Altered regulation of mesenchymal cell senescence promotes pathological changes associated with diabetic wound healing

Pathologic diabetic wound healing is caused by sequential and progressive deterioration of hemostasis, inammation, proliferation, and resolution/remodeling. Cellular senescence promotes physiological wound process; however, diabetic wounds exhibit low levels of senescent factors and accumulate senescent cells, which impair the healing process. In this study, we demonstrate that the number of p15 INK4B + senescent PDGFR-α + mesenchymal cells in adipose tissue transiently increases in early phases of wound healing in non-diabetic mice and humans. Transplantation of adipose tissue from diabetic mice into non-diabetic mice results in wound healing impairment and an alteration in the cellular senescence-associated secretory phenotype (SASP), suggesting that insucient induction of adipose tissue senescence after injury is a pathological mechanism of diabetic wound healing. These results give novel insight into how regulation of senescence in adipose tissue contributes to the wound healing process and provide the basis for the development of therapeutic approaches for wound healing impairment in diabetes.


Introduction
Adipose tissue is attracting attention for promoting wound healing [16][17][18][19][20] . Within subcutaneous adipose tissue, stromal vascular cells and their subsets release growth factors and cytokines critical for neovascularization and wound repair 20 . Subcutaneous tissue has garnered attention particularly in obesity and type-2 diabetes because increased in ammation can alter the outcome of wound repair through the production of large numbers of hormones and cytokines, such as tumor necrosis factor α (TNF-α), Interleukin 6 (IL-6), and plasminogen activator inhibitor 1 (PAI-1) 21 ; these proteins are the major components of the SASP. Senescent cell accumulation in the adipose tissue of patients with diabetes and obesity is associated with insulin resistance and systemic in ammation [22][23][24][25][26] . Furthermore, subcutaneous adipose tissue in obese subjects is characterized by an excessive amount of interstitial brosis and phenotypic changes in human pre-adipocytes, which may contribute to tissue deterioration 27 .
In this study, we investigated cellular senescence in skin and subcutaneous adipose tissue during the wound healing process using a Lepr db/db type-2 diabetic mouse model. We found that growth of subcutaneous adipose tissue reached the wound site in control mice; however, in the diabetic Lepr db/db mice, the adipose tissue remained below the granulating tissue, and thereby impaired wound healing. In addition, transplantation of Lepr db/db -derived adipose tissue into control mice impaired wound healing. In the adipose tissue of control mice, senescent cells increased at 2 days post-wound (DPW) and decreased at 8 DPW, whereas senescent cells in Lepr db/db mice remained low at 2 DPW and increased slightly at 8 DPW. The composition of the SASP factors differ between the control and Lepr db/db mice, resulting in the inhibition of broblast migration. Our results demonstrate that diabetic adipose tissue impairs transient senescence during pathological healing, which causes deteriorated wound healing, suggesting that cellular senesces of adipose tissue could be a therapeutic target for diabetic ulcers.

Results
Type-2 diabetic mice models have impaired wound healing C57BLKS/J Iar -Lepr db /+Lepr db (Lepr db/db ) and C57BLKS/J Iar -m+/+Lepr db (Lepr db/+ ) mice were used in this study as type-2 diabetic and control mouse models, respectively. Blood glucose level ≤ 200 mg/dL (Lepr db/+ ) or > 300 mg/dL (type-2 diabetic, Lepr db/db ) was used to con rm the absence or presence of diabetes, respectively (Fig. 1a). Full-thickness excisional skin wounds were created on the backs of Lepr db/db and Lepr db/+ mice, and the wound closure rate was evaluated at 2 DPW, 5 DPW, and 8 DPW. At each time-point, the wound closure rates were impaired to a greater extent in Lepr db/db mice than in Lepr db/+ mice ( Fig. 1b and 1c). Histological analysis, using hematoxylin and eosin (H&E) and Masson trichrome stain, indicated that adipose tissues were present immediately under the wound eschar region in Lepr db/+ mice at 2 DPW; however, in Lepr db/db mice, cell in ltration with collagen deposition was observed under the wound eschar region (Fig. 1d). At 8 DPW in Lepr db/+ mice, the adipose tissue under the wound eschar decreased with wound closure. By contrast, in Lepr db/db mice, increased cell in ltration with collagen deposition was beneath the wound eschar region, and the adipose tissues were observed under cell in ltration with collagen deposition at the wound closing region (Fig. 1d). Next, we used platelet-derived growth factor receptor α (PDGFR-α) and α smooth muscle actin (α-SMA) antibodies to identify mesenchymal stromal cells that produce ECM and play important roles in wound healing of subcutaneous tissue of the skin 28 . PDGFR-α is a cell surface receptor tyrosine kinase expressed in mesenchymal stromal cells in a variety of tissues and is also used as a marker for adipose progenitor cells [29][30][31][32] . Immunohistochemical analysis indicated that PDGFR-α-positive cells were located along the panniculus carnosus layer in the wound region, and α-SMA-positive cells were located on the panniculus carnosus layer at the wound edge in Lepr db/+ mice. However, Lepr db/db mice exhibited diffuse distribution of α-SMA-and PDGFR-α-positive cells on the wound region at 8 DPW (Fig. 1e).
Transplantation of adipose tissue derived from Lepr db/db mice into Lepr db/+ mice impairs wound healing To elucidate the role that adipose tissue plays in wound healing impairment, adipose tissue derived from Lepr db/db mice was transplanted into the excisional skin wound region of Lepr db/+ mice. The excisional skin wound was sutured closed after transplantation to stabilize the transplanted tissue (Fig. 2a). We evaluated the histological wound healing score at 2 DPW and 8 DPW using H&E staining and Masson trichrome staining [33][34][35] . The histological wound healing scores in Lepr db/db adipose tissue transplanted mice (Lepr db/db ATT mice) were signi cantly lower than those for Lepr db/+ adipose tissue transplanted mice (Lepr db/+ ATT mice) ( Fig. 2b and 2c). Immunohistochemical analysis indicated that Lepr db/+ ATT mice had increased levels of PDGFR-α-and α-SMA-positive cells along the boundary line between the wound region and the normal dermis at 8 DPW (Fig. 2d). By contrast, Lepr db/db ATT mice had nonlocalized PDGFR-α-and α-SMA-positive cells at 8 DPW, which is also observed for impaired wound healing in Lepr db/db mice (Fig. 2d). These results suggest that transplantation of adipose tissue derived from Lepr db/db mice into Lepr db/+ mice impairs wound healing.
To ascertain the role of cell movement in transplanted adipose tissue, we investigated the localization of CM-DiI-labeled adipose tissue-derived cells in wound regions. We identi ed CM-DiI-positive adipose cells in both the Lepr db/+ ATT and Lepr db/db ATT mice in layers below the panniculus carnosus layer at 2 DPW and 8 DPW, and the CM-DiI-positive cells remained under panniculus carnosus layer over time (Fig.   2e).
Senescence-related gene expression levels were transiently increased in subcutaneous adipose tissue during wound healing Senescent cells play bene cial roles in wound healing by expressing SASP factors including PDGF-AA, CCN1, VEGF, and Serpine1, which promote the production of ECM and prevent excessive brosis 9 . By contrast, accumulation of senescent cells is observed in diabetic ulcers 36 . We speculated that different types of cellular senescence occur between normal and diabetic wound healing in response to wounds. Hence, we investigated the time-dependent change in the expression of senescence-related factors during wound healing in Lepr db/db and Lepr db/+ mice. We harvested the skin and subcutaneous adipose tissue at pre-wound, 2 DPW, and 8 DPW. Skin and adipose tissue were divided by the panniculus carnosus layer.
In skin tissue, damage induced the expression of cellular senescence markers. Cdkn1a mRNA transcription levels increased at 8 DPW relative to those at pre-wound and 2 DPW in both Lepr db/db and Lepr db/+ mice ( Supplementary Fig. 1a). The Trp53 transcription level decreased at 2 DPW relative to that of the pre-wound level in Lepr db/+ mice, but the level in Lepr db/db mice was not signi cantly different ( Supplementary Fig. 1a). Cdkn2b is a marker of cell senescence 37,38 . The transcription level of Cdkn2b mRNA increased at 2 DPW relative to that of the pre-wound level in Lepr db/+ mice and decreased at 8 DPW relative to that of the level at 2 DPW in Lepr db/db mice ( Supplementary Fig. 1a). The mRNA transcription level of the SASP-related factor Serpine1 was essentially the same in both mouse models, but the Serpine2 transcription level increased at 8 DPW in Lepr db/db mice ( Supplementary Fig. 1a).
Next, we investigated the mRNA levels in subcutaneous adipose tissue during wound healing. The mRNA transcription levels of Cdkn2b and Trp53 in Lepr db/+ mice increased at 2 DPW relative to those at prewound, and the level of Cdkn2b transcription decreased at 8 DPW (Fig. 3a). In Lepr db/db mice, Cdkn2b and Trp53 transcription levels did not signi cantly change during wound repair, and increased levels of Cdkn1a, Serpine2, and Tgfb1 transcription, relative to 2 DPW, were observed at 8 DPW (Fig. 3a). The mRNA transcription levels of Serpine1 and Il6 did not change in Lepr db/+ mice during wound healing, but Lepr db/db mice exhibited a time-dependent decrease in Serpin1 and a transient increase in Il6 levels (Fig.  3a). To clarify these differences in gene expression patterns related to senescence during the wound healing process between Lepr db/+ and Lepr db/db adipose tissue, principal component analysis (PCA) was performed (Fig. 3b). The PCA plots indicate that Lepr db/+ mice had large changes in senescence-related gene transcription at 2 DPW, whereas at 8 DPW, the transcription of these genes returned to a cluster similar to what was observed pre-wound. By contrast, Lepr db/db mice had minimal changes in senescence-related gene transcription during wound healing (Fig. 3b). To identify the cell type exhibiting senescence in subcutaneous adipose tissue, immunohistochemistry of p15 INK4B , which is encoded by CDKN2B and is an INK4 class of cell-cycle inhibitors 37,39 , and PDGFR-α was performed. In Lepr db/+ adipose tissue, the percentage of p15 INK4B + in PDGFR-α+ cells increased at 2 DPW and 8 DPW, and the uorescence intensity of p15 INK4B in PDGFRα+ cells transiently increased at 2 DPW ( Fig. 3c and 3d).
However, in Lepr db/db adipose tissue, an increase in the percentage of p15 INK4B + in PDGFR-α+ cells was delayed at 8 DPW relative to that observed for Lepr db/+ mice ( Fig. 3c and 3d). In addition, the uorescence intensity of p15 INK4B in PDGFR-α+ cells in Lepr db/db mice increased at 2 DPW but did not decrease at 8 DPW ( Fig. 3c and 3d). These results suggest that cellular senescence rapidly occurs in Lepr db/+ adipose tissue after wounding, and cell senescence is transient. However, in Lepr db/db adipose tissue, cell senescence is delayed after wounding, and this delay may allow for accumulation of senescent cells in diabetic wound healing. Senescent PDGFR-α+ cells accumulate in subcutaneous adipose tissue during wound healing in diabetic patients To test whether the number of senescent cells increases in subcutaneous adipose tissue during wound healing in diabetic or non-diabetic patients, we performed immunohistochemical analysis of PDGFR-α+ cells and looked for expression of p15 INK4B and γH2A.X, which is a DNA damage-induced cell senescence marker 40 . Patient demographics are presented in Table 1. In non-diabetic patients, a negative correlation was observed between the percentage of p15 INK4B -positive cells in PDGFR-α-positive cells and the time post-wound ( Fig. 4a and 4c). By contrast, in diabetic patients, we observed a high positive correlation between the percentage of PDGFR-α-positive cells that were p15 INK4B -or γH2A.X-positive and the time post-wound ( Fig. 4b, 4c, and 4d). These results indicate that cell senescence gradually accumulates in diabetic adipose tissue during wound healing, but in non-diabetic patients, cell senescence occurs robustly in adipose tissue during the early stages of wound healing. SASP factors derived from Lepr db/db mice adipose tissue impair wound healing Overall, our results suggest that the difference in cell senescence between diabetic and non-diabetic adipose tissue in uences wound healing. Because adipose-derived cells stayed under the panniculus carnosus layer during wound healing, SASP factors could be contributing to wound healing. Therefore, we evaluated the differences in the expression levels of SASP factors in Lepr db/+ and Lepr db/db adipose tissue. Adipose tissue was collected from Lepr db/+ and Lepr db/db mice at pre-wound, 2 DPW, and 8 DPW, and organ culture was performed to harvest SASP-containing culture media. A proteome pro ler antibody array was used to investigate whether SASP factors were present in the culture media. At pre-wound, adipose tissue in Lepr db/+ mice had higher expression levels of IGFBP3 and IGFBP5, but Lepr db/db mice have higher expression levels of CCL6 and CCL11. At 2 DPW, both Lepr db/+ and Lepr db/db adipose tissue had increased expression levels of various secretory factors; however, the composition was different (Fig.   5a, 5b, and Supplementary Fig. 2). The Lepr db/+ adipose tissue exhibited higher expression levels of Adiponectin, Ang2, CCL2, CRP, CXCL3, IL1Ra, MMP9, VEGF, and CCN4 (Fig. 5a, 5b, and Supplementary   Fig. 2). By contrast, Lepr db/db adipose tissue had higher expression levels of IL11, CCL11, and MMP3 (Fig.  5a, 5b, and Supplementary Fig. 2). At 8 DPW, the levels of VEGF, IGFBP family members, and Serpin F1, which are important factors for wound healing 9,41 , increased in Lepr db/+ adipose tissue-derived conditioned media, but Lepr db/db conditioned media continued to have increased expression levels of CCL6, CCL11, and CXCL2 (Fig. 5a, 5b, and Supplementary Fig. 2). Finally, to test whether the conditioned media derived from Lepr db/+ or Lepr db/db adipose tissue promotes or inhibits wound healing, we performed wound scratch assays that were treated with these conditioned media. We found that the conditioned media derived from Lepr db/db adipose tissue delays wound closure relative to the results from Lepr db/+ adipose tissue ( Fig. 5c and 5d).

Discussion
Cellular senescence contributes to wound healing in both the normal and pathological healing processes [9][10][11][12][13][14][15]42 . Therefore, it is critical to understand the role that cellular senescence plays in wound healing impairment, and modulating this senescence presents a novel therapeutic approach to reduce this impairment. In this study, we observed a rapid and transient increase in senescent cells in control adipose tissue; however, the number of senescent cells in diabetic Lepr db/db adipose tissue gradually increased post-wound. In normal wound healing, the number of senescent cells in mice skin increases 3-6 DPW and decreases 9 DPW, and genetic depletion of p16 INK4A -expressing senescent cells delays wound healing 9 . Senescent broblasts and endothelial cells in the dermis accelerate wound closure by inducing myo broblast differentiation through the secretion of platelet-derived growth factor (PDGF) AA 9 . Our ndings demonstrate the importance of transient cell senescence during wound healing, especially in adipose tissue.
In wound healing, dermal adipose tissue contributes to wound healing 19 and to the initiation of cutaneous brosis through adipocyte-myo broblast transition 43 . We demonstrate that diabetic Lepr db/db mouse-derived adipose tissue transplantation into Lepr db/+ impaired the wound healing process, even though the donor was normal dermal tissue. Furthermore, we found that the expression levels of SASP factors differ between Lepr db/db -and diabetic Lepr db/db -derived adipose tissue, which affects wound healing.
In the diabetic wound, VEGF and CCL2 enhance wound healing by promoting angiogenesis 20 and regulating macrophages 44 , respectively. Secretion of VEGF increased at 2 DPW in adipose tissue from Lepr db/+ mice but not in tissue from Lepr db/db mice. CCL2 secretion also increased at 2 DPW in adipose tissue in Lepr db/+ mice. The levels of CCN4, a regulator of senescence 45 that enhances wound healing by regulating dermal broblast cell migration, proliferation, and ECM expression 46 , increased at 2 DPW in adipose tissue from Lepr db/+ mice. By contrast, diabetic Lepr db/db adipose tissue exhibited higher secretion of IL11, CCL11, and MMP3 at 2 DPW. IL11 is a member of the IL6 family of cytokines and binds to IL11 receptor subunit alpha (IL11RA), which is expressed on stromal cells and promotes tissue brosis 47,48 . In atopic dermatitis, the number of IL11-expressing cells is elevated in skin biopsy specimens relative to controls, and a signi cant correlation exists between IL11 and type I collagen deposition 49 .
MMPs, known SASP factors 1 , are important regulators of ECM degradation and deposition, and the timing and level of MMP activation are vital for determining whether successful wound healing or chronic non-healing is observed 50 . In normal wound healing, MMP9 expression increases concurrent with broblast migration into the wound area, and MMP3 levels increase at re-epithelialization 50 . In this study, the levels of MMP9 secretion increased in Lepr db/+ -derived adipose tissue at 2 DPW, and MMP3 levels increased at 8 DPW. However, in Lepr db/d -derived adipose tissue, MMP3 increased at both 2 DPW and 8 DPW, suggesting that dysregulation of MMPs in adipose tissue results in wound healing impairment in Lepr db/db mice. Furthermore, the secretion levels of IGFBP3 and Serpin F1 in Lepr db/+ adipose tissue increased at 8 DPW. IGFBP3 is one of 6 structurally related IFGBPs that bind to IGF peptides with high a nity, and the IGFBP3•IGF1 complex binds to brin clots and concentrates at wound sites to facilitate wound healing 41 . Serpin F1 is a pigment epithelium-derived factor (PEDF), and PEDFs contribute to the resolution of wound healing by causing regression of immature blood vessels and stimulating maturation of the vascular microenvironment, which promotes a return to tissue homeostasis after injury 51 . Our ndings suggest that Lepr db/+ adipose tissue-secreted SASP factors promote wound healing by inducing cell migration and proliferation and inhibit brosis by inducing broblast senescence, which results in resolution of wound healing. By contrast, Lepr db/db adipose tissue-secreted SASP factors induce chronic in ammation, which results in wound healing impairment. Although the detailed function of p15 INK4B is not as well understood as those for p16 INK4A and other senescence-related factors 55 , the function and protein structure of p15 INK4B are predicted to be similar to p16 INK4A , and p15 INK4B is upregulated in TGF-β-related cell senescence 56 . p15 INK4B binds to CDK4 and CDK6, preventing their binding to cyclins and thereby inhibiting cell cycle progression 56 . The antiproliferative action of TGF-β is also mediated through the inhibition of c-Myc expression. c-Myc inhibits the expression of p15 INK4B and p21 Cip1 in proliferating cells 57 , and suppression of c-Myc by TGFβ limits c-Myc availability and suppresses the activity of p15 INK4B and p21 Cip158,59 . TGF-β expression and activation are rapidly induced in response to injury, and TGF-β controls wound healing by acting as a potent chemoattractant for monocytes and broblasts 60 . Although the role of TGF-β in the control of diabetic wound healing remains to be fully explained, increased levels of TGF-β1 are linked with type-2 diabetes, and TGF-β1 contributes to diabetic wound healing 61 . The regulation of TGF-β-related cell senescence may be a therapeutic target for diabetic ulcers.
Our intriguing results suggest that the transient increase in the number of senescent PDGFR-α+ cells in adipose tissue is important for wound healing, and that diabetic wounds exhibit a decrease in cell senescence in the acute wound-healing phase that results in impairment of wound healing concurrent with accumulation of senescent cells. However, the study had several limitations. We analyzed subcutaneous adipose tissue; however, intradermal adipose tissue contributes to wound healing through regulation of adipocyte precursor proliferation and mature intradermal adipocyte repopulation in the skin after wounding 19 . The anatomical location of intradermal adipose tissue and subcutaneous adipose tissue is separated by the panniculus carnosus in mice, but human skin does not have a detectable panniculus carnosus. It is also di cult to con rm the existence of intradermal adipose tissue in humans because of the obvious inability to conduct lineage-tracing studies 62,63 . Even though intradermal and subcutaneous adipose tissue are not physically demarcated in humans, increasing evidence suggests that there is a functional distinction between these tissues 63 . Hence, careful interpretation of our results is needed, but at a minimum, we have demonstrated that cellular senescence in subcutaneous adipose tissue contributes to both normal and diabetic wound healing. Next, our study focused on PDGFR-a+ mesenchymal cell senescence in adipose tissue. Macrophages are one of the senescent cell populations in the diabetic wound 36 , and their abundance is correlated with senescent cell burden in adipose tissue [23][24][25][26] . Although PDGFR-a+ cells play important roles in adipose tissue homeostasis 64 , further study is needed regarding the role that macrophage senescence plays in adipose tissue during wound healing.
Finally, we used the Lepr db/db diabetic mouse model, which is widely used as a typical delayed healing model 65 , and our ndings are similar to what is observed in type-2 diabetic patients. To better understand adipose tissue-related mechanisms in diabetic wound healing, further study is needed using a type-1 diabetic model, which exhibits adipose tissue atrophy and brosis 66-68 . In summary, we demonstrate that transient mesenchymal cell senescence occurs in adipose tissue in physiological wound healing; however, accumulation of cell senescence occurs concurrently with expression of different components of the SASP in diabetic wound healing, suggesting that impairment of senescence in adipose tissue contributes to intractable wound healing in diabetes.

Human wound samples
Six wound tissue samples from ve diabetic patients (age: 68.7±9.0 years old; HbA1c(NGSP): 7.02±2.15%) and six wound tissue samples from six non-diabetic patients (age: 58.0±21.7 years old) were used in the study. The tissues used in the study were debrided and disposed during surgery for treatment, and no additional excisions were made for the study. Samples were recruited after agreement was obtained through informed consent. Detailed information about these subjects is given in Table 1.
The Ethical Review Board at the Sapporo Medical University in Japan approved the study.

Animals
The Committee of the Animal Experimentation Center at the Sapporo Medical University School of Medicine approved all animal protocols. Mice were fed a standard chow diet and were maintained on a 12 h light / 12 h dark cycle with free access to food and water at all times. Male C57BLKS/JIar-+Lepr db /+Lepr db (Lepr db/db ), male C57BLKS/JIar-m+/+Lepr db (Lepr db/ ), and female C57BL/6 mice (age>11 weeks; Sankyo Lab Service, Tokyo, Japan) were used in the experiments. At 10-11 weeks of age, blood samples were taken from the tail to measure blood glucose, which was con rmed to be above 300 mg/dL in the diabetic mice and below 200 mg/dL in the control mice. Nipro Stat Strip XP2 (Nova Biomedical, Tokyo, Japan) was used for blood glucose level measurements.

Wound model
Mice were anesthetized using mixed anesthetic agents (medetomidine, midazolam, butorphanol) 69 . To create the skin ulcer model, a 10 mm diameter, full-thickness circle excision was made on the back of Lepr db/db and Lepr db/ mice. Wounds were photographed with a digital camera (COLPIX S9700; Nikon, Tokyo, Japan). Images were analyzed by tracing the wound margin, and the enclosed pixel area was calculated using the Image J software, version 1.5 (National Institutes of Health, Bethesda, MD, USA). The wound areas were standardized by measuring the captured image. Using the original wound size for comparison, the percentage of wound closure was calculated as follows: day n area / day 0 area × 100 (%).

Adipose tissue transplantation
To perform adipose tissue transplantation, 100-120 mg of subcutaneous adipose tissue was collected from the back of Lepr db/db and Lepr db/+ mice and labeled with Cell Tracker Vybrant CM-DiI Cell-Labeling Solution (V22888; Thermo Fisher Scienti c, MA, USA). Labeling of adipose tissue was performed according to the protocol. Brie y, tissue was incubated with CM-DiI at 37°C for 5 minutes and at 4°C for 15 minutes. The labeled adipose tissue was transplanted onto the back of Lepr db/+ mice, and the wound was sutured with nylon thread.

Histological analysis and wound healing scoring
Wound tissue and adipose tissue were harvested with marginal skin and xed in 4% paraformaldehyde at 4℃ overnight. The following day, the tissue was cut into 5 mm sections and was para n-embedded. Each slide was stained with H&E and Masson trichrome, which were used for histological wound-healing scoring. For the scoring, an examiner assesses the progression of wound healing on a 12-point scale for in ammation, granulation, and collagen deposition 33,35 . In brief, each sample was given a score from 1 to 12: 1-3, none to minimal cell accumulation and granulation tissue or epithelial migration; 4-6, thin, immature granulation tissue dominated by in ammatory cells but with few broblasts, capillaries, or collagen deposition and minimal epithelial migration; 7-9, moderately thick granulation tissue, dominated by in ammatory cells and more broblasts and collagen deposition; and 10-12, thick, vascular granulation tissue dominated by broblasts and extensive collagen deposition. Because the wound was sutured, epithelialization was excluded from the evaluation criteria. All images were captured using a BZ-X700 uorescence microscope (KEYENCE, Osaka, Japan).
Immuno uorescence staining Para n-embedded sections were depara nized and rehydrated for immunostaining. Antigen retrieval was performed in a microwave oven (95-98℃ for 10 minutes) using citrate buffer (10 mM sodium citrate, pH 6.0). After cooling, the slides were washed twice with deionized water and once with 1X Trisbuffered saline with Tween-20 (TBST) for 5 minutes each. The sections were blocked with 1% bovine serum albumin (BSA) in TBST for 15 minutes at room temperature (RT) and were then incubated with primary antibodies overnight at 4°C or for 1 hour at RT. After washing three times with TBST for 5 minutes each, the sections were incubated with SignalStain Boost IHC Detection Reagent (HRP, Rabbit #8114; Cell Signaling Technology, Danvers, MA, USA) for 30 minutes at RT in the dark. The sections were then washed in TBST three times for 5 minutes each and treated with TSA Plus Working Solution (Fluorescein, Cyanine 3, and Cyanine 5; AKOYA BIOSCIENCES, Malborough, MA, USA) for 10 minutes at RT in the dark. For multiplex staining, stripping was performed in a microwave oven (95-98℃ for 10 minutes) using citrate buffer. After cooling, staining with different tyramide uorescent labels was performed according to the above procedure. Nuclei were labeled with Cellstain DAPI solution (1:1000, 4′,6-diamidino-2phenylindole; Dojindo, Kumamoto, Japan), and after further washes, the sections were mounted in VECTASHIELD (Vector Laboratories, Burlingame, CA, USA). The following primary antibodies were used: rabbit anti-p15 (1:500; ab53034; Abcam, Cambridge, UK), rabbit anti-PDGFR-a (1:1000 (mouse) and 1:500 (human); D1E1E, XP; Cell Signaling Technology), rabbit anti-a-SMA (1:500; D4K9N, XP; Cell Signaling Technology), and rabbit anti-phospho-histone H2A.X(Ser139) (1:480; Ser139, 20E3; Cell Signaling Technology). These primary antibodies were used after dilution with SignalStain Antibody Diluent (Cell Signaling Technology). All images were captured using a BZ-X700 uorescence microscope (KEYENCE).

RNA extraction and quantitative real-time PCR
Total RNA was isolated from skin wound tissues and subcutaneous adipose tissues using TRI Reagent (Molecular Research Center, Cincinnati, OH, USA) and was reverse-transcribed into cDNA using the iScript cDNA Synthesis Kit (1708891; Bio-Rad). Quantitative real-time PCR was performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) in a 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) with the following conditions: 95°C for 30 seconds and 40 ampli cation cycles of 95°C for 15 seconds and 60°C for 1 minute. Expression levels were normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) or beta-actin (ACTB) levels. The primer sequences used for the PCR analysis are shown in Supplementary Table 1.

Organ culture and SASP characterization
Subcutaneous adipose tissue under the wound for ve Lepr db/+ or Lepr db/db mice in each group [0 DPW (before wound), 2 DPW, and 8 DPW] was collected and put into phosphate-buffered saline (PBS) supplemented with 2% penicillin/streptomycin. The tissue was then washed and transferred to Dulbecco's Modi ed Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). The volume of the culture medium was 1 mL for every 60 mg of adipose tissue, the medium was changed once at 24 hours post-extraction, and the culture medium was collected at 48 hours post-extraction for use in the following assay. The culture medium from adipose tissue culture was characterized using the Proteome Pro ler Mouse XL Cytokine Array Kit (R&D Systems, Minneapolis, MN, USA). Developed lms were imaged and integrated density analysis was performed using Image J, version 1.5 (National Institutes of Health).

Cell preparation and in vitro wound-healing assays
Skin broblasts were collected from C57BL/6 mice. After euthanasia, skin was harvested and digested in Liberase TL (5401020001, Merck) for 120 minutes at 37°C. The digested skin slurries were ltered through a 100 µm cell strainer (EASYstrainer Cell; Greiner Bio-One, Kremsmuenster, Austria) and through a 70 µm cell strainer (Greiner Bio-One). Cells were suspended in DMEM supplemented with 10% PBS and 1% penicillin/streptomycin and were cultured in a T75 culture ask. Cells reached 80-90% con uence after incubation for 1-2 weeks, and the cells were passaged. Cells from passage 2 were used for the in vitro studies. In vitro wound healing was studied using 2-well Culture-Inserts (Ibidi, Bavaria, Germany). Mouse skin-derived broblasts were cultured in 2-well Culture-Inserts with adipose tissue-cultured media for 6 days, and phase contrast images were obtained every 24 hours and immediately after removing the 2-well Culture-Insert using Primovert and Axiocam208 microscopes (Carl Zeiss, Jena, Germany).

Statistical analysis
Mice were only excluded from the study if they had visible wounds from ghting. Statistical analyses were performed using R (The R Foundation for Statistical Computing, Vienna, Austria). Statistical signi cance between two groups was determined using an unpaired t-test. A one-way or two-way analysis of variance (ANOVA) was conducted to assess differences among three or more groups. Pairwise comparisons were made only when the ANOVA test identi ed a statistical signi cance. p-values for multiple comparisons were adjusted using the Tukey method. Statistical analyses were performed using EZR, which is a graphical user interface for R 70 . Two-sided p-values <0.05 were considered statistically signi cant. Quantitative data are presented as either the mean±standard error of the mean (SEM) or median with interquartile range (IQR) and 1.5 × IQR. Box-and-whisker plots and bar plots were generated using ggplot2, a plotting system for R based on The Grammar of Graphics (The R Foundation for Statistical Computing, Vienna, Austria). The R packages FactoMineR and factoextra were used to generate heat maps, Ward's hierarchical agglomerative clustering, and principal component analyses.     Cellular senescence in subcutaneous adipose tissue during wound healing in Leprdb/+ and Leprdb/db mice (a) Relative mRNA expression of senescence-related genes at pre-wound, 2 DPW, and 8 DPW in Leprdb/+ and Leprdb/db mice (n=6¬-7 for each group) and (b) principal component analysis (PCA) of the levels of senescence for senescence-related genes at pre-wound, 2 DPW, and 8 DPW for Leprdb/+ and Leprdb/db mice. (c) Representative images of PDGFR-α and p15INK4B immunostaining of adipose tissue at pre-wound, 2 DPW, and 8 DPW and (d) quantitative data for the percentage of PDGFR-α-and p15INK4B-positive cells and the uorescence intensity of p15INK4B in PDGFR-α-positive cells (n=3¬-4 for each group). Quantitative data are presented as the means and medians with IQRs and 1.5 times the IQR and are displayed as box-and-whisker plots. p-values were determined using one-way ANOVA adjusted by the Tukey method (*p<0.05 and **p<0.001).

Figure 4
Cellular senescence in subcutaneous adipose tissue during wound healing in diabetic patients (a-b) Representative images of PDGFR-α, p15INK4B, and γH2A.X immunostaining of adipose tissue during wound healing in diabetic or non-diabetic patients. (c-d) Correlation between the percentage of p15INK4B-or γH2A.X-positive cells in PDGFR-α-positive cells and the time post-wound (n=6 for each group). Correlations were examined statistically using Pearson's correlation coe cient, and 95% con dence intervals are shown with translucent ll corresponding to each marker color.