microRNA-155 inhibition restores Fibroblast Growth Factor 7 expression in diabetic skin and decreases wound inflammation

Treatment for chronic diabetic foot ulcers is limited by the inability to simultaneously address the excessive inflammation and impaired re-epithelization and remodeling. Impaired re-epithelization leads to significantly delayed wound closure and excessive inflammation causes tissue destruction, both enhancing wound pathogen colonization. Among many differentially expressed microRNAs, miR-155 is significantly upregulated and fibroblast growth factor 7 (FGF7) mRNA (target of miR-155) and protein are suppressed in diabetic skin, when compared to controls, leading us to hypothesize that topical miR-155 inhibition would improve diabetic wound healing by restoring FGF7 expression. In vitro inhibition of miR-155 increased human keratinocyte scratch closure and topical inhibition of miR-155 in vivo in wounds increased murine FGF7 protein expression and significantly enhanced diabetic wound healing. Moreover, we show that miR-155 inhibition leads to a reduction in wound inflammation, in accordance with known pro-inflammatory actions of miR-155. Our results demonstrate, for the first time, that topical miR-155 inhibition increases diabetic wound fibroblast growth factor 7 expression in diabetic wounds, which, in turn, increases re-epithelization and, consequently, accelerates wound closure. Topical miR-155 inhibition targets both excessive inflammation and impaired re-epithelization and remodeling, being a potentially new and effective treatment for chronic diabetic foot ulcers.

In murine models, whole-body over-expression of miR-155 leads to hypoglycemia, because miR-155 positively regulates insulin sensitivity and glucose uptake in insulin-sensitive cells, whereas complete deficiency of miR-155 results in hyperglycemia 11 . MiR-155 is expressed by immune cells 16 , including Th1 and Th17 17 , as well as other cells in inflammatory conditions 18 , and plays a pro-inflammatory role in cells, by targeting Cytotoxic T-Lymphocyte-associated protein (CTLA)-4 19 , suppressor of cytokine signaling (SOCS)1, and SH2-Containing Inositol-5′-phosphatase (SHIP)1 from the Toll-Like Receptor (TLR)-2 pathway 20 . Furthermore, inhibition of regulatory T cells by miR-155, both in control 17 and diabetic subjects 13 , promotes the exacerbation of inflammation, which is involved in the pathology of psoriasis 17 . Moreover, anti-inflammatory drugs such as L-arginine and ibuprofen 12 , resveratrol 21 , vitamin D 22 or M2000 23 have been shown to downregulate miR-155.
MiR-155 is also important for the function of skin cells involved in wound healing, including keratinocytes 24 , dermal mesenchymal stem cells 25 , mast cells 26 , melanocytes 27 , adipocytes 28 and fibroblasts 10 . Furthermore, miR-155 deficiency 29 and miR-155 inhibition 30 was shown to improve wound healing, in healthy and diabetic animal models, but the mechanism by which miR-155 impairs wound healing remains elusive.
In this work, we aimed to investigate the effect of specifically inhibiting miR-155 in diabetic skin on wound healing in a type 1 diabetic mouse model. Our diabetic mouse model confirmed a significant increase in miR-155 expression in the skin during wound healing and topical miR-155 inhibition improved wound closure via de-repression of FGF7 (fibroblast growth factor 7).

Results
Wound healing is compromised under diabetic conditions, in humans 3 and in rodents 31 . To confirm wound healing impairment, we used a well-established mouse model of diabetic wound healing 31 . Wound closure was monitored for 10 days post-wounding. Our results demonstrate wound healing impairment in diabetic mice (Fig. 1A).
To evaluate the effect of diabetes on skin miR expression during wound healing, we used skin samples collected at baseline (Day 0) and Days 3 and 10 post-wounding and profiled miR expression. The array screening results indicate that in diabetic mice (n = 6 in each group -days 0, 3 and 10), a large fraction of the detected miRs show more than two-fold difference in expression at days 0, 3 or 10 following wounding (Fig. 1B): 36, 29 and 17 miRs were upregulated and 44, 35 and 37 miRs were downregulated in diabetic skin at these time points.
Since a large fraction of the analyzed miRs were decreased in diabetic mouse skin, we further evaluated the mRNA expression levels of four proteins involved in miR processing, TAR RNA Binding Protein 1 (Trbp-1), DiGeorge Syndrome Chromosomal Region 8 (Dgcr8), Dicer and Drosha-2 (Fig. 3). Interestingly, mRNA levels of Drosha-2, Dgcr8 and Dicer were significantly increased (p < 0.001 for Drosha-2 and Dgcr8 and p < 0.01 for Dicer) in diabetic mouse skin, before and after wound induction, when compared to non-diabetic mice. While the overall miR expression was decreased at Day 0 in diabetic mice, the increased RNA transcript level of these Expression of 641 unique miRs, shown as heat map, in skin wound samples from the same control and diabetic mice, collected at baseline (day 0) and at days 3 and 10 post wounding. miR expression levels were normalized to the levels of U6 present on the arrays and control levels at day 0. MiRs changed by more than 2-fold (up or down) were included in the heat map. Data were analyzed using Student's t-test; *p < 0.05; **p < 0.01; ***p < 0.001. www.nature.com/scientificreports www.nature.com/scientificreports/ factors involved in miR processing (Drosha-2, Dgcr8, Dicer, Trbp1) may suggest an unaccommodated feedback from decreased miR action in diabetic mouse skin at Day 0.
Since miR-155 was one of the most differentially expressed miRs in diabetic and control skin and it has previously been shown to be important for wound healing 29 , we chose to investigate miR-155 in depth. Importantly, we also evaluated the expression of one in silico identified target of miR-155's, FGF7 (also known as keratinocyte growth factor) (Fig. 4A). As opposed to control mice, where FGF7 mRNA levels decrease during wound healing, in diabetic mouse skin, the levels of FGF7 mRNA inversely correlate with the expression of miR-155, increasing significantly during wound healing. To test miR-155 efficacy in downregulating FGF7 expression, we constructed an FGF7-luciferase-reporter vector containing the 3′ UTR (untranslated region) of the FGF7 gene ((FGF7 UTR) (Fig. 4B) and analyzed luciferase activity in response to different concentrations of miR-155 inhibitor in human HaCaT keratinocytes. FGF7 UTR mediated reporter-gene activity increased (FGF7 UTR: 1.65-fold ± 0.06 at 25 pmol inhibitor/well and 1.99-fold ± 0.12 at 35 pmol inhibitor/well compared with control, both p < 0.001) with miR-155 inhibition in a dose-dependent manner. Importantly, removing just one of the predicted two miR-155 target sites of the FGF7 UTR significantly decreases the response to miR-155 inhibition (Fig. 4B).
While the scratch wound in vitro assay is a simple wound healing model and does not reflect the multiple cellular interactions taking place in in vivo, we performed wound healing experiments in diabetic mice. To test the in vivo effect of miR-155 inhibition we measured wound closure kinetics over a period of 10 days (Fig. 5A). Dorsal wounds were induced on the back of diabetic mice and were subsequently topically treated with different concentrations of the miR-155 inhibitor, twice a day, until Day 3 (n = 6 in each group). Topical administration of the miR-155 inhibitor after wound induction significantly improved wound closure, especially with 2.5 nmol dose applied. Improvement in wound closure was visible as early as the first day of treatment and was persistent even after the 3 days of treatment until the end of the experiment, 10 days post-wounding. The 0.25 nmol dose had no effect on wound healing and 2.5 nmol and 10 nmol doses had a similar effect (data not shown). Moreover, the negative control inhibitor did not display altered wound healing kinetics compared with the saline control (data not shown), showing that the action of the miR-155 inhibitor is very likely to be sequence specific.

Figure 2.
The expression of miR-155-5p and other miRs are significantly increase in diabetic mouse skin. The expression of selected miRs was confirmed by RT-qPCR and normalized to the mean of TFIIB and U6 levels and control baseline. miR-155-5p and miR-126-5p were significantly up-regulated in diabetic skin at baseline and miR-127-3p, miR-411-5p, miR-31-3p and miR-31-5p are significantly down-regulated in diabetic wound skin at day 10 post-wounding. Data were analyzed by two-way ANOVA with Dunnets' post hoc correction; *p < 0.05; ***p < 0.001. www.nature.com/scientificreports www.nature.com/scientificreports/ HE and Herovici's staining at Day 10 post-wounding indicated anti-inflammatory effects of in vivo miR-155 inhibition (Fig. 5B). As opposed to diabetic mouse wounds treated with the negative inhibitor, where a clear immune cell infiltration was observed, diabetic mouse wounds treated with 2.5 nmol of miR-155 inhibitor presented a complete re-epithelization, with no traces of inflammation. We also observed a prevalence of young, unorganized collagen in control wounds (Fig. 5B), in clear contrast with miR-155 inhibitor treated wounds, where young collagen expression was residual, indicating that miR-155 inhibition accelerates wound maturation. To further evaluate immune cell infiltration at Day 10 post-wounding, we performed fluorescent immunohistochemical staining for CD3, to identify T-cells and CD68 to identify macrophages (Fig. 6) on control and miR-155 inhibitor treated wounds (n = 3 in each group). Macrophage wound infiltration was significantly decreased (32% of Ctrl.) in diabetic wounds treated with 2.5 nmol of miR-155 inhibitor in control diabetic wounds than (Ctrl.: 77.7 ± 3.1 vs MiR-155 inhibitor: 24.7 ± 1.8 cells/field; p < 0.0001) (Fig. 6B), while T-cell infiltration was 39% of Ctrl. following inhibition of miR-155 (Ctrl.: 22.0 ± 6.0 vs MiR-155 inhibitor: 8.7 ± 1.1 cells/field, p = 0.05) (Fig. 6C).
To elucidate the action of the miR-155 inhibitor on FGF7 expression, we performed fluorescent immunohistochemical staining for FGF7 (Fig. 7A), on samples collected at Day 10 post-wounding. We stained non-diabetic wounded skin and observed high levels of FGF7 associated with hair shafts, as well as staining throughout sub cutis and the epidermis. In diabetic skin, treated with the negative control inhibitor oligo, FGF7 levels were clearly suppressed (3.8% ± 2.5%, p < 0.001 compared to non-diabetic control), both in the hair shaft follicles and in the epidermis. Treatment with miR-155 inhibitor markedly increased FGF7 levels in hair shafts, in the subcutis and in the epidermis of diabetic mice (41.3% ± 19.6%, p < 0.01 compared to diabetic mice) (Fig. 7B). Thus, FGF7 is clearly de-repressed at the protein level by miR-155 inhibition, while the shorter Seed miR-155 inhibitor had no effect in vivo, under the tested conditions. www.nature.com/scientificreports www.nature.com/scientificreports/ We tested the action of the miR-155 inhibitor on FGF7 mRNA levels in skin from diabetic mice wounded and subsequently topically treated with inhibitor. Despite clear functional actions of the miR-155 inhibitor at Day 3, there was no effect of the miR-155 inhibitor on steady state FGF7 mRNA level or on other predicted mRNA targets (Ctla4, Hbp1, Stat3, Nfia) (data not shown). Similarly, despite clear effects on scratch-assay migration (Suppl. Fig. 1), the miR-155 inhibitor did not alter measured miR-155 levels or FGF7 mRNA levels (data not shown). We also evaluated the effect of miR-155 inhibitor on angiogenesis, by staining the vessels for CD31, and we did not observe changes between the negative control oligo and the miR-155 inhibitor (Suppl. Fig. 2).

Discussion
Skin injury triggers acute inflammatory responses, beginning with recruitment of neutrophils and monocytes to the site of injury that, in turn, secrete various inflammatory cytokines, chemokines and growth factors coordinating wound repair 32 . Excessive inflammation observed in DFU causes tissue damage through the release of increasing levels of various proteins involved in bacterial control, such as granzymes and perforins, that degrade the extracellular matrix 33 and inhibit re-epithelization 34 impairing wound healing. This excessive inflammation is partially controlled by miRs, including miR-155 35 , and its suppression, either directly using a miR-155 inhibitor 36 or indirectly using other substances 21,37 , has an anti-inflammatory effect in different conditions.
Our results show that miR-155 is over-expressed in diabetic mouse skin, when compared to non-diabetic control mice. Most importantly, we show, for the first time, that FGF7 mRNA expression is significantly decreased  www.nature.com/scientificreports www.nature.com/scientificreports/ during wound healing under diabetes conditions, contrary to what is observed in wound healing in healthy mice. Moreover, we demonstrate that topical application and inhibition of miR-155 for only 3 days has long-lasting and positive effects on diabetic wound healing.
Diabetic individuals show decreased levels of miR-155 expression in peripheral blood plasma 11,38 and mononuclear cells 39 , as well as in the kidney, heart, aorta and sciatic nerve 40 , when compared to age and sex-matched controls. However, miR-155 was shown to be increased in peripheral blood of diabetic patients with retinopathy 13 and in retinal cells of diabetic rats 41 . Moreover, circulating miR-155 levels are down-regulated in pre-diabetic individuals 38 and this decrease is not reversed by insulin treatment and glycemic control 42 . Variations between tissues or related to concomitant diabetes-associated pathologies may explain the differences in miR-155 expression observed in these studies, reinforcing the need to characterize miR-155 expression in the various tissues and conditions.
It has been reported that miR-155 has pro-inflammatory effects and that miR-155 inhibition leads to a reduction in inflammation, observed by decreased IL-1β and TNF-α levels 43 , and more regular collagen fiber arrangement 43 and faster diabetic wound healing in a rat model 30 . Non-diabetic mice lacking miR-155 show improved wound repair 29 through increase in M2 macrophage polarization and increase in type I collagen deposition. Accordingly, increased miR-155 expression is associated with M1 polarization which, in turn, has a negative effect on diabetic wound healing 44 . Our results also demonstrate that topical miR-155 inhibition reduces T-cell and macrophage wound infiltration, consequently reducing tissue inflammation and promotes collagen fiber rearrangement, accelerating wound maturation in diabetic wounds. It will be relevant to further explore the role of miR-155 on inflammatory responses and the extracellular matrix composition during wound healing.
Moreover, our results show, for the first time, that miR-155 impairs diabetic wound re-epithelization, by targeting FGF7, which is essential for keratinocyte migration, proliferation and consequently wound closure 45 . Early www.nature.com/scientificreports www.nature.com/scientificreports/ studies have shown that FGF7 over-expression favors wound healing in various ways that are not restricted to keratinocyte or fibroblast proliferation and migration, but also involve increased revascularization and antimicrobial effects 46 . Here, we demonstrate that topical inhibition of miR-155 leads to increased keratinocyte migration and faster wound closure, in a dose-dependent manner. We also show in vitro that miR-155 regulates keratinocyte migration through inhibition of the FGF7 3′ untranslated region (UTR).
However, when evaluating the temporal modulation of miR-155 and FGF7 it is apparent that other factors than miR-155 are likely to contribute to the regulation of FGF7 mRNA and protein levels. For example, in diabetic animals miR-155 peaks at day 0 and is significantly less expressed at day 10, while no difference in FGF7 mRNA abundance is observed at day 0 between diabetics and controls, and FGF7 transcript is significantly more abundant at day 10 in diabetics. When considering other miRNAs as regulators then miR-21 also has a target site in the FGF7 3′UTR, which may contribute to suppression of FGF7 protein at day 0 in diabetic skin (where miR-21 is upregulated). Moreover, other factors than microRNAs (such as inflammatory cytokines 47 ) control FGF7 mRNA and protein amounts. In addition, inhibition by microRNAs on their mRNA or protein targets is not instantaneous but rather slow, and the temporal effect of a microRNA also depends on the turnover of the protein, providing a possible explanation for discrepant observations between the miR-155 and FGF7 levels.
The FGF7 mRNA levels and FGF7 immunostainings in diabetic skin at day 10 following wounding did not correlate, but the mRNA of FGF7 in diabetic skin could be increased in a compensatory response. Not all microR-NAs act to cause complete degradation of their cognate mRNA targets, but instead primarily halts the translation of the mRNA into protein. In fact, we could not detect significant increase in the FGF7 mRNA by treatment with www.nature.com/scientificreports www.nature.com/scientificreports/ the miR-155 inhibitor (data not shown), despite clear increases at the FGF7 protein levels and responses of the FGF7 3′UTR to miR-155 inhibitor in reporter assays.
Our results suggest that topical administration of miR-155 inhibitors, at the wound site, may have significant therapeutic value in DFU treatment, especially if applied during the first days after wounding, where miR-155 inhibition will have an immunosuppressive effect and decrease tissue damage.

Methods
Multiple low-dose streptozotocin diabetic mouse model. C57BL Diabetes was induced as previously described 31 . Briefly, streptozotocin (STZ) (50 mg/kg) in saline solution was injected i.p., for 5 consecutive days. Seven days post STZ injection, blood glucose was measured to confirm the diabetic phenotype. Mice with blood glucose levels above 250 mg/dL (Accu-Chek glucometer, Roche, Basel/ Switzerland), were considered diabetic. Animals were treated with isophane (NPH) insulin (0.1-0.2 units), subcutaneously, as needed, to avoid weight loss. Animals were kept diabetic for 6 weeks prior to the wounding experiments. Like diabetic patients, STZ-induced diabetic mice also develop chronic low-grade inflammation leading to wound healing impairment 31 .
Wound healing model and treatments. Wound induction was performed as previously described 31 .
Briefly, control or diabetic mice were anesthetized with Ketamine/Xylazine (100/10 mg/kg, i.p.). After removing the dorsal hair, two 6 mm excisional wounds 2 cm apart were created using a punch biopsy tool (Miltex, Rietheim-Weilheim, Germany). The wound area was traced daily onto acetate paper to follow rates of wound closure for 10 days post-wounding. Wound size was determined with ImageJ version 1.46 (NIH Image, USA).
Diabetic mice were used for miR-inhibitor treatments. Both wounds were treated topically, twice daily up to day 3 post wounding, with a miR-155 inhibitor (5′-TcaCaaTuaGcaTuaA-3′) (0.25, 1, 2.5 or 10 nmol) or with a negative control oligo (5′-CaaTagGguCaaGauT-3′); locked nucleic acid (LNA) bases are written in capital letters while 2′-O-methyl RNA bases in small letters. Oligos were designed using LNA bases for every third nucleotide and the backbone was phosphorothioate substituted for enhanced binding strength and stability. A seed miR-155 inhibitor (5′-AGCAuTaA-3′) was synthesized to only target the seed sequence of miR-155, and a seed control oligo (5′-TCAAgAuT-3′) was also synthesized 48 . Animals were sacrificed at days 3 or 10 post-wounding and the wounded skin was harvested for analysis.
Skin wound homogenization and RNA extraction. Skin tissue (50-100 mg) was homogenized with 1 mL TRI Reagent (Sigma Aldrich, St. Louis, Missouri, USA) with a polytron homogenizer followed by purification as per manufacturers' instructions. The RNA pellet was dissolved in DEPC water (50 µL). RNA concentration and purity were assessed using the NanoDrop ND-1000 spectrophotometer (ThermoFisher Scientific, Waltham, Massachusetts, USA). Samples were stored at −80 °C until further analysis.
Profiling miRs in skin wounds. MiR expression was measured on pools of skin samples using Rodent TaqMan Low Density Array cards, v.2.0 for RT-primer pool A and v3.0 for RT-primer pool B, containing 641 unique murine miRs (ThermoFisher Scientific, Waltham, Massachusetts, USA) according to manufacturer's instructions. Each pool contained six samples from the same experimental group, with total RNA input of 600 ng per array card. Briefly, total RNA was reverse-transcribed using multiplex RT primer pool sets followed by quantitative PCR (qPCR) step with sequence-specific primers and probes on the TaqMan ® MicroRNA Arrays.
Expression data were obtained using the Viia 7 qPCR system (ThermoFisher Scientific, Waltham, Massachusetts, USA). Data was normalized against the stably expressed U6 snRNA available on the array and relative miR expression was calculated using the comparative ∆∆Ct method (2 −∆∆Ct ), against the expression of the same miR in control mice at day 0.
Measurement of miR and mRNA levels by RT-qPCR. miRs were detected using reverse-transcription quantitative PCR (RT-qPCR) as previously described 49 . Oligonucleotides used for reverse transcription and qPCR are shown in Supplementary Table 1. Fibroblast growth factor (FGF) 7 mRNA levels were quantified using qPCR primers (QT00172004, Qiagen, Hilden/Germany), with Quantitect Sybr 2x Master Mix (Qiagen, Hilden, Germany) in 10 µL reactions using the MX3005 qPCR system (Agilent, Santa Clara, California, USA). Transcripts were quantified using standard curve quantification, diluted skin cDNA served as input to generate the standard curve. The geometric mean of transcription factor (TF)IIB and U6 levels were used to normalize for variation in input template. The geometric mean of these two transcripts was unaltered.

Cell culture and transfections. Human clonal keratinocyte (HaCaT) cells were cultured in Dulbeccos
Modified Eagle Medium (DMEM, 20 mM glucose) (ThermoFisher Scientific, Waltham, Massachusetts, USA) supplemented with FBS (10%) and Penicillin/Streptomycin (1%). MiR-155 levels were measured by RT-qPCR and this microRNA is expressed in HaCaT cells although at moderate levels, with Ct levels from 1 ng cDNA starting at cycle 25-27. For scratch migration assays, HaCaT cells were seeded 30.000 cells per well in 48 well plates and allowed to adhere for 24 hrs, before transfection with miR inhibitors. Transfections consisted of inhibitor (25 pmol) and Lipofectamine 2000 (0.5 µL) (ThermoFisher Scientific, Waltham, Massachusetts, USA) per well and were prepared according to manufacturer instructions in triplicate. After 24 hrs the medium was changed