All the experimental protocols, animal surgery and treatments applied in the current study have been approved by the research committee and Institutional Animal Care and Use Committee at the First Affiliated Hospital of Nanchang University, respectively. The experiments were carried out in accordance with the approved guidelines. For each animal experiment, a power calculation (p<0.05) was performed to determine exactly sufficient number of mice to allow the observed effects to be legitimate. An allocation concealment method was used to provide randomization in allocating experimental units to control and treatment groups. The use of the inbred littermate mice in a specified experiment ensured that the potential confounders were minimized. No criteria were used for excluding animals (or experimental units) during the experiment, and no data were excluded during the analysis.

Randomization and blind assessment were used in all animal studies. Diabetes was induced in 12-week-old C57/BL6 mice (male and female are evenly distributed in each group, SLAC Laboratory Animal, Shanghai, China) by single intraperitoneal injection of 160 mg/kg streptozotocin (STZ) in 100µl normal saline after a 12-hour fasting, as described before (15). This dose used in fasting wildtype mice results in 100% development of diabetes, since STZ specially goes into beta-cells through glucose transporter 2 (Glut2) on beta-cells. Fasting is critical since circulating glucose can compete with STZ for Glut2 to compromise the effects of STZ. The control mice received equal volume of the normal saline, the solvent to dissolve STZ. One week later, all mice that had received STZ became hyperglycemic (fasting blood glucose >=350mg/dl). If any mice did not develop hyperglycemia (fasting blood glucose >=350mg/dl) one week after STZ, they were supposed to be excluded from this study. The STZ-treated mice then underwent a surgery to create a 6 mm-diameter full-thickness wound on the dorsal midline with a biopsy punch, with/without orthotopic injection with 100µl AAV viruses [10¹¹ genome copy particle (GCP)/ml], as described (16). Meanwhile, the wounds of the experimental mice were treated with 50 µl recombinant PlGF (10 µg, Ab207150, Abcam, Cambridge, MA, USA) or an equivalent volume of saline solution a Hamilton’s syringe under the clot covering the wound at a frequency of twice per week. Fasting blood glucose were performed at 10am after a three-hour fasting period and assessment of beta-cell mass was done as described (17, 18). For vessel density measurement, tomato-lectin (50 µl, Vectorlabs, Burlingame, CA USA) was injected to the mice through the tail vein 10 minutes before sacrifice. The vessel density was determined bases on the percentage of lection-positive area in total measured tissue area, as described before (10).

Transfection of human embryonic kidney 293 cells by AAV serotype 6 vectors were applied by as described before (19, 20). The backbone plasmid and the plasmid containing human CD68 promoter were both obtained from an Addgene plasmid (#32395 (21) and #34837 (22), Watertown, MA, USA). The sequence for siRNA for VEGFR1 was 5’-CCAGACACTGCATCCAA-3’, and the control scramble sequence was 5’-GCATTAACTAAAGGCUGCC-3’. The Transfection was performed with Lipofectamine 3000 reagent (Invitrogen, CA, Carlsbad, USA). Purification and titration of AAV vectors were performed with standard procedure and a dot-blot assay, respectively, as described before (19).

The injured skin tissue was dissected out and digested with 0.25% trypsin for 35 minutes to obtain a single cell fraction for flow cytometry. The flow cytometry-based cell purification was based on direct fluorescence from GFP (AAVs-originated) and Tomato (Lectin-originated) as well as immunofluorescence from VEGFR1 (by a PE-cy7-conjugated anti-VEGFR1 antibody, Becton-Dickinson Biosciences, San Jose, CA, USA) or CD68 (by a BV421-conjugated anti-CD68 antibody) or CD31 (by an APC-conjugated anti-CD68 antibody), as described before (15). The flow cytometry data were analyzed by Flowjo (version 11.0, Flowjo LLC, Ashland, OR, USA).

A RNeasy mini kit (Qiagen, Germantown, MD, USA) was used to extract RNA. Quantitative real-time PCR (RT-qPCR) were performed using commercial primers ordered from Qiagen. Data were collected and analyzed using 2-△△Ct method. Values of genes were obtained by sequential normalization against the internal control GAPDH and the corresponding experimental controls.

The extracted protein by RIPA buffer was subjected to mouse PlGF-2 (MP200) and VEGFR1 (MVR100) enzyme-linked immunosorbent assay (ELISA; R&D System, Los Angeles, CA, USA), as described before (10). The microplate was read at 450 nm within 30 minutes. The wavelength correction was set to 540 nm or 570 nm.

Ten minutes after tomato-lectin infusion through the tail vein, the mice were sacrificed. The pancreas and the injured skin were dissected out and fixed in 4% paraformaldehyde (PFA, Sigma-Aldrich, St. Louis, MO, USA) for 6 hours. After incubation in 30% sucrose for 24 hours, the samples were frozen and embedded. Fluorescent immunostaining was done as described before (10). Tomato-lectin was detected by direct red fluorescence and virus-transduced macrophages were detected by GFP. Immunofluorescent staining for glucagon and insulin was performed with a monoclonal antibody against glucagon (Ab10988, Abcam) and a guinea pig polyclonal antibody against insulin (Ab7842, Abcam), respectively. Hoechst 33342 (HO, Sigma-Aldrich) was applied to stain the nucleus.

Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) was used to obtain data for bioinformatic analysis (23). The data surf applied “diabetic foot” OR “diabetic foot disease “ AND “Homo sapiens”. After a careful assessment for relatedness, two gene expression profiles (GSE80178 and GSE134431) were selected (24). The GEO2R online analysis tool and “DESeq2” R package were applied for determining the differentially expressed genes (DEGs) and calculation for P-value, adjusted P-value and logFC. PlGF levels in both series were obtained and analyzed the difference among groups. Pathway enrichment analyses of DEGs were performed at Metascape (http://metascape.org).

Individual values were represented in the figures. Statistical analysis was carried out using a one-way analysis of variance (ANOVA) test followed by the Fisher’s Exact Test (GraphPad Software, version 9, Inc. La Jolla, CA, USA). A value of p<0.05 was considered statistically significant.

Since recent studies revealed a possible role of PlGF in DFD, we analyzed tissue PlGF levels at day 3 and day 7 after induction of DFD in mice (9, 10). PlGF levels in diabetic foot skin (or diabetic ulcer (DU) at day 0), did not differ from non-diabetic controls (or non-diabetic ulcer (NDU) at day 0) ( Figure 1A ). PlGF levels increased at both day 3 and day 7 after ulcer induction, regardless of the diabetic status ( Figure 1A ), which was consistent with the analysis on 2 human DFD datasets (GSE134431, Figure 1B ; GSE80178, Figure 1C ). However, the overall increases in PlGF levels were significantly attenuated in diabetes ( Figure 1A ). The analysis on human DFD dataset GSE80178 also showed significant difference in gene expression ( Figure 1D ) and vasculature development pathway ( Figure 1E ). Together, these data highlight the importance of angiogenesis and suggest a possible role of PlGF in the pathological process of DFD.

Next, we investigated the role of PlGF in angiogenesis and the wound recovery in DFD. Since PlGF has a unique receptor, VEGFR1, we analyzed the cell types that harbor VEGFR1 to respond to PlGF at the wound. In a previous publication, macrophages, endothelial cells and fibroblasts have been shown to express VEGFR1 (2). To confirm it, day 3 injured skin tissue was digested into single cell population and then sorted for VEGFR1+ cells versus VEGFR1- cells by flow cytometry ( Figure 2A ). Afterwards, the VEGFR1+ cells were further sorted for CD68 (a marker for macrophages) and CD31 (a marker for CD31). Therefore, the VEGFR1+ cells were in addition separated as VEGFR1+CD68+ cells (macrophages), VEGFR1+CD31+ cells (endothelial cells), and VEGFR1+CD68-CD31- cells (VEGFR1+ cells that are not macrophages or endothelial cells) ( Figure 2A ). We found that among all VEGFR1+ cells, either CD68+ cells or CD31+ cells are much more than CD68-CD31- cells, suggesting that most VEGFR1+ cells are either macrophages or endothelial cells ( Figure 2A ). RT-qPCR was then performed on the sorted cells, showing significantly higher VEGFR1 in all 3 VEGFR1+ populations than VEGFR1- population, significantly higher levels of CD68 in VEGFR1+CD68+ cell population than the other 3 cell populations, significantly higher levels of CD31 in VEGFR1+CD31+ cell population than the other 3 cell populations, and significantly higher levels of Vimentin (a marker for fibroblasts) in VEGFR1+CD68-CD31- cell population than the other 3 cell populations ( Figure 2B ), with the latter suggests that most fibroblasts are in the VEGFR1+CD68-CD31- cell population. Compared with results from Figure 2A , fibroblasts may represent a relatively smaller cell population that expresses VEGFR1.

The effects of PlGF on fibroblasts but not on macrophages during tissue repair have been investigated (14). Since macrophages are known to play a critical role in angiogenesis, and since we detected a great number of macrophages among all VEGFR1+ cells in the mouse DFD model ( Figure 2A ), here we focused on macrophages. We generated an AAV carrying siRNA for VEGFR1 under a macrophage-specific CD68 promoter (AAV-pCD68-siVEGFR1), which allows specific depletion of VEGFR1 in macrophages and allows us to assess the effects of PlGF on angiogenesis and recovery of the DFD injury through macrophages ( Figure 3A ). To ensure the specificity of this construct, isolated macrophages, endothelial cells and mouse embryonic fibroblasts (MEFs), 3 cell types that express VEGFR1, were infected with either AAV-pCD68-siVEGFR1 or control AAV-pCD68-SCR. We found that CD68 promoter allowed effective transduction of macrophages, but not endothelial cells or MEFs, based on the expression of the GFP reporter ( Figure 3B ). The levels of VEGFR1 in the infected cells were analyzed by RT-qPCR, showing significant reduction in VEGFR1 mRNA in AAV-pCD68-siVEGFR1-transduced macrophages, while the VEGFR1 mRNA in either endothelial cells or MEFs remained unchanged ( Figure 3C ). Similar results were obtained when VEGFR1 protein was analyzed by ELISA ( Figure 3D ). Thus, AAV-pCD68-siVEGFR1 specifically depletes VEGFR1 in macrophages.

The effects of PlGF on DFD and the role of macrophages were then assessed in a mouse model for DFD, in which C56/BL6 mice received STZ to induce diabetes and then received a surgical ulcer at dorsal midline after one week. At the time of surgical ulcer induction, PlGF or saline was orthotopically injected biweekly while AAVs were given locally to the wound only once, immediately after ulcer induction. We have previously shown that the AAV-mediated transduction in vivo can sustain much longer than the one month’s experimental time window (25–27). After another 4 weeks, the mice were sacrificed and analyzed ( Figure 4A ). We found that STZ treatment induced irreversible hyperglycemia in mice, while any treatments or treatment combinations neither correct hyperglycemia ( Figure 4B ), nor alter beta-cell mass ( Figures 4C, D ). These data suggest that the sustained hyperglycemia induced by STZ resulted from loss of beta-cells rather than from other reasons, i.e. insulin resistant, in which hyperglycemia is developed while the beta cell number can remain normal or even increase at the early stage (28). Moreover, PlGF treatment with/without macrophage-depletion of VEGFR1 does not affect diabetes.

Next, we examined the effects of PlGF through macrophages on the wound healing in DFD. We found that the wound was completely cured at 4 weeks after ulcer induction in saline-treated normoglycemic mice (saline) ( Figure 5A ). The ulcers did not cure in all STZ-treated mice. However, better recovery was detected in diabetic mice treated with PlGF ( Figure 5A ). This improvement of wound recovery by PlGF, was however, abolished by macrophage-depletion of VEGFR1 ( Figure 5A ). Angiogenesis was analyzed at the end of the experiment by vessel density (5 weeks after STZ or 4 weeks after PlGF/AAVs). We found that the vessel density was significantly reduced by STZ treatment in the injured tissue ( Figures 5B, C ). However, improved angiogenesis was detected in diabetic STZ-mice treated with PlGF ( Figures 5B, C ). This improved angiogenesis by PlGF was also abolished by macrophage-depletion of VEGFR1 ( Figures 5B, C ). Thus, PlGF promotes wound healing and angiogenesis, which are likely through VEGFR1 on macrophages.

Since we found that PlGF-induced improvement in angiogenesis and recovery of DFD was largely attenuated by macrophage-specific depletion of VEGF1R, we examined the changes in macrophages in response to PlGF and VEGFR1 depletion. First, PlGF in tissue was assessed in all experimental groups, showing significant reduction by STZ and significant increase by PlGF ( Figure 6A ). Surprisingly, PlGF levels were reduced by macrophage-depletion of VEGFR1 ( Figure 6A ). Next, macrophages (CD68+) and their M2 (CD68+CD163+) versus M1 (CD68+CD163-) polarization were analyzed by flow cytometry ( Figures 6B–D ). Interestingly, we found that STZ not only reduced the number of macrophages, but also reduced their M2 polarization ( Figures 6B–D ). Moreover, PlGF not only increased the number of macrophages, but also enhanced their M2 polarization ( Figures 6B–D ). Finally, the effects of PlGF on both macrophage number and their M2 polarization were both attenuated by macrophage-depletion of VEGFR1 ( Figures 6B–D ).