Generation of Functional Hepatocytes from Human Adipose-Derived MYC+ KLF4+ GMNN+ Stem Cells Analyzed by Single-Cell RNA-Seq Profiling.

Cell transplantation holds considerable promise for end-stage liver diseases but identifying a suitable, transplantable cell type has been problematic. Here, we describe a novel type of mesenchymal stem cells (MSCs) from human adipose tissue. These cells are different from previously reported MSCs, they are in the euchromatin state with epigenetic multipotency, and express pluripotent markers MYC, KLF4, and GMNN. Most of the genes associated with germ layer specification are modified by H3K4me3 or co-modified by H3K4me3 and H3K27me3. We named this new type of MSCs as adult multipotent adipose-derived stem cells (M-ADSCs). Using a four-step nonviral system, M-ADSCs can be efficiently Induced into hepatocyte like cells with expression of hepatocyte markers, drug metabolizing enzymes and transporters, and the other basic functional properties including albumin (ALB) secretion, glycogen storage, detoxification, low-density lipoprotein intake, and lipids accumulation. In vivo both M-ADSCs-derived hepatoblasts and hepatocytes could form vascularized liver-like tissue, secrete ALB and express metabolic enzymes. Single-cell RNA-seq was used to investigate the important stages in this conversion. M-ADSCs could be converted to a functionally multipotent state during the preinduction stage without undergoing reprogramming process. Our findings provide important insights into mechanisms underlying cell development and conversion. Stem Cells Translational Medicine 2018;7:792-805.

Limited supply of donor cells remains a major challenge for hepatocyte transplantation. This study found a novel type of mesenchymal stem cell (MSC) from human adipose tissue (named as adult multipotent adipose‐derived stem cells, M‐ADSCs), which has an active euchromatin state and expresses pluripotent markers MYC, KLF4, and GMNN. The functional hepatocytes were generated from M‐ADSCs using a four‐step nonviral system. Single‐cell RNA‐Seq techniques were used to investigate the important stages and molecular events for the first time. M‐ADSCs were converted to a functionally multipotent status in the preinduction stage without undergoing a reprogramming process. The induced cells occurred many times synchronization in this conversion.

Introduction

Chronic liver diseases affect over 500 million people worldwide 1 and currently, liver transplantation remains the standard of care for end‐stage liver diseases. The shortage of suitable liver donors, however, greatly limits the use of primary human hepatocytes as a mode of treatment 2. An alternative to whole organ transplantation is to take a cell‐based therapy approach 3, but this would require a cell that can regulate cholesterol production and phase I detoxification enzymes post‐transplantation 4, 5.

Pluripotent stem cells (PSCs) such as embryonic stem cells (ESCs) and induced pluripotent stem cells exhibit hepatic differentiation capacity. Numerous reports describe attempts to generate functional hepatocytes from PSCs 6, 7, 8, 9, 10, 11, 12. In most cases, these cells lack many desired characteristics required for a transplantable hepatic cell as they are typically heterogeneous in nature and lack functional activity 8, 13, 14. Other features such as the potential for teratoma formation and immunogenicity are also serious limitations for the clinical utility of them. Thus, there is an urgent need for a stem cell population with hepatic differentiation capacity that is safe, functional, and feasible for clinical applications.

Evidence indicates that bone mesenchymal stem cells have the potential to regenerate human liver in vitro and in vivo 15, 16. However, the frequency of the events appears low in vivo 17, 18. Among MSCs from different tissue sources, adipose‐derived mesenchymal stem cells (ADSCs) show particular promise due to easy accessibility, high proliferative capacity, and lack of immunogenicity. However, hepatocyte‐like cells derived from different MSCs display different hepatocyte‐like properties 19.

We have focused on MSCs from adult adipose tissue and used specific culture medium to select a novel population of multipotent stem cells we term adult multipotent adipose‐derived stem cells (M‐ADSCs). These cells express MYC, KLF4, and GMNN 20, and share a similar chromatin state with ESCs. We recently showed that M‐ADSCs give rise to functional neuronal cells and insulin‐secreting cells with high efficiency 21, 22. Here, using a four‐stage non‐virus method, we induced M‐ADSCs into functional hepatocyte like cells. To understand the molecular events underpinning the conversion of M‐ADSCs into functional hepatocytes, we profiled single cells by RNA‐Seq 23. Our study not only demonstrates a new cell source for the derivation of human hepatocytes for cell‐based therapy, but also defines important molecular events that control liver regeneration.

Cell Isolation and Culture

Human adipose tissue was obtained from patients undergoing liposuction according to procedures approved by the Ethics Committee at the Chinese Academy of Medical Sciences and Peking Union Medical College. Fresh liposuction tissue was collected, digested, and isolated according to our established method 24. Cells were cultured with M‐ADSC culture medium containing Dulbecco's modified Eagle medium (DMEM) DMEM/F‐12, MCDB‐201, 2% fetal bovine serum (FBS), 1 × insulin transferrin selenium, 10−8 M dexamethasone, 10−4 M ascorbic acid 2‐phosphate, 10 ng/ml EGF, 10 ng/ml PDGF‐BB, and 1 ng/ml Activin A. ES cell line chHES22 was obtained from our group at National Engineering and Research Center of Human Stem Cell 25. H9 was cultured according to the WiCell Research Institute instructions and protocols in the UCSB stem cell lab. In brief, cells were maintained in mitomycin C‐(Sigma‐Aldrich, St. Louis, MO) inactivated human embryonic foreskin fibroblast hs27 (ATCC, Manassas, VA) in hESC culture medium. The feeder‐free cultures were maintained on matrigel (BD Biosciences) with conditioned medium from hs27 at 24‐hour intervals.

HepaRG cells 26, 27, 28 were cultured in RPMI‐1640 Medium (Gibco, Raritan, NJ; C11875500BT), adding 10% HepaRGTM supplement (HPRG730, Invitrogen, Paisley, U.K.) and fetal bovine serum (Gibco,16000–044). A final concentration of HepaRG supplement is necessary for HepaRG cell to remain high P450 activity. Human primary umbilical vein endothelial cells (HUVECs) were isolated from the umbilical cord of a newborn and culture as described previously 29. In brief, veins were isolated and washed twice with phosphate buffered saline (PBS), and endothelial cells were flushed out after digestion with 0.1% collagen P (Roche) for 15 minutes at room temperature. HUVECs were seeded on gelatin‐coated plastic dishes in an Medium 200 Low Serum Growth Supplement Kit (Gibco, S‐003‐K) medium containing 2% fetal bovine serum (Gibco, Grand Island, NY), hydrocortisone 1 μg/ml, human epidermal growth factor 10 ng/ml, basic fibroblast growth factor 3 ng/ml, heparin 10 μg/ml, and 100 U/ml of penicillin with 100 μg/ml of streptomycin. The medium is replaced every 2 days. HUVECs in passages 3–4 were used in this study.

Chromatin Immunoprecipitation and Whole‐genome Sequencing

ChIP experiments using antibodies against H3K4me3 (Abcam, 8580) and H3K27me3 (Upstate, 07‐449) were performed using the EZ ChIP kit (Millipore, Billerica, MA) according to the standard protocol provided by Millipore. Ten nanograms of ChIP DNA were prepared for sequencing on the Illumina Genome Analyzer by standard procedures. Readings were aligned to the reference (Hg18) human genome. Visualization of the ChIP‐seq results was performed using the ChIP‐seq peak detection tools CisGenome 30 with a window size of 100 and a read cutoff of 7 reads.

Hepatocytes Induction

For initiation stage of hepatocytes induction, cells were seeded at 5 × 105 cells/well in six‐well plate in human M‐ADSC maintained medium. One day later, after a brief wash in D‐Hank's, cells were cultured in high glucose‐DMEM (H‐DMEM; Gibco) supplemented with 5 ng/ml Activin A (R&D, 338‐AC), 50 ng/ml Wnt3a (R&D, 5036‐WNP), and 0.5% FBS (Gibco) for the first day. The next day the medium was changed to H‐DMEM supplemented with 5 ng/ml Activin A and 0.5% FBS, and the cells are cultured for four additional days. Next step (stage 2), the cells were briefly washed with D‐Hank's and then cultured in H‐DMEM with 10 ng/ml FGF basic (R&D, 233‐FB), 10 ng/ml BMP‐4 (R&D, 314‐BP), and 1% FBS (Gibco) for 4 days. Then, the medium was changed to H‐DMEM with 10 ng/ml FGF‐4 (R&D, 235‐F4), 10 ng/ml HGF (R&D, 294‐HG), and 0.5% FBS for 4 days (stage 3). In the final stage of the induction, the cells were cultured in H‐DMEM with 10 ng/ml FGF‐4, 10 ng/ml HGF, 10 ng/ml OSM (Peprotech, 300–10), 100 nM DEX, and 0.5% FBS for 6 days.

RNA Isolation and Quantitative Reverse Transcription Polymerase Chain Reaction

Total RNA was extracted using Trizol RNA isolation reagent (Invitrogen) according to the manufacturer's instructions. One microgram of RNA was reverse transcribed into cDNA with M‐MLV Reverse Transcriptase (TaKaRa Bio, Beijing, China) according to manufacturer's instructions. Quantitative reverse transcription polymerase chain reaction (RT‐PCR) was performed with SYBR Green I Master Mix (TaKaRa) on ABI Step One Plus real‐time PCR system (Applied Biosystems, Foster, CA). The average of three independent analyses for each gene and sample was calculated and normalized to the GAPDH. Primers were confirmed by proper melting curves. Primers used for amplification are listed in Supporting Information Table S1.

Single Cells Preparation

For the isolation of single cells, cells were washed with PBS twice and then were treated by trypsin at 37°C for 2 minutes. The digestion was ended by FBS. Cells were then dissociated into single‐cell suspension. The resulting single cells were washed with bovine serum albumin (BSA)–PBS twice and prepared to be picked as single cell for subsequent analysis.

Preparation of Single‐Cell cDNAs

The single‐cell RNA‐seq method has been described previously 23, 31. Briefly, we use a capillary pipette to pick a single cell and transfer it into lysate buffer, and execute reverse transcription reaction directly on the whole‐cell lysate. Following this procedure, we use terminal deoxynucleotidyl transferase to add a poly (A) tail to the 3′ end of first‐strand cDNAs, and then carry out 20 + 9 cycles of PCR to amplify the single‐cell cDNAs.

RNA‐Seq Library Preparation, Sequencing, and Alignment

After generation of the target cDNA from a single cell, 200 ng cDNA (0.5–5 kb) was sheared into 150–300 base pair (bp) fragments. And a DNA library Prep Master Mix Set kit (NEB) was used to prepare the sequencing library according to the manufacturer's procedures. In brief, the fragmented cDNA was end‐repaired, dA‐tailed, adaptor ligated, and then subjected to 8–10 cycles of PCR amplification.

Electron Microscopic Analysis

The cells were put in a carrier and vitrified using a Leica EM PACT2 high pressure freezer, and subjected to a substitution process with a 2% osmium tetroxide: acetone solution at −90°C, −60°C, and −30°C for 8 hours each using a Leica EM AFS2. The substituted samples were washed with acetone and then embedded in 100% spurr resin polymerized at 60°C for 48 hours. The samples in the embedding block were then cut into 70 nm‐thick ultrathin sections using a Leica UC6 ultramicrotome with a diamond knife and stained with uranyl acetate and lead citrate. EM images were captured in FEI Sprit 120 kV electron microscope operated at 100 kV.

Immunofluorescent Staining

Cells or tissue sections were fixed with 4% paraformaldehyde for 10 minutes at 4°C, and then incubated with PBS containing 0.25% Triton X‐100 (Sigma‐Aldrich) for 10 minutes at room temperature. After blocked by 5% BSA in PBS for 1 hour at room temperature, cells were incubated with primary antibodies at 4°C overnight. Then, after washed three times with PBS, samples were incubated with appropriate fluorescence‐conjugated secondary antibody for 1 hour at room temperature in the dark. Nuclei were stained with DAPI (Roche, Mannheim, Germany). Primary and secondary antibodies were diluted with PBS containing 3% BSA. The list of antibodies and dilution ratios are available in the Supporting Information Table S2.

Flow Cytometry Analysis

Cells were harvested and washed twice in Hank's Balanced Salt Solution (HBSS, Sigma‐Aldrich) with 0.1% BSA, and then incubated with antibodies diluted in HBSS with 0.1% BSA at 4°C for 30 minutes in dark. For flow cytometry analyses, cells were permeabilized with Cytofix/Cytoperm Fixation/Permeabilization kit (BD) for 15 minutes and incubated with primary antibodies for 1 hour at 4°C or overnight, then washed by 1× BD Perm/Wash buffer and incubated with the secondary antibodies for 1 hour at 4°C in dark. After incubation, cells were washed three times and analyzed by the BD Accuri C6 (BD Biosciences). Antibodies used for fluorescence activating cell sort are available in the Supporting Information Table 2. Data were analyzed with CFlow sample analysis software.

Enzyme‐linked Immuno Sorbent Assay

To determine the secretion of human albumin, supernatants of cell culture were collected after 48 hours culture. Cells were seeded on 12‐well plates for 12 hours, and then maintained in medium for 48 hours until collection of supernatants. For transplantation experiments, animal serum was collected. Levels of human albumin and α‐1 antitrypsin were measured using the human albumin enzyme‐linked immuno sorbent assay (ELISA) Quantitation kit (Bethyl Laboratory) according to the manufacturer's instructions. Serum was diluted in a range from 10‐ to 10000‐fold to obtain values falling to the linear range of standard curve.

Assays for Glycogen Storage, CYP1A2 and Glutathione S Transferase Activity, CYP Induction, and Metabolism Assay

For the measurement of cytochrome P450 oxidase (CYP) induction, cells were cultured in medium receptively for 24 hours and then change to culture medium supplemented with 10 μM omeprazole, for additional 24 hours. For measurement of CYP metabolism activities, cells were incubated with substrate in 200 μl incubation medium at different concentrations for 3 hours at 37°C. To stop the reaction, 800 ml cold methanol was added and centrifuged. The supernatants were collected for measurement of metabolized compounds. Total cell protein amount was used to normalize the data. Substrates and metabolized products used for standard curves were commercially purchased.

LDL Uptake and Oil Red O Staining

For LDL uptake assay, 10 μg/ml DiI‐Ac‐LDL (Invitrogen, Paisley, U.K.) was incubated with cells for 4 hours at 37°C and observed by fluorescence microscopy. For lipid detection, cultures were fixed with 4% paraformaldehyde and treated with 60% isopropanol for 5 minutes. Then the isopropanol was removed and oil red O working solution was added and incubated for 15 minutes at room temperature. Then the oil red O was removed and cultures rinsed with until clear.

Liver Organogenesis In Vivo

HUVECs were isolated from umbilical veins and cultured in endothelial growth medium (Gibco). To generate hepatic‐like tissue in vivo, 1 × 106 M‐ADSC‐derived hepatoblasts (MA‐HEBs) or hepatocytes (MA‐HECs), 0.8 × 106 to 1 × 106 HUVECs and 2 × 105 M‐ADSCs were resuspended in sterile PBS, and then were mixed with 1 × Matrigel in equal volumes. After subcutaneous injection of the mixture in male BALB/c nude mice (6–8 weeks) of 7 days, 30 days, and 60 days, the vascularized tissues and peripheral blood were collected and detected by immunohistochemistry and biochemical analysis respectively. All of the animal experiments were approved by the ethics committee at the Chinese Academy of Medical Sciences and Peking Union Medical College.

Statistical Analysis

For most statistic evaluation, unpaired Student's t test was applied for calculating statistical probability in this study. For all statistics, data from at least three independent samples or repeated experiments were used. Data are presented as mean ± SEM.

Results

Characterization of Human M‐ADSCs

Flow cytometry analysis showed that M‐ADSCs uniformly express high levels of Flk1, CD29, CD44, CD105, and CD166 (≥95%) while expression of CD106, CD31, CD34, CD45, and HLA‐DR is low (≤5%) (Supporting Information Fig. S1A). In addition, the cells have high expression of pluripotent genes MYC, KLF4, and GMNN (Supporting Information Fig. S1B). Similar to ESCs, M‐ADSCs appear in a euchromatin state by electron microscopic analysis of the nucleus of high‐pressure frozen cells (Fig. 1A). In M‐ADSCs, pluripotency, germ layer specification, and differentiation‐related genes are modified by H3K4me3 or co‐modified by H3K4me3 and H3K27me3 (Fig. 1B–1D, Supporting Information Fig. S2A, S2B). Although hepatic and pancreatic‐related genes were modified by H3K27me3 (Supporting Information Fig. S2C, S2D), key genes involved in definitive endodermal (DE) and mesoendodermal specification were modified by H3K4me3 or co‐modified (Supporting Information Fig. S2A, S2B). These results suggest that M‐ADSCs are epigenetically multipotent.

Figure 1

Characterization of human M‐ADSCs. (A): Nuclear chromatin state analysis. Scale bar, 200 nm. (B): ChIP‐seq analyzed histone modification patterns of genes associated with pluripotency, germ layer initiation, lineage specification, and EMT. (C) and (D): Histone modification status of genes associated with pluripotency and ectoderm specification. Abbreviations: EMT, epithelial to mesenchymal transition; ESCs, embryonic stem cells; M‐ADSCs, multipotent adipose‐derived stem cells.

Induction of M‐ADSCs into Hepatic Lineage by a Four Step Non‐Virus Method

We previously showed that Activin A and Wnt3a could induce M‐ADSCs into DE via activation of the Activin/Nodal and Wnt signaling pathways 22. By mimicking liver embryonic development and employing the signaling pathways involved, including FGF and BMP 32, 33, 34, we established a non‐virus method to induce M‐ADSCs into mature hepatocytes over a 19‐day period (Fig. 2A). During the initiation step (0d–5d), the cells show a remarkable transition from a bipolar fibroblast‐like morphology to a round epithelial‐like shape. In the second stage (5d–9d), the gap between induced cells increased and the cells became larger. In the third stage (9d–13d), cells became typical epithelial cells and their number increased (Fig. 2B). Post‐induction, cells showed the typical morphology of primary human hepatocytes 32, 35.

Figure 2

Induction of multipotent adipose‐derived stem cells (M‐ADSCs) into hepatic lineage using four‐stage protocol. (A): A schematic representation of the four‐stage protocol from M‐ADSCs to mature hepatocytes. (B): Morphological changes of the induced cells. Scale bar, 200 μm. (C): Quantitative reverse transcription polymerase chain reaction analyzed the expression of key genes. Data represent the mean ± SEM (n = 3). (D): Immunofluorescence analyzed the protein levels of stage‐specific factors. Scale bar, 100 μm. (E): Immunofluorescence analyzed the protein levels associated with hepatocyte function at induced day 19. Scale bar, 100 μm.

During the differentiation process, expression of DE‐related genes (FOXA2 and SOX17) and the primitive streak marker MIXL1 increased significantly from day 5 (Fig. 2C). The endogenous activation of key genes involved in liver development including PROX1, HNF6, HNF1B, HNF4A, CEBPA, and ATF5 36, 37 , and critical factor for the expression of many major drug‐metabolizing CYP genes CAR and PXR 38, was observed. The hepatocyte progenitor markers AFP and ALB then began to increase (Fig. 2C, 2D). To generate mature hepatocytes, we cultured the induced hepatoblasts in low serum medium containing 10 ng/ml HGF, 10 ng/ml FGF4, and 20 ng/ml BMP4 for 6 days. AFP expression was decreased while ALB increased in the final stage (Fig. 2C). A set of genes involved in lipoprotein, cholesterol, fat, glucose, and drug metabolism such as PCK2, G6PC, ATT, CYP1A2, CYP7A1, and CYP3A4 34, 39, 40, were also significantly upregulated (Supporting Information Fig. S3). The DE markers, FOXA2 and SOX17 were double‐positive on day 5, hepatic specification markers HNF4A and AFP were double‐positive on day 9, and hepatoblastic markers AFP and ALB were double‐positive on day 13. Proteins associated with liver function such as KRT18, KRT19, ATT, PCK2, CYP1A2, and CYP3A4 were also activated (Fig. 2E). Bile duct, pancreatic, and intestine development is closely related to liver during embryonic development 41, 42. As the induction process progressed, the expression of bile duct marker gene CK 19 and HNF1B, pancreatic marker gene PDX1, intestine lineage marker gene CDX2 and hepatoblast marker gene AFP were significantly reduced, and the expression of hepatocyte marker gene ALB increased significantly, suggesting further differentiation and maturation of induced cells into hepatocytes (Fig. 2C and Supporting Information Fig. S3). Pancreatic marker PDX1 is indeed low (detectable in only 1 of 21 cells) and the endoderm progenitor cell marker NGN3 and intestine lineage marker CDX2 were undetectable in all 19 day‐induced single cells, as well the proteins were negative (Fig. 2E, Supporting Information Table S3). Those data indicate that M‐ADSCs were converted into hepatocyte like cells.

M‐ADSC‐Derived Hepatocytes Exhibit Typical Characteristics of Human Hepatocytes

Human HepaRG cells were previously shown to exhibit many characteristics of primary human hepatocytes 27, 28, and were used as a positive control in this study to characterize the function of MA‐HECs. Notably, comparable mRNA levels of mature markers (ATT and PCK2), phase I CYP enzymes, phase II enzymes (MGST1), and phase III transporters (MRP2), major drug metabolism‐associated enzymes and transporters CYP1A2, CYP2C9, and CYP3A4 that account for 60% of human drug oxidation 43, were detected in MA‐HECs and HepaRG cells, in contrast to human embryonic foreskin fibroblast cells cells (Fig. 3A). This indicates that the key network of enzymes for drug metabolism was successfully established in MA‐HECs.

Figure 3

Functional characterizations of multipotent adipose‐derived stem cell (M‐ADSC)‐derived hepatoblasts and hepatocytes. (A): Quantitative reverse transcription polymerase chain reaction analysis of the expression level of hepatic maker genes and metabolic enzymes (n = 3 per treatment group; *p < .05 and **p < .05). (B): Positive cells were quantified by flow cytometry. (C): Enzyme‐linked immuno sorbent assay measured secretion of albumin (ALB). (D): Glycogen storage was confirmed by anthrone‐sulfuric acid colorimetry. (E): Glutathione S transferase activity was detected in M‐ADSC‐derived heptocytes (MA‐HECs). (F): Cells were stimulated by 10 μM omeprazole for 24 hours. Activity of CYPs was detected. (G): Lipid accumulation in MA‐HECs as shown by oil red O staining. Scale bar, 100 μm. (H): Alkaline phosphatase (ALP) expression in MA‐HECs as shown by ALP staining. Scale bar, 100 μm.

To corroborate this in greater detail, we showed that approximately 83% and 100% of MA‐HEBs and MA‐HECs respectively, expressed both ALB and alpha‐1‐antitrypsin (AAT), the double positive rate of ATT and ALB in induced terminal cells was 99.9% (Fig. 3B), while PDX1 and CDX2 were negative in MA‐HECs (Fig. 2E), suggesting an efficient and specific conversion. During the induction process, ALB secretion, glycogen storage ability, and CYP enzymes expression increased significantly (Supporting Information Fig. S4A–S4C). The level of ALB secretion in MA‐HECs was significantly higher than that of HepaRG cells, as measured by ELISA (Fig. 3C). Moreover, MA‐HECs showed a remarkable capacity for glycogen storage, a property also seen in HepaRG cells (Fig. 3D). In MA‐HECs, glutathione S transferases activity was also significantly higher than that of HepaRG cells (Fig. 3E). Cytochrome P450 (CYP450) enzymes of hepatocytes are major enzymes accounting for drug metabolism. When stimulated by 10 μM omeprazole for 24 h, the activities of CYP1A2, CYP3A4, and CYP2C9 in MA‐HECs were similar or higher than that of HepaRG cells (Fig. 3F). MA‐HECs also stained positively for alkaline phosphatase (ALP) and oil red O staining (accumulation of fatty droplets) (Supporting Information Fig. S4G, S4H). G banding analysis showed that all of M‐ADSCs and their‐derived HECs had a normal human karyotype (Supporting Information Fig. S4D, S4E). These data indicate an efficient conversion from M‐ADSCs to functional human hepatocytes.

M‐ADSC‐Derived Hepatocytes Can Regenerate Functional Liver Tissue In Vivo

During early liver organogenesis, newly generated hepatic cells form a condensed, vascularized tissue mass called the liver bud. Such large‐scale morphogenetic changes depend on the complex interactions between multiple cells before blood perfusion occurs 11, 40. This led us to propose that transplantation of endothelial cells together with hepatic cells into nude mice might recapitulate liver‐bud formation.

To test this notion, HUVECs were isolated, and confirmed by the phenotype of endothelial cell markers CD31 (99.4%) and CD144 (98.4%) (Supporting Information Fig. S5A–S5C) and the capacity to form blood vessels and take up Dil fluorescence labeled acetylated low‐density lipoprotein (LDL) (Supporting Information Fig. S5D, S5E). We then used matrigel to mix in vitro HUVECs and M‐ADSCs with MA‐HEBs or MA‐HECs and transplanted them into nude mice. Seven days after transplantation, a solid uplift could be observed at the transplant site, and a vascularized tissue structure was found after dissection (Fig. 4A). HE staining showed that MA‐HEBs + M‐ADSCs + HUVECs transplantation group (HEB group), MA‐HECs + M‐ADSCs + HUVECs transplantation group (HEC group), and M‐ADSCs + HUVECs transplantation group (control group) were vascularized, but in the MA‐HEBs mixed transplantation group was much greater than the other two groups and the organized structure was more compact and much higher in the MA‐HECs mixed transplantation group than that of the other groups (Fig. 4A). All three transplanted groups expressed the endothelial marker CD31 and the mesenchymal marker CD105. ALB, AAT, CYP7A1, and CYP3A4 were highly expressed in the MA‐HECs mixed transplantation group (Fig. 4B, 4C). The levels of human serum ALB in the hepatocytes from the mixed transplantation group were twice that of ALB in the MA‐HEB mixed transplantation group. No differences were found in the levels of ALP or AST between the two groups (Fig. 4D). These results suggest that both MA‐HEBs and MA‐HECs can generate functional liver tissue in vivo.

Figure 4

In vivo transplantation of mixed cells into nude mice. (A): hematoxylin and eosin staining (HE) staining showed the vascularization. Scale bar, 200 μm. (B): Quantitative reverse transcription polymerase chain reaction showed the expression of key hepatic genes (n = 3 per treatment group; *p < .05 and **p < .05). (C): Immunofluorescence analysis of key proteins and enzymes associated with hepatocyte function. Scale bar, 200 μm. (D): Secretion of ALB, ALP, and AST detected by enzyme‐linked immuno sorbent assay. Abbreviations: ALB, albumin; ALP, alkaline phosphatase; AST, aspartate transaminase.

Transcriptional Profiles Indicate Key Differentiation Stages from M‐ADSCs to Hepatocytes

Recently, several single‐cell RNA‐Seq techniques have been developed, making it feasible to analyze the transcriptome of human adult stem cells and their derivatives 23, 31, 44, 45. Using this technology we set out to analyze the transcriptome of 90 individual M‐ADSCs and their‐derived cells. Using the Illumina HiSeq2000 sequencer, we generated 155.2 Gb of cleaned data from 90 single cells, with, on average, 17.1 million reads per cell with read length of 101 bp. On average, we detected the expression of 6,328 (26.8%) of 23,616 RefSeq genes with FPKM >0.1. Thus, approximately half of the known human genes and transcripts were expressed in the M‐ADSC‐derived cells. Cells that clustered together were at the same developmental stages in almost all cases, showing reproducibility from cell to cell in each stage of differentiation (Fig. 5A). Moreover, the differentiation order was also accurately captured from un‐induced M‐ADSCs to late blastocysts, as neighboring stages clustered together in the analysis, as expected. RNA‐Seq data of the different stages from M‐ADSCs differentiate into hepatic progenitor cells and hepatocytes indicating that the greatest changes in gene expression were seen between the induced 0d and 0.5d stages, 3d and 5d stages, and also between the induced 5d and 9d stages (Fig. 5B). We then selected all genes that were differentially expressed between any two consecutive stages (fold change >2 or < 0.5, p < .01) generating six distinct group clusters (Fig. 5C, Supporting Information Table S4–S9).

Figure 5

Global expression patterns of known RefSeq genes during multipotent adipose‐derived stem cell differentiation into hepatic cells. (A): Unsupervised hierarchical clustering of the transcriptomes of single cell. All RefSeq genes expressed in at least one of the samples with Reads Per Kilobase Million (RPKM) ≥ 0.1 were used for the analysis. (B): Clusters of genes showing representative expression patterns. Differentially expressed genes were classified by the SOTA function in the clValid package (fold change > 2 or < 0.5, p < .01). The top Gene Ontology terms and corresponding enrichment P values are shown on the right side. (C): Single‐cell RNA‐seq profiles.

Gene expression changes are highly related to culture medium changes. 0.5d after induction medium treatment, gene expression profile in MSCs changed significantly, with many genes upregulated such as genes involved in biological functions and development. From induction day 0.5 to day 5, genes associated with three germ layer specification changed. The greatest change from induced 5d to 9d stages may be consistent with hepatic lineage specialization. From induction day 9 to day 13, many genes were upregulated, including genes involved in lipid transport, vitamin metabolic process, lipid biosynthetic process, regulation of cholesterol storage, liver development, lipoprotein catabolic process, and negative regulation of cholesterol storage (Fig. 5C). Gene expression related to epithelial to mesenchymal transition was low first and then increased, suggesting that the induced cells underwent mesenchymal to epithelial transition at an early stage of induction and turnover at the induced late stage.

M‐ADSCs Acquired a Differentiation Multipotent Status in Preinduced Stage

To identify the key biological events corresponding to these changes, we carried out Gene Ontology (GO) analysis for the differentially expressed genes and found that the highly expressed genes KLF4 and GMNN in M‐ADSCs began to decrease in induction day 1 and became undetectable from day 5. MYC decreased during the induction process (Fig. 6A), but other pluripotent genes, such as Nanog, SOX2, OCT4, and Lin28, were not activated. At induction day 0.5, expression of genes associated with cell cycle, proliferation, and neurogenesis began to decrease. Genes involved in the regulation of epithelial development increased significantly (Fig. 6B). Eomes, a key gene involved in mesoendoderm development, was undetectable in M‐ADSCs, but the Eomes positive cells accounted for 95% (38/40) in 0.5d induced cells (Supporting Information Table S10). Its downstream gene T was also activated. The phenotype of the cell population changed from an uninduced Mychigh KLF4high GMNNhigh Eomesneg state to a Mychigh KLF4high GMNNlow Eomeshigh state at induced 0.5 day stage, suggesting that the induced cells were converted into a mesoendoderm stage.

Figure 6

Multipotent adipose‐derived stem cells (M‐ADSCs) were converted from an epigenetically multipotent status to a functionally multipotent status. (A)–(E): Single‐cell RNA‐seq detection and Gene Ontology (GO) analysis of relevant genes. (F): GO analysis of differentially expressed genes from induction day 5 to day 9. (G): The schema showing the key stages of differentiation of ADSCs into hepatocyte like cells.

From induction day 0.5 to day 1, genes involved in collagen fibril organization, skeletal system development, sensory organ development, vasculature development, axon guidance became activated; genes associated with cytoskeleton organization, M phase response to endogenous were downregulated (Fig. 6C). From induction day 1 to day 3, most genes involved in germ layer development were activated (Fig. 6D). Genes involved in ear development, sensory organ, blood vessel development, mesenchymal cell differentiation, Wnt receptors were downregulated (Fig. 6D, 6E). The key liver development gene ATF5 became activated at induced 3d stage (Fig. 6A). From induction day 0.5 to day 3, genes associated with germ layer specification were gradually activated, indicating that M‐ADSCs were converted from an epigenetically multipotent status to a functionally multipotent status. The phenotype of the cell population changed from the Mychigh KLF4high GMNNlow Eomeshigh state to the Mycmean KLF4low GMNNlow Eomesmean state, and then to the Mycmean KLF4neg GMNNlow Eomeslow ATF5mean state at induced 0.5 day to 3 day stages, suggesting that induced cells began to specialize into hepatic lineage.

From induction day 3 to day 5, Mesodermal lineage differentiation associated genes and mesenchymal markers were significantly downregulated (Fig. 6A). Expression of histone deacetylase was downregulated during induction and became undetectable at day 5 (Fig. 6A). The key liver development gene ATF5 was activated at 3 day, and highly expressed at 5d and accompanied by a decrease of Eomes (Fig. 6A, Supporting Information Table S10). Genes associated with early liver specialization such as Hes1 and CEBPB, were also significantly upregulated (Fig. 6A). Chromatin and histone modification‐related genes were decreased, and Wnt and JAK–STAT signal pathways were suppressed (Fig. 6E). In addition, genes involved in mesoderm and ectoderm specification were gradually downregulated while genes associated with early liver specialization began to increase. The phenotype of the cell population converted from the Mycmean KLF4low GMNNlow Eomesmean state to a Mycmean KLF4neg GMNNneg Eomeslow ATFhigh state, indicating that the induced cells had progressively specialized into cells of the hepatic lineage.

During day 5 to day 9, genes associated with liver development and function, such as lipid biosynthesis and cholesterol storage, were upregulated while genes involved in the negative regulation of cell differentiation were downregulated (Fig. 6F). In order to better present key differentiation evens, we have added a schema to explain the process using some specific genes (Fig. 6G). Throughout the process of differentiation of ADSCs into hepatocyte like cells, it appears that ADSCs began to differentiate into the liver lineage, hepatoblasts, and hepatocyte earlier than we expected. The boundaries between phases liver specification, hepatoblast production, and mature stage of hepatocyte are blurred and overlap.

Synchronization During Induction of Liver Specialization

Recently, Yan et al. revealed that ESCs also exhibit significant heterogeneity 46. We compared the transcriptome of M‐ADSCs with that of ESCs. Heterogeneity was found among M‐ADSCs (Fig. 7A). However, none of the M‐ADSCs analyzed clustered with ESCs, excluding the possibility that three‐germ layer differentiation potential was due to cell contamination. Interestingly, even single cell‐derived M‐ADSCs possessed different transcriptomes (Fig. 7B). Discrete degree analysis showed that ESCs, M‐ADSCs and single cell‐derived M‐ADSCs all had a high degree of dispersion (Fig. 7C). These results suggest that cell homogeneity is relative while heterogeneity is definitive.

Figure 7

Heterogeneity of M‐ADSCs and synchronization during induction of liver specialization. (A) and (B): Single‐cell RNA‐Seq analysis of heterogeneity among M‐ADSCs versus ESCs, and single cell‐derived MADSCs. (C): Dispense analysis of ESCs, M‐ADSCs and single cell derived M‐ADSCs (CCs). (D): Discrete degree analysis during differentiation. (E): GO analysis of differentially expressed genes from induction day 9 to day 13. (F): Induction efficiency measured by flow cytometry analysis marked by AFP and HNF4A. (G): Single‐cell quantitative reverse transcription polymerase chain reaction analysis of HNF4A and AFP expression during hepatic induction. ESC, embryonic stem cells; M‐ADSCs, multipotent adipose‐derived stem cells.

Cell discrete degree analysis showed that M‐ADSCs are a relatively heterogeneous population and at induction day 3 the discrete degree was reduced, suggesting that M‐ADSCs underwent synchronization (Fig. 7D). Although transcriptome changes from day 9 to day 13 were not significant (Fig. 7E), key genes related to liver specialization, such as HNF4A and AFP were significantly upregulated. The HNF4A and AFP positive cells were 92.1% and 93.3% of the detected cells respectively at day 9, and further increased to 99.8% and 99.9% at day 13 (Fig. 7F). The expression levels of HNF4A and AFP peaked at day 13 (Fig. 7G), suggesting that the induced cells underwent synchronization.

Discussion

While liver transplantation is limited by the availability of suitable donor organs. Hepatocyte transplantation, however, is not constrained by this limitation and may provide an attractive alternative 47. Adult human stem cells may play an important role in liver regeneration 48, 49 but is questioned by low efficiency and poor hepatocyte function. Human induced hepatocytes (hiHeps) were generated from fibroblasts by lentiviral overexpression of hepatic fate conversion factors, HNF1B, HNF4A, and HNF6 along with the maturation factors ATF5, PROX1, and CEBPA 36, 37. We reveal that human functional hepatocytes are readily and reproducibly generated from M‐ADSCs through the endogenous activation of all six factors with our four‐step non‐virus induction system. A set of genes involved in lipoprotein, cholesterol, fat, glucose and drug metabolism, including CAR, PXR, PCK2, G6PC, ATT, CYP1A2, CYP7A1, and CYP3A4, were also activated. M‐ADSCs‐derived hepatocytes display similar functional characteristics of hiHeps 36, 37. MA‐HEBs and MA‐HECs can form vascularized liver‐like tissue, secrete ALB and express metabolic enzymes in vivo. More importantly, the key differentiation stages from M‐ADSCs to hepatocytes are indicated through analyzing transcriptional profiles via single‐cell sequencing techniques.

The hPSCs have been used advantageously to produce hepatocytes for disease modeling 50 and for developmental studies. However, generation of cells displaying all the functional characteristics of mature hepatocytes has been proved to be difficult. Indeed, hPSC‐derived hepatocytes uniformly express fetal markers such as AFP and lack key metabolic activity associated with adult cells such as cytochrome p450, especially CYP3A4 51. Human HepaRG cells exhibit many characteristics of primary human hepatocytes 27, 28. We used HepaRG cells as a positive control to confirm the function of M‐ADSC‐derived hepatocytes by using similar evaluation indicators and methods of paper 36, 52. We confirmed that the key network of drug metabolism was successfully established. Both MA‐HEBs and MA‐HECs could form vascularized liver‐like tissues and secrete ALB and metabolic enzymes in vivo, but not expressed the pancreatic and intestinal lineage markers. These results confirmed a specific and efficient conversion from M‐ADSCs to functional hepatocytes.

A particular challenge in studies with MSCs is the variability of outcomes depending on the tissue of derivation and the culture methodology, raising concerns regarding the validity or reproducibility of the differentiation capacity of these cells. Here, we isolated MSCs from human adipose tissue and selected them using specific medium. The cells represent a MSC population different from previously reported cells 53, 54. They express the pluripotency markers MYC, KLF4, and GMNN and exhibit a nuclear euchromatin state similar to ESCs. In addition, most of the genes associated with germ layer specification present in M‐ADSCs are modified by H3K4me3 or co‐modified by H3K4me3 and H3K27me3, suggesting that they are epigenetically multipotent. Recently, we demonstrated that M‐ADSCs could give rise to functional neuronal cells and insulin‐secreting cells with high efficiency 21, 22. In this study, we induced M‐ADSCs into functional hepatocytes by a non‐viral system, inferring that M‐ADSCs have the capacity to undergo three‐germ layer differentiation.

Functional heterogeneity among stem cells isolated from adult tissues appears to be unavoidable. Consequently, the transcriptome of whole populations may generate data that could lead to biased conclusions. The genetic basis of cell fate determination of adult stem cells can be better characterized by analyzing gene expression at the single‐cell level. Recently, several RNA sequencing (RNA‐Seq) techniques have been developed 23, 31, 44, 45. Primary hESC outgrowth has been shown to exhibit dramatically different transcriptomes 46. Significant heterogeneity was also found among M‐ADSCs. Even single clone derived, M‐ADSCs also possess different transcriptomes. So which molecules could serve as markers to identify adult stem cells with three germ layer differentiation potential requires further studies.

Our study is the first to use single‐cell RNA‐seq to examine the underlying molecular mechanisms during hepatic differentiation and identified important events. We found that early induction medium made M‐ADSCs synchronize. Genes related to three‐germ layer specification were upregulated and then downregulated, suggesting that M‐ADSCs were converted from an epigenetically multipotent status to a functionally multipotent status. Although the induced cells did not undergo a typical DE stage as in the case of ESCs during early hepatic induction, the mesendodermal marker Eomes and T were activated, suggesting that M‐ADSCs were induced into mesendodermal stage. During our induction, MYC, KLF4, and GMNN were reduced and other pluripotent genes SOX2, OCT4, and Lin28 were not activated, which suggested that M‐ADSCs could be converted to the mesoendoderm stage without undergoing the reprogramming process. With the induction progress, genes involved in mesoderm and ectoderm specification were gradually downregulated while genes associated with early liver specialization began to increase, indicating that the induced cells became progressively specialized into cells of the hepatic lineage. Throughout the process of differentiation of M‐ADSCs into hepatocyte like cells, it appears that M‐ADSCs began to differentiate into the liver lineage, hepatoblasts, and hepatocyte earlier than we expected. The boundaries between phases liver specification, hepatoblast production and mature stage of hepatocyte are blurred and overlap. However, M‐ADSCs could give rise to ALB and ATT double positive hepatocytes with an efficiency of nearly 100%.

The single cell transcriptome underwent significant changes in the mid‐induced phase, with activation of genes related to liver development and function. Although in the later stages of induction, there was moderate change in the transcriptome, expression of key genes related to liver specification increased significantly. HNF4A and AFP were close to 100% at day 13, and then significantly declined. These results indicate that induced cells underwent further change to become mature liver cells. Although we have basically established a system of metabolizing enzymes in our M‐ADSCs‐derived hepatocyte like cells, these cells may not be mature enough. How to optimize the culture system, make the M‐ADSCs‐derived hepatocyte like cells more mature, and can maintain stable function under in vitro culture conditions, is worth further investigation for us.

Conclusion

In conclusion, the above results revealed that M‐ADSCs in our specific induction system could give rise to a homogenous population of terminally differentiated cells through multiple processes of synchronization without undergoing a reprogramming process. The successful application of single‐cell sequencing techniques will provide an additional valuable tool to investigate models of cell development and regenerative medicine.

Author Contributions

H.L. and L.Z.: conception and design, provision of study material, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; H.C. and T.L.: provision of study material, collection and/or assembly of data, data analysis and interpretation; Q.H., X.D.S., and R.C.Z.: conception and design, financial support, manuscript writing; X.Y., H.F, P.Z., and L.F.: collection and/or assembly of data, data analysis and interpretation; S.G., R.B., X.C., and A.K.: manuscript writing, final approval of manuscript; J.L.: financial support.

Disclosure of Potential Conflicts of Interest

The authors indicated no potential conflicts of interest.

Table S1. List primer sequences used in qRT‐PCR.

Table S2. Antibodies used in this article.

Table S3. The expression levels of PDX1, NGN3 and CDX2 in 19d‐induced cells.

Table S10. The expression levels of Eomes in 0d to1d‐induced cells.

Click here for additional data file.

Table S4‐S9. The GO analysis of six distinct groups of Stage‐specific expression patterns during hepatic lineage induction. Data were showed in Excel.

Click here for additional data file.

Figure S1. Characterization analysis of human M‐ADSCs. (A) Flow cytometry analyzed the immunophenotypic of M‐ADSCs. (B) qRT‐PCR detected expression of pluripotent related genes in M‐ADSCs compared with HEF and ESCs (n = 3 per treatment group).

Figure S2. Histone modification of genes in human M‐ADSCs. (A) Histone modification of genes associated with mesendoderma specification. (B) Histone modification of genes associated with definitive endoderm. (C) Histone modification of genes associated with hepatic lineage specification. (D) Histone modification of genes associated with pancreatic specification.

Figure S3. qRT‐PCR analysis of the expression level of hepatic maker genes and metabolic enzymes.

Figure S4. Functional characterization of M‐ADSC‐derived hepatic cells. (A) Dynamic secretion level of ALB during hepatic induction. (B) Dynamic glycogen storage ability of induced cells. (C) Dynamic GST activity was detected during the induction process. (D) and (E) G banding analysis demonstrated karyotype (44, XX). Scale bar, 100 μm.

Figure S5. Characterization of human HUVECs. (A) Morphology of HUVECs. (B) and (C) Phenotype analysis of HUVECs by Immunofluorescent staining and flow cytometry assay. (D) HUVECs have the capacity to form blood vessels, confirmed by blood vessel formation assay in vitro. (E) Detection the Acetylated low‐density lipoprotein (Dil‐Ac‐LDL) phagocytosis by fluorescently labeled test. Scale bar, 100 μm.

Click here for additional data file.