Light-inducible antimiR-92a as a therapeutic strategy to promote skin repair in healing-impaired diabetic mice

MicroRNAs (miRs) are small non-coding RNAs that post-transcriptionally control gene expression. Inhibition of miRs by antisense RNAs (antimiRs) might be a therapeutic option for many diseases, but systemic inhibition can have adverse effects. Here we show that light-activatable antimiRs efficiently and locally restricted target miR activity in vivo. We use an antimiR-92a and establish a therapeutic benefit in diabetic wound healing. AntimiR-92a is modified with photolabile protecting groups, so called ‘cages'. Irradiation activates intradermally injected caged antimiR-92a without substantially affecting miR-92a expression in other organs. Light activation of caged antimiR-92a improves healing in diabetic mice to a similar extent as conventional antimiRs and derepresses the miR-92a targets Itga5 and Sirt1, thereby regulating wound cell proliferation and angiogenesis. These data show that light can be used to locally activate therapeutically active antimiRs in vivo.

M icroRNAs (miRs) are small non-coding RNAs that post-transcriptionally regulate gene expression by binding to targeted mRNAs and thereby inducing degradation or blocking its translation. MiRs have important functions in different pathophysiological processes and diseases. Therefore, targeting miRs by application of specific miR inhibitors might have great therapeutic potential. MiRs can be inhibited by different types of antisense RNAs (antimiRs) including phosphothioate-modified antisense RNAs that are linked to cholesterol for enhanced uptake of the oligonucleotide or locked nucleotide acids DNA mixmers 1 . Such antisense oligonucleotides are relatively easily taken up by detoxifying organs such as the liver or kidney, but uptake in other organs such as the muscle or brain tends to be limited 2,3 . Local delivery or activation may be necessary to augment the biological functions of antimiRs in the target tissue and reduce systemic toxicity. Local activation might also avoid unwanted side effects of antimiRs, since miRs have diverse functions in different tissues. Several targeting strategies have been experimentally used including the linking of miRs or antimiRs to aptamers 4,5 , nanoparticle-or microparticle-mediated delivery 6,7 and cell type-specific delivery by viral vectors 8 as well as attempts of local delivery by mechanical tools, for example, catheters 9 . In addition, we and others have developed photoactivatable antimiRs by attaching photolabile protecting groups (cages) to the nucleobases that temporarily inhibit duplex formation with the target miR, thereby allowing an excellent on/off behaviour upon irradiation [10][11][12][13] . However, the therapeutic in vivo use of light-activatable antimiRs has been unclear. Therefore, we tested whether light-activatable antimiRs directed against miR-92a can be used to locally augment impaired wound healing in diabetic mice. Inhibition of miR-92a was previously shown to improve angiogenesis and recovery after ischaemia 9,14,15 ; however, its regulation and function during wound healing, a process that is dependent on the angiogenic response, is unknown. Here we show that light-activatable antimiR-92a efficiently downregulate miR-92a expression leading to target gene derepression in the murine skin, thereby improving diabetic wound healing by stimulating cell proliferation and angiogenesis.

Results
Light-inducible antimiR activation in the murine skin. In order to transfer the concept of activation with light in vitro to nontransparent tissue, we established an ex vivo skin explant culture model to define suitable wavelengths, light dose and concentration of applied caged antimiRs. First, we used isolated skin tissue of adult male mice and measured light intensity behind a skin barrier by testing light-emitting diodes (LEDs) of different wavelengths and currents. These findings were compared to light intensities without the skin barrier. Although longer wavelengths show, as expected, a better penetration, 385 nm light efficiently enters and penetrates the murine skin ( Supplementary Fig. 1). Since wavelengths of this range can effectively remove photolabile-protecting groups 10 but are not expected to induce direct photo damage as compared to shorter wavelengths of the ultraviolet spectrum, light of 385 nm was used for the subsequent experiments. Caged antimiR-92a, developed in earlier studies 10 , was injected intradermally into murine skin explants (1 mg diluted in 50 ml PBS). After irradiation for 10 min (385 nm, 300 mA) and subsequent cultivation for 48 h, levels of miR-92a were analysed by quantitative real-time PCR. In comparison to the samples treated with non-irradiated caged antimiR-92a, miR-92a levels are significantly downregulated upon light induction (Fig. 1a). The efficiency of miR-92a inhibition by light-induced caged antimiRs is similar as compared to conventional constitutively active antimiRs, which are used as positive control (Fig. 1a). A caged control antimiR with no target-specific sequences, which is used as negative control, does not inhibit miR-92a expression neither in the presence nor absence of light (Fig. 1a).
For in vivo studies, we synthesized antimiRs with a slightly changed architecture, in accordance with first antagomiR studies by Krutzfeldt et al. 16 (Supplementary Fig. 2). AntimiRs were composed of 2 0 -OMe-modified nucleotides and contained in total six phosphorothioates at the 3 0 -and 5 0 -end to allow stability against endonucleases and exonucleases. We attached a cholesterol moiety on the 3 0 -end of the antimiRs for better in vivo delivery. To temporarily abolish antimiR activity, we attached six 1-(2-nitrophenyl)ethyl photolabile-protecting groups to the N 6 of A and N 4 of C nucleotides ( Supplementary Fig. 2), respectively. Additionally, we synthesized a non-caged antimiR-92a as a positive control and a caged antimiR as a negative control that does not affect miR-92a levels ( Supplementary Fig. 2). The purity of the compounds was confirmed by gel electrophoresis and analytical high-performance liquid chromatography and we demonstrated the efficient removal of cage groups after light activation in vitro ( Supplementary Fig. 3).
These antimiRs were then injected using the same concentrations, injection strategy and light activation as described above. The in vivo activation of caged antimiR-92a leads to a significant reduction of miR-92a levels in the murine skin with an efficiency that is comparable to constitutively active antimiR-92a, while the controls have no effects (Fig. 1b). Additionally, by measuring miR-92a level in the kidney and liver, we show that locally light-activated antimiR-92a does not significantly affect the kidney or liver miR-92a levels (Fig. 1c,d). These data demonstrate that caged antimiRs can be successfully activated by light in vivo. Local activation results in reduced systemic inhibitory effect of the antimiRs and as such may reduce adverse effects in other organs.
miR-92a inhibition improves wound healing in diabetic mice. Cutaneous wound healing is a highly orchestrated physiological process involving the interplay of keratinocytes, fibroblasts, endothelial cells, extracellular matrix remodeling and the immune system 17 . Inflammation, proliferation and remodeling, the three different healing phases, that overlap in time and space, are regulated not only by several cytokines and growth factors but also by various miRs 18 . The formation of new blood vessels by proliferation and migration of endothelial cells (a process termed angiogenesis) promotes the wound-healing response 19,20 . AntimiR-92a augments angiogenesis in muscle tissues after ischaemia 9,14 ; however, its role in the complex process of wound healing is unknown. Therefore, we measured miR-92a level in acute wounds from Bl/6 wild-type mice and healingimpaired diabetic wounds from db/db mice (time points were chosen in accordance with the estimated peak of angiogenesis) as well as in acute human wound tissue and from non-healing human venous leg ulcers. These analysis revealed that miR-92a is strongly upregulated in chronic wounds in comparison to acute wounds, indicating that miR-92a may represent a therapeutic target to rescue impaired chronic wound healing (Fig. 2a).
Next, we tested the therapeutic applicability of light-activated antimiR-92a. To this end, we used diabetic mice as a relevant model of impaired wound healing. Non-healing, chronic wounds are a particularly common problem in diabetes. Db/db mice are deficient for leptin receptor activity by carrying a homozygous point mutation in the gene for this receptor and are a wellestablished model to test therapeutic improvements for chronic wound healing 21 . To study skin repair with or without functional ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15162 miR-92a in diabetic mice, 13-week-old mice were wounded by using 6 mm punch biopsies and wounds were treated with caged antimiRs (2 mg diluted in 25 ml PBS) at days 0, 4 and 7. Knockdown efficiency was quantified via qPCR 6 and 11 days post injury. Light-activated caged antimiR-92a therapy leads to a marked downregulation of miR-92a levels whereas no effect is seen in control groups ( Supplementary Fig. 4).
In order to analyse the distribution of caged antimiRs, we injected Cy3-labelled caged antimiR-92a and isolated the wound tissue 6 days post injury. (Supplementary Fig. 5). Positive cells are detected in the hyperproliferative epithelium and within the dermal compartment. Moreover, double staining shows that Cy3-labelled antimiRs are taken up by F4/80-positive macrophages ( Supplementary Fig. 5c).
The wound-healing kinetics were observed over a period of 11 days and the wound size was quantified macroscopically at defined days post injury. While in the early healing phase, no differences are observed among the different groups macroscopically, both the conventional antimiR-92a and the caged antimiR-92a with light activation exhibit a smaller wound size 11 days post injury (Fig. 2b). In contrast to control antimiRtreated wounds, light-induced antimiR-92a or constitutively active antimiR-92a-treated wounds lose scabs 11 days postwounding (Fig. 2c), indicative of accelerated re-epithelialization. These findings were further confirmed by microscopic analysis of haematoxylin and eosin-stained paraffin sections isolated at days 6 and 11 post injury. Although almost all wounds are closed at day 11 postwounding, miR-92a inhibitor-treated wounds reveal faster wound closure kinetics, estimated by a shorter distance between epithelial tips measured at day 6 postwounding (Fig. 2d). Furthermore, wounds after antimiR-92a treatment show a more dense and cell-rich granulation tissue, equipped with a thick and hyperproliferative epithelium at day 11 postwounding (Fig. 2e,g). In contrast, control wounds have a more fragile appearance and are only covered by a thin epithelial layer, often detached from the underlying granulation tissue (Fig. 2g). Finally, wound contraction, estimated by measuring the distance between the edges of the panniculus carnosus, are significantly stronger upon miR-92a inhibition at day 11 post injury ( Fig. 2f), indicating that caged antimiR-92a harbour great therapeutic potential to improve wound healing in diabetic mice.
To rule out toxic side effects of repetitive caged antimiR treatments and the released 2-nitrosoacetophenon side product upon light activation, plasma levels for the liver transaminases aspartate aminotransferase and alanine aminotransferase as well as for the alkaline phosphatase were measured as biomarkers for liver injury. Furthermore, creatinine and urea plasma levels were determined to control for kidney function. For all used biomarkers, no significant upregulation is detected in comparison to untreated db/db mice ( Supplementary Fig. 6), indicating that caged antimiRs and the released photolysis side product have no obvious toxic short-term effects on organs, such as the liver and kidney.
AntimiR-92a enhances cell proliferation and angiogenesis. In order to analyse the underlying mechanism for improved healing upon miR-92a inhibition, we first characterized the woundhealing response by measuring angiogenesis, wound cell proliferation, myofibroblast differentiation and the inflammatory response post injury. Indeed, a significant higher amount of CD31 þ blood vessels within the granulation tissue is observed in mice that are treated with light-activated antimiR-92a and constitutively active antimiR-92a as compared to all the control groups at days 6 and 11 post injury (Fig. 3a,b). Furthermore, staining for the proliferation marker Ki67 reveals improved wound cell proliferation. Quantification of positive stained cells unveils that upon inhibition of miR-92a proliferation is improved  ARTICLE within the epithelial and the dermal compartment at day 6 post injury (Fig. 3c,d). Besides angiogenesis and re-epithelialization, myofibroblast differentiation and the inflammatory response are two other essential events for successful healing. However, for both, we excluded an influence of miR-92a inhibition estimated by staining for the myofibroblast marker a-smooth muscle actin (aSMA; Supplementary Fig. 7a,b) and the pan inflammatory cell marker CD45 ( Supplementary Fig. 7c,d).
Validated pro-angiogenic miR-92a targets are the integrin subunit a5 (Itga5) that mediates cell-matrix interactions and cell migration 22,23 via binding to fibronectin and the class III histone deacetylase sirtuin (Sirt)-1 (ref. 14). The upregulation of both was quantified by immunohistochemical stainings against Itga5 and Sirt1 of wound sections. While for Itga5 only during the early healing response an upregulation is detected (Fig. 3e,f), Sirt1 expression is significantly induced upon effective miR-92a inhibition at days 6 and 11 post injury, generally highlighting the functional inhibition of miR-92a upon caged antimiR-92a treatment following light activation (Fig. 3g,h). Besides the angiogenesisrelated functions of Itga5 (ref. 14) and Sirt1 (refs 24,25) that might explain the healing-supportive effects of miR-92a inhibition, both proteins are also shown to control re-epithelialization.
Itga5 is expressed on endothelial and epithelial cells during the wound-healing response. An increase of Itga5 during the early phase of the wound-healing response is contributing to keratinocyte migration over the fibronectin-rich matrix. In line, we observe an upregulation of Itga5 during the early woundhealing phase in miR-92a-inhibited wounds probably mediating an improved keratinocyte migration and finally forcing the observed improved wound-closure kinetics at day 6 post injury ( Fig. 3e,f). In addition, Itga5 abundance in dermal wound tissue is reflecting the angiogenic response since Itga5 is expressed on endothelial cells. In accordance with those findings, we find an upregulation of the Itga5 signal in dermal wound tissue in all groups at day 11 post injury with a slight stronger signal in antimiR-92a-treated samples (Fig. 3e).
The second miR-92a target Sirt1 was shown to promote corneal epithelialization 26 and improved keratinocyte proliferation during skin repair 27 . In line, a strong upregulation of Sirt1 upon miR-92a inhibition in the epithelium covering the wound area is observed (Fig. 3h), probably leading to the thick and hyperproliferative epithelium in miR-92a-inhibited wounds and thereby promoting an accelerated healing response (Fig. 2). Together, we find that antimiR-92a-treated wounds reveal faster healing kinetics based on a stronger angiogenic response and increased wound cell proliferation mediated by the derepression of the miR-92a targets Itga5 and Sirt1.
In conclusion, here we show that caged antimiRs are effectively activated by light in non-transparent living tissue. To the best of our knowledge, the concept of ultraviolet uncaging of oligonucleotides in the skin of living mice has only been shown once before for photoregulation of aptamer activity and biodistribution 28 and to uncage green fluorescent protein-coding plasmids 29 . Furthermore, there has been no approach to spatially confined therapeutic use of antimiRs by light activation in vivo. Inhibiting miR-92a by light-activatable antimiRs may harbour great therapeutic potential to treat chronic skin repair in healing impaired diabetic patients avoiding unwanted adverse effects in other tissues that might be expected by systemic delivery. However, given the low potential of cytotoxicity of systemic miR-92a treatment, our study can be rather seen as 'poof-ofconcept' experiment paving the way to testing cytotoxic antimiRs for treating malignant diseases or antimiRs that could be valuable for inducting regeneration, which often have the risk for promoting tumor growth. In addition, future experiments will show if the concept of light activation can be transferred to the application of well-perfused inlying organs-eventually necessitating visible or near-infrared light deprotection-and thereby expand the therapeutic benefit for different antimiR therapies in the future, including acute myocardial infarction or cancer.
After the cleavage from the solid support, the Cy3-labelled caged antimiR was prepurified on a MultoKrom 100 C18 column (250 mm Â 4.6 mm) using a binary solvent system (buffer C and solvent D) and a two-step gradient on an Agilent 1200 series system: Step 1: 5-20% solvent D in in 6.25 min; Step 2: 20-90% solvent D in 52.5 min.
Product-containing fractions were pooled and further purified via anion exchange chromatography with a Dionex DNAPac PA100 column (250 mm Â 4.0 mm) using a ternary solvent system containing acetonitrile (solvent E), 25 mM Tris*HCl pH 8 (buffer F) and 25 mM Tris*HCl 500 mM NaClO 4 (buffer G) and a linear gradient from 0% buffer G to 70% buffer G in 18.5 min while the content of solvent E was kept constant at 30%.
For all oligonucleotides, all product-containing fractions were concentrated, analysed by LC-MS (Bruker micrOTOF-Q), the purest fractions were pooled and desalted via NAP25 columns (GE Healthcare), lyophilized and dissolved in PBS buffer (80 ng ml À 1 ).
For in vivo treatment, antimiRs were injected intradermally. Ten minutes after injection, the area was irradiated for 10 min (l ¼ 385 nm, 300 mA), while the remaining part of the body was protected with aluminum foil.
Measurement for toxicity of light-inducible antimiRs. Blood of light-inducible antimiR-treated mice (13-week-old male BKS(D)-Leprdb/JOrlRj (db/db) mice) was collected and plasma was isolated via centrifugation (15 min, 1,500g). As biomarkers for liver injury, plasma levels for transaminases aspartate aminotransferase and alanine aminotransferase as well as for alkaline phosphatase were quantified. Creatinine and urea plasma levels were determined to proof kidney functionality. Measurements occurred in accordance with standard procedures in the central laboratory of the university hospital Frankfurt am Main, Germany. Commercial cobas tests were used in accordance with the manufacturer's protocols (Roche Diagnostics).
Preparation of human wound tissue samples. Human tissue samples were obtained from patients presenting with healing (day 14-30 days postsurgery, n ¼ 6) and non-healing (venous leg ulcers, n ¼ 6) skin wounds. The tissue was fixed in 4% paraformaldehyde and paraffin embedded. RNA was isolated using the miRNeasy FFPE Kit (Qiagen) according to the manufacturer's instructions from 20 mm paraffin-embedded tissue sections. The use of human material was approved from the local ethics committee and informed consent of patients was received (University Hospital Cologne).
Quantitative real-time RT-PCR. Total RNA was isolated using the miRNeasy Mini Kit (Qiagen) according to the manufacturer's protocol. For miR-92a detection, 10 ng RNA was reverse transcribed and quantified by TaqMan real-time PCR using the TaqMan microRNA Assay (Assay name: hsa-miR-92, assay ID: 000430, Applied Biosystems). RNU6 was used as the endogenous control (Assay name: U6 snRNA, assay ID: 001973). A StepOnePlus device was used for detection (Applied Biosystems).
Morphometric analysis. The macroscopic wound area was quantified based on photographs taken at various time points post injury and was calculated as the percentage of the initial wound area with a diameter of 6 mm. For scaling, a ruler was placed next to the wound. The extent of granulation tissue formation, as a measure for wound quality and stability, was determined on 4 mm haematoxylin and eosin-stained paraffin tissue sections from the central portion of the wound using a Zeiss microscope equipped with a bright field filter (Axio Observer.Z1, Zeiss). Additionally, the distance between the ends of panniculus carnosus was determined as a measure of wound contraction and the distance between ends of epithelial tips was measured for the wound closure rate. For the quantitative analysis of antibody-stained tissue sections, images were generated with the laser scanning confocal microscope LSM 780 (Zeiss) at a magnification of 20 Â or 25 Â (oil). The percentage of the positive stained CD31, Itga5 and aSMA area was calculated by using the ImageJ software and normalized to the total wound area.
Quantitative analysis of Sirt1 and Ki67 staining. By using the ImageJ Software, the Sirt1 and Ki67-positive-stained area was measured and normalized to all Dapi-positive-stained nuclei within the wound area, including the hyperproliferative epithelium. CD45-positive cells were counted and normalized to the total wound area.
Statistical analysis. Significance of difference was analysed using analysis of variance one-way test analysis with Newman-Keuls Multiple Comparison Test. All data are presented as mean±s.e.m.; a P value of r0.05 was considered significant.
Data availability. The authors declare that all data supporting the findings of this study are available within the paper and its Supplementary Information files.