Induction of human hemogenesis in adult fibroblasts by defined factors and hematopoietic co-culture

Abstract

Transcription factor (TF)-based reprogramming of somatic tissues holds great promise for regenerative medicine. Previously we demonstrated that the TFs GATA2, GFI1B, and FOS convert mouse and human fibroblasts to hemogenic endothelial-like precursors that generate Hematopoietic Stem Progenitor (HSPC)-like cells over time. This conversion is lacking in robustness both in yield and biological function. Herein, we show that inclusion of GFI1 to the reprogramming cocktail significantly expands the HSPC-like population. AFT024 co-culture imparts functional potential to these cells and allows quantification of stem cell frequency. Altogether, we demonstrate an improved human hemogenic induction protocol that could provide a valuable human in vitro model of hematopoiesis for disease modeling and a platform for cell-based therapeutics.

INTRODUCTION

Hematopoietic stem cells (HSCs) are thought to generate all the cellular elements of the blood in a hierarchical manner [1], though recent work in this field suggests a more complicated process in the mouse system [2, 3]. Multiple studies demonstrate an endothelial origin for multipotent HSCs, notably showing their emergence from a specific subset of cells called hemogenic endothelium (HE) through a process of cell budding termed the endothelial to hematopoietic transition (EHT) [4–8]. Due to their ability to repopulate the entire hematopoietic system upon transplantation in both mice and humans, HSCs represent the currently established standard for stem cell therapy. The source material required for these applications, however, remains in limited supply because HSCs notoriously die or differentiate ex vivo [9]. To this end, several studies exist that employ different methods to expand these cells ex vivo or generate them de novo [10–13]. This issue hinders their use for a multitude of in vitro applications, such as drug testing platforms and disease modeling systems. Allogeneic transplants involve an additional hurdle, in that they carry multiple risks of graft-versus-host disease and graft rejection due to poor HLA matching and a lack of ethnic diversity for sufficient matching material [14].

A paradigm shift in stem cell biology emerged once Yamanaka and colleagues demonstrated that overexpression of a defined set of TFs could reprogram differentiated somatic cells to induced pluripotent stem cells (iPSCs) [15, 16]. Ectopic TF overexpression to alter cell identity translated to the field of hematopoiesis; multiple studies using different starting mouse or human cell populations, TF combinations, or culture conditions to obtain various types of in vitro derived blood products de novo [11, 12]. Several of these studies focused on using either pluripotent [17–24] or somatic [25–30] cells with varying levels of success. Encouragingly, recent studies obtained transplantable cells using either human iPSCs [31] or human/mouse endothelial cells with a vascular niche co-culture system [32, 33].

Previously we demonstrated that overexpression of TFs – Gata2, Gfi1B, and Fos (GGF) – was sufficient to induce a hemogenic program in mouse embryonic fibroblasts (MEFs) while the addition of Etv6 increased efficiency. Together these factors produce hematopoietic cells through a process that mimics developmental hematopoiesis. During reprogramming the transduced cells appear to traverse from an endothelial precursor and then undergo an EHT to yield clonogenic progenitors [29]. The endothelial precursor cells bore a Prom1+, Sca1+, CD34+, CD45- cell surface phenotype that we also found in cells of the aorta-gonad-mesonephros (AGM) and placenta. These cells gave rise to engraftable hematopoietic cells after maturation demonstrating that in vitro reprogramming can inform normal hematopoiesis [34]. Recent data demonstrates that GGF overexpression in iPSCs that are then used to form teratomas leads to an increased production of long-term HSCs [35]. We have transferred these findings to the human system and found that GGF reprogrammed fibroblasts appear to undergo the same type of process and can develop into short term repopulating cells [30]. The yield and overall in vitro/ex vivo function of these human derived cells, however, requires further improvement.

In this study, we sought to optimize our hemogenic induction process in adult human fibroblasts to produce greater yields of functional HSC-like cells. To this end we identified an additional TF – GFI1 – that expands the yield of functional hematopoietic progenitors when in concert with GGF reprogramming (the TF cocktail is now termed 3GF for GATA2, GFI1, GFI1B and FOS). The cell surface phenotype of the derived cells correlates to that found on human HSCs. Gene expression profiles also show generation of hematopoietic cells through a hemogenic developmental process. AFT024 stromal co-cultures permitted the derivation and quantification of functional HSC-like cells capable of forming colonies composed of various hematopoietic lineages. Additionally, cells reprogrammed with the new cocktail function at a much earlier timepoint as seen through our in vitro assays and short term multilineage engraftment. Altogether we demonstrate an enhanced human hemogenic induction strategy by using an additional TF as well as an in vitro maturation system.

METHODS

Human Dermal Fibroblast, AFT024, and 293T Cell Culture

Human dermal fibroblasts (HDFs) used for all experiments were obtained from ScienCell. Cells were plated in 10cm tissue culture dishes in D10 media (Dulbecco’s Modified Eagle Medium; Thermo Fisher Scientific) containing 10% fetal bovine serum (FBS; Benchmark), 1mM L-Glutamine and penicillin/streptomycin (P/S) (10 ug/ml; Thermo Fisher Scientific) in 37°C. 293T cells for viral production were also cultured in standard D10 media at 37°C. AFT024 cells used for long term culture (LTC) and limiting dilution analysis (LDA) experiments were cultured in D10 media supplemented with 50μM 2-ME at 32°C for expansion. The day prior to the experiments, AFT024 were mitotically inactivated via irradiation as previously described [36], and placed in 37°C.

Molecular Cloning, Lentivirus Production, and tdTomato (tdT)-HDF Generation

The coding regions of each candidate TF (Table S1A) were individually cloned into the pFUW-TetO vector where expression is controlled by the minimal CMV promoter and the tetracycline operator [37]. Lentiviral vectors carrying each of the chosen reprogramming factors were generated by calcium phosphate transfection into the 293T packaging cell line with a mixture of the viral plasmids of choice as well as the constructs that instruct viral packaging and the VSV-G protein (pMD2.G and psPAX2). For transgene activation, cells were co-transduced with lentiviral vectors containing the reverse tetracycline transactivator M2rtTA, which is controlled by the constitutively active human ubiquitin C promoter. Viral supernatants were collected 36, 48, and 60 hours after 293T transfections and then filtered (0.45μm) and stored in −80°C. Lentivirus carrying the pSin-tdTomato vector (constitutively driven by the EF2 promoter) was generated as described above and used to transduce low passage HDFs. The top 10% of tdTomato⁺ (tdT⁺) cells were sorted and cultured to establish the tdT-HDF line in D10 media.

Viral Transduction and Reprogramming Cell Culture

HDFs were transduced with a viral cocktail consisting of 33.33% D10 media, 33.33% viral supernatant containing M2rtTA, and the remaining 33.33% containing equal portions of each factor within the GGF or 3GF TF sets to ensure equal multiplicities of infection of each individual viral particle as well as 8μg/ml of Polybrene. Additional factors were tested in combination with GGF in the same manner (Table S1B). Control transductions with mOrange in pFUW resulted in >95% efficiency. HDFs on Day −1 were plated at a density of 1.5 × 10⁵ – 3.0 × 10⁵ on 0.1% gelatin coated 10cm dishes or 6-well plates with D10 media. After 24 hours HDFs were transduced 3 times every 12 hours over 24 hours (at 0, 12, and 24 hours). 12 hours after the final transduction, the media was switched to D10 supplemented with 1ug/ml doxycycline (DOX) to begin transgene activation. On Day 4, transduced HDFs were dissociated with trypLE Express (Thermo Fisher Scientific) and split 1:2 onto 0.1% gelatin coated 6-well or 12-well plates with Myelocult media (H5100; Stem Cell Technologies) supplemented with hydrocortisone (HC) (10⁻⁶ M; Stem Cell Technologies), the cytokines SCF, FLT3L, and TPO (all R&D systems, 25ng/ml as previously described [38]), 1ug/ml DOX, and 50μM 2-ME. Myelocult media was changed every 4 days for the duration of the cultures.

FACS Analysis and Sorting

Cells from standard reprogramming, Colony Forming Unit (CFU), or LTC experiments were first harvested using trypLE express at specified days and washed with PBS supplemented with 5% FBS and 1mM EDTA. Flow cytometry analysis was performed on a 5-laser LSRII with Diva software (BD Biosciences) and analyzed with FCS Express 6 Flow Research Edition (Win64). Cells were stained with PE/CY7-hCD45 (2D1), FITC-hCD235a (GA-R2), APC-hCD41 (MReg30), BV421-hCD14 (M5E2), BV421-hCD34 (581), APC-hCD45 or FITC-hCD45 (2D1), PE-hEPCR (RCR-401), or APC-hCD49f (GoH3) (all Biolegend), as well as PE-hACE (BB9), FITC-hCD90 (5E10), PE-hCD49f (GoH3) (BD Biosciences), APC-hCD90 (5E10, Thermo Fisher), or hACE-Biotin (BB9, R&D Scientific Corporation) with APC-Cy7 Streptavidin (BioLegend). 4,6-diamidino-2-phenylindole (DAPI, 1ug/mL, Sigma) or Propidium Iodide (PI, 1ug/ml, R17755, Invitrogen) was added prior to analysis to exclude dead cells. Sorting for RNA sequencing, transplants, LTC, and CFU assays was performed with APC-CD49f alone, PE-CD49f alone or BV421-CD34 and PE-CD49f using DAPI or PI to exclude dead cells.

Live imaging

Reprogrammed GGF and 3GF cells were analyzed by live staining at Days 14, 20, 28, and 35. Myelocult media was removed and 300ul of 1 × PBS with 5% FBS and incubated with PE-hEPCR 1:20 and incubated for 15 minutes at 37°C. The antibody mix was then aspirated, cells were washed with 1 × PBS with 5% FBS, their supplemented Myelocult media was replenished, and they were subsequently imaged on a Leica DMI 4000 B using Leica LAS AF software. For CFU live stains, colonies were collected and washed with 1 × PBS with 5% FBS. They were then resuspended in 200ul trypLE express, incubated at 37°C for 5 minutes, triturated, and washed again with 1 × PBS with 5% FBS. Cells were then resuspended in 200ul of 1 × PBS with 5% FBS, loaded into TC treated, sterile μ-Slide VI 0.4 ibiTreat chamber slides (Ibidi, #80606) and imaged on a Leica DMI6000 Inverted scope using Leica LAS AF software.

1° CFU, LTC-Initiating Cell (LTC-IC) and Cobblestone Area Forming Cell (CAFC) Assays

Reprogrammed tdT-HDFs were harvested with trypLE at Day 15, 20, and 25 of reprogramming, washed in 1 × PBS, suspended in 500ul of DMEM, and dispersed in 3ml of cytokine-supplemented methylcellulose (h4435, Stem Cell Technologies). The cell suspension was then drawn into an 18G needle with a 5ml syringe and plated 1ml per well of a non-TC treated 6-well plate (Costar, #3736). Empty spaces between the wells were filled with sterile H₂O and the plates were then incubated at 37°C in 5% CO₂. For LTC assays, 12-well plates were first coated with 0.1% gelatin. Expanded AFT024 stromal cells were harvested and seeded at a density of 3 × 10⁵ – 3.5 × 10⁵ cells/ml in D10 media supplemented with 50μM 2-ME. 1ml of the cell suspension was plated in each well of the gelatinized 12-well plates. Cells were then grown overnight at 32°C with 5% CO₂. The next day cells were irradiated with 2000 rads. 20 – 30k Day 15 CD49f⁺ sorted cells were then placed in each well with 4ml of supplemented Myelocult media (with previously described concentrations of HC, SCF, FLT3L, TPO, DOX and 2-ME and incubated at 37°C in 5% CO₂. Plates were then observed for colony growth and morphology, with weekly half-media changes for up to 5 weeks. For 1° CFU assays using Lin⁻CD34⁺ cord blood (CB) HSCs, 250 cells were plated per 1ml of methylcellulose (h4435, Stem Cell Technologies), observed over 2 weeks, and colony types/total colony numbers were counted. 2,000 Lin⁻CD34⁺ CB HSCs were plated per well of a 12-well plate for AFT024 LTC cultures as well.

Colony imaging, LTC-IC CFU, and Cytospins

Selected wells from the LTC assay for reprogrammed cells as well as Lin⁻CD34⁺ CB HSCs were harvested and plated in CFU assays as previously described. For LTC-IC CFU plating for CB HSCs from LTC assays, 1 well from these cultures was taken and separated into 90% (therefore representative of 1,800 of the initially seeded Lin⁻CD34⁺ CB HSCs) and 10% (200 initial cells) samples. Cobblestone and CFU colonies were imaged on a Leica DMI 4000 B automated inverted scope using Leica LAS AF software. After colony derivation in CFU from the LTC assays, colonies were collected in 1 × PBS supplemented with 5% FBS, washed, and resuspended in 200ul trypLE. Colonies were then incubated in trypLE in 37°C for 5 minutes, triterated to make single cell suspensions, washed, resuspended in 200ul 1 × PBS with 5% FBS, and loaded into cytospin prepared slides. Samples were spun at 250 rpm for 3 minutes, stained with Hematoxylin and Eosin (H&E), and then imaged on a Leica DM5500 upright scope using Leica LAS AF software.

AFT024 in vitro CAFC Limiting Dilution Analysis (LDA)

AFT024 stroma was cultured as previously described [36] and harvested to a concentration of 350,000 cells/ml in D10 supplemented with 50μM 2-ME. In 0.1% gelatin coated 96-well plates, 100μl of this suspension was plated and allowed to grow overnight at 32°C with 5% CO₂. The following day cells were irradiated and the media was replaced with 100μl of fresh supplemented Myelocult. On Day 15 of reprogramming, CD49f⁺ tdT-HDF cells reprogrammed with either GGF or 3GF were sorted and seeded in all 12 wells of Row A of the prepared 96-well plate with 20,000 cells in 100ul per well. Using a multichannel pipet, 100ul of the now 200ul cell suspension was taken to Row B, mixed with the 100μl already present in the wells, and then serially diluted down to Row H with the dilutions as follows: Row A: 10,000 cells per well; Row B: 5,000 cells; Row C: 2,500 cells; Row D: 1,250 cells; Row E: 625 cells; Row F: 312.5 cells; Row G: 156.25 cells; Row H: 78.125. 24 hours after seeding, 100μl of fresh supplemented Myelocult was added to each row. Half media changes were performed weekly, and wells with emerging cobblestone colonies were counted after 5 weeks of LTC. Stem cell frequency was then calculated using Poisson statistics [36] and extreme limiting dilution analysis (ELDA) [39]. The same process was used for LDA analysis of Lin⁻CD34⁺ CB HSCs, but instead cell densities started at 1,000 cells per well in Row A, 500 in Row B, 250 in Row C, 125 in Row D, 62.5 in Row E, 31.25 in Row F, 15.625 in Row G, and 7.8125 in Row H for 1° LDA. For 2° LDA, cell densities were 250 cells per well in Row A, 125 in Row B, 62.5 in Row C, 31.25 in Row D, 15.625 in Row E, 7.8125 in Row F, 3.90625 in Row G, and 1.953125 in Row H. CB Lin⁻CD34⁺ cells were also grown in supplemented Myelocult media, but without added DOX.

mRNA, cDNA, and library sample preparation

3GF reprogrammed HDFs were reprogrammed to Day 15 and D25 and subsequently sorted in triplicate for 2 populations at both time points: CD49f+CD34- and CD49f+CD34+. 10⁵ cells for CD49f+CD34- replicates and at least 3 × 10⁴ cells for CD49f+CD34+ replicates were sorted into 1 × PBS with 5% FBS, pelleted, and RNA was subsequently isolated using the NucleoSpin RNA XS extraction kit (Clontech, 740902.50). cDNA was synthesized and amplified using the SMART-seq v4 Ultra Low Input RNA Kit for Sequencing (Takara Bio USA, 634889). Amplified cDNA was then purified using the Agencourt AMPure XP Kit (Beckman Coulter, A63880). The concentration of the derived cDNA was quantified using a Qubit fluorometer. The quality of the derived cDNA samples was determined using the Agilent High Sensitivity DNA Kit (Agilent, 5067–4626) and an Agilent 2100 Bioanalyzer. cDNA libraries were then created using the Nextera XT DNA Library Preparation Kit (Illumina, FC-131–1024) and the Nextera XT Index Kit (Illumina, FC-131–1001) and subsequently sequenced on an Illumina HiSeq 4000 with 25M 100-nt reads per sample.

RNAseq Analysis

Reads were mapped to the human genome (GRCh37.75) using STAR to avoid high mapping error rates, low mapping speed, and read length limitation/mapping biases [40]. Reads mapping to annotated genes were counted using featureCounts [41]. Read count normalization and pairwise differential expression analyses between groups of samples were performed using DESeq2, which is used for differential analysis of count data (in our case normalized read counts) with shrinkage estimation for fold changes to improve the accuracy and stability of our results [42]. GAGE, a gene set analysis method that can manage datasets with different sample sizes or experimental designs, was used to perform gene set enrichment tests between pairwise groups of samples [43]. Enrichment tests were performed for custom gene sets extracted from prior literature and for gene sets associated with specific gene ontology (GO) terms [44, 45]. Samples were additionally analyzed using principal component analysis (PCA) and hierarchical clustering in R. DESeq2 normalized read counts for all relevant samples were plotted using Graphpad Prism 7 software. RNAseq data from Gomes et al [30] was used for comparative purposes. Comparisons of TFs up-regulated in 3GF cells as compared to GGF cells were done using Biojupies [46].

NSG Mouse Transplants

After sorting CD49f⁺ cells from 3GF reprogrammed HDFs cultured on 0.1% gelatin, cells were washed in 1 × PBS and then transplanted (3.0 × 10⁵ cells) into NOD-scid IL2Rg^null(NSG) 0–2 day old pups via intrahepatic injection. Mouse peripheral blood (PB) was analyzed 4, 8, and 16 weeks post-injection for engraftment of human-derived cells. To distinguish levels of engraftment cells were stained for mouse PacBlue-mCD45 (30-F11, Biolegend), and PE-Cy7-hCD45 (2D1, ebioscience). Within the hCD45 compartment we looked at cells stained with APC-eflour 780-hCD3 (UCHI1), PerCP-Cy5.5-hCD8 (RPA-T8), PE-hCD19 (HIB19) and Alexa488-hCD14 (all from Biolegend). Engraftment levels were compared to levels found from CB derived CD34⁺ hematopoietic progenitors isolated using Diamond CD34 Isolation Kits (Miltenyi) using the same injection method as previously described (1.0 × 10⁵ cells/pup).

Statistical analysis

Data was analyzed with Graphpad Prism 7 software using the nonparametric Mann-Whitney test for samples not assuming a normally distributed data set. Bars represent mean and error bars represent standard error of the mean (SEM). Statistically significant differences are as follows: * = p < .05, ** = p < .01, *** = p < .001, and **** = p < .0001. Limiting dilution frequencies were determined using Poisson statistics and ELDA software [39].

RESULTS

GFI1 added to the GGF cocktail expands the derived HSPC pool

We recently demonstrated that three TFs, GATA2, GFI1B, and FOS, were sufficient to induce a hemogenic program in human fibroblasts [30]. Transplantation of CD49f⁺ cells sorted at day 25 of reprogramming in immunocompromised mice resulted in short term multilineage engraftment. These results were encouraging but the overall yield and function of the reprogrammed cells needed improvement. We elected to ask if other TFs from previously published reprogramming strategies together with our factors would improve outcome. These factors were individually cloned into the doxycycline-inducible pFUW cassette used in our initial studies [37] (Table S1A) [17, 18, 23, 26, 29, 32–34, 47]. These TFs were arranged into 12 distinct combinations (Table S1B) and used in concert with GGF to induce hemogenesis in adult HDFs.

Reprogrammed cells were screened at day 30 to determine levels of CD34 induction, a known HSC marker [29, 48] (Figure 1A). We found that 3 TF combinations significantly improved conversion to CD34⁺ cells (Figure 1B). C2 contains the polycistronic STEMCCA pluripotency reprogramming cassette [49] and C10 contains a combination of shRNAs to p53, which has been shown to improve reprogramming efficiency upon repression of p53 [26, 50]. To avoid both reprogramming to pluripotency and altering the p53 network of the reprogrammed cells, these cocktails were not used further. C12, however, contained a group of TFs: FOSB, GFI1, RUNX1c, and SPI1 (FGRS), which enhanced GGF reprogramming without induction of pluripotency (Figure 1C). This cocktail of factors has been recently shown to successfully reprogram human and mouse endothelial cells to acquire a hematopoietic fate upon continuous co-culture with a vascular niche [32, 33].

Figure 1. GFI1 expands the CD34⁺ progenitor pool induced by GGF reprogramming.

(A) Scheme displaying the reprogramming process for the initial TF screen. SFT = SCF, Flt3L, and TPO (each at 25ng/ml). (B) Flow cytometric quantification of day 30 reprogrammed cells identifies 3 TF combinations that significantly expand the CD34⁺ progenitor population compared to GGF reprogrammed cells out of the 12 tested (n = 2 – 3). (C) Day 31 cellular morphology during reprogramming culture reveals a massive proliferation of round hematopoietic-like cells when GGF reprogramming is supplemented with FGRS. Expansion of the CD34⁺ and CD45⁺ cell pools are also observed in the representative flow cytometry dot plots. (D) Scheme for N+1 and N-1 experiments to identify the TF or combination of TFs from the FGRS cocktail that permits the observed CD34⁺ expansion. (E) Results from the N+1 and N-1 experiments reveal that GFI1 is the factor in the FGRS combination responsible for the CD34⁺ cell expansion (n = 6). (F) Representative flow plots of GGF and 3GF reprogrammed cells reveals the significant induction of CD34⁺ cells (Auto represents an empty channel). (G) Representative reprogramming culture images throughout GGF and 3GF reprogramming recapitulate the massive proliferation of round hematopoietic-like cells present when GFI1 is used with GGF. Data are represented as mean ± SEM. Nonparametric Mann-Whitney T-test for samples not assuming a normally distributed data set. *P < .05, **P < .01, ***P < .001.

N-1 or N+1 experiments using the GGF and FGRS cocktails (Figure 1D) revealed GFI1 as the factor that improves the yield of CD34⁺ cells as seen when GFI1 alone is added to the GGF cocktail, or when GFI1 is removed from the 7 factor combination (GGF+FGRS) (Figure 1E–F). Reprogramming with new reprogramming cocktail (now termed 3GF for GATA2, GFI1, GFI1B, and FOS) results in a significant expansion of rounded hematopoietic-like cells in culture as compared to GGF cells based on cellular morphology (Figure 1G). In-house developed software “GPSforGenes” [30] shows that the TF combinations for GGF and 3GF, as well as GATA2 + FOS or GATA2 + GFI1 + FOS were all highly expressed both in CD34⁺ HSPCs and placental tissue (Figure S1A–D). This program takes gene expression data from human tissues in the GeneAtlas U133A database, and demonstrates that the TFs we selected for hemogenic induction are highly expressed in a combinatorial manner in multiple hematopoietic tissues and lineages. It is interesting that without the inclusion of GFI1B, expression in CD71⁺ early erythroid lineages cells is lost (Figure S1C–D).

With initial experiments suggesting that GFI1 improves reprogramming efficacy based on CD34 induction, we next characterized the HSC surface phenotype of the 3GF vs. GGF derived cells. To this end, cell surface marker expression developmental human HSC markers CD34, CD49f, and angiotensin converting enzyme (ACE) were analyzed at several time points during reprogramming. ACE enriches for HSCs, and also marks all cells in the developing embryo destined to adopt a hematopoietic fate [51–53]. Likewise, CD49f (also known as integrin α6) also enriches for HSCs [53, 54]. 3GF reprogramming results in expanded yields of all populations of hematopoietic progenitors (in this case ACE⁺, CD49f⁺, or ACE⁺CD49f⁺ cells) as compared to GGF (Figure 2A). 3GF also results in expanded populations of CD34⁺ throughout reprogramming, as well as the CD49f⁺CD34⁺ and ACE⁺CD34⁺ populations (Figures 2B – D). Likewise, 3GF reprogramming induces an expansion of CD49f⁺ACE⁺CD34⁺ progenitors (Figure 2E). Additionally, the bulk ACE⁺ subset at D35 of GGF or 3GF reprogramming demonstrates robust populations of both CD49⁺CD34⁻ (red box) and CD49f⁺CD34⁺ (green box) cells (Figure S2A).

Figure 2. 3GF reprogramming expands the hematopoietic progenitor pool based on cell-surface immunophenotype.

A) Flow cytometric quantification of ACE⁺, CD49f⁺, or CD49f⁺ACE⁺ cells (n = 7 – 12). (B) Representative flow plots of the CD34⁺ population over the course of reprogramming showing a significant expansion of these cells with 3GF quantified on the right (n = 13 – 24). (C) Representative flow cytometry contour plots of CD49f⁺CD34⁺ cells in GGF and 3GF cells and their quantification (n = 14 – 24). (D) Representative flow plots of ACE⁺CD34⁺ cells in GGF and 3GF cells and their quantification (n = 7 – 12). (E) Quantification of flow data for CD49f⁺ ACE⁺CD34⁺ cells in GGF and 3GF reprogramming (n = 7 – 12). Data are represented as mean ± SEM. Nonparametric Mann-Whitney T-test for samples not assuming a normally distributed data set. *P < .05, **P < .01, ***P < .001, ****P < .0001.

Previous studies have identified Endothelial Protein C Receptor (EPCR, CD201) as a marker of expanded CD34⁺ progenitors from CB [55] as well as functional HSCs in murine BM [56]. Interestingly, GGF and 3GF reprogrammed cells throughout the reprogramming process both stain positive for EPCR, which by morphology appears to mark both endothelial-like cells as well as the rounded HSC-like cells that emerge in these prolonged cultures In particular, the EPCR population in 3GF reprogramed cells significantly expands over time (Figure 3A). As with the other HSC markers (Figure 2), greater yields of CD49f⁺CD34⁺EPCR⁺ cells are obtained in the 3GF population, with the majority of EPCR⁺ cells emerging at day 27, which corresponds to a late-intermediate time point in the reprogramming. Notably, virtually all of the CD49f⁺CD34⁺ cells are EPCR⁺ (Figure 3B and C). Within the bulk EPCR⁺ subset of cells at D35 of GGF and 3GF reprogramming, distinct populations of CD49f⁺CD34⁻ and CD49f⁺CD34⁺ cells can be observed as well (Figure S2B).

Figure 3. GGF and 3GF induced cells express the stem cell marker EPCR.

(A) Live staining of EPCR in GGF and 3GF cells throughout reprogramming. (B) Representative flow plots from D27 of rtTA control and 3GF cells stained for CD49f, CD34, and EPCR. (C) Quantification of CD49⁺CD34⁺EPCR⁺ cells (n = 4 – 12). Data are represented as mean ± SEM. Nonparametric Mann-Whitney T-test for samples not assuming a normally distributed data set. *P < .05, **P < .01.

Expression profiling of 3GF cells as they undergo reprogramming

Transcriptomic analyses via RNAseq of both 3GF and GGF reprogrammed cells were compared to determine if 3GF cells also underwent a developmental process as cells transition through endothelial intermediates to form hematopoietic cells while potentially remaining distinct from GGF cells [30]. To this end, 4 different populations of 3GF reprogrammed cells were sequenced in triplicate: 1) D15 3GF CD49f⁺CD34⁻; 2) D15 3GF CD49f⁺CD34⁺; 3) D25 3GF CD49f⁺CD34⁻; and 4) D25 3GF CD49f⁺CD34⁺.

Molecular profiling of these 4 separate 3GF populations shows strong similarities to previously acquired datasets of GGF reprogrammed cells [30]. The first observation in this analysis was that dimension 1 in the PCA analysis likely accounted for technical variation between the GGF and 3GF experiments, and therefore in comparative PCA analyses and hierarchical clustering only dimension 2 and 3 were used (Figure 4A – B). Comparison of dimension 2 and 3 for GGF and 3GF reprogramming reveals a similar developmental trajectory taken by 3GF cells, and a strong separation of HDF negative controls from the reprogrammed populations (Figure 4A). Hierarchical clustering of GGF and 3GF cells, again after removal of dimension 1 to correct for batch effects, shows a close relationship of D25 CD49f⁺CD34⁺ cells and, interestingly, clustering of GGF D25 CD49f⁺CD34⁻ and 3GF D15 CD49f⁺CD34⁺ cells (Figure 4B). Previous data shows that the GGF D25 CD49f⁺CD34⁻ population clusters most closely to known HSCs from published work [54], suggesting that the 3GF D15 CD49f⁺CD34⁺ population may most closely resemble endogenous HSCs in the 3GF reprogramming system.

Figure 4. Comparative RNA sequencing analyses of GGF vs. 3GF cells.

(A) PCA of dimension 2 (13.6%) and dimension 3 (8.5%) between GGF and 3GF reprogrammed cells. The blue curved arrow represents the hypothetical trajectory of cells as they traverse from endothelial intermediates defined by the data as CD49f⁺CD34⁻ towards hematopoietic cells defined as CD49f⁺CD34⁺. (B) Hierarchical clustering analysis of GGF and 3GF cells after depletion of dimension 1. (C) Up-regulated TFs between GGF vs. 3GF populations at different time points and surface phenotype determined using Biojupies [46] and plotted to reflect statistically significant differential expression. Known TFs identified from the human CD34⁺ hematopoietic cell database are labeled by blue bars [57], TFs seen to be up-regulated in E9.5 HE and E9.5 artery HE are shown in red [58], and hematopoietic TFs up-regulated as hPSC-derived CD34⁺ cells undergo mesodermal specification, EHT, and eventual blood production are shown in green [6]. The patterned purple bar for SRF reflects TFs identified in both Gomes et al., 2002 [57] and Gao et al., 2018 [58].

Interactive notebooks containing bioinformatics data from this RNAseq data were created using Biojupies software [46]. We focused on comparative analyses of up-regulated TFs between GGF and 3GF populations at the same day and sorting purity. We observed a variety of TFs up-regulated specifically in the 3GF populations in each comparison. These TFs (labeled in blue, red, green or purple) are found in the literature and support acquisition of a more hematopoietic TF landscape [6, 57, 58] (Figure 4C). Notably, this analysis demonstrates clear up-regulation of MAZ, RCOR1, and ZKSCAN1 in each 3GF population. MAZ is known to be involved in the activity of the SCL 1b promoter in CD34⁺ primitive myeloid cells [59]. RCOR1, another TF enriched in HSCs suggested to play an important role in HSC specification from the AGM, was also found to be up-regulated in 3GF cells [60]. ZKSCAN1 was found in some hPSC-derived hemogenic endothelium cells that engrafted immunocompromised mice [31], as well as in other transcriptomic analyses of HSCs [61, 62]. These data suggest that the TF profile of 3GF cells more closely resembles native HSCs than GGF cells.

Using a list of up-regulated genes in CD34⁺CD38⁻CD45RA⁻CD90⁺CD49f⁺ HSCs as compared to CD34⁺CD38⁻CD45RA⁻CD90⁻CD49f⁻ MPPs [54], there is a statistically significant upregulation of these genes in the 3GF dataset as D15 CD49f⁺CD34⁻ cells mature to D25 CD49f⁺CD34⁺ cells, p = 0.047 (Figure 5A, red dots are genes from Notta et al., 2011 [54] elaborated in Table S2A). Interestingly, comparative analysis of D15 CD49f⁺CD34⁺ and D25 CD49f⁺CD34⁺ with highlighted endothelial genes [30] shows a significant down-regulation of this gene list, p = .004 (Figure 5B, blue dots are endothelial genes elaborated in Table S2B). Comparative analyses of these previously published gene lists highlight and support the theorized developmental trajectory initiated by 3GF reprogramming, where hematopoietic cells emerge from endothelial intermediates as they mature throughout the induction process either through time or with activation of CD34.

Figure 5. Differential gene expression analysis of 3GF cells.

(A) DESeq2 MA plot showing comparative analysis of the 3GF D15 CD49f⁺CD34⁻ to D25 CD49f⁺CD34⁺ populations using a CD34⁺CD38⁻CD45RA⁻Thy1⁺CD49f⁺ HSC gene list from Notta et al., 2011 (red dots; [54]). (B) DESeq2 MA plot showing comparative analysis of the 3GF D15 CD49f⁺CD34⁺ to D25 CD49f⁺CD34⁺ populations using an endothelial gene list from Dr. Filipe Pereira (blue dots; [30]). (C) Quantification of normalized read counts for select endothelial and hematopoietic genes across the 4 RNA sequenced 3GF populations. Bar graphs in green represent genes commonly associated with endothelial identity. Bar graphs in red represent genes commonly associated with hematopoietic identity. Data are represented as mean ± SEM (n = 3).

To globally assess expression differences between the maturation stages captured in our 3GF reprogramming, D15 CD49f⁺CD34⁻ cells were used as a baseline for subsequent GO term analysis. From this analysis a consistent down-regulation of key pathways pertaining to the cell cycle, including M phase, mitotic cell cycle, DNA packaging, and microtubule cytoskeleton organization GO terms is observed (Figure S3A). This suggests that the cells generated later in the reprogramming (as well as D15 CD49f⁺CD34⁺ cells) shut down the machinery required for the cell cycle, which is consistent with known inactive cell cycle machinery in endogenous quiescent HSCs that usually remain in the G₀ phase in the BM [63]. Significantly up-regulated GO terms in this analysis include skeletal system/vasculature development, acute inflammatory response, activation of the immune response, polysaccharide metabolic process, aminoglycan metabolic process, and glycoprotein catabolic process. Up-regulation of the skeletal system/vasculature development terms could indicate enrichment for a response to HSC-support factors typically found in the endogenous HSC niche [64]. Unsurprisingly, the terms for the inflammatory and immune response indicate an activation of hematopoietic-type genes, as HSCs are involved in these pathways [65]. Perhaps most interestingly, several terms involved with metabolism and catabolism of aminoglycans, polysaccharides, and glycoproteins are seen (Figure S3A). This suggests that 3GF reprogrammed cells are primed to utilize inductive signals from a stromal niche layer to further mature.

Key individual endothelial and hematopoietic genes possess interesting expression patterns throughout 3GF reprogramming. Regarding endothelial genes, we see von Willebrand Factor (vWF), a factor known to be expressed in both endothelial cells and HSCs is expressed in each analyzed population, with a predominance of vWF in the CD49f⁺CD34⁺ hematopoietic cells. Conversely, there is a predominance of ETS2 and FOXC2 in the CD49f⁺CD34⁻ cells, both genes known to play key roles in HE. Interestingly, there is a down-regulation of CXCL5 and ANGPTL4 across time and throughout each population. JAG1, however, appears to increase as time goes on and as 3GF cells acquire CD34. (Figure 5C, green bars represent endothelial genes). This further indicates the endothelial identity these cells take on, and show that the cells may be constructing an intrinsic endothelial niche within the reprogramming. There is also an induction of key hematopoietic genes throughout the reprogramming, several of which confirm what is observed in the flow data (Figures 2 and 3). CD34 expression limited to the CD34⁺ sorted populations validates the accuracy of the sorting and sequencing. Interestingly, although there is expression of our other markers in all populations (as expected), there is a predominance of ACE, EPCR, and ITGA6 (CD49f) in the CD49f⁺CD34⁺ populations. These markers are known to purify for functional human HSCs [51, 52, 54, 55, 66], further confirming the HSC-like identity that is induced in 3GF reprogramming. Interestingly, there is an increased expression of F11R, EPCAM, and CD9 in the D15 CD49f⁺CD34⁺ cells as compared to the D25 CD49f⁺CD34⁺ cells, suggesting that the 3GF D15 CD49f⁺CD34⁺ population possesses greater functional potential than the more mature counterparts. CD9 complexes with c-kit in CB CD34⁺ cells, and may regulate hematopoietic progenitor proliferation and differentiation [67]. Unsurprisingly, each population expresses RUNX1, further supporting the thought that the derived cells are hemogenic and theoretically undergoing EHT. Interestingly, HGF has been found to act as a mobilizer of HSCs to the PB [68] (Figure 5C, red bars represent hematopoietic genes). What it does in endothelial populations (where it appears to be primarily expressed), however, remains unknown.

AFT024 LTC imparts functional potential to 3GF reprogrammed cells

tdT-3GF cells reprogrammed to day 15, 20, and D25 displayed hematopoietic morphology in reprogramming cultures, but when harvested and plated in CFU assays no colonies formed (Figure 6A). Interestingly, although the derived 3GF cells clearly displayed a cell surface phenotype highly similar to endogenous human HSCs (Figures 2 and 3), their in vitro functional potential required further optimization. We theorized that the reprogramming process required a separate maturation step in the form of a co-culture system as has been shown to be crucial for other reprogramming strategies [29, 32–34, 69].

Figure 6. In vitro maturation on AFT024 stroma imparts functional potential on 3GF reprogrammed cells.

(A) Images of 3GF reprogrammed tdT-HDFs at days 15, 20, and 25 of culture on gelatin with DOX and after an additional 3 weeks in CFU culture. CFUs failed to form. (B) Experimental scheme for LTC experiments on AFT024 stroma. Initially transduced cells were cultured on gelatin with DOX. After 15 days CD49f⁺ sorted cells were placed on AFT024 DOX; was continued for varying lengths: 1 week = 1 week of DOX and 4 weeks NO DOX, 2 weeks = 2 weeks DOX and 3 weeks NO DOX, etc. (C) Images of 3GF reprogrammed tdT-HDFs after 5 weeks on AFT024 with the designated lengths of DOX exposure. (D) Resultant CFU from 5 weeks of LTC on AFT024 with DOX. (E) CD45 staining of CFU. (F) Cytospins of CFU. (G) Number of colonies derived from 4,000 or 20,000 initially seeded CD49f⁺ 3GF reprogrammed cells after 5 week AFT024 LTC with continuous DOX (n = 3 – 6). Error bars represent SEM. Nonparametric Mann-Whitney T-test for samples not assuming a normally distributed data set. ***P < .001.

As a positive control, 250 Lin⁻CD34⁺ CB HSCs per ml were plated directly into CFU assays and CFU-GEMM, BFU-E, and CFU-GM counted after 2 weeks (Figure S4A). FACS analysis of these colonies revealed that the majority population was CD235a⁺CD14⁻. A population of CD235a⁺CD14⁺ cells also emerged, which may indicate the presence of erythro-myeloid precursors in these cultures. Populations of CD235a⁻CD14⁺ and CD41⁺ cells also existed within these colonies (Figure S4B and C). Quantification of these colonies showed an average of 71.33 (standard deviation (SD) of 19.442) colonies per 250 initially seeded cells, signifying a colony forming potential of roughly 1 in 3.5 Lin⁻CD34⁺ cells. From these counts, there is an average of 9.333 CFU-GM (SD of 5.339), 6.44 BFU-E (SD of 2.506), and 55.556 CFU-GEMM (SD of 14.293) (Figure S4D).

Extensive early experiments to test in vitro maturation systems using OP9-DL1 [34] or E4EC [32, 33] monolayers resulted in a loss of reprogrammed cells that did not display functional potential (data not shown). We elected to try co-culture with the mouse fetal liver (FL) cell line AFT024 that we have previously shown to support both mouse and human HSCs [36, 70, 71]. On day 15 3GF reprogrammed tdT-HDFs were sorted for CD49f⁺ cells and plated onto confluent monolayers of AFT024 with varying lengths of DOX exposure for varied activation of the transgenes (Figure 6B). Strikingly, when day 15 CD49f⁺ sorted cells were cultured on AFT024 for 5 weeks with continuous DOX exposure, cobblestone-like colonies emerged (Figure 6C and S5A). When these cells were harvested and plated in CFU assays hematopoietic colonies formed that possessed myeloid and erythroid cells and stained positive for human CD45 (Figure 6D–F). The frequency of CFU formation was approximately 1 in 250 cells that initially seeded the AFT024 LTC (Figure 6G).

Cobblestone-like colonies with GGF reprogrammed cells also emerged in AFT024 LTCs (Figure S5B) but the quality and quantity of the subsequent CFU was negligible (Figure S5C and D). This establishes AFT024 as a stromal co-culture niche that imparts the required signals to permit 3GF reprogrammed cells to mature and adopt hematopoietic functional potential.

AFT024 LTC permits quantification of stem cell frequency

The above results prompted the use of LDA and the AFT024 in vitro maturation system to determine the stem cell frequency of both GGF and 3GF reprogrammed cells, with the CAFC capacity correlating with repopulation activity [36, 72] (Figure 7A). This system allowed for distinct identification of positive cobblestone colonies as compared to tdT-HDFs that did not form colonies (Figure S5A). Quantification of cobblestone-like colony formation demonstrated a significantly greater CAFC frequency of 1/4020 in 3GF reprogrammed cells as compared to 1/7465 in GGF cells (p = 0.0270) (Figure 7B). Using ELDA software, the likelihood ratio test of a single-hit model a score test of heterogeneity were statistically significant (p = 0.00106 and 0.0015, respectively). Positive control experiments using Lin⁻CD34⁺ CB HSCs also shows the formation of large cobblestone-like colonies after 5 weeks of LTC on AFT024. Cytospin of these cells reveal a majority of myeloid cells as well as some erythroid cells (Figure S4E). 1° LDA of CB HSCs shows a steady decrease in HSC frequency over 5 weeks, suggesting that short-term progenitors in these purified cells may expand and exhaust, revealing a CAFC frequency that may correlate with long-term repopulating HSCs as previously shown [70], (Frequencies: W1 – 1/11.8; W2 – 1/19.5; W3 – 1/23.3; W4 – 1/55.3; W5 – 1/85.4) (Figure S4F and Table S3).

Figure 7. 3GF reprogramming generates a greater frequency of HSC-like cells capable of short-term multilineage engraftment.

(A) Experimental scheme for GGF and 3GF LDA using AFT024 monolayers. (B) ELDA of GGF and 3GF cells after 5 weeks of AFT024 LTC, with a 95% CI of 1/2740 – 1/5899 (frequency of 1/4020) for 3GF and 1/4834 – 1/11527 (frequency of 1/7465) for GGF. (C) Experimental scheme for intrahepatic transplants of D15 CD49f⁺ sorted 3GF cells into newborn NSG mice. (D) Representative flow cytometry contour plots of hCD45 chimerism in CB controls and 2 mice that showed engraftment using D15 3GF cells. (E) Representative flow cytometry contour plots showing the multilineage distribution of week 8 hCD45⁺ cells from CB controls and D15 3GF IH transplants. CD3⁺ T cells, CD19⁺ B cells and CD11c⁺ myeloid cells. PBMC = peripheral blood mononclear cells and were used as a staining control (non-transplanted).

To determine if hematopoietic populations could be isolated from LTC on AFT024, FACS analysis was performed for several of the aforementioned markers on day 15 CD49f⁺ cells sorted onto either 0.1% gelatin or AFT024 coated plates for 5 weeks with DOX supplementation. Though some populations show a significant decrease in cell yield (CD49f⁺, ACE⁺ and CD49f⁺CD34⁺) while others remain the same (CD34⁺, ACE⁺CD34⁺, CD49f⁺ACE⁺ and CD49f⁺ACE⁺CD34⁺). The more mature CD38⁺ population significantly decreases in cells grown on AFT024 (Figure S5E). This suggests that some progenitor populations are maintained in AFT024 co-cultures, and that these populations can be isolated and sorted for downstream applications. Additional positive control experiments using Lin⁻CD34⁺ CB HSCs after AFT024 LTC show the continued derivation of multilineage colonies in 2° CFU assays after 5 weeks of LTC on AFT024 (Figure S4G), with a majority of CD45⁺ cells composed primarily of CD14⁺ myeloid cells. Interestingly, the CD235a⁺CD14⁺ population is more abundant in these cells after AFT024 maturation, potentially indicating that this in vitro system may also support their expansion (Figure S4H and I). From these counts, there is an average of 35.333 CFU-GM (SD of 5.164), 2.00 BFU-E (SD of 2.098), and 20.667 CFU-GEMM (SD of 4.227) (Figure S4J). 2° LDA after 5 weeks of AFT024 co-culture shows a sustained, higher stem cell frequency as compared to 1° LDA assays, signifying the maintenance of true HSCs in vitro with this AFT024 cell line (Frequencies: W1 – 1/9.04; W2 – 1/7.0; W3 – 1/12.6; W4 – 1/24.61; W5 – 1/31.31) (Figure S4K and Table S3).

3GF cells engraft with short-term multilineage potential

To determine the engraftment potential of 3GF cells, derived CD49f⁺ cells grown on gelatin were sorted at three different time points (D12, D15, and D18) of the reprogramming process and subsequently transplanted into the liver of newborn immunocompromised mice (P0 – P2) (Figure 7C). Similar to what AFT024 supplies for the reprogrammed cells in vitro, the mouse microenvironment was theorized to provide necessary in vivo niche signals to support the further maturation of the reprogrammed cells similar to other published work [25]. The mothers were supplied with 1mg/ml of DOX in the drinking water for 2 weeks to maintain activation of the transgenes as the reprogramming cells interact with the in vivo niche.

Cells sorted from D12 or D18 reprogramming cultures for CD49f did not result in detectable hCD45 chimerism. D15 CD49f⁺ 3GF cells engrafted 2 out of 8 mice. These 2 mice, labeled Mouse ID:1 and ID:4, had detectable levels of hCD45 chimerism in their peripheral blood on week 4 of analysis post-transplant. At week 8, however, Mouse ID:4 began to lose hCD45 engraftment, while mouse ID:1 hCD45⁺ cells appeared to expand. Lin-CD34+ CB transplants are shown as a positive control (Figure 7D). After these time points, however, the hCD45⁺ chimerism became undetectable (on weeks 12 and 16 of analysis, as well as in subsequent organ harvesting experiments, data not shown), suggesting short-term engraftment. Multilineage contribution of the hCD45⁺ cells from Mouse ID:1 at week 8, however, reveals multilinaege engraftment with a predominance of CD3⁺ T cells, with smaller levels of CD11c⁺ myeloid cells and CD19⁺ B cells (Figure 7E). This indicates that D15 3GF CD49f⁺ cells grown on gelatin appear to possess limited hematopoietic functional potential. As such, it will be of great interest to expose 3GF reprogrammed cells to the AFT024 in vitro maturation system and subsequently transplant purified reprogrammed cells based on cell surface immunophenotype as previously discussed.

DISCUSSION

A replenishable source of engraftable, autologous human blood cells can provide a potential foundation to study and ultimately cure a multitude of hematologic disorders. To address this issue, several studies sought to generate various blood products de novo. Previous reprogramming strategies, in both human iPSCs and somatic cells, remain limited in the identity of their final derived cells, or have practical issues with either their starting cell populations or TF cocktails [17–28, 32]. Thus far, the low efficiency and poor engraftment capabilities of cells derived through iPSC differentiation restricts the utility of this method, although recent breakthroughs reveal possible derivation of engraftable cells from human iPSCs utilizing both TF overexpression and directed differentiation [31]. In this study, we optimized our hemogenic induction process without going through pluripotency to yield HSC-like cells that parallel endogenous HSCs in their cell surface phenotype, gene expression profile, and functional potential. The new 3GF cocktail, as well as co-culture on AFT024 stroma, improves the yield and functional output of the derived cells.

Previous work [30] as well as our current data demonstrates that we can induce the same developmental program in both mouse and human fibroblasts to derive hematopoietic cells. Addition of GFI1 to the reprogramming cocktail (Figure 1) highlights the importance of the axis formed by GFI1 and GFI1B in regulating human hematopoiesis and EHT, likely via RUNX and other pathways [73]. Both factors act as transcriptional repressors that recruit histone-modifying genes to the promoter and enhancer regions of their target genes [74]. HSCs, MPP1 and MPP2 cell populations all express GFI1 and GFI1B, with GFI1B expression highest in the earliest HSC compartment while as cells differentiate GFI1 expression increases [75, 76]. GFI1 targets several previously reported TFs used in HSC reprogramming, such as Hoxa9, Pbx1, Meis1, and PU.1 [77]. Knockout studies show that GFI1 prevents HSC proliferation, and plays a critical role in maintaining HSC self-renewal capacity [78]. GFI1B knockout results in massive HSC proliferation and maintenance of self-renewal, but concomitant removal of GFI1 and GFI1B impairs HSC survival [76]. GATA2 directly activates GFI1B via promoter and distal enhancer element binding in ChIP-seq studies of GATA2 in primary mast cells. Data in this ChIP-seq set also suggests that GATA2, GFI1, and GFI1B exist in a regulatory triad, with GFI1 and GFI1B acting in a mutually inhibitory manner, GFI1 inhibiting GATA2, and GATA2 activating GFI1B [79]. Other studies show the absolute requirement of GFI1 and GFI1B together in EHT, where GFI1 in particular specifically defines the subset of HE that gives rise to emergent HSCs. A set of EHT genes bound and repressed by GFI1 and GFI1B also show co-binding with RUNX1, as well as a prevalence of RBPJ binding motifs, signifying a potential role for NOTCH signaling [80]. This suggests that these factors, along with NOTCH and RUNX1 signaling [81], orchestrate the derivation of hematopoietic cells from HE.

Our analyses focused in part on the markers CD49f and ACE, together with the known human HSC marker CD34 [53, 82], all shown to enrich for the functional population of HSCs (Figure 2). ACE marks the para-aortic splanchnopleura and the AGM regions that possess all the hematopoietic potential of the developing human embryo [52], as well as derived hemangioblasts from human pluripotent stem cells [51, 83]. CD49f expression in human cells enriches for the engraftable HSC population [54]. Recent studies show that EPCR marks the functional subset of CB CD34⁺ progenitors exposed to UM171 [84], a small molecule shown to expand HSCs ex vivo [85]. Interestingly, our derived cells stain positive for this newly defined CB HSC marker, yet stain negative for GPI-80, a glycophosphatidylinositol-anchored surface protein that marks human FL HSPCs [86] (Figure 3 and data not shown). This suggests possible subpopulations of phenotypic HSPCs marked by either EPCR or GPI-80, and that our cells more closely resemble those derived from CB.

RNA sequencing of 4 different 3GF populations (D15 CD49f⁺CD34⁻, D15 CD49f⁺CD34⁺, D25 CD49f⁺CD34⁻, and D25 CD49f⁺CD34⁺) reveals that inclusion of GFI1 to the reprogramming process results in acquisition of an endothelial fate that leads to emergence of a hematopoietic profile as cells mature or activate CD34 (Figures 4, 5 and S3). While these cells follow the same developmental trajectory as GGF reprogrammed cells, they remain distinct from their GGF counterparts based on day of analysis or cell surface marker sorting (Figure 4B). The up-regulated TF profiles, however, appear to differ as 3GF reprogramming demonstrates up-regulation of multiple TFs known to play important roles in HSC biology (Figure 4C). Interestingly, 3GF cells begin to down-regulate machinery needed for cellular replication, suggesting that more mature and/or CD34⁺ cells become more quiescent compared to the D15 CD49f⁺CD34⁻ population (Figure S3A). Notably, normalized read count comparisons reveal a significant reduction of CDK6 in CD49f⁺CD34⁺ vs. CD49f⁺CD34⁻ populations within both D15 and D25 subsets (Figure S3B). CDK6 is a known regular of cell cycle progression absent in human long-term HSCs but found in abundance in short-term HSCs [87]. Adult HSCs are known to remain quiescent in the BM to maintain their function [63, 88]. Through our reprogramming, it is possible that the early cells we derive resemble those that emerge from the AGM and/or the placenta. The reprogrammed cells up-regulate the machinery required for processing various glycosaminoglycans (GAGs) and proteoglycans, suggesting that they are primed to take in these signals to further mature as they become hematopoietic (Figure S3A). It is widely known that a variety of HSC in vitro niche systems, such as AFT024, provide key GAGs and proteoglycans that help maintain biological function in vitro [9, 70, 71, 89–92]. All of this aligns with our hypothesis that the reprogrammed cells are ready to process signals provided by tissues such as the FL, and thus need to expand and further mature to become functional after exposure to these signals.

To date, the full HSC in vivo niche remains incompletely understood, complicating attempts to reconstruct this complex signaling system in vitro. Several known signals play a large role in inducing and sustaining hematopoiesis, such as the NOTCH [81], mTOR [93], and Wnt/β-catenin pathways [94–96]. Using a vascular niche known to express physiological levels of key angiocrine signals such as NOTCH, BMP, and c-KIT, engraftable HSC-like cells emerge from reprogrammed endothelial cells [32, 33]. Other work demonstrates improved expansion of iPSC-derived hematopoietic progenitors on endothelial monolayers overexpressing the NOTCH ligands JAG1 and DL4 [69]. Previous work from our lab demonstrated the importance of NOTCH activation in hemogenic precursor Prominin1⁺Sca1⁺CD34⁺CD45⁻ (PS34CD45⁻) cells isolated from mouse placentas. In 0.1% gelatin or OP9 co-culture, these PS34CD45⁻ cells cannot engraft immunodeficient mice. Only after 4 days of stromal co-culture on OP9-DL1, which provides a canonical NOTCH signal, can these PS34CD45⁻ cells give rise to all blood lineages and engraft primary and secondary immunodeficient mice [34].

The stromal cell line AFT024, derived from murine FL, can support both mouse [36] and human [90] hematopoiesis in vitro. Shown to express key signals for sustaining hematopoiesis such as delta-like (dlk) [97] – which constitutes a non-canonical ligand for NOTCH [98] – and dermatopontin [9], this cell line represents a component of the in vivo stem cell niche. AFT024 supports the ex vivo maintenance of human CD34⁺CD38⁻ HSPCs significantly more efficiently than other human derived cell lines in a contact-dependent manner, as supported by our data as well (Figures 6 and S4), highlighting the plethora of signals these cells specifically express to support hematopoiesis in vitro [70, 71, 91]. Conversion of Pro-B cells to HSC-like cells required maturation with the mouse in vivo niche prior to full functionality [25], demonstrating the importance of an instructive niche to assist and/or complete the reprogramming process, regardless of the starting cell population or the TF panel used in the reprogramming. It will be of great interest to further analyze our reprogrammed cells after AFT024 LTC. More comprehensive purification of the reprogrammed cells will permit a stronger determination of stem cell frequency in AFT024 LDA assays. Additionally, key analyses after AFT024 LTC will include RNAseq and transplantation of isolated reprogrammed cells at various levels of phenotypic HSC purity.

CONCLUSIONS

All together our results demonstrate that inclusion of GFI1 to the original GGF TF cocktail, as well as co-culture on AFT024 hematopoiesis-supporting stromal layers, are sufficient to generate HSC-like cells from human fibroblasts capable of multilineage functionality. This process remains dynamic, and travels through a HE-like intermediate characterized by previously identified markers that identify endothelium with hematopoietic potential. Our results further support the notions that hematopoietic specification is a step-wise process [99], that it traverses through endothelial intermediates [100] and that it requires variety of signals that can be provided by stromal cells for successful EHT and maturation [101]. To summarize, we show that manipulations to the previously established reprogramming cocktail and strategy in mouse and human fibroblasts can induce hemogenesis that leads to the production of HSC-like cells capable of multilineage function. The optimization of this process can also provide an in vitro platform for drug testing and hematopoietic disease modeling with the goal of identifying putative treatments for hematopoietic disorders and eventual avenues for autologous HSC transplants.

Supplementary Material

Supp FigS1. Figure S1. TF expression analyses in human cells and tissues. Information related to Figures 1 and 2.

The TF combinations of either (A) GATA2, GFI1B and FOS, (B) GATA2, GFI1, GFI1B and FOS, (C) GATA2, GFI1 and FOS, or (D) GATA2, and FOS are most enriched in hematopoietic tissues and organs such as CD34⁺ HSPCs or placenta among 33 tissue and cell types present in the GeneAtlas U133A database. The “GPSforGenes” program [30] was used for co-expression analysis of the chosen TFs within these tissues to identify those most enriched in these combinations (best fit = 1).

Supp FigS2. Figure S2. CD49f⁺ACE⁺ or CD49f⁺EPCR⁺ cells within the positive and negative CD34 populations.

(A) Representative FACS plots for D35 GGF or D35 3GF cells demonstrating a CD49f+CD34- (red box) or CD49f+CD34+ (green box) population after gating for ACE+ cells. (B) Representative FACS plots for D35 GGF or D35 3GF cells demonstrating a CD49f+CD34- (red box) or CD49f+CD34+ (green box) population after gating for EPCR+ cells.

Supp FigS3. Figure S3. GO term analysis and CDK6 quantification. Information related to Figures 4 and 5.

(A) GO term analysis using 3GF D15 CD49f+CD34- expression as a baseline compared to the 3 remaining RNA sequenced 3GF cells. The top 15 up- or down-regulated GO terms are listed to the right of the heat map. (B) Quantification of normalized read counts for CDK6 in 3GF cells. Bars in black represent CD49f⁺CD34⁻ cells. Bars in grey represent CD49⁺CD34⁺ cells. Data are represented as mean ± SEM (n = 3). Nonparametric Mann-Whitney T-test for samples not assuming a normally distributed data set. **P < .01, ***P < .001.

Supp FigS4. Figure S4. In vitro analysis of Lin⁻CD34⁺ CB cells. Information related to Figures 6 and 7.

(A) Representative images of CFU-GEMM, BFU-E, CFU-GM, and cytospins of CFU colonies derived 250 freshly seeded CB Lin⁻CD34⁺ cells. (B) Representative flow plots of CB 1° CFU CD45 subpopulations labeled with CD235a, CD14, and CD41. (C) Quantification of CD45⁺ cells and subsequent subpopulations from panel B (n = 7). (D) Colony counts of 1° CFU assays from 250 initially seeded CB cells per ml (n = 9). (E) Representative images of CB cells after 5 weeks of LTC on AFT024 and subsequent cytospins. (F) 1° LDA of CB cells seeded on AFT024 monolayers in 96well trays quantified for CAFCs across 5 weeks. (G) Representative images of CFU-GEMM, BFU-E, CFU-GM, and cytospins of CFU colonies formed from CB cells seeded at a density of 2000 cells per well on AFT024 monolayers in 12well trays and split into 10% (200 cells) or 90% (1800 cells) samples. (H) Representative flow plots of 2° CFU CB cells for CD45, and then CD45 subpopulations labeled with CD235a, CD14, and CD41. (I) Quantification of CD45⁺ cells and subsequent subpopulations from H (n = 12). (J) Colony counts of 2° CFU assays from 200 initially seeded CB cells per ml after AFT024 LTC (n = 6). (K) 2° LDA of CB cells seeded on AFT024 monolayers in 96well trays for 5 weeks (CAFC counted weekly) after a prior 5 weeks of LTC on AFT024. Data are represented as mean ± SEM.

Supp FigS5. Figure S5. Observed differences of GGF LDA and comparative flow analysis of 3GF LDA from gelatin and AFT024 LTC. Information related to Figures 6 and 7.

(A) Representative images of 3GF reprogrammed tdT-HDFs after 5 weeks of AFT024 LTC to demonstrate positive and negative CAFCs used for determining stem cell frequency LDAs. (B) Seeding of D15 CD49f⁺ sorted GGF cells on AFT024 stromal co-culture for 5 weeks with continued DOX exposure also results in the formation of CAFCs. (C) 3 week CFU colonies from GGF reprogrammed cells after 5 weeks of AFT024 LTC with continued DOX exposure. (D) Colony counts of GGF and 3GF cells after seeding 20,000 D15 CD49f⁺ sorted cells on AFT024 LTCs (n = 6 – 9). (E) Flow quantification of 3GF cells after 5 weeks of LTC on AFT024 monolayers or 5 weeks of LTC on gelatin-coated plates (n = 6 – 11). Data are represented as mean ± SEM. Nonparametric Mann-Whitney T-test for samples not assuming a normally distributed data set. *P < .05, ****P < .0001.

Abbreviations:

TF

Transcription Factor

HSPC

Hematopoietic Stem Progenitor Cell

HSC

Hematopoietic Stem Cell

HE

Hemogenic Endothelium

EHT

Endothelial to Hematopoietic Transition

iPSC

Induced Pluripotent Stem Cell

GGF

GATA2, GFI1B, FOS

3GF

GATA2, GFI1, GFI1B, FOS

MEF

Mouse Embryonic Fibroblast

AGM

Aorta-Gonad-Mesonephros

HDF

Human Dermal Fibroblast

FBS

Fetal Bovine Serum

P/S

Penicillin/Streptomycin

LTC

Long Term Culture

LDA

Limiting Dilution Analysis

TdT

tdTomato

DOX

Doxycycline

HC

Hydrocortisone

CFU

Colony Forming Unit

LTC-IC

Long Term Culture-Initiating Cell

CAFC

Cobblestone Area Forming Cell

CB

Cord Blood

H&E

Hematoxylin & Eosin

ELDA

Extreme Limiting Dilution Analysis

GO

Gene Ontology

PCA

Principal Component Analysis

SEM

Standard Error of the Mean

NSG

NOD-scid IL2Rg^null

FGRS

FOSB, GFI1, RUNX1, SPI1

ACE

Angiotensin Converting Enzyme

EPCR

Endothelial Protein C Receptor

vWF

von Willebrand Factor

SD

Standard Deviation

FL

Fetal Liver

GAG

Glycosaminoglycan

PBMC

Peripheral Blood Mononuclear Cell