Close this search box.

Direct conversion of cardiac fibroblasts into endothelial-like cells using Sox17 and Erg – Nature Communications


All mouse procedures and usage follow the ethical guidelines of and were approved by the Institutional Animal Care and Use Committee of the University of North Carolina in Chapel Hill. Neonatal and adult cells isolated and used in the described methods are isolated from C57BL/6J mice (Jackson Strain#000664) or Tcf21-CreERT225 bred with Rosa26-tdTomato (#007914). SCID mice (Charles River #236) are used in the in vivo cell engraftment assay. Mice are housed in a facility maintained at an average of 70 °F and 50% humidity at a 12 h light cycle.

Standard 5-day tamoxifen injection protocol (IP injection of 75 mg/kg in corn oil) was used to induce Tcf21CreERT2 in adult Tcf21CreERT2 TdTomato mice (6–12 weeks old). Mice underwent surgery one to 2 weeks after the fifth injection of tamoxifen. The GFP and SEG cohorts contain mice that underwent tamoxifen injection and surgery at the same points to ensure consistency between groups.

Myocardial infarction surgery and injection of samples

Mice were first anesthetized with a combination of ketamine (100 mg/kg) and xylazine (10 mg/kg), and then intubated and ventilated using RoVent Jr. Standard Ventilator. The heart was accessed by cutting the intercostal muscle. Once exposed, the left anterior descending artery was ligated with a 7-0 silk suture. Immediately after ligation, 20 µL of iECs suspended in sterile PBS (1 million cells per 20 µL), 20 µL of sterile PBS, or 20 µL of retrovirus (GFP or SEG) is injected in three sites surrounding the infarcted area using insulin syringes with 29 G needles (BD #305935). The chest of the animal is then sutured closed, and the animal recovers on a 37 °C heating pad.

To prepare the retrovirus for injection, the virus was harvested from platE cells, filtered, and then ultracentrifuged (23,000 × g for 2 h). Resulting pellet was then resuspended in PBS with polybrene (8 μg/mL).

Four hearts were isolated for each experimental group (PBS control group or iEC cells) seven days after surgery. Each experiment group contains hearts from 2 males and 2 females. Tissue isolated from the PBS group was used to confirm the absence of GFP signal during immunofluorescence staining. Five hearts were isolated for each experimental group (PBS control or SEG iECs) containing males and females for the 4-week study and retroviral injection.

Lectin perfusion and analysis of myocardial infarctions

Mice were sacrificed using approved UNC IACUC methods. The heart was accessed by opening the chest cavity with scissors. The right atria was then cut, and the heart was then slowly perfused with 6 mL of PBS through a 27 G needle. The heart was then perfused with 6 mL of 10 µg/mL of DyLight 594 tomato lectin (Vector DL-1177-1) and left to incubate for 3 min. After incubation, the heart is then perfused with 8 mL of 4% paraformaldehyde and then excised into a 15 mL tube that contains 4% paraformaldehyde. Samples were sectioned at a 50-micron thickness. Three sections that were at least 100 microns apart were imaged per mouse. The myocardial scar region per section was imaged using Zeiss LSM900 using both Z-stack and tile scans. The myocardial scar region was identified based on morphology as well as the differential autofluorescence when compared to normal cardiac regions in the GFP channel. The resulting images were then flattened to a maximum intensity image and then randomly assigned a new name for blinded quantification. The resulting images were then quantified in a blinded manner using the FIJI polygon selection tool in combination with the region of interest manager to first identify the scar area and then trace the outlines of the perfused vessels. The total area of traced lectin vessels found in the section’s scar area was then divided by the section’s total scar area to get the percent lectin perfused scar area. For normalization of lectin perfusion to the non-scar area, representative images of the scar area for two sections per animal were collected along with a representative non-scar left ventricle region with normal lectin perfusion for each collected section. Percent lectin volume was calculated by dividing the volume of lectin measured by Imaris surfaces by the volume of tissue in the image (measured using the autofluorescence in the GFP channel).

Imaging and analysis of day 7 Tcf21 TdTomato lineage tracing

At Day 7 post-MI surgery, mouse hearts are harvested, perfusion fixed with 4% paraformaldehyde, and sucrose dehydrated, and then cryosectioned at 50 microns. Sections are then stained for GFP and PECAM1 using the immunofluorescence protocol stated below. Three sections containing GFP-positive cells were then imaged per heart, and all GFP-positive TdTomato-positive cells found in the section were imaged as Z-stacks using Zeiss LSM900. A maximum intensity image of the Z-stack image was then generated using ImageJ for downstream analysis. Marker-positive cells were then counted using QuPath multiplex analysis based on the classification of cells using DAPI, GFP, TdTomato, and PECAM126.

Neonatal cardiac fibroblast and endothelial cell isolation

In brief, hearts of neonatal mice (Day 0–Day 3) are isolated, pooled, minced, and then transferred to a 50 mL conical tube. 10 mL of 0.05% Trypsin is then added to the minced tissue, and the tube is then incubated at 37 °C for 10 min. The trypsin is then removed and 5 mL of 0.5 mg/mL Collagenase Type 2 (Worthington #LS004177) in HBSS is added to the tissue. After a 5 min incubation at 37 °C, the tube is vortexed, and the supernatant is removed and combined with 5 mL of complete media (DMEM + 10% FBS + 1% Penicillin/Strepavidin + 1% NEAA). The collagenase step is then repeated four more times, and the supernatant from each step is combined. The resulting solution of digested tissue is then passed through a 40 micron mesh sieve, and the flow through is then centrifuged for 5 minutes at 300 × g. If endothelial cell depletion is performed, the resulting cell pellet is resuspended in MACS buffer (0.5% BSA 2 mM EDTA in PBS), and Biotin-tagged PECAM1 antibody (BD #553371) is added to the resuspended cells (the amount of PECAM1 antibody scales with the number of isolated hearts). After a 30 min incubation at 4 °C on a rotator, the cell suspension is washed with 4 mL of MACS buffer and centrifuged for 5 min at 300 × g. The pellet is then resuspended in MACS buffer and Anti-biotin microbeads (Milltenyi Biotec #130-090-485) are added to the cell solution. After a 30 min incubation at 4 °C on a rotator, the cell solution is then run through a magnetic bead column (Milltenyi Biotec #130-042-401) and the flow through is collected. For endothelial cell depletion, the column is then washed with 4 mL of MACS buffer. Fibroblast isolation then follows endothelial depletion using the same methodology except for the magnetic column step, and the use of Biotin-labeled THY1 antibody (Thermo Fisher #13-0902-85). When added to the magnetic column, the initial and three 1 mL wash flow-throughs are discarded. After the wash steps, the cells still attached to the column beads are flushed with the provided plunger and then centrifuged for 5 min at 300 × g. The pelleted cells are then counted and resuspended in fibroblast media (DMEM + 20% FBS + 1% Penicillin/Strepavadin + 1% NEAA). If the depleted endothelial cells are used, they are flushed from the depletion column, pelleted, and then resuspended in EGM2 media (Lonza #CC-3162). All cells are plated on collagen-coated plates.

Neonatal cardiomyocyte isolation

This procedure follows the same initial steps as fibroblast isolation. After the collagenase digestion, the digested cells are passed through a 70 micron mesh cell strainer. The filtered cells are then centrifuged at 100 × g for 5 min. The supernatant is then removed, and the pellet is resuspended in complete media. The resuspended cells are then plated on a 0.1% gelatin plate and cultured for 90 min. The plated cell suspension is then removed and centrifuged for 5 minutes at 100 × g. The pelleted cells are then counted and then mixed with iECs at a ratio of 1 to 1.

Adult fibroblast isolation

Adult tissues (heart, lung or skeletal muscle) are isolated from 3 to 4-month-old adult animals and washed in PBS. The washed tissue is then minced and digested in collagenase solution (1 mg/mL Collagenase Type 2 + 1 mg/mL Collagenase Type 4 (Worthington #LS0004189) in HBSS) for 30 min at 37 °C and is vortexed for 30 s every 5 min. This solution is then passed through a 40 micron mesh sieve and then centrifuged for 10 minutes at 400 × g. The supernatant is then removed and the pellet is resuspended in 10 mL of ACK solution (Thermo Fisher #A1049201) and incubated at room temperature for 5 min. The cells are then centrifuged for 10 min at 400 × g. The pelleted cells are then resuspended in fibroblast culture media and cultured on 0.1% gelatin-collated plates for 3 h. The culture media is then removed, and the attached cells are washed with PBS twice, and new fibroblast media is added. On the next day, the media is then changed. The cells are monitored daily and then undergo MACS isolation (detailed above) when confluent, usually 4 days post-isolation.

Retroviral constructs and infection of fibroblasts

Eight of the ten tested retrovirus constructs were cloned into the pMX-puromyocin backbone. pMX-Tal1 (#131601) and pMX-Klf2 (#50786) were purchased from Addgene.

platE (Cell Biolabs #RV-101) cells were used to create the appropriate retroviruses for reprogramming with Nanofect (Alstembio #NF100) or Lipofectamine 2000 (Thermo Fisher #11668019) used to transfect the platE cells using the manufacturer recommended Nanofect or Lipofectamine protocols that has been previously described27. Retrovirus collected on Day 2 and Day 3 post-transfection are filtered using 0.45 micron PES syringe filtered, pooled, pelleted, and resuspended in neonatal fibroblast media with polybrene (EMD Millipore #TR-1003-G). To control for batch effects, the viruses used in compared conditions are always produced on the same day using the same transfection and infection reagents.

Immunofluorescence staining of tissue and cultured cells

All tissue is fixed in 4% paraformaldehyde and then washed with PBS. Heart tissue is then dehydrated in 15% and then 30% sucrose and then frozen in O.C.T. Compound. 10 or 50 micron sections are then sectioned from the frozen tissue using a cryostat.

Tissue sections and cultured cells are stained using the methodology using a previously published protocol. Samples are first blocked with 1% normal donkey serum in TSP (0.5% Triton X-100 and 0.1% Saponin in PBS) for one hour at 37 °C. Primary antibody solutions are then prepared using fresh blocking solution, and then the samples are incubated in primary antibody solution for one hour at 37 °C. Samples are then washed with TSP three times for 5 min each. Secondary antibody solution is then prepared using fresh blocking solution with secondary antibodies added at 1:200. Secondary antibody solution is then added, and samples are incubated for one hour at 37 °C. Samples are again washed three times for five minutes each. Samples are then stained with Hoescht to visualize nuclei and then washed one more time. PBS is then added to cultured cell samples. Tissue samples are then mounted with Prolong Diamond (#P36961) and covered with a coverslip.

Two-dimensional images were imaged using an EVOS7000 microscope. Three-dimensional images were obtained using Zeiss LSM900 at 1024 × 1024 resolution. Three-dimensional images were analyzed using Imaris and display images were deconvolved using AutoQuant.


anti-CD31 (1:100, R&D #AF3628); anti-CDH5 (1:100, R&D #AF1002); anti-GFP (1:500, Aves #GFP-1020). Donkey anti-Goat 568 (1:200, Thermo Fisher #A-11057); Donkey anti-Chicken 647(1:200, Jackson Immunoresearch (#703-605-155); Donkey anti-Chicken 488 (1:200, Jackson Immunoresearch #703-545-155). All primary antibodies were validated with controls before usage.

Ac-LDL uptake and NO production

Ac-LDL uptake studies were performed using Cell Applications Dil-Ac-LDL Kit and protocol. Cells were treated with Dil-Ac-LDL for 4 h at 10 µg/mL in EGM2 media. Cells were washed with EGM2 media, and then nuclei were stained with Hoescht. Samples were immediately imaged using identical microscope settings during the same session and then quantified using CellProfiler using the same object identification parameters.

Nitric oxide production was detected using DAF-FM diacetate (AAT Bioquest #16298). Cells were treated with DAF-FM at a concentration of 5 mM in EGM2 media for 30 min. Cells were then rinsed with media and then incubated for an additional 15 min with Hoescht. Cells were again washed with media and then imaged using the identical microscope settings. The average pixel intensity was then measured for each image using ImageJ and then the values were normalized to the average pixel intensity of the control samples.

TNFalpha stimulation and THP1 binding assays

Cells were treated with recombinant mouse TNFalpha at a concentration of 10 ng/mL in EGM2 media for four hours; untreated cells just received fresh EGM2 media. If collected for qPCR analysis, cells were then washed with PBS and then RNA was isolated using TRIzol extraction (Ambion #15596018). cDNA was then created from isolated RNA and then qPCR was performed on created cDNA for targeted genes.

THP1 cells were fluorescently labeled with CellTrace Far Red (Thermo Fisher#C34572) following the provided packaging instructions. After CellTrace labeling, cells were washed in THP1 culturing media (RPMI1640 + 10% FBS + 0.05 mM 2-mercaptoethanol), centrifuged at 300 × g for 5 min, and then resuspended in EGM2 media. During the THP1 washing steps, TNF-alpha-treated cells are washed with EGM2 media and labeled with Hoescht dye in EGM2 media for 5 min. The labeled THP1 cells are then added to TNF-alpha-treated cells at a concentration of five hundred thousand cells per well for a 24-well plate. After a 30 min incubation, all wells are washed with EGM2 media three times. All wells are then imaged using identical microscopy settings. The number of TNF-alpha (Hoescht-positive) and THP1 (CellTrace-positive) cells per field was quantified using CellProfiler. The total number of THP1 and Hoescht-positive cells for each replicate was totaled for all of the collected fields and used to obtain the ratio of THP1 cells per nuclei in order to take into account potential differences of TNF-alpha stimulated cells per condition. The TNF-alpha stimulated ratios were then normalized to the appropriate non-stimulated condition.

CellProfiler quantification of images

CellProfiler was used in the unbiased quantification of the number of cells positive for particular cell markers. When comparing different conditions or reprogramming cocktails, all samples collected for the same replicate were analyzed using identical settings that take into account signal strength and background signal to control for batch effects. All images within a replicate are acquired using identical microscope settings. A similar pipeline was used in the quantification of the number of cells positive for a particular marker. In brief, the nuclei are first identified using IdentifyPrimaryObject module. Next, the protein of interest is identified using IdentifyPrimaryObject module. The settings of this module are set by performing test images on at least three images that are randomly selected by CellProfiler. The identified nuclei are then masked by the identified marker objects using the MaskObject module, and a majority of the identified nuclei must be masked by the marker objects in order to be counted as a nucleus that is positive for the marker of interest. At least 10 adjacent images are analyzed by this pipeline per replicate of a particular test condition.

Flow cytometry analysis

Cells are detached from the cell culture plate using Accutase (#A6964-100 mL), and an equal volume of EGM media is added to the detached cells. The cells are then centrifuged at 300 × g for 5 min, and the pelleted cells are resuspended in 100 µL of FACS Buffer (0.5% BSA in PBS) with 10 µL of APC-conjugated PECAM1 antibody (R&D FAB3628A) or 5 µL of PE-conjugated CDH5 antibody (Thermo Fisher #12-1441-82). Cells are incubated in antibody solution for 30 min at 4 °C. The samples are then washed with 1 mL of FACS buffer and centrifuged at 300 × g for 5 min (repeated three times). After the last wash, the cells are filtered through a 40 micron strainer, briefly treated with DAPI to label dead cells, and then analyzed using Thermo Fisher Atunne NxT. Collected data is then analyzed using FlowJo (Supplementary Fig. 14 for gating strategy).

Shear stress flow experiments

Day 6 iECs were detached from well plates using accutase, centrifuged, and then were seeded in individual wells of Ibidi μ-Slide VI 0.4. On day 7, cells were washed with EGM2 media, and then their media was changed to flow media (EBM2 media with 2% FBS and 1% Penicillin/Strepavadin). Ibidi μ-Slides were attached to Masterflex Ismatec Reglo Digital Pump. The flow samples were slowly ramped up to final shear stress condition using the following protocol: 0.5 dynes/cm2 (15 min), 1 dyne/cm2 (15 minu), 3 dynes/cm2 (15 min), 5 dynes/cm2 (30 min), 10 dynes/cm2 (30 min), 15 dynes/cm2 (30 min), 20 dynes/cm2 (1 h), 25 dynes/cm2 (1 h), and 30 dynes/cm2 (1 h). Flow condition samples were then subjected to 35 dynes/cm2 of shear stress for 24 h using a pulsatile pump in a standard 37 °C cell incubator.

After flow conditions, samples were fixed using 4% PFA in PBS. The immunofluorescence protocol described above was then used to stain the samples for GFP and CDH5. After staining, the samples were imaged using EVOS7000 microscope. Images from each sample were then analyzed for alignment to flow using the Directionality plugin in FIJI on the GFP images (at least 9 images per sample). The Directionality of each image was calculated using the Local Gradient Orientation function, and the absolute value of the Direction of the GFP channel was recorded per image. The recorded values per sample were then averaged. The four replicates for each condition were then compared using unpaired t-test.

qPCR of TNFalpha-stimulated samples

After 4 h TNF-alpha stimulation, cells were washed with PBS, and then the RNA was extracted from the samples using standard TRIzol extraction protocol (Thermo Fisher #115596018). cDNA was then created for each sample using SuperScript IV Vilo Mastermix (Thermo Fisher #11756050). qPCR of samples was then performed on samples in duplicate using SYBR green master mix (Thermo Fisher #4309155). The average CT value of each gene was then taken and only used if values were within a half-cycle difference. Delta CT was then calculated using beta-actin as a reference gene. The change in expression due to TNF-alpha stimulation was calculated using an unstimulated sample.

qPCR primer sequences (5’–3’)














Bulk RNAseq sample collection and analysis

RNA from samples was isolated using standard TRIzol extraction protocol (Thermo Fisher #15596018). Libraries were prepared using the Kapa mRNA Stranded Kit. Samples were sequenced using paired-end sequencing at a 2 × 50 read length using an Illumina Novaseq 6000 platform by UNC HTSF core. Raw reads were demultiplexed using bcl2fastq. Samples were checked for quality control using fastqc and multiqc, and high GC content was trimmed using BBDuk. Transcript expression was then quantified using Salmon28. DESeq2 was used to obtain differential gene expression between control and reprogrammed samples at each timepoint29. Gene Ontology enrichment analysis and enrichment analysis of the differentially expressed genes to previously published datasets were performed using ClusterProfiler30.

Single-cell RNAseq library preparation and sequencing

Cells from two replicates for each condition were detached from cell culture plates using Accutase or 0.05% Trypsin and then centrifuged at 300 × g for 5 min at 4 °C. Replicates were then pooled and then the pooled replicates for each timepoint were multiplexed using the 3’ CellPlex Kit Set A (10x Genomics #1000261) with the Day 3, Day 7, 2 Weeks, and 4 weeks timepoints as separate samples. After the final wash step, the pooled multiplexed cells were immediately used in the Chromium Next GEM Single Cell 3’ Reagent Kit v3.1 (Dual Index) library preparation protocol (10x Genomics #1000268). The prepared gene expression and multiplexed libraries were then paired-end sequenced using NovaSeq 6000 platform by UNC HTSF core.

RNA was extracted from unused cells from scRNAseq sample preparation, and the overexpression of Sox17, Erg, and Etv2 was independently confirmed by qPCR before samples were further analyzed.

qPCR Primer Sequences (5’–3’):

Etv2 F cag agt cca gca ttc acc ac Etv2 R agg aat tgc cac agc tga at

Sox17 F gaa tcc aac cag ccc act g Sox17 R tag gga aga ccc atc tcg gg

Erg F acc tca ccc ctc agt cca aa Erg R tgg tcg gtc cca gga tct g.

Single cell RNAseq data analysis

The cellranger multi pipeline was used to demultiplex the sequenced samples (using the preassigned min-assignment-confidence of 0.9). The standard Seurat workflow was then used to filter (nFeatures RNA > 200, and < 25), merge, and cluster the demultiplexed samples31. Once the clusters were assigned, FindAllMarkers was used to find the genes associated with each cluster with the specifications of min.pct above 25% and logfc.threshold of 0.25. ClusterProfiler was used for both GO Term Analysis and comparison to published gene sets30. Annotation of Tabula Muris heart data was performed by using CellID R-package32.

Pseudotime analysis was performed by using published Monocle3 pseudotime workflow by converting the Seurat object to the cell_data_set object function found in SeuratWrappers33. Sample-specific trajectories were isolated using choose_cell function and gene modules describing the trajectories were identified using graph_test in combination with find_gene_modules functions (adapted from Monocle3 vignette workflow).

Cell cycle states were identified using tricycle R-package with the SeuratWrappers workflow34. Cell signaling for each cluster was estimated using CellChat35. Gene regulatory network analysis was performed using SCENIC36. In silico cell perturbation analysis was performed using CellOracle24.

H3K27ac CUT&Tag

Samples were detached from cell culture plates using Accutase, centrifuged, counted and then 100,000 cells were used in CUT&Tag V.3 protocol starting at the fresh cells step37. Samples were then prepared using the published CUT&Tag protocol using an H3K27ac antibody (Abcam #ab4729). Samples were then paired-end sequenced using NextSeq2000 platform by UNC HTSF core.

Data was analyzed following the published CUT&Tag data analysis protocol. The quality control statistics at each step of the analysis pipeline were routinely compared between replicates and sample types. Peaks were assigned using the SEACR package, which returned the top 0.01 fraction of stringent peaks per sample. Differential peaks were assigned by comparing the assigned peaks of each condition (Etv2, SEG, or primary endothelial cells) to the control condition using DESeq2 R-package29. Deeptools were used to generate peak heatmap for each sample38. HOMER package was used for transcription factor motif analysis of the differentially expressed peaks per sample type (p-value cut-off of 0.01 for Venn diagram comparison)39. The differentially expressed peaks were annotated using ChIPseeker R-packge, and ClusterProfiler was used for enrichment analysis of the annotated genes30,40.

Statistics and reproducibility

GraphPad Prism was used to perform all of the statistical analyses. Comparisons were evaluated using two-sided t-tests, the Mann–Whitney test (for non-normal data), or One-way ANOVA with Tukey’s multiple comparison tests with a p-value of 0.05 being the minimum cut-off for statistical significance. For the screening data, ratio-paired t-tests were used to take virus titer variability and differences in starting samples into account. For all other t-tests, an un-paired t-test with Welch’s correction was used. No statistical method was used to predetermine the sample size. No data were excluded from the analyses. Sample blinding was used for the analysis of in vivo data. When possible, all quantifications of immunofluorescence data were automated using CellProfiler or QuPath.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.