Photobiomodulation drives pericyte mobilization towards skin regeneration

Photobiomodulation is being widely applied for improving dermal or mucosal wound healing. However, the underlying cellular and molecular processes that directly contribute to its effects remain poorly understood. Pericytes are relevant cells involved in the wound microenvironment and could be one of the main targets of photobiomodulation due to their plasticity and perivascular localization. Herein, we investigate tissue repair under the photobiomodulation stimulus using a pericyte labeled (or reporter) transgenic mice. Using a model of two contralateral back wounds, one the control and the other photoactivated daily (660 nm, 20 mW, 0.71 W/cm2, 5 J/cm2, 7 s, 0.14 J), we showed an overall influx of immune and undifferentiated cells and higher mobilization of a potent pericyte subpopulation (Type-2 pericytes) in the photoactivated wounds in comparison to the controls. Doppler analysis showed a significant increase in the blood flow in the photoactivated wounds, while marked vascular supply was observed histologically. Histochemical analysis has indicated more advanced stages of tissue repair after photoactivation. These data suggest that photobiomodulation significantly accelerates tissue repair through its vascular effects with direct recruitment of pericytes to the injury site.

Photobiomodulation is being widely applied for improving dermal or mucosal wound healing. However, the underlying cellular and molecular processes that directly contribute to its effects remain poorly understood. Pericytes are relevant cells involved in the wound microenvironment and could be one of the main targets of photobiomodulation due to their plasticity and perivascular localization. Herein, we investigate tissue repair under the photobiomodulation stimulus using a pericyte labeled (or reporter) transgenic mice. Using a model of two contralateral back wounds, one the control and the other photoactivated daily (660 nm, 20 mW, 0.71 W/cm 2 , 5 J/cm 2 , 7 s, 0.14 J), we showed an overall influx of immune and undifferentiated cells and higher mobilization of a potent pericyte subpopulation (Type-2 pericytes) in the photoactivated wounds in comparison to the controls. Doppler analysis showed a significant increase in the blood flow in the photoactivated wounds, while marked vascular supply was observed histologically. Histochemical analysis has indicated more advanced stages of tissue repair after photoactivation. These data suggest that photobiomodulation significantly accelerates tissue repair through its vascular effects with direct recruitment of pericytes to the injury site.
Wounds induced by burns, chronic disorders, and in consequence of radiotherapy/chemotherapy treatments affect millions of people. Delayed or aberrant wounds result in quality of life impairment and long hospitalization periods, leading to high treatment costs and overloaded health systems 1 . Accumulating scientific evidence suggests that photobiomodulation (PBM) therapy is potentially useful in medical applications. PBM is particularly attractive in superficial injuries-where it can be readily accessible to a light source-and modulate pain, inflammatory and immune processes, being recognized as a powerful tool in wound healing and tissue repair [2][3][4][5][6][7] .
The mechanisms underlying PBM therapy are mostly related to the generation of energy and reactive oxygen species in cells and tissues, thus modulating cell signaling and metabolism in an injured milieu. PBM is also known to activate critical endogenous molecules such as Fibroblast Growth Factor, Vascular Endothelial Growth Factor, Bone Morphogenetic Protein 4, Transforming Growth Factor β, and NF-κB 3,5,[8][9][10] and is recognized to stimulate neovascularization [11][12][13] . However, the lack of understanding of the primary cellular and molecular components involved in the photoactivation process hinders adequate evidence of the therapy and its broader application. tioned and observed by confocal microscopy in order to verify the colocalization of labeled cells with Dapi during the wound healing process. Undifferentiated cells (Nestin + ), total pericytes population (NG2 + ) and Type-2 pericytes subpopulation (Nestin + , NG2 + , and Dapi colocalization) have accumulated more in the photoactivated wounds than in the controls on the 3rd day of treatment (p = 0.033; p = 0.033 and p = 0.028, respectively) ( Fig. 3a,b). These cells were observed inside and around vessels and frequently seemed to be detaching from them (Fig. 3a). The NG2 + cells were mainly found in delimiting these vessels. Total cell quantification (Dapi fluorescence) was significantly higher in the PBM-treated wounds, meaning more significant input of cells in the wounds under the PBM stimulus than in the controls (p = 0.034) (Fig. 3b). The same features were observed in the samples analyzed at the 7 day-time interval (Fig. 4a,b). Otherwise, the NG2 + cells were no longer seen around the vessels but were dispersed in the wound bed in the dermis region in both groups (Fig. 4a). In the photoactivated wounds, the cells were clearly attracted to the wound bed's upper dermal area, immediately underneath the epithelium (Fig. 4a). Additionally, at the 3rd and 7th-day post-surgery, the NG2 + and Nestin + /NG2 + cells appeared colocalized with Dapi (pinkish and yellowish components, respectively) in areas correspondent to the bulge region of the hair bulb (Sup. Fig. S4). Sometimes, these cells looked to be detaching from those skin appends (Sup. Fig. S4).

Type-2 pericytes and undifferentiated cells accumulated in
PBM has stimulated angiogenesis and promoted advanced tissue repair. At the 12 h time interval, the mean score of fibrin clot formation and angiogenesis degree were similar between both groups (p = 0.664) (Sup. Fig. S5). However, the inflammatory infiltrate was significantly increased in the photoactivated wounds than in the controls (p = 0.019) (Sup. Fig. S5). At 12 h, no difference in mast cell count was observed between PBM-treated and control wounds (p = 0.164) (Fig. 5a,b). Despite this, PBM led to a significant reduc- www.nature.com/scientificreports/ Remaining wound area (%) for total closure from day 2 until day 7 (c). Macroscopic appearance of control and PBM-treated wounds at days 3 and 7. Note that PBM-treated wounds are humid and with less rolled edges-3rd day; smaller, more uniformly healed and thinner eschar than the control at day 7 (d). Data are shown as the mean ± SD, n = 5 for each group. N.S. = non-significant. Scale bar = 5 mm. www.nature.com/scientificreports/  www.nature.com/scientificreports/ tion of eosinophil counts in this early inflammatory phase of wound repair (p = 0.020) (Fig. 5c,d). No differences concerning collagen fibers deposition were observed (p = 0.986) (Fig. 6a). At day 3 time interval, the epidermis area and thickness were significantly higher in the control group compared to the photoactivated wounds showing a premature but disorganized epithelialization process (p = 0.012 and p = 0.009, respectively) (Sup. Fig. S6). In the PBM-treated groups, the epithelium was thinner but continuous (Sup. Fig. S6a). The granulation tissue area was also higher in control concerning the PBM-treated wounds (p = 0.022) (Sup. Fig. S6b). Otherwise, the scores for the degree of angiogenesis (where blood vessels were defined as a tube-like structure with a lumen, whether or not it contained red blood cells) were significantly higher in the photoactivated wounds than in the controls (p = 0.033) (Sup. Fig. S6b). CD31 staining has shown vessels in the controls mostly located at the wound edge (Fig. 7a). Small caliber vessels showing the expression of the CD31 marker or a colocalization of the NG2 and CD31 were identified in the photoactivated samples' wound bed after 3 days of treatment (Fig. 7a). No differences between groups were observed for the mean score of fibrin clot formation (p = 0.864), epidermal differentiation (p = 0.880), and in the amount of inflammatory infiltrate (p = 0.851) (Sup. Fig. S6b). Eosinophil counts were similar in both groups (p = 0.421) (Fig. 5d). However, a higher number of mast cells were found in the PBM-treated wounds after 3 days (p = 0.031) (Fig. 5b). Signs of dermis reconstruction characterized by the deposition and reorganization of collagen fibers were observed mostly in the control group than in the PBM-treated wounds (p = 0.040) (Fig. 6b).
After 7 days of treatment, histological analysis and macroscopy have indicated that PBM-treated wounds presented more advanced tissue repair stages than the control (Fig. 8). PBM-treated wounds have shown more blood vessels, including larger caliber vessels and vascular congestion, than those observed in the control group (p < 0.001) (Fig. 8b). CD31 staining has shown few vessels within the wound bed of the controls (Fig. 7c), while an exuberant arteriole and small caliber vessels could be identified in the photoactivated samples (Fig. 7d). Once again the colocalization between the NG2 and CD31 marker could be observed in vessel structures (Fig. 7d). The granulation tissue area was significantly smaller than in the control wounds (p = 0.028) (Fig. 8b). At this time point, fibrin clot formation and inflammatory infiltrate were reduced considerably in the PBM-treated wounds than in the controls (p = 0.027 and p = 0.004, respectively) (Fig. 8b). Eosinophils quantification was similar in both groups (p = 0.404) (Fig. 5d). However, the number of mast cells was still higher in the PBM-treated samples than in the controls (p = 0.004) (Fig. 5b). PBM-treated wounds presented more significant epidermal differentiation (p = 0.015) and epithelium thickness (p = 0.001), clearly showing granular and spinous layers of www.nature.com/scientificreports/ the regenerated epithelium (Fig. 8b). Epidermal indentation could also be detected in some of the PBM-treated samples but not in the controls (Fig. 8a). As such, the control wounds have shown more prevalence of the basal layer, which indicates a more proliferative status of the epithelium (Fig. 8a). Moreover, even on the 7th day, fragmented epithelium showing ulcerative regions could be noted only in the control groups (Fig. 6c). The Masson trichrome mostly stains type I collagen fibers. Accordingly, increased collagen fibers distribution, packing, and orientation were observed in the PBM-treated groups on day 7th compared to the controls (p = 0.004) (Fig. 6c). A higher number of cells expressing α-smooth muscle actin (α-SMA)-which characterizes the myofibroblast phenotype-were seen around blood vessels but also as isolated cells dispersed within the interstitial tissue, particularly in the PBM-treated wounds (p < 0.001) (Fig. 8a,b).
Together, these data indicate that photobiomodulation significantly accelerates tissue repair through its vascular effects with direct recruitment of pericytes to the injury site (Sup. Fig. S7).

Discussion
The absorption of visible light and infrared radiation may cause biological effects on the skin 25 . However, the beneficial mechanisms of light exposure have not been fully elucidated. In this study, we have investigated the ability of PBM at 660 nm in driving endogenous cells to collaborate in tissue repair. Using a reporter model for pericytes, our study has demonstrated that PBM promotes remarked vascular support and significantly accelerates the skin repair through an overall cellular input scenario with the direct involvement of Type-2 pericytes.
The association of pericytes and PBM in the skin was only barely explored. Using immunohistochemical techniques, Medrado et al. showed an increased number of NG2 + and α-SMA + cells-both well-recognized markers for pericyte cell targeting-in the PBM-treated groups with dosimetry parameters similar to the ones of the current study 26 . These cells showed a parabasal position, and some of them seemed to be detaching from the periphery of capillaries and venules when stimulated by PBM 26,27 . There is also accumulated evidence showing that PBM at 660 nm can increase local blood circulation 6,28,29 . In the present study, we have observed that, under the PBM stimulus, not only Type-2 pericytes but also undifferentiated cells (Nestin + ) identified by confocal microscopy were clearly attracted to the wound dermis upper region, close to the epidermis. Many of these cells seemed to be accumulating in or detaching from blood vessels dispersed into the dermis region, particularly at the wound healing process's early stages. This is the first time that these cells' colocalization is shown in the www.nature.com/scientificreports/ injury site of photoactivated wounds using state-of-art techniques. However, it remains unclear whether PBM contributes to the direct activation of pericytes and endothelial cells or is indirectly responsible by the recruitment of these and other cells due to the marked vascular changes promoted by the light. A particular immune component of the wound healing process is the mast cell (MC). MCs represent up to 8% of the total number of cells within the dermis and are known to infiltrate at neovascularization sites at the beginning of the wound healing process 30,31 . In truth, the MCs topographic distribution within small vessels' vicinity guarantees provisions of the essential mediators needed for normal angiogenesis, such as histamine, serotonin, tumoral necrosis factor-α, kinins, and tryptases/proteases 32,33 . Herein, we have quantified Toluidine Blue O-stained MCs population in the PBM-treated groups, both intact or degranulated, and observed a progressive increase in MCs counts compared to the controls throughout the experimental times (12 h, 3 and 7 days). In this line, a previous study has also found a significant increase in MCs count after 3 days of injury in wounds treated with PBM and a concomitantly higher expression of vascular endothelial growth factor-a potent vasoactive growth factor 34 . Others have shown that red and infrared wavelengths promote immediate MCs' degranulation in the human gingival tissue 35,36 . Recently, Kimizuka et al. has demonstrated that the increased microvascular permeability induced by PBM is mediated by MCs 6 . Our findings are in line with the current literature, which partially explains the acceleration of the tissue repair in the photoactivated tissues as a consequence of MCs degranulation leading to the early increase of vascular permeability.
PBM treatment is known to stimulate the transient receptor potential vanilloid 1, 2, and 4 present in the plasmatic membrane of MCs 9,37-39 . PBM-activation of transient receptor potential vanilloid-2, for example, allows Ca 2+ ions to enter the cell, thus feedbacking MCs degranulation 37,38 . However, in 2005, Tóth et al. demonstrated by immunoelectron microscopy that brain pericytes also present transient receptor potential vanilloid-1 40 . This knowledge can contribute to the understanding regarding pericytes recruitment during PBM stimuli in the skin healing or repair. Accordingly, increased blood flow and increased angiogenesis observed in the PBMtreated wounds on the 3rd and 7th day may correlate to the sustained higher number of MCs during almost all experimental times, and with the consequent cellular input to the injury site. Nevertheless, the hypothesis of direct activation of pericytes by PBM is also reasonable once, in addition to physically stabilize blood vessels and contribute to the vascular development, maturation and remodeling, pericytes may regulate the permeability and the blood flow 41 . Further work is needed to explore the crosstalk between pericytes and mast cells under the PBM scenario considering their shared vascular role.
Considering that MCs and eosinophil cells are developmentally and functionally similar, we also have looked for eosinophil cells participation in the wound healing process. We have observed a significant suppression in the eosinophils count in the first 12 h after PBM. After 3 and 7 days, these cells' counts were restored but slightly greater in the PBM-treated than in the control groups. Eosinophils are, in fact, important to degrade collagen, but they also stimulate dermal fibroblast DNA synthesis and matrix production 42 . However, MCs and eosinophil cells can be influenced by paracrine mechanisms. For instance, for eosinophils, histamine may act as a chemoattractant through histamine receptor (H1R) 43 and inhibit chemotaxis through histamine receptor (H2R), i.e., when histamine concentration is high 44 . Although our results do not fully support the true interaction between MCs and eosinophil cells under the PBM influence, the inverse ratio of these cells may be contributing to the premature collagen deposition in the controls (on the 3rd-day time interval) and to the belated collagen deposition in the PBM-treated wounds (on the 7th-day time interval).
A light on the PBM's influence in the process of epithelialization of the wounds is worth noting. We have observed that PBM-treated groups presented more continuous and differentiated epithelialization, but only at the 7th day-time interval. Earlier, at the 12 h and 3rd-day intervals, the epithelium was thinner, and the wounds presented macroscopically humid. Despite it, secondary infection was found in none of the groups. More recently, it has been demonstrated that infrared laser induces unique immunologic cascade in the MCs via reactive oxygen species generation and creates an immunostimulatory milieu for dendritic cells, augmenting the adaptive immune response and playing a critical role in the host defense 6 . We hypothesize that PBM-treated groups presented a gradual keratinocyte epithelialization process, which may have allowed enough time to connective tissue reconstruction and remodeling.
In the current study, the PBM-treated groups kept expressively vascularized after 7 days, and less granulation tissue was formed. The myofibroblast population in the PBM-treated groups was increased compared to the controls, as represented by the marked expression of α-SMA. Macroscopically, these groups presented thinner eschars and were more uniformly healed in comparison to the controls. Indeed, tissue regeneration harmoniously progressed toward the center of the wound in most of the PBM-treated samples. Notably, in the PBM-treated groups at the 7th day time interval, samples were repaired almost entirely, and some were presenting restored skin attachments, such as the hair follicles. As aforementioned, together with undifferentiated cells, Type-2 pericytes have accumulated in the upper region of the dermis of PBM-treated wounds. Curiously, despite numerous of these cells seemed to be detaching from blood vessels, we have also noticed some yellowish and pinkish cells around the hair follicles, particularly in the wound edges from both groups. The skin and its appendages undergo continuous renewal and maintain reservoirs of multipotent stem cells that are promptly activated upon skin injury 45 . A classic stem cell niche is localized in the bulge region of the hair bulb, and after wounding, these cells give rise to the epidermis, follicles, and sebaceous glands 46 . Even though the NG2 protein can be expressed by hair follicles cells 47,48 , the association of these cells' pericytic origin needs further attention.
One of the mainstream regulators of PBM-induced cascades is the Transforming Growth Factor-β1, which is directly activated from its latent state by light 2,3 . Besides many other functions, Transforming Growth Factor-β induces chemotaxis of inflammatory and immune cells to the wound site, drives stem cell differentiation, and mediates collagen type III replacement with collagen type I 49  www.nature.com/scientificreports/ Transforming Growth Factor-β1, although many other signals can participate 51 . The association of the increased number of undifferentiated cells and Type-2 pericytes with the marked vascular changes can be critical elements involved in the PBM therapy process, thus contributing to the structural replenishment of the wounds very close to the original skin. Finally, the present study model has limitations concerning the Nestin + cells' identity once green fluorescent protein levels may not accurately mirror the Nestin gene activity 52 . Otherwise, our transgenic mice lineage may be considered one of the most appropriate animal models to understand the molecular and cellular components involved in the PBM-triggered cascades. Noteworthy, the study was performed with test and control groups in the same animal due the phenotype's rarity. Although there is an appeal that PBM may have a systemic effect, we have observed that the local stimulus is critical. Future studies should investigate the mechanisms underlying pericytes angiogenic potential under photoactivation and whether their ablation affects angiogenesis and PBM outcomes.

Materials and methods
Ethical issues and animal procedures. The Animal Care and Use Committee at Universidade Federal de Minas Gerais approved animal handling and procedures in this work (#19/2018), and all experiments were performed according to the ARRIVE guidelines. Our Nestin-GFP + transgenic mice colony was maintained homozygous for the transgene on the C57BL/6 genetic background 23 . Nestin-GFP + mice were crossbred with NG2-DsRed + mice to generate Nestin-GFP + /NG2-DsRed + double-transgenic mice. Accordingly, this model presents endogen fluorescence for two proteins: Nestin-GFP + /undifferentiated cells (marked in green) and NG2-DsRed + /pericytes (marked in red). All animals were housed in controlled conditions for temperature (24 °C) and under a 12:12-h light-dark cycle and fed ad libitum.
Eight to ten-week-old male or female Nestin-GFP + /NG2-DsRed + mice (n = 15; 5 per group) were used in this study. Animals were anesthetized [100 mg/kg (Ketamine) + 10 mg/kg (Xylazine) via intraperitoneal-IP] and two full-thickness excisions (including the panniculus carnosus) were performed on the back of the mouse, one on each side of the animal midline, by using a 4.0 mm biopsy punch. A 0.5 mm thick silicone splint was then placed around the wounds and sutured with six 5.0 nylon stitches to prevent premature wound closure 53 (Sup. Fig. S1). Next, the left wound received the PBM treatment, while the other was kept as the control.
Photobiomodulation setting parameters. PBM was performed by using an indium-gallium-aluminum-phosphide (InGaAlP) customized diode laser, in continuous operation and punctual and contact modes. The dosimetry parameters were as follows: 660 nm, 20 mW, 0.028 cm 2 spot area, 0.71 W/cm 2 , 7 s, 5 J/ cm 2 , 0.14 J total energy per point 5 . The photoactivation was performed transoperatively and every day up to 7 days (Sup. Fig. S1). Noteworthy, unless wound measurements needed to be done, animals were only manually restrained for the photoactivations. PBM exposure was avoided in the control wound by using a black cardboard paper over it. The output power was checked with a power meter (Lasercheck, Coherent Inc., Santa Clara, California, USA) before all photoactivations. A double microporous tape (Nexcare, 3 M, Saint Paul, Minnesota, USA) was then applied on the silicon splint as an occlusive patch and replaced every day to minimize wound dehydration in both groups. Each mouse was individually caged to prevent stitches biting. Histological assessment. The mice were euthanized after 12 h, 3, and 7 days (Sup. Fig. S1f). Samples were collected 2.0 mm beyond the wounds and fixed in paraformaldehyde 4% for 48 h and processed for both optimal cutting temperature compound or paraffin embedding. Optimal cutting temperature compound was done in 20 µm cryosections (CM3050 S, Leica Microsystems, Morrisville, North Carolina, USA), and the slides were observed by confocal microscopy. Paraffin blocks were serially sectioned using a microtome at 6 μm sections (Leica). Five consecutive histological fields in five different sections were selected to analyze re-epithelialization and inflammatory process grading (40 × or 100 ×). According to a previous study 57 , the inflammatory process intensity, fibrin clot formation, degree of angiogenesis, epithelium differentiation, and indentation were given a score of 0 to 5 based on its level of abundance in the wound beds according to a previous study. www.nature.com/scientificreports/ tissue area, epithelium area and thickness were quantified by the Fiji software (NIH). All samples were analyzed with an optical microscope (Axioskop 40 Zeiss, Göttingen, Lower Saxony, Germany), and one blinded examiner undertook all histological analysis. Detailed staining protocols are described below. . During color image deconvolution, images with Masson's trichrome stain were deconvolved, and the green component was identified as the collagen fibers. The area of the green collagen fibers was measured using the "threshold" tool. The threshold was manually adjusted until the entire green area was highlighted in red. Then, the threshold was measured as the percentage of the stained area 58 .

Masson's trichrome stain for collagen fibers and quantification.
Toluidine blue staining for mast cell identification. Sections were stained with 1% Toluidine Blue O (89640, Sigma Aldrich, Saint Louis, Missouri, USA) in distilled water for 2 min, to the identification of mast cells 59 . For each histological section, five consecutive histological fields at × 400 magnification were selected for the analysis. These fields were located in the dermis and submucosa regions of the wound beds. Mast cell counting was performed according to Sawasaki et al. and was blindly performed by the same examiner 36 .
Sirius red staining for eosinophil identification and quantification. Sirius red staining is a method used to access the presence of eosinophils 60 . Briefly, slides were incubated in Harris hematoxylin for 2 min, rinsed in tap water, and then 100% ethanol. Subsequently, slides were immersed in an alkaline (pH 8-9) Sirius Red solution (CI 35780, Sigma Aldrich) for 2 h. Ten random foci were selected from low power in an optic microscope (Axioskop 40 Zeiss; Carl Zeiss) and were assessed by one blinded examiner. Eosinophils were counted per viewing field (× 400 magnification) and averaged for each wound bed.

α-Smooth muscle actin (α-SMA) immunohistochemical analysis. Immunohistochemistry was
performed for α-SMA detection at day 7 post-surgery in the wound sections. Briefly, the slides were deparaffinized and dehydrated, followed by incubation with 3% hydrogen peroxide and 1% bovine serum albumin (BSA). The sections were then incubated with a polyclonal rabbit anti-mouse α-SMA primary antibody (MM1, Novocastra, Newcastle, UK, 1:200) at room temperature for two hours. The immunolabeling was visualized through incubation in 3,3-diaminobenzidine (DAB) solution (Dako, Carpinteria, California, USA). Then, the sections were stained with Mayer's hematoxylin and covered. Negative controls were obtained by omission of the primary antibody, which was substituted for 1% PBS-BSA. Images of 10 consecutive fields were taken in × 200 magnification using an optic microscope (Axioskop 40 ZEISS; Carl Zeiss, Gottingen, Germany). Immunostaining for α-SMA was quantified by color deconvolution plugin on ImageJ Version 1.52 (NIH). The red component was identified as α-SMA staining, and a threshold was manually adjusted and finally measured as the percentage of the stained area in the wound beds. CD31 immunofluorescence analysis. Immunofluorescence was performed for CD31 detection at day 3 and 7 post-surgery in the wound sections. Briefly, cryosectioned slides were rehydrated twice in PBS, followed by incubation with Triton X-100 0.5% bovine serum albumine 2% in PBS for 1 h and a half for permeabilization and blocking. The sections were then incubated with a monoclonal rat anti-mouse conjugated CD31 primary antibody (Brilliant Violet 421 nm, Abcam, Cambridge, UK, 1:100) overnight at 4 °C.

Confocal analysis. Skin sections were counterstained with Dapi and mounted on slides using Fluoroshield
Mounting Medium (Abcam) and examined under confocal microscopy (Nikon Eclipse Ti with an A1R confocal head). Slides were examined under objectives Plan Apo 10 × and 20 × . A large scan image was performed for each sample under the 10 × objective. Each channel was counted separately, except for Type-2 pericytes that were counted in the merged images. A threshold was manually adjusted, and the fluorescence intensity (mean gray values) was calculated using the Fiji software (NIH).

Statistical analysis.
Statistical significance between groups was calculated using paired (for the remaining wound area and Doppler analysis comparisons) or unpaired Student's t-test or Mann-Whitney test (for image analysis) using the GraphPad Prism 8.0 Software (GraphPad Prism Version 8.0c for Mac, GraphPad Software, La Jolla, California, USA). Data were expressed as means ± standard deviation. The differences among the groups were considered significant when p < 0.05 in all experiments.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files). www.nature.com/scientificreports/