Chromatin-modifying agents convert fibroblasts to OCT4+ and VEGFR-2+ capillary tube-forming cells

Rationale The human epigenome is plastic. The goal of this study was to address if fibroblast cells can be epigenetically modified to promote neovessel formation. Methods and results Here, we used highly abundant human adult dermal fibroblast cells (hADFCs) that were treated with the chromatin-modifying agents 5-aza-2'-deoxycytidine and trichostatin A, and subsequently subjected to differentiation by activating Wnt signaling. Our results show that these epigenetically modified hADFCs increasingly expressed β-catenin, pluripotency factor octamer-binding transcription factor-4 (OCT4, also known as POU5F1), and endothelial cell (EC) marker called vascular endothelial growth factor receptor-2 (VEGFR-2, also known as Fetal Liver Kinase-1). In microscopic analysis, β-catenin localized to cell-cell contact points, while OCT4 was found to be localized primarily to the nucleus of these cells. Furthermore, in a chromatin immunoprecipitation experiment, OCT4 bound to the VEGFR-2/FLK1 promoter. Finally, these modified hADFCs also transduced Wnt signaling. Importantly, on a two-dimensional (2D) gel substrate, a subset of the converted cells formed vascular network-like structures in the presence of VEGF. Conclusion Chromatin-modifying agents converted hADFCs to OCT4+ and VEGFR-2+ capillary tube-forming cells in a 2D matrix in VEGF-dependent manner.


Introduction
Myocardial ischemia and cardiovascular diseases (CVD) are among the most significant causes of death and disability in the United States [1][2][3]. Despite the progress made in understanding PLOS  proliferate in response to growth factor stimulation [19][20][21][22]57,58]. Similarly, fibroblasts can proliferate upon activation [59][60][61]. Interestingly, one recent study has described the ability of fibroblast cells to acquire EC morphology and incorporate into newly formed blood vessels in the aftermath of injury [62]. Therefore, we tested the hypothesis that by partially erasing the "epigenetic memory" of fibroblasts cells using chromatin-modifying agents, it might be possible to reprogram these cells into blood vessel forming cells. By subsequently adding a differentiation agent (specifically a GSK-3β inhibitor) to these cells, we converted them into OCT4 + and VEGFR-2/FLK1+ cells that formed tube-like structures in a two-dimensional (2D) matrix.

Treatment of cells with epigenetic modifier
Stock solutions of Aza, a DNMT inhibitor, was dissolved at 1:1 ratio of acetic acid and water; TSA, an HDAC inhibitor, and TDG, a GSK-3β inhibitor, were prepared in sterile DMSO and stored in aliquots at -20˚C. Growing hADFCs at passage 3 (p3) were detached using cell dissociation solution. Cells were re-suspended in complete DMEM media with FBS to inactivate the trypsin/EDTA. The number of cells was counted using a hemocytometer, and equal numbers of cells re-plated into eight 10 cm dishes containing 10 mL complete DMEM media. The final working concentrations of Aza, TSA, and TDG were prepared prior to use at optimized concentrations of 50 nM, 25 nM, and 60 nM, respectively. Experiments were repeated four times unless otherwise indicated, with triplicate samples in each experiment.
Extraction of messenger RNA (mRNA) and quantitative reverse transcription (qRT)-polymerase chain reaction (PCR) assay Extraction of mRNA, cDNA preparation, and qRT-PCR methods have been previously described [63][64][65][66]. Total mRNA was prepared by washing control hADFCs and those treated with Aza + TSA or Aza + TSA + TDG once with 1× cold PBS, then solubilizing cells in Trizol. The mRNAs were quantified by reading the optical density at a 260 nm wavelength using a spectrophotometer. Gene-specific primers were synthesized and supplied by IDT-DNA (Coralville, IA): OCT4: For-  Table 1.

Biochemical methods
The immunoblotting methods have been previously described [63][64][65][66]. In brief, cell extraction buffer was prepared by mixing 25 mM Tris pH 7.4, 0.1% Triton X-100, 0.25% NP-40 detergent, 150 mM Sodium Chloride, 1 mM EDTA, and 25 mM Sodium Fluoride, 1 mM Sodium Orthovanadate, and 10 mM Sodium Pyrophosphate pH 7.5. The proteins were separated using a glass plate SDS-PAGE apparatus, then transferred to a nitrocellulose membrane. The membrane was incubated with the appropriate antibodies, washed with 1× TBST, then reacted with highly sensitive chemiluminescent (ECL) detection reagent (Thermo-Fisher, Rockford, IL). X-ray films were exposed to the nitrocellulose membrane and developed using an automated Kodak X-ray film developer.  Cell staining and microscopy The microscopy methods have been previously described [63][64][65][66]. In brief, the hADFCs were plated on sterile coverslips, incubated in media without additional compounds; with TSA and Aza; or with TSA, Aza, and TDG. The cells were fixed with 4% paraformaldehyde (PFA), washed three times with 1× TBS pH 7.4, permeabilized with 0.5% Triton X-100 in 1× TBS, and washed three times with 1× TBS. To visualize the fibroblastic phenotype of cells, hADFCs were stained with PHD-TRITC dye. Microscopic images were captured using an Olympus BX51 microscope with an Olympus 12.5MP DP71 CCD camera using an UPlanFL N dry 20×/0.75NA objective at room temperature, and images saved as TIFF files.

Chromatin immunoprecipitation (ChIP) experiment
The ChIP experiments were carried out as previously described [64,66]. The Magna ChIP kit was purchased from Thermo-Fisher Scientific (Rockford, IL). The chromatin samples were pre-cleared, then subjected to immunoprecipitation with anti-human OCT4 antibody bound to beads. The chromatin immunoprecipitate complexes were analyzed by PCR using primer pairs (For-5'-GGGAAATAGCGGGAATGTTG-3' and Rev-5'-GCGAAATGCCCAGAACTC-3'; NCBI reference sequence: NC_000004.12) that amplify an~850-bp DNA sequence in the region of the human VEGFR-2/FLK1-promoter/enhancer. The PCR product was resolved by 4% agarose gel in 1× TAE buffer, thereafter visualized by staining with 0.2 μg/ml ethidium bromide (EtBr), and images captured using a Polaroid gel documentation system.

Statistics
Data obtained from at least three independent experiments (n = 3) are expressed as means ± standard errors of the mean (S.E.M.). Comparisons between two means were performed with an unpaired Student's t-test. The differences among groups were evaluated by analysis of variance (ANOVA). P<0.05 was considered statistically significant.

Gene expression profile of hADFCs
In pilot experiments, we optimized the doses of epigenetic modifiers Aza (50 nM) and TSA (25 nM), and differentiation factor TDG (60 nM) in hADFCs. At these concentrations, hADFCs were highly proliferative and the compounds were non-toxic (S1 Fig). We also optimized specific time points for this study (Fig 1A). The dose and the experimental groups are listed in Fig 1B. The experimental groups included: a) day-2 (d2) control, untreated; b) day-3 (d3), received one dose of Aza + TSA; c) day-4 (d4), received second dose of Aza + TSA; and d) day-5 (d5), received third dose of Aza + TSA as well as TDG. All cells were grown in complete media, not deprived of growth factor or serum. Initial pilot experiments were conducted to test if the exposure of hADFCs to chromatin modifying agents induced expression of Next, to test the hypothesis that epigenetic modification might be adequate to reprogram fibroblast cells to blood vessel forming endothelial-like cells, we performed qRT-PCR to assay to examine the expression of pluripotency gene OCT4, and EC genes VEGFR-2/FLK1 and LPP3. Control (d2) cells showed insignificant levels of OCT4, VEGFR-2/FLK1, and LPP3, whereas expression of these gene increased significantly in d3 and d4 (Aza + TSA), and d5 (Aza + TSA + TDG) cells ( Fig 1C). Importantly, expression of the EC genes was not detectable in d2 cells, but increased significantly in d3, d4, and d5 cells ( Fig 1C). However, expression of the 18S RNA control remained unchanged in all conditions ( Fig 1C). Although Sox2, Nanog Chromatin modification mediate the expression of β-catenin, OCT4, VEGFR-2/FLK1, Brachyury, N-cadherin, TIE-2, CD31, and LPP3 proteins in hADFCs To address the hypothesis that epigenetic modification might be sufficient to convert hADFCs to immature or endothelial-like cells, we performed Western blot analysis. Cell extracts prepared from these cells showed basal levels of β-catenin in d2 control cells (Fig 2A). However, the level of β-catenin increased in d3, d4, and d5 cells (Fig 2A). We observed a very high level of β-catenin at d5 cells. Because of the inhibition of GSK-3β, d5 cell extract also likely contains non-phosphorylated (i.e., stabilized) β-catenin protein, so the increase in levels may indicate enhanced expression, stabilization, or both. TDG alone did not increase the level of β-catenin polypeptide species in control hADFCs (S4 Fig). In d2 control cells, OCT4, VEGFR-2/FLK1, Brachyury, N-cadherin, TIE-2, CD31, and LPP3 proteins were not detectable (Fig 2B-2H). However, we observed significant increases in OCT4, VEGFR-2/FLK1, Brachyury, N-cadherin, TIE-2, CD31, and LPP3 proteins in d3 and d4 cells, and the highest levels were observed in d5 cells (Fig 2B-2H). To determine if VEGFR-2/FLK1 protein was localized at the cell surface, these cells were subjected to cell surface labeling with anti-VEGFR-2/FLK1 antibody and analyzed by Fluorescence activated cell sorting (FACS) assay (S5 Fig). As expected, cell surface expression of VEGFR-2/FLK1 was increased steadily from d3-d5, while isotype matched control did not (S5A- S5D Fig). We also detected expression of mesodermal transcription factor Brachyury ( Fig 2D). Although VE-cadherin was undetectable (S6 Fig), on d4 and d5 the modified cells expressed N-cadherin ( Fig 2E). Moreover, the expression of endothelial cell markers TIE-2 ( Fig 2F), CD31 (Fig 2G), and LPP3 ( Fig 2H) were increased. In contrast, Western blot analyses failed to detect the expression of Nanog, KLF4 and Sox2 proteins in these cell lysates (S3 Fig). We used an anti-tubulin (Fig 2I) antibody as well as Ponceau-S staining (Fig 2J) of the nitrocellulose membrane to confirm equal loading of proteins across the lanes. These data show that chromatin modification can mediate expression of β-catenin, OCT4, VEGFR-2/ FLK-1, Brachyury, N-cadherin, TIE-2, CD31 and LPP3 proteins in hADFCs. untreated cells; (b) day-3 (d3) dish received one dose of Aza + TSA for 24 hours before the cells were used; (c) day-4 (d4) received a second dose of Aza + TSA for 24 hours before use; (d) day-5 (d5) received a third dose of Aza + TSA and TDG 24 hours before use. C) Total mRNAs prepared from these cells were subjected to q-RT-PCR using gene-specific primers for 18S RNA (control), VEGFR-2/FLK1, OCT4, and LPP3. Experiments were carried out three times, with triplicate mRNAs. * p<0.05 or ** p<0.01 vs. control 18S RNA from same group. In epigenetically modified fibroblasts β-catenin mostly localizes to the cell-cell junctions, while OCT4 localizes to the nucleus Next, we used epifluorescence microscopic imaging to examine the distribution of β-catenin, OCT4, N-cadherin, and von Willebrand factor (vWF) in control hADFCs and those treated with Aza and TSA with or without TDG as described in Fig 1A. β-catenin was diffusedly distributed in d2 and d3 cells, while its localization increased in cell-cell junctions in d4 and d5 cells (Fig 3A-3D). We observed OCT4 staining primarily in the nucleus (Fig 3E and 3F, green), with some localization in the cytoplasm in d3, d4, and d5 cells (Fig 3E and 3F, green, arrows). Costaining of d5 cells with DAPI showed increased OCT4 in the nucleus (Fig 3I-3L, magenta arrows). D5 cells had increased OCT4 compared to d4 cells. These cells were also stained with phalloidin-TRITC (red) to visualize actin cytoskeleton (Fig 3A-3L). In these cells, VE-cadherin was not detectable (S7 Fig), while N-cadherin was diffusedly distributed throughout the cells and OCT4 in the nucleus (Fig 3M-3P and S8 Fig). The normal staining pattern and the presence of actin stress fibers indicated that these cells appeared to be adherent phenotype, rather than anchorage-independent cancer cells. Importantly, d3, d4 and d5 cells expressed vWF (Fig 3Q-3T). Together, these data confirm that OCT4 was expressed in epigenetically modified fibroblast cells and reveal that it does translocate to the nucleus. In addition, at d3, d4, and d5 cells also expressed endothelial markers N-cadherin, TIE-2, CD31, LPP3 and vWF proteins (Fig 2). Fibroblasts epigenetically modified into OCT4+ and VEGFR-2+ cells OCT4 binds to the VEGFR-2/FLK1 promoter/enhancer region and transcriptionally activates the expression of VEGFR-2/FLK1 in epigenetically modified cells Because we found high levels of OCT4 expression in epigenetically modified cells (Figs 1 and 2), and OCT4 regulates transcription of critical pluripotency/stemness genes, we examined if OCT4 regulated gene expression in these hADFCs. Accordingly, we performed the ChIP assay to identify sites bound by OCT4 in the VEGFR-2/FLK1-promoter/enhancer DNA region (Fig 4A and  S9 Fig). DNA sequence analysis identified five putative OCT4-binding sites (G/CA/TNATGC) within the 860-bp genomic segment upstream of the VEGFR-2/FLK1 transcription start site ( Fig  4B). Using the ChIP assay, we did not observe any binding of OCT4 to the VEGFR-2/FLK1 promoter/enhancer DNA segment in d2 (control) or d3 cells (Fig 4C, lanes a and b). In contrast, OCT4 bound to the VEGFR-2/FLK1 promoter/enhancer DNA segment in d4 and d5 cells ( Fig  4C, lanes c and d), while OCT4 did not bind to the LPP3 promoter/enhancer (S10 Fig). To address the relationship of OCT4 binding to the VEGFR-2/FLK1-promoter/enhancer DNA segments, we knocked-down OCT4 using OCT4-shRNA and proteins extracts were analyzed by Western blotting (Fig 4D and 4E). The timeline of Aza + TSA treatment and OCT4-shRNAmediated knockdown in hADFCs is shown (Fig 4D). Accordingly, OCT4-knockdown of d2 or d3 hADFCs decreased the expression of OCT4 and VEGFR-2/FLK1 proteins. These data show that OCT4 binds to and transcriptionally activates the expression of the VEGFR-2/FLK1 gene.

Epigenetically modified fibroblast cells form capillary tube-like structures
Next, we subjected hADFCs onto 2D Matrigel assay and examined whether the epigenetically modified hADFCs behaved as ECs. Only ECs are known to form branching capillary tube-like structures in 2D gel [66,68]. The Matrigel assay is considered an in vitro correlate of angiogenesis [66,68]. Thus, to assess the potential of our epigenetically modified cells to form branch point structures in vitro, we plated these cells onto growth factor reduced 2D Matrigel and quantified the formation of branch points (Fig 5A). Plating of control fibroblast cells (d2) did not result in any network-like arrangements (Fig 5A and 5B). In contrast, d3, d4, and d5 cells formed branching, network-like structures. Arrowheads indicate branch points (Fig 5B-5E). To characterize and identify the tube forming cells, d5 Matrigels were embedded in paraffin, sectioned and stained either with VEGFR-2/FLK1 and vWF (Fig 5F-5I), or VEGFR-2/FLK1 with OCT4 (Fig 5J-5M). We show that d5 cells in 2D Matrigel not only formed branching points, but also retained expression of OCT4 and VEGFR-2/FLK1 proteins. In the absence of VEGF, however, these cells did not form branch point structures (S11 Fig). Thus, epigenetically modified fibroblasts cells transitioned into endothelial-like cells and the increase in VEGF receptor expression allowed the cells to respond to VEGF to form capillary-like structures.
We treated the fibroblast cells with optimized doses of Aza and TSA for 3 days. If the cells were treated with these chemical compounds only once over 2 days, there was no significant amplify the~860-bp PCR product upstream of the transcription start site (TSS) is shown. B) Primers used for amplifying OCT4 binding region in VEGFR2/FLK1 promoter. C) Representative ethidium bromide (EtBr) stained agarose gel showing PCR product obtained from anti-OCT4 ChIP experiment using chromatin from a) d2 control untreated cells; b) d3 cells treated once with Aza + TSA; c) d4 cells treated twice with Aza + TSA; d) d5, treated with a third dose of Aza + TSA and TDG. D) Strategy, timeline of Aza + TSA treatment and OCT4-knockdown. E) Total cell extracts prepared from: d2 control hADFCs; d2 cells receiving Aza and TSA; d3 cells receiving second dose of Aza and TSA; d2/kd, d2 cells receiving Aza and TSA were subjected to OCT4-shRNA-knockdown (d2/kd); d3 cells receiving second dose of Aza and TSA were subjected to OCT4-shRNAknockdown (d3/kd) were subjected to WB with indicated antibodies. GAPDH was used to judge equal loading of proteins across the lanes. The numbers below the OCT4 and VEGFR-2 WB panels represent WB signal intensities. Experiments were repeated three times.  induction of OCT4. However, if the cells were allowed to grow in the presence of these two compounds (Aza and TSA) and let these cells cycle for at least one round, before an additional stimulation with Aza and TSA, the cells expressed higher levels of β-catenin and OCT4. Importantly, if we let these cells cycle for two to three rounds (equivalent to d4 and d5 cells), the expression of OCT4, VEGFR-2/FLK1, Brachyury, N-cadherin, TIE-2, CD31 and LPP3 increased significantly. These results suggest that for effective epigenetic modification to occur, the cells must be cycling actively at least for 3 to 4 rounds. Thus, we followed the same experimental strategy and timeline (Fig 1A) throughout this study. In pilot experiments, we observed expression of OCT4, Sox2, Nanog, and Klf4 genes, however, their corresponding protein products were not detectable by Western blot. Although, the level of expression of OCT4 was increased at d4 and d5, these cells did not acquire a pluripotent stem cell state. In other words, the epigenetic change that occurred in these cells were incomplete, heterogeneous, or both. One could surmise that for these cells to become pluripotent stem cells there could be a threshold level of expression of OCT4 that is required to convert fully these cells to pluripotent stem cell state. Alternatively, there might be other unknown mechanisms that prevent OCT4 Fibroblasts epigenetically modified into OCT4+ and VEGFR-2+ cells from exerting its pluripotency function, e.g., by associating with a transcription repressor complex [70][71][72][73].
Fibroblasts are abundant in many organs throughout the body [74][75][76]. For instance, skin is a very good source of fibroblast cells [76]. Fibroblast cells can also be harvested from blood cells [76]. Literature show that human umbilical cords, which are often discarded in the operating rooms, can be used as a source of neonatal fibroblast cells. Fibroblast cells do not usually express key stem cell TFs such as OCT4 or Nanog. OCT4 expression is only found in stem cells and iPS cells. Accordingly, control hADFCs did not express OCT4 mRNA or protein; however, OCT4 and its effector VEGFR-2/FLK1 were expressed at very high levels in hADFCs treated twice with Aza + TSA as well as those treated additionally with all three compounds. These data indicate the ability of TSA and Aza to mediate the activation of OCT4 and VEGFR-2/ FLK1 genes. Because OCT4 is a stem cell TF, it likely binds to many promoter/enhancer sites to activate various genes. We also observed increased expression of Sox2, Nanog and KLF4 genes, but their corresponding protein products were not detectable by Western blot (S3 Fig). If not pluripotency, what is the significance of increased OCT4 expression? One could surmise that there might be more than one pool of OCT4 protein complexes in the nucleus, and some OCT4-protein could be sequestered by the transcriptional repressor proteins. Thus, the level of OCT4 protein that is transcriptionally competent in chromatin modified hADFCs may not be sufficiently high enough to exert pluripotency. We treated cells with TDG to activate Wnt signaling, because Wnt signaling is required for differentiation of ECs [20,40,44]. It appears that TDG induced the highest level of β-catenin accumulation and allowed sustained expression of OCT4 and VEGFR-2/FLK-1. Increased β-catenin level is likely driven by several transcription factors including canonical T-cell factor (TCF)/lymphoid enhancer factor (LEF-1) transcriptional network. In this regard, we currently do not fully understand the role of TDG. To examine whether these cells can transition into vascular ECs, we chose to focus on the VEGFR-2/FLK-1 receptor protein. Genetic studies have shown that without VEGFR-2/FLK-1, mouse embryos die in utero due to a lack of vascular ECs [40]. Accordingly, the binding of VEGF to VEGFR-2/FLK1 modulates several aspects of blood vessel formation, including EC development, migration, proliferation, survival, differentiation, and tube formation [20,40,42]. Thus, our data indicate that epigenetic modification of fibroblasts can lead to the formation of OCT4+ and VEGFR-2+ cells in vitro.
We examined levels of VEGFR-2/FLK1 to address whether fibroblast cells can acquire a mesodermal phenotype, because VEGFR-2/FLK1+ mesodermal cells are known to give rise to ECs, smooth muscle cells, cardiomyocytes, and hematopoietic stem cells. VEGFR-2/FLK1 expression, which was sustained in our modified hADFCs, is necessary for EC identity and required to enable ECs to form blood vessels during development and wound healing [20,40,42]. VEGFR-2/ FLK1 is considered a bona fide marker for mesodermal cells, ECs, and cardiomyocyte cells. We acknowledge that in epigenetically modified fibroblast cells, VEGFR-2/FLK1+ cells may include several types of mesodermal precursor cells rather than ECs alone. However, staining for β-catenin demonstrated the ability of these cells to form cell-cell junctions, much like ECs. These results indicate that we were able to partially remove "epigenetic memory" in these cells, which resulted in the expression of OCT4, VEGFR-2/FLK1, and LPP3. As these cells also expressed N-cadherin, CD31, TIE-2, LPP3 and vWF, and formed tube like-structures, we posit that epigenetic reprogramming might be one of the ways to produce vascular endothelial cells from fibroblasts. Complete reprogramming of somatic cells into iPS cells require a genome-wide "relaxed" chromatin state, characterized by Histone-H3 trimethyl Lysine-4 (H3K4me3) enrichment and decreased methylation epigenome. This event concurrently induces re-expression of OCT4, SOX2, and KLF4 stemness genes, preceded by gene promoter demethylation. Currently it is unknown, if only a subset of hADFCs might be epigenetically modified. Nevertheless, our results are important and significant, as they show the potential of Aza, TSA, and TDG to reprogram fibroblast cells into immature cells without the use of viral vectors. Importantly, the capacity to reprogram a large number of functional ECs from fibroblast cells, and the potential of these ECs to re-vascularize ischemic cardiovascular organs, will be of significant value in regenerative medicine.

Conclusion
Here, we showed that hADFCs can be epigenetically modified to express high levels of OCT4, VEGFR-2/FLK-1, N-cadherin, TIE-2, CD31, LPP3 and vWF proteins, and these cells formed capillary tube-like structures in 2D gel. Their ability to form vascular capillary-like structures highlights the potential of autologous modified endothelial-like cells to form new blood vessels, suggesting they might aid in neovascularization and rescuing ischemic cardiovascular tissues. However, further studies will be required to address if modified cells form blood vessels in clinically relevant animal experiments. (2 x 10 5 ) were detached non-enzymatically from culture dishes, neutralized by washing twice with 1x PBS, incubated with isotype matched control IgG (2.0μg/ml) or with anti-VEGFR-2/ FLK1 antibody, thereafter incubated with donkey anti-mouse IgG conjugated to Fluorescein isothiocyanate (FITC). (EPS) S6 Fig. VE-cadherin is not detectable in control d2, and chromatin modified d3, d4, and d5 hADFCs. A) Equal amount of total proteins prepared from d2, d3, d4, d5, and control human umbilical vein endothelial cells (HUVECs) were separated by SDS-PAGE, thereater analyzed by Western blotting with indicated antibodies. The membrane was intentionally overexposed to reveal minor non-specific signals present in d4 and d5 lanes. The fast moving anti-VE-cadherin antibody reactive polypeptide species are likely non-specific signals. B) The nitrocellulose membrane was stripped and reprobed with anti-GAPDH to estimate equal loading of proteins across the lanes. The Molecular weights are given in kiloDalton (kDa). nih.gov/; and (American Heart Association, http://www.heart.org/HEARTORG/; GRNT25710129) and by the University of Illinois at Chicago Center for Clinical and Translation Science Award (UL1RR029879) to K.K.W. J.B was supported by AHA pre-doctoral fellowship Mid-West affiliate (15PRE22760004) and CCTS-PECTs (1UL1TR002003). V.M. was supported by NIH-T32-HL07829 pre-doctoral training grant from the NIH. The UIC, RRC Core Imaging Facility, at the University of Illinois Chicago. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.