TWIST1-Reprogrammed Endothelial Cell Transplantation Potentiates Neovascularization-Mediated Diabetic Wound Tissue Regeneration

Endothelial progenitor cells (EPC) are the key cellular effectors that have the potential to differentiate into endothelial cells (EC) during postnatal neovascularization (1). Neovascularization is often compromised due to reduced number of EPC in diabetic conditions thereby limiting the endogenous enrichment or autologous EPC transplantation therapies (2–4). De novo EC generation from nonendothelial precursor cells could be a promising strategy to improve neovascularization to increase the EPC repertoire. Studies have been attempted to modulate the fate of induced pluripotent stem cells (iPSC) or embryonic stem cells toward EC by differentiation using specific growth factors (5) or direct reprogramming of fibroblast/somatic cells toward endothelial lineage by overexpressing endothelial-specific transcription factors (6–8). Han et al. (9) depicted conversion of only 4% of murine skin fibroblasts into EC by forced expression of defined EC-specific transcription factors. These approaches of EC’s generation are limited by their low reprogramming/transdifferentiation efficiency. Similarly, multiple E26 transformation–specific (ETS)-related transcription factors were used to reprogram somatic cells into EC, which are known potent regulators of vascular development and angiogenesis. However, stable proliferative EC were not obtained because of the lack of precise temporal control on gene overexpression (10,11). On the contrary, another independent study reported that overexpression of a single ETS-related transcription factor, ETV2, was sufficient to reprogram human fibroblasts into EC (12,13). However, it was concluded that too low or high levels of ETV2 expression led to compromised endothelial reprogramming. Therefore, specific optimization of combination or expression levels of target cell–specific transcription factors is required for successful endothelial reprogramming of cells. Nonetheless, Takahashi et al. (14) also revealed that reprogramming efficiency of iPSC from somatic cells was even lower i.e., ∼0.1%, thereby necessitating the use of other stem cell types for EC generation. Interestingly, adult stem cells such as mesenchymal stem/stromal cells (MSC) have been suggested to be a reliable source for induced generation of various other cell types apart from its trilineage—chondrocyte, adipocyte, or osteoblast—due to its unique characteristic of plasticity; i.e., they can be transdifferentiated into nonobvious lineages using molecular or pharmacological approaches. However, transdifferentiation of MSC into induced EC (iEC) has been explored using EC-specific growth factors only till date (15–17). Also, MSC are the most abundantly available adult stem cell types that can be harvested for various types of human tissues: bone marrow, adipose, umbilical cord, blood, etc. Thus, the reprogramming of MSC into induced EC (iEC) using molecular tools could be a promising strategy for the generation of patient-specific EC for autologous transplantation therapy. Recently, we demonstrated that inhibition of cyclo-oxygenase 2 (Cox-2) enzyme using specific inhibitor potentiated transdifferentiation of Wharton jelly–derived MSC (WJ-MSC) into EC in vitro, and its transplantation demonstrated an improved vascularization-mediated translational efficiency in vivo (18). The current study used the meta-analyses of microarray data sets along with molecular and cellular tools for gaining insights into the transcriptional regulation occurring during transdifferentiation of WJ-MSC into iEC followed by developing a novel strategy for generating reprogrammed EC (rEC) by forced overexpression of a single transcription factor in WJ-MSC. Furthermore, in vivo murine diabetic nonhealing wound models were used to evaluate the translational efficiency of these rEC to enhance neovascularization during skin tissue regeneration.

The raw data files of mRNA expression data sets for three microarray studies available in the National Center for Biotechnology Information Gene Expression Omnibus (NCBI-GEO) (https://www.ncbi.nlm.nih.gov/geo) and ArrayExpress (https://www.ebi.ac.uk/arrayexpress/) were retrieved to compare the gene expression profile in different human cell types such as 1) MSC (E-MEXP-3071), 2) EPC (GSE37045 and GSE12891), and 3) human umbilical vein EC (HUVEC) (GSE25979) using Affymetrix Hugene-1.0.stv1, Affymetrix hgu133plus2, and Affymetrix HuEx-1-0-st, respectively. Since these microarrays were performed using different platforms consisting of different numbers of genes, we performed an initial data quality check, preprocessing, and normalization of each data set separately using the R programming language (https://www.r-project.org) installed with various statistical packages from the Bioconductor suite (https://www.bioconductor.org). Principal component analysis was performed using Partek Genomics Suite (https://www.partek.com) to identify the close relationship between the sample groups used in the study. The differentially expressed genes (DEGs) were determined between each group by the limma package (Linear Model for Microarray Analysis). Next, an empirical Bayesian approach was used to calculate t-statistic for each group followed by the use of Benjamini-Hochberg algorithm for multiple testing error to calculate the corrected P values. Selection of up- and downregulated genes with log fold change by 2 were sorted from each group and subjected to annotation using the “annotate” package. To find the pathways associated with the top DEGs, we imported the up- and downregulated genes from each array in Protein ANalysis THrough Evolutionary Relationship (PANTHER) software to classify them based on their participation in signaling pathways relevant to vascularization (19). Heat maps were generated using expression values of upregulated and downregulated genes with the help of Gitools software to identify genes showing increased or decreased expression profile during the transdifferentiation of MSC to EPC and EPC to mature EC. Furthermore, self-organizing map clustering was performed using Partek Genomics Suite software to identify the gene clusters with a consistent increase or decrease in gene expression profile (20). Protein-protein interaction network was also generated using STRING v9.1 software (https://string-db.org) to determine the existing information about the interactions between selected genes (21,22) for further in vitro analysis.

A well-characterized human WJ-MSC (HiMedia, Mumbai, India) was cultured to perform experiments within passages five to seven in MSC expansion medium (HiMedia) as previously described (18).

WJ-MSC were subjected for transdifferentiation into EC by culturing in endothelial growth medium (EBM-2; Lonza) for 7 days (iEC-D7) and 14 days (iEC-D14) as previously described (18). Morphological characterization of WJ-MSC and iEC was performed using bright-field microscopy at ×10 magnification (Olympus). Separately, the transdifferentiation of WJ-MSC into iEC-D7 was also performed in the absence/presence of BMP2 (10 ng/mL; HiMedia) along with BMP2 receptor inhibitor, DMH1 (100 nmol/L; (HiMedia) (15–17).

The WJ-MSC and iEC-D7 and -14 were incubated with DiI Ac-LDL (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine acetylated LDL) (5 µg/mL) containing respective medium for evaluation of its rate of uptake as described previously (18). In another set of experiments, the effect of BMP2 during transdifferentiation of WJ-MSC into iEC was evaluated by comparing the level of DiI Ac-LDL uptake in these cells.

For evaluation of expression levels of endothelial-specific gene markers in WJ-MSC and iEC-D7 and -D14, quantitative real-time PCR (qPCR) was performed as previously described (18) using gene-specific primers of KDR, VEGFR1, CD31, VE-Cadherin, CD146, VWF, ANGPT2, TIE-2, EDN1, EDNRB, NRP1, NRP2, eNOS, and CD105. The gene expression values were normalized to the housekeeping gene, eukaryotic 18S rRNA, and values represented as fold change relative to WJ-MSC as control.

Cultured WJ-MSC and iEC-D7 and -D14 were washed, fixed, and permeabilized before separate incubation with primary antibodies against KDR, eNOS, and/or CD31 followed by Alexa Fluor 488–conjugated secondary antibody and mounted with DAPI containing mounting media as previously described (18). The role of BMP2 on nuclear translocation of inhibitor of differentiation 1 (ID1) in WJ-MSC and iEC-D7 was separately analyzed by immunostaining using confocal microscopy as previously described (18).

In silico analysis of the promoter region of Id1 depicted the presence of three putative SMAD-complex binding sites (BS). Three constructs of the Id1 promoter region were designed and cloned from WJ-MSC DNA using human gene–specific primer sequences containing restriction sites (XhoI, BamHI, and HindIII; New England Biolabs) by PCR amplification, which were termed Wt-Id1, 2,316 base pairs (Wt-Id1-bp); ∆BS1-Id1-1,131 bp; ∆BS1-S2 Id1-618 bp; and ∆BS1-S2-S3 Id1-318 bp. These amplified constructs were subjected to restriction digestion followed by DNA ligation (T4 DNA Ligase; Invitrogen) into the pMCS-Green Renilla Luc plasmid constructs (Thermo Scientific) (23). Similarly, in silico analysis of Kdr promoter up to 5 kb upstream of transcription start site confirmed the presence of two enhancer-box (E-box) sequences (TWIST1 BS). Three constructs of KDR promoter region were designed containing either two TWIST1 BS (KDR WT-1,560 bp), one BS (KDR ∆BS1-1,338 bp), or none (KDR ∆BS1-S2-721 bp) and were successfully cloned using restriction enzymes (XhoI and BamHI) in pMCS-Green Renilla Luc vector as described above (23). Also, four E-box sequences on the Sfrp4 promoter region up to 5 kb upstream of transcription start site were identified. These four constructs, SFRP4-WT-2,097 bp, SFRP4-∆BS1-1,881 bp, SFRP4-∆BS1-S2-1,668 bp, and SFRP4-∆BS1-S2-S3-1,341 bp, were cloned using restriction enzymes (XhoI and BamHI) and T4 DNA ligase into the pMCS-Green Renilla Luc plasmid as previously described (23).

For overexpression of TWIST1, WJ-MSC were transfected with pCMV-TWIST1-Flag (OriGene) and/or control vector, pCMV-EGFP-N1 (Addgene), using Xfect (Clontech). Concurrently, for silencing of the TWIST1 expression, WJ-MSC were similarly transfected with piLenti-TWIST1-siRNA-GFP (abm) and/or control vector, piLenti-Scr-siRNA-GFP (abm). 24 h posttransfection, the medium was changed into either MSC expansion medium or EGM-2 medium as described above. Transfection efficiency was monitored by evaluating the GFP expression followed by confirmation of transgene overexpression and/or silencing using Western blot analysis for TWIST1, FLAG, or GFP as previously described (24).

WJ-MSC at a cell density of 10⁴cells/well were plated in a 96-well plate and incubated for 24 h at 37°C. Cells were transfected with pMCS-Green Renilla Luc vector subcloned with the desired promoter region and internal control vector cytomegalovirus (CMV)-Red Firefly Luc control plasmid (pCMV-Red Firefly Luc) using Lipofectamine 3000 (Invitrogen). Twenty-four hours posttransfection, WJ-MSC were induced for transdifferentiation in EGM-2 medium in the absence/presence of BMP2 and/or DMH1 at 37°C for 72 h. Cells were washed and lysed followed by detection of luminescence according to the manufacturer’s protocol (Thermo Scientific) using a multimode reader (PerkinElmer) as described previously (23). Similarly, cells were transfected with pMCS-Green Renilla Luc vector subcloned with the desired promoter region of KDR and SFRP4 and internal control vector, pCMV-Red Firefly Luc, along with pCMV-TWIST1-Flag overexpression and/or silencing piLenti-TWIST1-siRNA-GFP vectors and/or their respective controls to determine the role of TWIST1 in the promoter-reporter activity of KDR and SFRP4 as described above (23).

Cellular protein extracted from WJ-MSC and iEC was subjected to immunoblotting with primary antibodies against inhibitor of differentiation 1 (ID1), twist family BHLH transcription Factor 1 (TWIST1), kinase insert domain receptor (KDR), secreted frizzled receptor protein (SFRP4), and β-tubulin as previously described (25). Separately, cell extracts of WJ-MSC and iEC were coimmunoprecipitated with ID1 antibody and immunoblotted with TWIST1 antibody (Abcam) as previously described (23).

For confirmation of the direct binding of TWIST1 on KDR and SFRP4 promoter region during transdifferentiation, WJ-MSC were transfected with pCMV-TWIST1-Flag and/or pCMV-EGFP-N1 and subjected to transdifferentiation. Cells were then washed and cross-linked with 1% formaldehyde followed by lysis using Bioruptor (Diagenode) for 20 cycles to get fragmented DNA. The supernatant was collected and subjected to immunoprecipitation using the anti-TWIST1 antibody according to the manufacturer’s protocol. The purified DNA was used as a template for qPCR analysis with primers specific for Twist1-binding sequences on KDR and SFRP4 promoter as previously described (25,26).

WJ-MSC and iEC transfected with pCMV-GFP vector were fixed with 4% paraformaldehyde and incubated with KDR-phycoerythrin or SFRP4-phycoerythrin along with GFP-FITC antibody to evaluate the percentage of WJ-MSC and iEC positive for these EC markers using flow cytometry (BD FACSCanto II), and data were analyzed using BD FACSDiva software as previously described (18).

TWIST1 gene insert (610 bp) was subcloned from pCMV-TWIST1-Flag plasmid vector into pLVX-AcGFP-C1, lentiviral vector, as described in Supplementary Material. Lentiviral particles were prepared by transfecting pLVX-AcGFP-C1/pLVX-AcGFP-TWIST1/piLenti-Scr-shRNA-GFP or piLenti-TWIST1-shRNA-GFP lentiviral plasmid along with second-generation packaging plasmids: psPAX2 and pMD2.G in HEK-293T cells using Lipofectamine 3000 (Invitrogen). Lentiviral particles were collected and filtered as previously described (18). Next, WJ-MSC were transduced with lentiviral particles and subjected for puromycin (0.5 µg/mL) selection. The GFP-labeled WJ-MSC were subjected to transdifferentiation in EBM-2 medium for 7 and 14 days to generate reprogrammed EC—rEC-D7 and rEC-D14, respectively—and utilized for further in vitro experiments: DiI Ac-LDL uptake, qPCR analysis, flow cytometry analysis, and immunofluorescence microscopy for deciphering TWIST1-mediated KDR and SFRP4 gene expression at RNA and protein levels as well as for transplantation studies in vivo (18,25).

WJ-MSC and rEC (pLVX-AcGFP-TWIST1) and control iEC (pLVX-AcGFP-C1/piLenti-Scr-shRNA-GFP) and negative control (piLenti-TWIST1-shRNA-GFP) at day 7 or 14 were plated at a cell density of 5 × 10³ cells/well to evaluate Twist1-mediated effect on proliferation potential using BrdU incorporation assay (25). These cells were separately plated at a density of 1 × 10³ cells/well in the Boyden chamber–chemotaxis assay with cells at the upper chamber and VEGF-A (10 ng/mL; Invitrogen) as chemoattractant in the lower chamber as previously described to determine the role of Twist1 on cellular migration (25). Next, these cells were transplanted on the top of Chick chorioallantoic membrane (CAM) as previously described (18) to evaluate the effect of Twist1 on the number and length of blood vessel formation using angiocount software (27).

Type 1 and type 2 diabetic mice were used to evaluate the in vivo fate of transplanted human rEC and/or iEC. Type 1 diabetes was generated in male or female C57bL/6J mice by administrating streptozotocin at a dose of 70 mg/kg intraperitoneally for five consecutive days, whereas, for type 2 diabetes, transgenic db/db mice were used (28). Diabetes generation was confirmed by monitoring blood glucose levels at regular intervals for 2 weeks. Both type 1 and 2 diabetic mice were grouped separately and used for transplantation studies. Stable human rEC (pLVX-AcGFP-TWIST1) and/or iEC expressing different control vectors (pLVX-AcGFP/piLenti-Scr-shRNA-GFP) as wound controls or negative control (piLenti-TWIST1-shRNA-GFP) were transplanted intradermally onto the wounds (1 × 10⁶ cells/wound) as previously described (25,29,30). Animal experimentation and biosafety protocols were approved by the institutional animal ethics committee (approval no. IICT/IAEC/49/2018) and institutional biosafety committee (approval no. IICT/IBSC/05/2018). The rate of wound closure was determined using morphometric and histological analysis as previously described (18). The wound sections were subjected to hematoxylin-eosin and Sirius red staining to evaluate the reepithelialization, granulated tissue formation, and collagen deposition (25,29). Further, engraftment and paracrine effect of transplanted cells were determined by qPCR analysis of human-specific and mouse-specific endothelial marker expression, respectively, in the regenerated wound tissues of type 1 and type 2 diabetic mice as mentioned previously (18,29). Separately, wound tissue sections were subjected to immunostaining with TWIST1 and GFP along with human-specific KDR and mouse/human-specific SFRP4 antibody for determining the quantifiable colocalization (Pearson coefficient) (25,29).

Mice were anesthetized and placed under the laser Doppler instrument (LDI2-IR; Moor Instruments, Axminster, U.K.) to determine the microcirculatory blood flow in regenerated wound area. The results were depicted as mean flux intensity, which was calculated by averaging flux intensity of each wound per mice group in both type 1 and type 2 diabetic mice (31). Also, the regenerated wound tissue sections were evaluated for the colocalization of human-specific endothelial markers such as CD31 (platelet endothelial cell adhesion molecule [PE-CAM]) with GFP to determine the fate of the transplanted rEC or iEC using confocal microscopy (25).

Data represented as mean ± SEM from experiments that were performed at least thrice. For assessment of the difference between groups and their respective controls, statistical significance was evaluated using one-way or two-way ANOVA followed by appropriate analysis such as post hoc, etc., or Student paired t test (Prism, version 6.05; GraphPad). Photomicrographs and blots from experiments were reproduced at least thrice with similar results.

All data generated or analyzed during this study are included here in the published article (and Supplementary Material).

We retrieved raw mRNA expression data files of E-MEXP-3071, GSE37045, GSE12891, and GSE25979 from ArrayExpress and NCBI-GEO and performed the quality check followed by normalization of the data sets as described in the workflow (Supplementary Fig. 1A). Next, principal component analysis of the data sets revealed the gene expression profile to be homogenous among the similar cell types but significantly distinct among the different cell types (Supplementary Fig. 1B). DEGs in the data sets with a log fold change ≥2, identified using the limma package based on their adjusted P values ≤0.05, depicted 271, 1,107, and 546 upregulated and 336, 871, and 515 downregulated genes in MSC, EPC, and HUVEC, respectively (Supplementary Fig. 1C). These DEGs from each array were classified based on their role in the different cell signaling pathways using PANTHER, resulting in the identification of 983 genes out of 1,554 pathway hits spanning 90 PANTHER pathways that are differentially expressed in all the three data sets (Supplementary Fig. 1C). Among these, 30 pathways relevant to the differentiation process were selected that yielded a heat map of 265 genes depicting a consistent increase or decrease in their expression levels exhibiting a hierarchical relationship between MSC, late endothelial progenitor cell (lEPC), and HUVEC (Supplementary Fig. 2A). Next, self-organizing map clustering of these 256 genes depending upon the expression pattern resulted in selection of clusters 11, 16, and 21 that depicted consistent increase but clusters 15, 20, and 25 with a consistent decrease in gene expression profile from MSC to lEPC and lEPC to HUVEC (Supplementary Fig. 2B). Lastly, these clusters consisting of 72 genes were represented using a heat map that depicted a consistent increase or decrease in expression in MSC to lEPC and lEPC to mature EC, HUVEC (Fig. 1A).

A random selection of 41 out of 72 genes from in silico data sets was validated using qPCR analysis in the prior well-characterized WJ-MSC and transdifferentiated iEC-D7 and iEC-D14 (18). The results of in vitro studies depicted an increased expression of BMP1, BMP2, KDR, SERPINE1, RHOB1, IFNGR1, SMAD1, IL1A, BIRC5, ITGB3, and RGS5 and decreased expression of ID1, SFRP4, and BIRC3 (Fig. 1B–E), which was corroborated well with the in silico observations and thereby incited us to identify the plausible molecular targets that can modulate the transdifferentiation of WJ-MSC into iEC. Next, the protein-protein interaction analysis of these identified genes using STRING v9.1 online software tool displayed regulatory connections among them suggesting cross talk between different pathways, as genes participating in one pathway were observed to regulate the expression of other genes by their activation or deactivation in other pathways (Supplementary Fig. 3). Among these, during transdifferentiation of WJ-MSC into iEC, a consistent eightfold and ∼twofold high expression of the growth factor BMP2 (Fig. 1C) and its downstream effector, Smad1 (Fig. 1D), suggests a Smad-mediated regulation of an atypical transcription factor, ID1. ID1 lacks the DNA-binding domain and regulates its downstream effector genes via its cofactors belonging to the basic helix-loop-helix (bHLH) transcription factor family (32) that, in turn, may transcriptionally upregulate the angiogenic gene, KDR (Fig. 1B), and downregulate the antiangiogenic gene, SFRP4 (Fig. 1C).

Transdifferentiation of WJ-MSC into iEC-D7 in presence of BMP2 led to a marked increase in DiI Ac-LDL uptake compared with control iEC, which was perturbed in presence of its receptor inhibitor, DMH1, thereby suggesting a plausible role of BMP2 in this process (Fig. 2A). Further, qPCR analysis of endothelial-specific markers depicted a significant increase in mRNA expression of CD31, VWF, VE-Cadherin, KDR, and eNOS by more than ∼3.5 fold in BMP2-mediated transdifferentiated iEC compared with control, WJ-MSC, which was mitigated in presence of DMH1 (Fig. 2B). A differential expression pattern of other endothelial-specific genes, ANGPT2, NRP1, NRP2, TIE2, etc., was observed in these treated iEC (Fig. 2C), thus suggesting modulation of selective endothelial-specific gene expression by BMP2.

BMP2 binds to its receptors, BMPRI and BMPRII, and activates its downstream signaling via sequential SMAD activation to regulate its target genes (33). Interestingly, literature also suggests tBMP2-mediated regulation of its downstream effectors in a SMAD-independent manner (34). Thus, it becomes pertinent to decipher the downstream signaling pathway of BMP2 during transdifferentiation of WJ-MSC into iEC to be SMAD dependent or independent. To evaluate this, we determined the global gene expression profile of all the SMADs along with BMP2 receptors in WJ-MSC and iEC in the absence or presence of BMP2. BMP2 could significantly increase Smad2 expression in WJ-MSC, while in iEC, BMP2 treatment led to a significant increase, by four to fivefold, in SMAD1, SMAD4, SMAD5, and SMAD8 as well as inhibitory SMADs—SMAD6 (∼s7-fold) and SMAD7 (∼15-fold)—suggesting the occurrence of SMAD-dependent signaling downstream to BMP2 (Fig. 2D). Literature suggests various BMP2-mediated SMAD-regulated target genes such as ID1, which is known to inhibit cell differentiation (32). BMP2 treatment led to a significant increase in ID1 expression in WJ-MSC (Fig. 3A). In contrast, a decreased expression of ID1 was also observed during transdifferentiation of WJ-MSC into iEC that were further decreased in presence of BMP2 (Fig. 3A), which led us to explore the transcriptional regulation of ID1 by BMP2. The presence of four SMAD complex BS was identified on the promoter region of ID1 (Supplementary Fig. 4A), which were cloned in pMCS-Green Renilla Luc vector by sequential deletion of BS (Supplementary Fig. 4B). Next, promoter-reporter luciferase activity was determined as described in research design and methods. The results depicted a significant decrease in the promoter activity of the Wt-ID1 promoter construct in iEC compared with WJ-MSC (Fig. 3B). As expected, treatment with BMP2 further led to a decrease in the luciferase activity in iEC but not in WJ-MSC. The sequential deletion of BS1 led to a further increase in promoter activity in WJ-MSC as well as iEC compared with the Wt-ID1 group, suggesting that the BS1 site is essential for the negative regulation of ID1 during WJ-MSC transdifferentiation into iEC. However, there was no significant change in promoter activity after the deletion of both BS1 and BS2 by the presence of BS3 and the deletion of BS1, BS2, and BS3 by the presence of BS1′. These data suggest that BMP2-mediated increased expression of inhibitory SMADs—SMAD6 and SMAD7—downregulates the ID1 transcription. Next, immunoblot analysis depicted an increased protein expression of ID1 in WJ-MSC only but not in iEC in the presence of BMP2 treatment that was mitigated by DMH1 (Fig. 3C). As mentioned earlier, ID1 imparts downstream regulation by interacting with the bHLH family of transcription factors such as HEY1, HIF-2α, and/or TWIST1, which may regulate the expression of target genes, KDR and SFRP4. Among these, HEY1 is known to act as a negative transcriptional regulator of its target genes, whereas HIF-2α has a proven role under hypoxic conditions. However, TWIST1 has been reported in the literature to act both as a negative and a positive regulator of its target gene transcription (35,36). Excitingly, a concomitant increase in TWIST1 expression with decreased levels of ID1 in iEC (Fig. 3C) suggested a negative correlation between ID1 and TWIST1-mediated regulation of its downstream effectors during WJ-MSC transdifferentiation into iEC. Coimmunoprecipitation of WJ-MSC and iEC with ID1 and immunoblot with TWIST1 showed binding of ID1 with TWIST1 in WJ-MSC but not in the transdifferentiated iEC (Fig. 3D), suggesting a decreased expression of ID1 in iEC renders the TWIST1 free for its downstream transcriptional regulation of its target genes. However, in WJ-MSC, ID1 sequesters TWIST1 and thereby makes it unavailable for its downstream regulation. ID1 is a nuclear protein, and immunofluorescent staining revealed that BMP2 induced marked increase in localization of ID1 in the nucleus of WJ-MSC but not iEC, an effect that was perturbed by DMH1 (Fig. 3E). Additionally, TWIST1 mRNA expression was observed to be significantly increased during the transdifferentiation of WJ-MSC into iEC-D7 and iEC-D14 (Supplementary Fig. 4C). Next, Western blot analysis also depicted a marked increase in TWIST1 and KDR but a decrease in SFRP4 expression in iEC-D7 and iEC-D14 as compared with WJ-MSC (Supplementary Fig. 4D). Also, flow cytometry analysis depicted high KDR and low SFRP4 in iEC compared with WJ-MSC (Supplementary Fig. 4E). These data provide critical insights into the role of TWIST1 as a downstream transcriptional modulator of the BMP2-ID1 signaling pathway during transdifferentiation of WJ-MSC into iEC.

To further evaluate the TWIST1-mediated transcriptional regulation of KDR and SFRP4 during the transdifferentiation process, we used the gain- and loss-of-function approach by overexpression and/or silencing of TWIST1 gene that was confirmed by immunoblotting the expression of the tags Flag or GFP, KDR, SFRP4, and TWIST1 in transfected WJ-MSC (Supplementary Fig. 4F). As described in research design and methods, the presence of putative TWIST1 binding sites on the promoter region of identified target genes KDR (Supplementary Fig. 5A and B) and SFRP4 (Supplementary Fig. 5C and D) led us to evaluate the promoter-reporter assay. A significant increase in Wt-KDR promoter activity was observed in iEC by 10-fold in the presence of TWIST1 overexpression (Supplementary Fig. 6A). Further, a sequential deletion of BS1 (ΔBS1-KDR) resulted in a fivefold increase in KDR promoter activity compared with control vector–transfected iEC groups (Supplementary Fig. 6A), suggesting TWIST1 positively regulates KDR promoter activity from two distinct binding sites. Similiarly, TWIST1 overexpression led to a significant increase in luciferase activity in WJ-MSC by ∼twofold, which was far less compared to iEC (Supplementary Fig. 6B). This effect was significantly perturbed in the presence of TWIST1 silencing in both iEC (Supplementary Fig. 6A) and WJ-MSC (Supplementary Fig. 6B). Furthermore, the deletion of both the binding sites significantly abolished the promoter activity (Supplementary Fig. 6A and B). Next, promoter-reporter assay for the Wt-SFRP4 promoter depicted a significant decrease in luciferase activity in the presence of Twist1 overexpression in iEC (Supplementary Fig. 6C) as well as in WJ-MSC (Supplementary Fig. 6D). Interestingly, the sequential deletion of BS1 (ΔBS1-SFRP4) led to a robust decrease in promoter activity in iEC (Supplementary Fig. 6C). However, subsequent deletion of both BS1 and BS2 (ΔBS1-S2-SFRP4) and BS1, BS2, and BS3 (ΔBS1-S2-S3-SFRP4) did not show any significant change in promoter activity, suggesting BS1 and BS2 are essential sites for TWIST1 binding. A similar effect was also observed in WJ-MSC but with lesser fold change (Supplementary Fig. 6D). Interestingly, the silencing of TWIST1 could not regain the SFRP4 promoter activity, suggesting the involvement of other cofactors along with TWIST1 for transcriptional regulation of the SFRP4 gene. Next, for evaluation of the direct interaction between TWIST1 and its putative binding sites on KDR and SFRP4 promoter, chromatin immunoprecipitation analysis was performed that depicted an efficient differential binding of TWIST1 at binding sites—BS2 on KDR promoter (Supplementary Fig. 7A) but both BS1 and BS2 on SFRP4 promoter (Supplementary Fig. 7B)—thereby correlating well with the promoter-reporter luciferase activity. These data suggest that TWIST1 plays a crucial role by transcriptional regulation of KDR and in part of SFRP4 by promoting KDR while inhibiting SFRP4 expression during transdifferentiation of WJ-MSC into iEC.

Next, to evaluate the role of TWIST1 in determining the fate of WJ-MSC toward endothelial transdifferentiation, TWIST1 was subcloned in a lentiviral vector as described in research design and methods. WJ-MSC were transduced with TWIST1 overexpressing/silencing lentiviral particles followed by puromycin selection and subjected for endothelial differentiation of these rEC. TWIST1 overexpression in rEC was confirmed by immunoblotting that depicted a shift in molecular weight of TWIST1 (27 kDa) fused with GFP (27 kDa) gene to 54 kDa as compared with GFP alone (Supplementary Fig. 8A). Next, the qPCR analysis also depicted a significant increase and decrease in TWIST1 expression in rEC at days 7 and 14 transduced with TWIST1 overexpression and silencing (negative control) lentiviral particles, respectively (Supplementary Fig. 8B). Immunofluorescence microscopy depicted GFP expression in all the transduced cell groups, while coimmunostaining of GFP and DAPI confirmed the nuclear localization of overexpressed TWIST1, a transcription factor, as compared with empty vector control– and piLenti-TWIST1-shRNA-GFP–transduced iEC (negative control) groups (Supplementary Fig. 8C). TWIST1 overexpression led to an increase in KDR expression (∼85%) and a decrease in SFRP4 expression (∼35%), while its silencing reversed this phenomenon in TWIST1 rEC as analyzed using flow cytometry assay (Supplementary Fig. 8D). To substantiate the endothelial transdifferentiation of our rEC, we carried out various physiological and molecular assays to confirm the role of TWIST1, which has not yet been implicated in similar cell fate determination to the best of our knowledge. The rEC showed a marked increase in DiI Ac-LDL uptake as compared with control iEC, which was further perturbed in iEC with TWIST1 knockdown (negative control) as compared with WJ-MSC (control) at day 7 (data not shown) and day 14 (Supplementary Fig. 9A). TWIST1-mediated reprogramming of WJ-MSC into rEC was further confirmed by qPCR analysis of the expression profile of endothelial marker genes. Our results indicated a significant increase in expression of several endothelial markers such as KDR, VEGFR1, VE-cadherin, ANGPT2, eNOS, VWF, CD31, EDN1, NRP1, TIE-2, and CD146 in rEC-D14, whereas there was a significant increase in expression of SFRP4 upon TWIST1 silencing in iEC (Supplementary Fig. 9B and C). This observation was further supported by immunofluorescent staining of these cells, which depicted a marked increase in KDR expression in rEC and iEC at day 14 as compared with WJ-MSC (control), whereas TWIST1 knockdown perturbed the transdifferentiation of WJ-MSC into iEC by decreased expression of KDR compared with control iEC (Supplementary Fig. 10A). Similarly, SFRP4 expression was decreased in rEC compared with WJ-MSC (control), but there was an increase in SFRP4 expression when TWIST1 was silenced (Supplementary Fig. 10B), confirming that TWIST1 negatively regulates SFRP4 during transdifferentiation of WJ-MSC into rEC. These data indicate that TWIST1 reprogramming resulted in enhanced endothelial lineage transdifferentiation of WJ-MSC.

The functional activation of rEC was determined by evaluating cell proliferation and migration assays after 7 and 14 days of reprogramming. There was a significant decrease in proliferation potential of rEC compared with WJ-MSC at days 7 and 14 (Supplementary Fig. 11A), thereby confirming the transdifferentiation of cells during which proliferation is halted as expected. The Boyden chamber–chemotaxis assay revealed a significant enhancement of migratory potential in rEC as compared with WJ-MSC that were perturbed by TWIST1 silencing in iEC after 7 and 14 days of transdifferentiation (Supplementary Fig. 11B). As late EPC is known to incorporate into new blood vessels but not the early EPC, we evaluated the vasculogenic potential of iEC-D14 by CAM angiogenic assay. Stable rEC or iEC at day 14 were mixed with matrigel and loaded on the top of CAM as described in research design and methods. After incubation of 72 h, the morphological analysis of matrigel depicted a higher vasculogenic potential of rEC as compared with control iEC with empty vector/scrambled siRNA, as well as iEC with TWIST1 knockdown (Supplementary Fig. 11C). Together, these observations suggested that rEC with enhanced vasculogenic capacity can be of therapeutic potential for hypovascularity-associated diseases like chronic nonhealing diabetic wounds.

To evaluate the in vivo vasculogenic potential of transplanted TWIST1 rEC, full-thickness excisional splint wound healing model was generated in streptozotocin-induced type 1 (Supplementary Fig. 12A) as well as db/db type 2 diabetic mice (Supplementary Fig. 12B). The stable TWIST1 rEC and/or iEC (1 × 10⁶ cells/wound) were transplanted intradermally at the periphery of the wound. Laser Doppler flowmetry was performed to evaluate the microcirculatory blood flow in regenerated wounds. The results showed a higher mean flux intensity (>300) in the wounds transplanted with rEC (TWIST1 overexpressing) as compared with other positive (vector controls) and negative (TWIST1 silencing) control iEC–transplanted type 1 and 2 diabetic mice groups (Fig. 4A). Quantitation of the percent mean flux intensity depicted a significant increase in the TWIST1 rEC–transplanted group as compared with iEC-transplanted groups (Fig. 4B), suggesting enhanced vascularity at the wound bed of type 1 and 2 diabetic mice transplanted with TWIST1-reprogrammed iEC.

The morphometric analysis of regenerated wound on postsurgery day 14 depicted an enhanced wound healing in rEC-transplanted type 1 and 2 diabetic mice groups as compared with other groups (Fig. 4C). The rate of wound closure was ∼78% and 84% in Twist1 rEC–transplanted type 1 and 2 diabetic mice groups, respectively, as compared with other iEC control or negative control groups (Fig. 4D). This was further corroborated by histological analysis of regenerated wounds by hematoxylin-eosin (Fig. 5A and B) and collagen staining (Fig. 5C and D), which revealed an increased granulation and collagen deposition in the TWIST1 rEC–transplanted group as compared with control iEC–transplanted groups. Quantitative analysis depicted no significant change in reepithelialization as evident from hematoxylin-eosin staining (Fig. 5B) but a significant decrease in collagen deposition (Fig. 5D) in TWIST1-silenced iEC–transplanted (negative control) as compared with iEC-transplanted (control vector) type 1 (data not shown) and type 2 diabetic mice groups.

For evaluation of the engraftment of transplanted cells, coimmunostaining of GFP with TWIST1 was evaluated by confocal microscopy, which depicted a higher colocalization of TWIST1 with GFP in TWIST1 rEC–transplanted type 1 (Supplementary Fig. 13A) and type 2 (Fig. 6A) diabetic mice group as compared with iEC-transplanted (control vector) groups (quantitative analysis) (Fig. 6B and Supplementary Fig. 13B). The higher engraftment of TWIST1 rEC in regenerated wounds was further supported by qPCR analysis of human-specific endothelial marker genes—VEGFR2 (KDR), CD31, TIE-2, VWF, VE-Cadherin, and NRP2—suggesting direct engraftment of rEC at the wound bed leading to increased neovascularization in the type 1 (Supplementary Fig. 13C) and type 2 (Fig. 6C) diabetic wounds. To determine whether the transplanted TWIST1 rEC regulate wound microenvironment in a paracrine fashion, we performed qPCR analysis for mouse-specific endothelial marker genes. CD31, Tie-2, vWF, VE-Cadh, and Nrp2 were also observed to be higher in TWIST1 rEC–transplanted type 1 (Supplementary Fig. 13D) and type 2 (Fig. 6D) diabetic mice groups as compared with groups transplanted with other control vectors, iEC and/or TWIST1-knockdown iEC (negative control). Next, immunostaining with human-specific KDR antibody and GFP revealed a higher colocalization at the wound bed of the TWIST1 rEC–transplanted group compared with control/empty vector iEC–transplanted groups in both type 1 (Supplementary Fig. 14A) and type 2 (Fig. 7A) diabetic mice. Quantification of colocalization revealed no significant colocalization of GFP with KDR in the regenerated wound bed of type 1 (Supplementary Fig. 14B) and type 2 (Fig. 7B) diabetic mice transplanted with TWIST1-silenced iEC as compared with scrambled shRNA iEC. Similarly, a decreased colocalization of GFP with SFRP4 was observed in the regenerated wound bed of TWIST1 rEC–transplanted type 1 (Supplementary Fig. 14C) and type 2 (Fig. 7C) diabetic mice. An increased colocalization of GFP with SFRP4 in the type 1 (Supplementary Fig. 14D) and type 2 (Fig. 7D) diabetic wounds was evident from the quantified images of TWIST1-silenced iEC–transplanted groups as compared with control groups, thereby corroborating with the in vitro observation of TWIST1-mediated transcriptional upregulation and downregulation of KDR and SFRP4 expression, respectively. Finally, TWIST1 rEC transplantation–mediated enhanced neovascularization of the diabetic wounds was determined by immunostaining with human-specific CD31. A higher CD31 colocalization with GFP at the wound of type 1 (Supplementary Fig. 15A) (quantification of colocalization) (Supplementary Fig. 15B) and type 2 (Fig. 8A and B) diabetic mice transplanted with TWIST1 rEC as compared with control vector iEC– and/or TWIST1-silenced (negative control) iEC–transplanted groups. Finally, detection of α-SMA (mural cells, green) and CD31 (engrafted rEC, red) in regenerated skin tissues in TWIST1 rEC–transplanted type 1 (Supplementary Fig. 15C and D) and type 2 (Fig. 8C and D) diabetic mice depicted potentially higher blood vessel maturation (yellow) as compared with other transplanted groups. These observations depicted the fate of the transplanted TWIST1 rEC toward neovascularization for improved vascularity-mediated diabetic wound tissue regeneration.

Neovascularization in hypovascularity-related nonhealing diabetic wounds can be enhanced by exogenous EC transplantation therapy to improve the vascularity. However, prior literature suggests that transplantation of mature EC fails to improve vascularity as well as functional improvement of the injured tissue (37,38). Recent studies on the clinical and therapeutic application of autologous EPC transplantation depicted limited improvement in neovascularization and tissue repair (39). A reduced levels as well as dysfunctional EC with an impaired proliferative, and mobilization capability, integrin profile, low differentiation and incorporation into new blood vessels were observed in Type 1 and Type 2 diabetes (40–42) thereby, limiting the source of EPC to be autologous in origin for transplantation. Thus, isolation/generation of EPC or iEC from the repertoire of adult stem cells, with higher proliferative and differentiation potential toward EC, is warranted to undertake the unmet medical needs. To address that, we attempted to accelerate the process of transdifferentiation of adult human WJ-MSC into iEC by identifying molecular targets that can promote the process. In silico analysis led us to identify molecular target BMP2, a growth factor that activates its downstream target ID1 via SMAD1-dependent signaling. BMP2 has been sporadically reported to play a role in tumor vasculogenesis via SMAD1, ERK1/2–dependent ID1 expression which enhanced the proliferative and tube forming capability of the cells and its inhibition using a BMP2 antagonist suppressed angiogenesis in tumor cells in vivo (34). Similarly, our study revealed that BMP2 activated canonical SMAD signaling in WJ-MSC but during transdifferentiation into iEC this gets interrupted by the presence of inhibitory SMADs that led to a decreased expression of ID1, a BMP2 target gene, thereby suggesting the involvement of coregulation by other transcription factors in the downstream signaling. Inhibition of ID1 has also been reported to regulate embryonic stem cell differentiation into EC (43). Hetero-dimerization of ID1 is sufficient to prevent the binding of bHLH family proteins on DNA (44). It not only blocks the transcription of its target gene but also inhibits the function of the bHLH transcription factor by sequestration of ubiquitous E proteins (32). Our in vitro mechanistic studies proved our hypothesis that ID1 sequesters off TWIST1 and inhibits the latter in transcriptionally activating its target genes in WJ-MSC, while, during transdifferentiation of WJ-MSC into iEC due to activation of inhibitory SMADs, ID1 expression is decreased and is unavailable to interact with TWIST1, thereby enabling the transcriptional regulation of TWIST1-dependent target genes. Further, to generate proof of concept, we generated reprogrammed iEC by stable overexpression of Twist1 using a lentiviral vector in WJ-MSC that was further induced in the presence of endothelial-specific growth factor–containing medium. As a similar approach to increase the EC repertoire, others in the field have used different types of stem and somatic cells such as iPSC (5), amniotic cells (10), and fibroblasts (9,12,13,45). Next, our study identified and used TWIST1, which is a master transcriptional regulator of mesodermal development and has been reported to promote ocular angiogenesis (46). The molecular mechanism of TWIST1 differs during the regulation of normal and tumor angiogenesis (47). Twist1 knockdown adversely affected embryonic vascular growth in Xenopus (48). Singh et al. (49) reported Twist1 overexpression in Giant cell tumor stromal cells induced expression of endolethial marker genes such as Vegfa and Vegfr1, whereas its knockdown depicted decreased expression levels of these endothelial marker genes, using in vitro studies. Similarly, we also observed an enhanced mRNA expression of endothelial markers such as VEGFR1, VEGFR2, CD31, VWF, TIE-2, and CD146, etc., in rEC. TWIST1-mediated biological functions are diverse, which is achieved by its homo/heterodimers interaction with other bHLH coregulators such as E2 proteins (35,36,47). It transcriptionally regulates its downstream targets by recognizing consensus enhancer-box (E-box) sequence “CANNTG” on the promoter region of its target genes (50). The activation or suppression of Twist1 target genes is determined by its stimulatory or inhibitory partner that modulates chromatin modifiers or RNA polymerase 2 transcription complex formation (35,36,50). Our findings, too, showed TWIST1 as both positive transcriptional regulator of the angiogenic gene VEGFR2 (KDR) and negative transcriptional regulator of antiangiogenic gene SFRP4.

MSC have been used for transdifferentiation into endothelial lineage cells using growth factors (15–18) and anti-inflammatory drugs (18), and their subsequent transplantation enhanced neovascularization-mediated wound healing in a full excisional splint wound healing model generated in C57BL/6J mice. Aguilera et al. (8) transdifferentiated MSC into EC for 14 (EC-D14) and 30 (EC-D30) days and reported an insignificant difference in wound healing profile in mice transplanted with EC-D14 or EC-D30. This led us to culture iEC under transdifferentiation up to day 14 for iEC transplantation. Our results clearly describe that transplantation of TWIST1 rEC potentiated neovascularization at the diabetic wound site by enhanced microcirculatory blood flow and enhanced KDR and decreased SFRP4 protein expression in the regenerated skin tissue at the diabetic wound bed. This study suggests use of transplantation of TWIST1-reprogrammed EC as a plausible autologous adult stem cell–based therapy for improving neovascularization in hypovascularity-related diseases or tissue injury.

In conclusion, these observations set the stage for future clinical application using small molecule therapeutic approaches to target this signaling pathway to modulate KDR and SFRP4 or transgenically modifying adult stem cells using CRISPR-Cas9 technology instead of lentiviral vectors to generate rEC for transplantation therapy.

Funding. A.D. acknowledges the funding provided by the Council of Scientific and Industrial Research (CSIR), Ministry of Science & Technology, Government of India, for Niche Creating High Science Projects under Healthcare theme: CSIR-IICT MLP0052 (PROMPT), MLP0053 (GRAFT). A fellowship provided by UGC-JRF/SRF to K.K. is gratefully acknowledged (manuscript communication number: IICT/Pubs./2019/379).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.D. conceived the idea and designed the study. K.K. performed the experiments. K.K. and A.D. analyzed the data and wrote the manuscript. A.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.