h. The successful uptake of both types of exosomes by the HUVECs when compared with the blank dye group, demonstrating their potential utility in cellular communication and therapeutic applications (Figure 2K).

A modified McFarlane flap model was utilised on the backs of experimental mice subjected to 1 week of SR (Figure 3A). Considering MT critical function in regulating circadian rhythms and sleep–wake cycles, we performed a rhythmic analysis of MT concentrations in flap and blood plasma post‐SR. The results indicated that mesor (mean) and amplitude of MT were decreased, and the acrophase (peak timing of the rhythm) of MT was delayed in both flap and blood plasma post‐SR (Figure 3B–E). This also demonstrated the success of modelling SR in mice. In addition, changes in MT levels and disturbances to the circadian rhythm adversely affect crucial physiological processes for skin repair. This included influencing the area of flap survival, blood flow establishment and skin barrier function.

Among the treatments evaluated, the group receiving exosomes derived from MEXOs demonstrated a significant enhancement in flap survival compared with groups treated with MT or EXOs alone (Figure 3F,G). Additionally, the treatment showed no signs of acute systemic toxicity (Figure S1I). Laser Doppler imaging showed the highest blood flow in the MEXOs‐treated group (Figure 3H,I). Infrared thermography (IRT), used to monitor skin temperature, indicated a decrease in temperature in the necrotic distal regions following SR. Treatment with MEXOs notably increased skin temperature and reduced the extent of hypothermia (Figure 3J,K). Furthermore, the expression of CD34 and α‐SMA in the micro‐vessels of the flap was elevated post‐MEXO treatment, suggesting enhanced angiogenesis (Figure 3O,R). Histological analysis via haematoxylin and eosin (HE) staining demonstrated a significant reduction in skin and epidermal thickness (Figure 3,P), with an increase in collagen fibre density (Figure 3N,Q). The expression levels of genes involved in hyaluronic acid synthesis and encoding aquaporin, as well as those of genes crucial for the formation of the cutaneous stratum corneum barrier, such as loricrin and filaggrin (Figure 3S), were found to be able to restore the expression levels of these critical barrier‐forming genes. These results illustrate that SR disrupts skin barrier function and impairs angiogenesis, while MEXO treatment not only ameliorates these disruptions but also promotes flap angiogenesis and improves overall skin barrier function.

In order to address the potential confounding effect of stress induced by the SR procedure, which resulted in flap necrosis, plasma and flap corticosterone levels were measured. While the administration of SR led to a significant elevation in corticosterone levels in comparison with the control group (Figure S2A), the levels remained similarly elevated in the EXOs‐ and MEXOs‐treated SR groups with the exception of the clock rhythmicity recovery (Figure S2J,K). This finding suggests that the therapeutic effects of MEXOs on flap survival are not primarily attributable to stress reduction, but rather to the specific regulation of clock rhythmicity. Additionally, the present study demonstrates that SR significantly induces ferroptosis in mouse flaps, an effect ameliorated by ferroptosis inhibitors (Figure S2A–D). Specifically, WB analysis (Figure 4A,B) and qPCR (Figure 4C) revealed an increase in ACSL4, a marker of ferroptosis and a decrease in the levels of SLC7A11 and GPX4, both of which are key anti‐ferroptosis markers, along with lower levels of FTH1, involved in iron sequestration. This was further demonstrated by labelling vascular endothelial cells with CD34 and co‐localising using the FTH1 ferroptosis marker (Figure 4L,M). The metabolic balance between oxidative and antioxidative processes, crucial for cellular homeostasis, was assessed by measuring the expression of SOD1 and HO1, both of which were found to be significantly decreased following SR. The immunohistochemical results of SLC7A11 were also consistent with the above results (Figure 4D,H). Additionally, high iron content in the tissues was evident from Prussian blue staining (Figure 4E,I). The intensity of dihydroethidium (DHE) and 4‐hydroxynonenal (4‐HNE), markers for oxidative stress and lipid peroxidation, respectively, was found to increase after SR (Figure 4F,J,G,K). The glutathione (GSH), malondialdehyde (MDA) and Fe²⁺ level, markers for ferroptosis, showed a decrease in GSH levels, with concurrent increases in MDA and Fe²⁺ levels in flap after SR (Figure 4Q–S).

Given the potential of MEXOs to specifically regulate circadian rhythms, we examined the expression patterns of ferroptosis‐related genes in skin flaps and observed consistent circadian fluctuations. Specifically, SR was found to disrupt these rhythms, decreasing the mesor and amplitude while delaying the acrophase of SLC7A11, GPX4 and FTH1 (Figure S). Since SR appears to concurrently exacerbate ferroptosis and induce circadian rhythm disruption, we propose the concept of 'clock rhythmic ferroptosis' to describe this phenomenon. Notably, MEXOs treatment not only significantly attenuated ferroptosis but also restored circadian rhythmicity, with a superior efficacy compared with both MT and EXOs treatments. 2L–N

Based on the above evidence, we hypothesise that MT intervention exerts its therapeutic effect by altering USP4 levels in MEXOs. This mechanism would explain why MEXOs is more efficacious than treatment with EXOs alone or direct MT application. Meanwhile, we also observed significantly reduced USP4 expression in both human necrotic flap tissue (Figures 4N and S2H,I) and a mouse model of SR‐induced flap necrosis (Figure S2F,G), via IF staining co‐localised with the endothelial marker CD34. Notably, MEXOs treatment enhanced USP4 expression in vascular endothelial cells more potently than EXOs (Figure 4O,P). This supports that SR triggers flap ferroptosis by down‐regulating USP4, while MEXOs exert a protective effect by up‐regulating it. MEXOs represent a promising therapy for SR‐induced clock rhythmic ferroptosis, potentially through modulating USP4 in endothelial cells.

To evaluate the protective effects of MEXOs on vascular endothelial cell function, HUVECs were subjected to oxygen glucose deprivation (OGD) and exposed to hydrogen peroxide (H₂O₂) to simulate the in vivo conditions of mice (Figure 5A). The viability and functional responses of HUVECs were assessed using a series of in vitro assays, including cell migration, scratch assays and angiogenesis (tube formation) assays. The results demonstrated that HUVECs treated with MEXOs showed a significant increase in cell migration than MT and EXOs (Figure 5B,C). The area covered by migrating cells in scratch assays was notably larger in the MEXOs‐treated group, indicating enhanced wound closure capabilities (Figure 5D,E). Furthermore, angiogenesis assays revealed that MEXOs significantly boosted the angiogenic capacity of HUVECs, leading to more extensive and complex tube formation (Figure 5F,G). Additionally, Brdu staining confirmed that all three treatments—MT, EXOs and MEXOs—enhanced the proliferative capacity of the cells. Notably, MEXOs exhibited the most significant improvement (Figure 5H,I). These findings underscore the therapeutic potential of MEXOs in enhancing endothelial cell function.

The experiments conducted on HUVECs under conditions of OGD and H₂O₂ exposure revealed that these stressors significantly intensified ferroptosis events in the cells, as evidenced by WB analysis (Figure 5J,K). Treatment with MEXOs effectively reversed the ferroptosis. This reversal was characterised by a decrease in ACSL4 and increases in SLC7A11, GPX4 and FTH1, all of which are crucial for reducing oxidative stress and inhibiting ferroptosis. This finding is further corroborated by the diminished levels of oxidative stress, as evidenced by DHE staining (Figure 5M,P) and DCFH‐DA probe labelling (Figure 5L,O,T,U). Apoptosis flow cytometry also showed that MEXOs significantly reduced late apoptosis following oxidative stress (Figure 5R,S). Additionally, biochemical assays showed a decrease in GSH levels and increases in MDA and Fe²⁺ levels after ferroptosis induction. These changes were effectively mitigated by treatment with MEXOs, which restored GSH levels and reduced MDA and Fe²⁺ levels (Figure 5V). The results demonstrate that MEXOs have the potential to inhibit HUVECs ferroptosis.

Further analyses focused on the expression of USP4 within the cellular environment, particularly in response to treatment with MEXOs. Our results indicate a significant elevation in the expression of USP4 in the MEXOs‐treated group. This finding was substantiated through WB and corroborated by IF staining (Figure 5N,Q). The observed increase in USP4 suggests a mechanistic pathway through which MEXOs exert their protective effects. Specifically, by elevating USP4 levels, MEXOs protect the function of HUVECs and suppress ferroptosis.

Based on in vitro and in vivo experiments, we confirmed that exosomes derived from USP4‐overexpressing BMSCs (MEXOs [Flag‐USP4]) successfully deliver functional Flag‐USP4 protein into target cells, increase total USP4 levels and facilitate its interaction with endogenous ARNTL (Figure S3A–J). Then, the study generated MEXOs with reduced USP4 levels (MEXOs‐shUSP4). Comparison of flap treated with these modified exosomes and those treated with a control shRNA (MEXOs‐shNC) revealed a significant decrease in USP4 protein expression in the MEXOs‐shUSP4 group, both with and without SR (Figure 6S,T). Treatments with MEXOs‐shUSP4 resulted in poorer survival outcomes compared with MEXOs‐shNC. This was evident from analyses of the necrotic area of the flap, blood flow signals and skin temperature (Figure 6A–F). Furthermore, IF staining indicated that angiogenesis within flap was decreased following treatment with MEXOs‐shUSP4 (Figure 6K,L). The study also explored the structural effects of MEXOs‐shUSP4 on skin and epidermal thickness as well as the arrangement of collagen fibres through HE staining and Masson's trichrome staining (Figure 6G–J). The expression of genes related to skin barrier function was found to decrease after treatment with MEXOs‐shUSP4, further suggesting a diminished recovery in skin barrier function (Figure 6X). These data collectively demonstrate that the ability of MEXOs to promote flap angiogenesis and skin barrier function recovery is dependent on USP4.

Furtherly, WB indicated that MEXOs‐shUSP4 treatment led to heightened ferroptosis activity, evidenced by increased ACSL4 and reduced GPX4 and FTH1 in vivo (Figure 6S,T). It also demonstrated levels of ferroptosis in vascular endothelial cells by using IF double staining (Figure 6U,V). Additionally, Prussian blue staining showed a significant escalation in iron ion release within these tissues (Figure 6M,P). The silencing USP4 markedly decreased the expression of HO1 and SOD1, weakening cellular defences against oxidative stress (Figure 6S,T). The intensities of DHE and 4‐HNE were also found that MEXOs‐shNC more effectively mitigated these markers than MEXOs‐shUSP4 (Figure 6N,O). Quantitative assessments of GSH, MDA and Fe²⁺ revealed decreased GSH levels and increased MDA and Fe²⁺ levels following treatment with MEXOs‐shUSP4 (Figure 6W). These results suggest that MEXOs inhibit ferroptosis, enhancing flap viability and skin barrier function recovery by delivering USP4.

The results revealed a marked decrease in USP4 expression in HUVECs treated with MEXOs‐shUSP4 compared with those treated with MEXOs‐shNC. This reduction was further exacerbated following treatment with the Erastin and was partially reversed by the ferrostatin‐1 (Figure S). Then, we evaluated endothelial functionality through a series of assays: cell migration, scratch tests and tube formation assays. The results demonstrated that MEXOs‐shUSP4 significantly impaired cell migration and angiogenesis when compared with the MEXOs‐shNC (Figure S). Furthermore, the inhibition of USP4 negated the enhanced cellular viability typically conferred by MEXOs‐shNC, as evidenced by Brdu assays (Figure S). 4Q–T 4A–F 4G,H

Additionally, WB analysis substantiated that MEXOs‐shUSP4 was less effective at inhibiting ferroptosis events in HUVECs, evidenced by increased ACSL4 and decreased SLC7A11, GPX4 and FTH1 levels. These changes were further validated through treatments with ferroptosis inducers and inhibitors. Additionally, MEXOs‐shUSP4 treatment significantly reduced the expression of HO1 and SOD1, indicating a diminished resistance to oxidative stress (Figure S4Q,R). DHE staining and DCFH‐DA probe labelling demonstrated that the antioxidative stress and anti‐lipid peroxidation capacities of MEXOs‐shUSP4 were compromised compared with those of MEXOs‐shNC (Figure S4I–L,O,P). The results of the apoptosis flow cytometry assay demonstrated that MEXOs‐shUSP4 diminished the capacity of MEXOs to impede late apoptosis (Figure S4M,N). The levels of GSH, MDA and Fe²⁺ suggesting that MEXOs contribute to the inhibition of ferroptosis in HUVECs by delivering USP4, which enhances cellular resistance against oxidative challenges (Figure S4U). All of those results proved that MEXOs inhibit ferroptosis by shuttling USP4, protecting the function of HUVECs.

To elucidate the molecular mechanisms by which USP4 regulates circadian rhythm and ferroptosis, we first employed immunoprecipitation–mass spectrometry (IP–MS) to identify its interacting proteins (Figure 7A,B). This approach revealed ARNTL, a core circadian clock component34 and a known repressor of ferroptosis, as a potential binding partner.35 Subsequent co‐IP assays in HUVECs confirmed a robust endogenous interaction between USP4 and ARNTL, which was reciprocally validated (Figure 7C–F). This interaction was further corroborated in HEK 293T cells using epitope‐tagged proteins, where Flag‐USP4 co‐precipitated with Myc‐ARNTL and vice versa (Figure 7G).

We next investigated the functional consequence of this interaction. Knockdown of USP4 in HUVECs led to a reduction in ARNTL protein levels, which could be rescued by the proteasome inhibitor MG132, implicating USP4 in regulating the proteasomal degradation of ARNTL (Figure 7H,I). The stabilisation of ARNTL was dependent on the deubiquitinase activity of USP4, as overexpression of wild‐type USP4, but not a catalytically inactive mutant (C311A), increased ARNTL protein abundance (Figure 7J,K). Consistent with this, cycloheximide (CHX) chase assays demonstrated that USP4 overexpression hindered, while its silencing accelerated, ARNTL degradation (Figure 7L,M). Mechanistically, USP4 was found to deubiquitinate ARNTL, as its silencing in HUVECs enhanced ARNTL ubiquitination (Figure 7O). Treatment with MEXOs, which deliver USP4, down‐regulated ARNTL ubiquitination and degradation, an effect that was reversed when MEXOs carrying shUSP4 were applied (Figure 7P). This activity was crucial, as wild‐type USP4, but not the C311A mutant, reduced ARNTL ubiquitination in HEK 293T cells (Figure 7Q).

These in vitro findings were consistent in a mouse model of skin flap after SR. MEXOs‐shUSP4 significantly decreased ARNTL protein levels without affecting its mRNA, confirming post‐translational regulation (Figure 7N). Furthermore, ARNTL was more highly ubiquitinated and degraded in flaps treated with MEXOs‐shUSP4 after SR (Figure 7S,T). Beyond stabilising ARNTL levels, USP4 was critical for maintaining its circadian expression rhythm. Silencing USP4 dampened the rhythm, reducing the mesor and amplitude while delaying the acrophase of ARNTL expression in the flap post‐SR (Figure 7N), which consequently led to circadian fluctuations in ferroptosis‐related proteins (Figure 7U). In conclusion, USP4 interacts with and stabilises ARNTL through deubiquitination, thereby regulating its circadian oscillation and executing clock‐rhythmic control over ferroptosis.

In human flap tissue, USP4 robustly co‐localised with ARNTL in vascular endothelial cell within viable areas, but this interaction was markedly disrupted in necrotic regions. Accompanying this disruption, the expression of the ferroptosis marker FTH1 was also markedly diminished (Figure 8A–D). These findings suggest that human flap survival involves mutual regulation between ARNTL and USP4, influencing ferroptosis levels. To establish the central role of ARNTL, we performed two key experiments. First, ARNTL knockdown under normal sleep conditions recapitulated the phenotype of clock rhythmic ferroptosis and flap necrosis, thereby demonstrating its sufficiency in driving this process. The shARNTL group exhibited a same necrotic flap area, along with decreased perfusion and skin temperature, relative to the SR group (Figure S5A–F). Consistent with this, associated biomarkers confirmed enhanced ferroptosis and circadian rhythm disruption (Figure S5G–L). Second, we found that overexpressing ARNTL could effectively rescue the flap necrosis phenotype even in mice with USP4 impairment (induced by MEXOs‐shUSP4). ARNTL overexpression significantly improved flap survival compared with MEXOs‐shUSP4 treatment alone, as demonstrated by a reduced necrotic area, enhanced blood flow signals and elevated skin temperature (Figure 8E–J). IF staining further confirmed that ARNTL promoted angiogenesis after MEXOs‐shUSP4 treatment (Figure 8O,P) and increased the co‐localisation of USP4 and ARNTL expression in vascular endothelial cells (Figure 8X,Y). HE and Masson's trichrome staining revealed that ARNTL treatment restored impaired skin barrier function and collagen deposition, accompanied by a marked alleviation of oxidative stress in the flap tissue (Figure 8K–N and 8S–V). Moreover, ARNTL overexpression counteracted ferroptosis induced by MEXOs‐shUSP4, as indicated by decreased iron ion levels (Figure 8Q,R) and the restoration of clock rhythmic ferroptosis (Figure 8W).

Circadian rhythms are self‐sustaining oscillatory phenomena that align with the natural 24‐h cycle.36 The transcription factor ARNTL, integral to the mammalian circadian clock, orchestrates clock‐controlled gene expression. To dissect circadian rhythm dynamics in flap, we analysed clock genes including the positive regulator Clock; negative regulators Cry1, Cry2, Per1, Per2 and Per3; and feedback loop genes Rev‐erbα and Rorβ (Figure 8Z). Treatment interventions significantly altered their rhythmic parameters with 24‐h cycle. Notably, Clock exhibited increased mesor and amplitude in the MEXOs‐shUSP4 group, but this effect was mitigated in the MEXOs‐shUSP4 + ARNTL group, reducing circadian oscillation. Conversely, MEXOs‐shUSP4 markedly decreased mesor and amplitude in negative regulators (Cry2, Per1, Per2, Per3) and feedback loop genes (Rev‐erbα, Rorβ), which was rescued by ARNTL overexpression. Acrophase remained unaltered for most genes except Cry1 and Rev‐erbα. Different clock gene types exhibited distinct responses to SR. Critically, MEXOs—which stabilise ARNTL via USP4 delivery—improved the circadian rhythm of downstream clock genes, reduce the clock rhythmic ferroptosis of skin flap.

Functional rescue experiments confirmed the critical role of the USP4–ARNTL axis in conferring protection. Overexpression of ARNTL in HUVECs effectively reversed the impairments in cell migration, angiogenesis and proliferation induced by MEXOs‐shUSP4 under OGD and H₂O₂ stress (Figure S6A–H). Concurrently, ARNTL overexpression mitigated the elevated oxidative stress, lipid peroxidation, apoptosis and ferroptosis levels in HUVECs caused by USP4 deficiency (Figure S6I–R). These results demonstrate that exosomes containing USP4 protect HUVECs function primarily by stabilising ARNTL to inhibit ferroptosis.

To elucidate how USP4 is up‐regulated in exosomes, we investigated the regulatory role of N6‐methyladenosine (m6A) modification.37 We found that MT treatment significantly enhanced global m6A levels in BMSCs (Figure S7A, B). Methylated RNA IP (MeRIP)‐qPCR specifically confirmed the enrichment of m6A‐modified USP4 mRNA, which was more pronounced after MT treatment (Figure S7C). Mechanistically, the m6A 'writer' METTL3 interacted with USP4 mRNA to facilitate its methylation (Figure S7E). This modification, in turn, recruited the nuclear reader YTHDF1,38 which binds to USP4 mRNA and promotes its translation, as evidenced by RNA IP (RIP)‐qPCR (Figure S7D). Consequently, the expression levels of USP4, YTHDF1 and METTL3 were consistently elevated in both MT‐treated BMSCs and the exosomes they secreted (Figure S7F–K). These findings establish that MT treatment enhances the m6A methylation of USP4 mRNA via the METTL3/YTHDF1 axis, thereby boosting USP4 protein expression in BMSCs and their exosomes.

This retrospective study, which covered a period of 10 years, analysed data from 344 patients who underwent flap surgery. Investigators collected demographic characteristics and clinical baseline data from the hospital electronic case system to partially adjust for potential confounders, which have the capacity to directly impact flap survival. It is important to note that, due to the constraints of the retrospective study design, specific anxiety/depression scales were not routinely available for all cases. Therefore, in our multivariate model, we included 'sedative use' as a key clinical proxy for anxiety management. Upon admission, the patient was evaluated by a nurse using a predetermined scale (the PSQI), with the results (normal or insomnia) documented on the admission evaluation form.

The PSQI serves as an effective instrument for assessing sleep quality and patterns. Recent expert consensus on standard insomnia research assessments has endorsed it as a primary metric for global sleep and insomnia symptoms.7^,8 Comprising seven components, the questionnaire evaluates sleep quality and distinguishes 'poor' from 'good' sleep. These include subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication and daytime dysfunction over the past month. The evaluating nurses underwent consistent training in administering this scale, and the outcome was documented as a binary variable within a structured data field. This study was conducted in accordance with the Declaration of Helsinki II and received approval from the Ethics Committee of our affiliated hospital (Protocol KY2025 ‐ R392).

Male C57BL/6J mice (6–8 weeks old, 20–30 g) were selected to avoid estrogen's impact on flap necrosis.77 Animals were obtained from our university's experimental centre and handled per national guidelines (Protocol wydw2019‐0729). They were kept under controlled conditions (21–25°C, 50–60% humidity, 12‐h light/dark cycle) and provided with food and water ad libitum. Under pentobarbital anaesthesia (50 mg/kg, i.p.), a 1.5 × 4.5 cm² caudal‐based flap was surgically created on the dorsum.78 After removing visible blood vessels, the flap was repositioned and sutured, followed by iodophor disinfection. This model replicates human ischemic flap pathophysiology.79

The SR group underwent a 7‐day intervention post‐surgery: sleep was suppressed via gentle tactile stimulation during 6‐h daily sessions (8:00 AM–2:00 PM), while controls slept freely.80 Mice were euthanised on day 7, and flap tissues (divided into proximal I, intermediate II and distal III regions) and right atrial blood were collected at 4‐h intervals across six time points. Flap survival area, blood perfusion and skin temperature were assessed on day 7 (2:00–5:00 PM). A 1 × 1 cm² sample from region II was excised, rinsed with PBS and processed as required.

Macroscopic properties of the flap, including appearance, colour and hair condition, were meticulously observed and documented on day 7 post‐operation. On the same day, the percentage of flap survival area was quantitatively assessed utilising Image J software (National Institutes of Health, Bethesda, MD). To determine the percentage of flap survival, the following equation was employed: % survival = survival area ÷ total area × 100%.

The blood supply and vascular network integrity of the flap were visualised using laser Doppler blood flow (LDBF). On day 7 post‐surgery, a laser Doppler instrument from Moor Instruments (Axminster, UK) was employed to assess the blood supply and map the small vessel networks within the flap. Blood flow was quantitatively evaluated in terms of perfusion units (PU), utilising the moor LDI Review software (version 6.1; Moor Instruments). The blood flow measurements were conducted three times for each mouse and used for subsequent statistical analysis.

The flap temperature was monitored using IRT. On day 7 after the operation, the flap temperature was photographed using a smartphone‐based camera (FLIR, USA) according to the previous study protocol. The relative temperatures in area II were quantified using FLIR supporting software. Each mouse was measured three times, and the average was used for statistics.

Histopathologic analysis was conducted on tissue samples harvested from area II of the flap s on day 7 post‐operation. The specimens were fixed in 4% paraformaldehyde for a full 24 h. The embedded tissues were then sectioned and subjected to staining processes critical for detailed histological examination: HE staining, which highlights cellular and nuclear features, and Masson's trichrome staining, which differentiates collagen fibres within the connective tissue, thereby allowing assessment of the epidermal thickness and the integrity of collagen structures. Additionally, the presence of iron deposits within the tissue was evaluated using the Prussian Blue Iron Stain Kit (YEASEN), following the manufacturer's protocols. The stained sections were meticulously examined under a Nikon microscope.

Histological sections of skin tissues from area II of the flap in each experimental group underwent a series of preparative steps. Initially, sections were deparaffinised in xylene and progressively rehydrated through a series of graded ethanol solutions. After rehydration, we performed antigen retrieval by heating the sections in 10.2 mM sodium citrate buffer at 95°C for 10 min. Subsequent steps included washing the sections, quenching endogenous peroxidase activity with 3% H₂O₂, and blocking with 10% bovine serum albumin (BSA) in PBS for 60 min to prevent nonspecific antibody binding. The sections were then incubated overnight at 4°C with a primary antibody targeting SLC7A11 (Abcam, ab307601). Then, sections were exposed to an HRP‐labelled secondary antibody. The presence of the target antigen was then visualised using a DAB detection Kit, producing a brown precipitate at the site of the antigen‐antibody reaction. The stained samples were examined under a Nikon microscope to assess the localisation and expression levels of SLC7A11.

Flap tissue sections were deparaffinised and rehydrated as previously described. Antigen retrieval was performed with 10.2 mM sodium citrate buffer at 95°C for 10 min. After cooling to room temperature, the sections were blocked for 60 min using 10% BSA. They were then incubated overnight at 4°C with the following primary antibodies: ACSL4 (Affinity; DF12141), FTH1 (Proteintech; 10727‐1‐AP), CD34 (Immunoway; YM8525), α‐SMA (Immunoway; YM0011), 4‐HNE (Abcam; ab275359), USP4 (Proteintech; 66822‐1‐Ig) and anti‐FLAG (Proteintech; 66008‐4‐Ig). Subsequently, a secondary antibody was applied for 1 h at 37°C, followed by nuclear staining with DAPI (Abcam). Imaging of positive cells was conducted using a Nikon fluorescence microscope.

Excised skin tissues were processed in a 30% sucrose solution until sinking, which served as the criterion for sufficient penetration. Subsequently, the tissues were embedded in OCT compound. The embedded tissues were then sectioned into 20 µm slices using a cryostat and allowed to thaw. To reduce autofluorescence, sections were incubated with a quenching reagent for 5 min. ROS detection involved staining the sections with DHE (Beyotime) at 37°C for 30 min. Subsequently, sections were stained with DAPI (Abcam) to visualise nuclei. The final examination of the sections was performed using a Nikon fluorescence microscope, capturing images of positively stained cells.

MSCs were isolated from mouse bone marrow using established protocols.13 Under sterile conditions, the femurs and tibiae from the lower limbs were dissected, and attached muscle along with adipose tissues were carefully removed. The marrow cavity was repeatedly flushed with culture medium via a syringe to harvest the bone marrow. The collected rinsing fluid was passed through a 70 µm sterile filter (Corning) to eliminate debris and then subjected to centrifugation at 200 g for 5 min to pellet the cells. The resulting cell pellet was resuspended and seeded at 1 × 10⁶ cells/mL into a 25 cm² culture flask. Cells were maintained in DMEM medium (Gibco) containing 10% foetal bovine serum (FBS; Gibco) and 1% penicillin‐streptomycin (Gibco) at 37°C with 5% CO₂. After 24 h, non‐adherent cells were removed, and the adherent population was continuously cultured until reaching 80–90% confluence, with the medium refreshed every other day. BMSCs at passages 3–5 were used for subsequent experiments.

Then, BMSCs were pretreated with MT (1 µmol/L) in exosome‐depleted 10% FBS‐ (System Biosciences) containing culture medium (Gibco) for 48 h. The surface biomarkers of BMSCs were identified using flow cytometry with FITC‐and PE‐conjugated antibodies. CD90 and CD44 were confirmed as positive markers, while CD45 and CD34 were negative markers (BD Biosciences), consistent with the manufacturer's instructions.

HUVECs were cultured in an incubator using DMEM supplemented with 10% FBS and 1% penicillin/streptomycin solution. The HUVECs underwent OGD by being cultured in a sugar‐free medium. After being placed in a sealed chamber and flushed with the gas mixture (95% N₂/5% CO₂) for 15 min, the cells were maintained in the sealed chamber within a 37°C incubator for 4 h. Subsequently, the medium was changed to normal medium, and the cells were cultured in a standard incubator for another 6 h.78 Finally, H₂O₂ (300 µmol/L) was added to the HUVECs to induce an oxidative stress environment for 24 h.

Exosomes were isolated and purified according to the classical differential centrifugation/ultracentrifugation protocol, guided by the MISEV2018 guideline. When BMSCs reached 80% confluency, the medium was replaced with exosome‐depleted 10% FBS, with or without MT, for 48 h. The conditioned medium was then collected and processed through sequential centrifugations: first at 300×g for 10 min, then at 2000×g for 30 min at 4°C, to eliminate dead cells, apoptotic bodies and debris. The resulting supernatant was further cleared by centrifugation at 10 000×g for 30 min and filtered through a .22‐µm Steritop™ filter (Millipore). The filtered solution was ultracentrifuged at 100 000×g for 70 min at 4°C using a CP100NX ultracentrifuge (Eppendorf Himac Technologies) to obtain the exosome pellet. After supernatant removal, the pellet was washed with sterile PBS and subjected to a second round of ultracentrifugation under the same conditions. Finally, the purified exosomes were resuspended, aliquoted and either used directly in subsequent assays or stored at −80°C.

Exosome size distribution was analysed using a NanoSight NS300 instrument (Malvern Panalytical Ltd.), which determines particle size via nanoparticle tracking analysis of Brownian motion in suspension. Morphological assessment was performed using a Tecnai 12 transmission electron microscope (Philips).

WB analysis was conducted to identify exosomal markers, specifically CD81 (Abcam; ab219209), CD9 (Abcam; ab236630) and TSG101 (Abcam; ab125011). Additionally, calnexin (Proteintech; 10427‐2‐AP), a negative marker for exosomes, was also assessed to evaluate the purity of the exosome preparations.

Exosomes suspended in PBS were incubated with Dil solution (Beyotime) to facilitate the detection of exosome uptake. The Dil‐labelled exosomes were then co‐cultured with HUVECs for 24 h. After incubation, HUVECs were washed with PBS to remove any unbound exosomes and fixed with 4% paraformaldehyde. The internalisation of Dil‐labelled exosomes was then imaged using a Nikon fluorescence microscope.

Constructs expressing Flag‐tagged USP4, Flag‐tagged USP4 C311A, Myc‐tagged ARNTL and HA‐tagged ubiquitin were generated by cloning the respective open reading frames into expression vectors with an N‐terminal Flag or Myc tag. Lipofectamine 3000 reagent (Invitrogen) was used according to the manufacturer's protocol to improve transient transfection. To achieve stable knockdown of USP4, lentiviral vectors encoding shRNA targeting USP4 were constructed by GENE, with scrambled shRNA lentiviral constructs serving as negative controls.

Recombinant adeno‐associated virus 2 (AAV2) vectors were engineered for both ARNTL knockdown and overexpression. For knockdown, ARNTL‐specific shRNA sequences were designed, synthesised and cloned into an AAV2 vector downstream of a U6 promoter. A separate CMV promoter driving ZsGreen expression was included in the same vector for transduction tracking. For overexpression, the ARNTL coding sequence was inserted into an AAV2 vector under the control of a CMV promoter. All constructs were verified by sequencing prior to viral packaging.

Total RNA from skin tissues was extracted with TRIzol reagent (Invitrogen) as instructed. This RNA was used for complementary DNA (cDNA) synthesis with HiScript III RT SuperMix (Vazyme) on a VeritiPro™ Thermal Cycler (Thermo Fisher Scientific, USA). Quantitative PCR (qPCR) then employed Taq Pro Universal SYBR qPCR Master Mix (Vazyme) on a C1000 Touch Thermal Cycler (BIO‐RAD, USA). The relative gene expression levels were calculated using the 2^–ΔΔCt method, with ACTB serving as an internal control gene to normalise the expression levels of the target genes.

HUVECs, HEK 293T cells and skin tissues from flap area II were lysed using RIPA buffer containing protease inhibitors (Beyotime). Prior to IP, antibodies were diluted as specified and pre‐incubated with protein A/G agarose beads (Elabscience) for 2 h at 4°C. The lysates were then subjected to overnight IP at 4°C. The next day, protein A/G beads were added for an additional 2‐h incubation to capture immunocomplexes. After five washes with RIPA buffer to remove non‐specific proteins, the bead‐bound complexes were eluted by boiling and analysed by SDS‐PAGE and WB.

In the IP/MS analysis, bound proteins were eluted by boiling and resolved by SDS‐PAGE. After electrophoresis, proteins were visualised by staining the gels with Coomassie Brilliant Blue (Solarbio). This staining provided clear visualisation and precise excision of the target protein bands. Subsequently, the excised gel slices were analysed by MS.

The assessment of m6A modifications in total RNA was performed using an EpiQuik m6A RNA Methylation Quantification Kit (Epigentek) following the manufacturer's instructions. The procedure begins by binding 200 ng of total RNA to the strip wells coated with a high‐affinity RNA binding solution. The m6A residues in the RNA are then specifically recognised and captured by an m6A‐specific antibody, followed by the addition of a detection antibody to facilitate the visualisation of the bound complex. This is achieved colorimetrically by reading the absorbance at a wavelength of 450 nm using a microplate spectrophotometer.

Following RNA isolation with RNAiso Plus (TaKaRa) for dot blot analysis, samples were spotted onto nylon membranes (Millipore, INYC00010) and crosslinked by UV irradiation. To assess RNA quality, membranes were stained with methylene blue and destained with ethanol. Prior to immunodetection, a 2‐h block was performed using 5% milk in TBST. The membranes were probed overnight at 4°C with an anti‐m6A primary antibody (Proteintech; 68055‐1‐Ig). After incubation with a secondary antibody for 2 h at room temperature, m6A signals were visualised using the Omni‐ECL™ Femto Light Chemiluminescence Kit (EpiZyme).

To analyse protein‐RNA interactions, RIP was performed using a commercial kit (BersinBio) per the manufacturer's protocol. Briefly, 10⁷ BMSCs were harvested, PBS‐washed and lysed in IP buffer supplemented with protease and RNase inhibitors. After centrifugation, 10% of the lysate was aliquoted as input and stored at −80°C. The remainder was incubated overnight at 4°C with antibodies against YTHDF1 (Proteintech; 17479‐1‐AP) or METTL3 (Proteintech; 15073‐1‐AP), using normal rabbit IgG as a negative control. Subsequently, magnetic protein A/G beads were added to capture the complexes for 1 h at 4°C. Following incubation, the beads were treated with Proteinase K at 55°C for 30 min, and RNA was extracted using RNAiso Plus. RNA enrichment was assessed by quantitative PCR and normalised to the input.

Following isolation with RNAiso Plus, total RNA was subjected to m6A analysis using the EpiQuik™ CUT&RUN m6A RNA Enrichment Kit (Epigentek). Antibody‐bead complexes were prepared by incubating the anti‐m6A antibody (Proteintech; 68055‐1‐Ig) or control IgG with Magna ChIP Protein A/G Magnetic Beads at 4°C for 90 min. From 200 µg of total RNA per IP, 20 µg was reserved as an input control at −80°C. The remainder was incubated overnight at 4°C with the complexes. Subsequent elution was performed twice with a buffer containing 20 mM m6A monophosphate sodium salt. The input and eluted RNA were then purified again using RNAiso Plus, and enrichment was assessed via qPCR following the established protocol.

Cellular ROS levels in HUVECs were determined using the oxidation‐sensitive fluorescent probe DCFH‐DA (Beyotime) as per the manufacturer's protocol. The cells were subsequently stained with Hoechst 33342 (Beyotime). Finally, positive signals were captured using a Nikon fluorescence microscope or flow cytometry.

The angiogenic activity of HUVECs was evaluated using Confocal Dishes (Corning) precoated with 100 µL/well of Matrix‐Gel™ Basement Membrane Matrix (Beyotime). Initially, HUVECs were stained with the cell‐permeable dye calcein AM for 30 min and then replated onto the precoated Confocal Dishes. The cells were incubated for 4 h at 37°C in a cell culture incubator to allow for capillary‐like tube formation. This formation was subsequently captured under a Nikon fluorescence microscope.

To assess the in vitro migration activity of HUVECs, cell migration assays were conducted using Transwell inserts with an 8.0 µm polycarbonate membrane (Corning). To assess migration under various treatments, HUVECs were seeded into the upper chambers of a Transwell system and incubated for 24 h at 37°C. Following fixation with 4% paraformaldehyde and staining with crystal violet (Beyotime), the cells that had migrated through the membrane were imaged using a Nikon fluorescence microscope for quantitative and qualitative assessment.

HUVECs were cultured in a six‐well plate until they reached 90% confluence. To create artificial wounds, a sterile 10 µL pipette tip was used to scrape the cell monolayer. Subsequently, the scraped cells were washed with PBS to remove cellular debris. Cell migration into the wound area was monitored using a Nikon fluorescence microscope and images were captured at 0 and 24 h post‐scraping. The migration distance, indicating the gap between the two edges of the wound, was quantitatively measured using ImageJ software.

The proliferative capacity of HUVECs was assessed using the BeyoClick™ EdU Cell Proliferation Kit (Beyotime) according to the manufacturer's instructions. Cells were incubated with the EdU reagent and then stained with Hoechst 33342 (Beyotime). EdU‐positive cells were finally imaged under a Nikon fluorescence microscope.

The concentrations of lipid peroxidation products, specifically MDA, were quantified in flap and cell lysates using lipid peroxidation detection kits (Solarbio). Additionally, the relative concentration of GSH in the same samples was measured using a GSH assay kit (Solarbio). The iron concentration in flap and cell lysates was determined using an iron assay kit (Abcam).

Blood from the right ventricle of mice was collected from each group starting at 9:00 AM, at 4‐h intervals following the last treatment, over a 24‐h period. The plasma samples were either immediately analysed or stored at −80°C for future analysis. MT and corticosterone levels in the plasma were quantified using mouse‐specific enzyme‐linked immunosorbent assay (ELISA) kits (Elabscience). In addition, the mice flap tissue for MT, Corticosterone, ARNTL (Coibo), SLC7A11 (Coibo), GPX4 (Cusabio) and FTH1 (Cusabio) assays using mouse‐specific ELISA kits according to the time scheme described above.

Following dissection from flap area II, skin tissue samples were homogenised in ice‐cold RIPA lysis buffer (Beyotime) containing protease inhibitor cocktail (Beyotime) and phosphatase inhibitor cocktail III (Beyotime). After centrifugation of the homogenates at 20 000×g for 30 min at 4°C, the protein concentration in the supernatant was measured with a BCA protein assay kit (Thermo Fisher Scientific). For analysis, 20 µg of total protein per sample was electrophoresed on 10% SDS‐polyacrylamide gels and then electroblotted onto PVDF membranes (Millipore). Following the TBST wash, a 2‐h block was performed using 5% milk in TBST. The membranes were incubated overnight at 4°C with primary antibodies against SLC7A11 (Abcam; ab307601), ACSL4 (Abcam; ab155282), GPX4 (Abcam; ab125066), FTH1 (Zenbio; R23306), HO1 (Abcam; ab68477), SOD1 (Abcam; ab308181), VEGFA (Abcam; ab214424), USP4 (Santa Cruz Biotechnology; sc‐376000), ARNTL (Proteintech; 14268‐1‐AP), anti‐Flag (Abcam; ab205606), anti‐Myc (Abcam; ab206486), Ub (Cell Signaling; 3936S) and GAPDH (Proteintech; 10494‐1‐AP). Subsequently, the membranes were incubated with HRP‐labelled secondary antibodies at room temperature for 2 h. Following detection with an Omni‐ECL™ Femto Light Chemiluminescence Kit (EpiZyme), images were acquired using a fluorescence imaging analysis system. The acquired images were then processed with Image J software (NIH) to determine protein expression levels.

All statistical analyses were performed using SPSS 25.0 and GraphPad Prism version 8.0, with results presented as means ± SEM. Normality of data distribution was assessed using the Shapiro–Wilk test, and homogeneity of variances was verified using Levene's test. For comparisons between two groups, if data met assumptions of normality and homogeneity, Student's t‐test was applied; otherwise, the non‐parametric Mann–Whitney U test was used. For comparisons among more than two groups, one‐way or two‐way analysis of variance (ANOVA) was employed when data were normally distributed with homogeneous variances, followed by Bonferroni's post hoc test for multiple comparisons. If data did not meet parametric assumptions, the Kruskal–Wallis test with Dunn's post hoc analysis was applied. The number of samples for each analysis is detailed in the figure legends. A p value < .05 was considered statistically significant.