Mechanism of paracrine communications between hepatic progenitor cells and endothelial cells

Abstract

Hepatic progenitor cells (HPCs) are facultative tissue-specific stem cells lining reactive ductules, which are ubiquitously observed in chronic liver diseases and cancer. Although previous research mainly focused on their contribution to liver regeneration, it turned out that in vivo differentiation of HPCs into hepatocytes only occurs after extreme injury. While recent correlative evidence implies the association of HPCs with disease progression, their exact role in pathogenesis remains largely unknown. Our previous research demonstrated that HPCs expressing angiogenic paracrine factors accumulate in the peritumoral area and are positively correlated with the extent of intratumoral cell proliferation and angiogenesis in the livers of patients with liver cancer. Given the crucial roles of angiogenesis in liver disease progression and carcinogenesis, we aimed to test the hypothesis that HPCs secrete paracrine factors to communicate with endothelial cells, to determine molecular mechanisms mediating HPCs-endothelial interactions, and to understand how the paracrine function of HPCs is regulated. HPCs promoted viability and tubulogenesis of human umbilical vein endothelial cells (HUVECs) and upregulated genes known to be involved in angiogenesis, endothelial cell function, and disease progression in a paracrine manner. The paracrine function of HPCs as well as expression of colony stimulating factor 1 (CSF1) were inhibited upon differentiation of HPCs toward hepatocytes. Inhibition of CSF1 receptor partly suppressed the paracrine effects of HPCs on HUVECs. Taken together, our study indicates that inhibition of the paracrine function of HPCs through modulation of their differentiation status and inhibition of CSF1 signaling is a promising strategy for inhibition of angiogenesis during pathological progression.

Graphical Abstract

1. INTRODUCTION

Chronic liver disease and cirrhosis are the 11^th leading cause of death in the United States [1]. The progression of chronic liver disease is associated with long-term inflammatory and fibrogenic processes that promote anomalous angiogenesis and cirrhosis [2, 3]. Conversely, recent studies suggest a potential contribution of angiogenesis to the pathogenic progression of fibrogenic and inflammatory diseases such as biliary fibrosis, non-alcoholic fatty liver disease (NAFLD), and alcoholic liver disease [3–6]. Furthermore, angiogenesis plays a critical role in tumor development, progression, and metastasis [7]. Since most cases of hepatocellular carcinoma (HCC) are associated with chronic liver diseases including cirrhosis [8], a better understanding of how endothelial cells are modulated during the progression of chronic liver disease may provide novel therapeutic targets.

Reactive ductules consist of epithelial cells referred to as hepatic progenitor cells (HPCs) [9]. Accumulation of reactive ductules, frequently observed in several chronic liver diseases, correlates with pathological progression and cancer development [10–13]. HPCs are facultative tissue-specific stem cells that are predominantly derived from the biliary compartment in the postnatal liver, although hepatocytes can also contribute to reactive ductules in a context-dependent manner [9, 14–16]. The contribution of HPCs to liver parenchymal regeneration has been assumed based on the fact that HPCs are bipotential and can differentiate into hepatocytes or biliary epithelial cells in culture [17]. However, we and other groups have demonstrated that in vivo conversion of HPCs into mature hepatocytes only occurs after extreme liver injury [14, 18, 19]. Therefore, although HPCs are detected in many liver diseases, the exact role of HPCs in the context of pathogenic progression as well as underlying molecular mechanisms linking the differentiation status of HPCs to their pathogenic function remain largely unclear.

Recently, several correlation studies implied a paracrine role of HPCs in pathogenesis. For instance, HPCs in the livers of pediatric patients with NAFLD express multiple adipokines [20], and vascular endothelial growth factor (VEGF) expression by HPCs correlates with angiogenesis in the context of chronic viral hepatitis and primary biliary cirrhosis [21]. Also, it has been reported that the expression levels of HPCs markers such as SRY-Box transcription factor 9 and cytokeratin19 in the livers of patients with HCC positively correlate with poor prognosis and increased recurrence [22, 23]. Furthermore, we previously demonstrated that peritumoral reactive ductules expressing paracrine factors such as angiopoietin 1 (ANGPT1), platelet-derived growth factor C (PDGFC), VEGF, and vascular endothelial growth factor D (VEGFD) accumulate in the livers of young patients with hepatoblastoma and HCC, and that paracrine factor expression by reactive ductules positively correlates with intratumoral cell proliferation and the extent of angiogenesis [24]. These findings suggest a potential role of HPCs in modulating endothelial cell biology in the setting of chronic liver disease and cancer.

Taken together, we hypothesize that HPCs crosstalk with endothelial cells by secreting paracrine factors and that differentiation of HPCs toward hepatocytes leads to the loss of the progenitor cell characteristic, i.e. paracrine ability to stimulate endothelial cells. The present study identified novel molecular mechanisms underlying HPCs-endothelial communication and revealed a previously unrecognized relationship between the differentiation status of HPCs and their paracrine function.

2. MATERIALS AND METHODS

2.1. Cell culture and preparation of conditioned media

We used clonally expanded HPCs previously established from forkhead box L1-expressing mice in this study [17]. HPCs culture medium was prepared by adding the conditioned medium derived from embryonic day 14.5 mouse liver cells to the same volume of hepatic colony-forming medium consisting of Dulbecco’s modified Eagle medium/F-12 (11320-033, Invitrogen, MA, USA), 10% fetal bovine serum (FBS; ES009B, Millipore, MA, USA), 2.5 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; 15630-080, Gibco, MD, USA), 10 mM nicotinamide (N0636, Sigma, MO, USA), 100 unit/mL penicillin/streptomycin (15140122, Invitrogen, MA, USA), 2 mM L-glutamine (25030-081, Gibco, MD, USA), and 1X nonessential amino acid solution (11140-050, Invitrogen, MA, USA) [17, 25]. 40 ng/mL hepatocyte growth factor (HGF, 100-39, Peprotech, NJ, USA), 20 ng/mL epidermal growth factor (AF-100-15, Peprotech, NJ, USA), 20 mM Y-27632 (1254/10, R&D Systems, MN, USA), 1X insulin-transferrin-selenium-X (51500-056, Gibco, MD, USA), and 50nM dexamethasone (D1756, Sigma, MO, USA) were also added. To induce hepatocytic maturation of HPCs, cells were cultured on collagen 1-coated 24-well plates. After reaching 100% confluency, cells were washed with phosphate-buffered saline (PBS) twice and cultured in the hepatocytic differentiation medium consisting of Dulbecco’s Modified Eagle Medium (DMEM, 10569010, Gibco, MD, USA), 10% FBS (SH30090.03E, Hyclone Fisher, MA, USA), 100 unit/mL penicillin/streptomycin (15140122, Invitrogen, MA, USA), 2 mM L-glutamine (25030-081, Gibco, MD, USA), 1X nonessential amino acid solution (11140-050, Invitrogen, MA, USA), 100 nM dexamethasone (D1756, Sigma, MO, USA), 20% Matrigel (3524230, BD Biosciences, CA, USA), 20 μg/mL oncostatin M (495-MO, R&D system, MN, USA), 10 μg/mL HGF (100-39, Peprotech, NJ, USA), and 10 μg/mL fibroblast growth factor 4 (FGF4, 5846-F4, R&D Systems, MN, USA) [17]. For overexpression experiments, lentiviral vectors expressing mouse hepatocyte nuclear factor 4 alpha (Hnf4α, NM_008261) (LV528762, Applied Biological Materials, Richmond, Canada) and red fluorescent protein (RFP) (LV084, Applied Biological Materials, Richmond, Canada) under the control of the cytomegalovirus promoter were used for viral packaging and transduction of HPCs per the manufacturer’s protocol. RFP- and Hnf4α-transduced HPCs were selected by maintaining in 5 μg/mL puromycin (A1113803, Thermo, MA, USA) for 3 passages. HPCs overexpressing RFP or Hnf4α were either maintained in the HPCs culture medium (undifferentiated) or treated with the hepatocytic differentiation medium (differentiated) for 1 week. For preparation of HPC-conditioned medium (HPC-CM), undifferentiated or differentiated HPCs were washed with PBS and incubated in the serum-free vehicle medium consisting of DMEM (41965-039, Gibco, MD, USA) and 20 mM Y-27632 (1254/10, R&D Systems, MN, USA) for 48 hours. All conditioned media were prepared from cells with > 95% confluency. Human umbilical vein endothelial cells (HUVECs) were purchased from the American Type Culture Collect (CRL1730, ATCC, VA, USA).

2.2. Isolation of mouse primary hepatocytes

Mouse primary hepatocytes were isolated from C57BL/6 mice (000664, The Jackson Laboratory, ME, USA). 6-week-old male mice were perfused with Earle’s balanced salt solution (EBSS; 14155063, Thermo, MA, USA) supplemented with 5 nM ethylene glycol-bis(β-aminoethyl)-N,N,N′,N′-tetraacetic acid for 3 minutes followed by EBSS with Ca²⁺/Mg²⁺ (24010043, Thermo, MA, USA) supplemented with 10mM HEPES for 2 minutes. Next, the liver was digested in EBSS with Ca²⁺/Mg²⁺ supplemented with 50 mg/mL collagenase IV and 10 mM HEPES for 8 minutes and dispersed on a petri dish, and isolated cells were resuspended in the complete medium (10% FBS-supplemented DMEM). The hepatocyte-enriched fraction was collected using low-speed centrifugation (50g, 1 min) as described previously [17]. All animal experiments were approved by the Institutional Animal Care and Use Committee of the Cincinnati Children’s Hospital Medical Center.

2.3. Functional analysis of differentiated HPCs

To assess albumin production, HPCs and mouse primary hepatocytes were cultured in the HPCs culture medium and the complete medium, respectively, for 24 hours (40,000 cells per well in a 24-well plate). Next, cells were incubated in the serum-free vehicle medium consisting of DMEM (41965-039, Gibco, MD, USA) and 20 mM Y-27632 (1254/10, R&D Systems, MN, USA) for 24 hours, and the media were collected and assayed using the Mouse Albumin ELISA Kit (ab108792, Abcam, Cambridge, UK). The concentration of urea secreted into the culture medium and the amount of glycogen in cell lysates were determined using the Urea Assay Kit (DIUR-100, Bioassay Systems, CA, USA) and the Glycogen Assay Kit (ab65620, Abcam, Cambridge, UK), respectively, per the manufacturer’s instruction. To assess low-density lipoprotein (LDL) uptake, the LDL Uptake Assay Kit (Cell-Based) (ab133127, Abcam, Cambridge, UK) was used. Cells were treated with LDL-DyLight for 4 hours and photographs were taken using the ECLIPSE Ti microscope (Nikon USA, NY, USA). The fluorescence intensity was quantified using Image J [26].

2.4. Treatment of HUVECs with HPC-CM

To perform the time-course cell viability assay in the setting of serum/growth factor deprivation, HUVECs were incubated in 100% HPC-CM or 100% vehicle serum-free medium for 24, 48, and 72 hours, respectively. For all subsequent experiments for assessing gene expression, cell viability, cell proliferation, and apoptosis, HUVECs were treated with 50% HPC-CM, i.e. 1:1 mix of HPC-CM and HUVECs culture medium consisting of F-12 medium (30–2004, ATCC, VA, USA), 10% FBS (S009B, Millipore, MA, USA), 0.1 mg/mL heparin (H3393, Sigma, MO, USA), and 0.05 mg/mL endothelial cell growth supplement (E2759, Sigma, MO, USA) for 48 hours. HUVECs were also treated with 1:1 mix of vehicle serum-free medium and HUVECs culture medium as a control. To inhibit colony stimulating factor 1 (CSF1) signaling, HUVECs were treated with 1:1 mix of HPC-CM and HUVECs culture medium in the presence or absence of 30nM pexidartinib (B5854, ApexBio Technology, TX, USA).

2.5. Cell viability, proliferation, and apoptosis assays

We performed the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay to assess cell viability of HPC-CM-treated HUVECs. HUVECs (2,000 cells per well) were seeded in a 96-well plate in the presence of 100 μL culture medium per well. Upon attachment, HUVECs were treated with HPC-CM or vehicle serum-free medium for 48 hours. 20 μL MTT solution (5 mg/mL in PBS) was added. After 4 hours of incubation, media were removed, formazan was resuspended in 150 μL dimethyl sulfoxide (DMSO), and optical density was measured at 560 nm. Bromodeoxyuridine (BrdU) incorporation and terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) assays were conducted to determine increased cell proliferation and reduced apoptosis in response to HPC-CM treatment, respectively. The BrdU Cell Proliferation Assay Kit (2750, Millipore, MA, USA) and the TUNEL Assay Kit - FITC (ab66108, Abcam, Cambridge, UK) were used per the manufacturer’s manual.

2.6. Tubulogenesis assay

The In Vitro Angiogenesis Assay Kit (ECM625, Millipore, MA, USA) was used to determine the ability of HUVECs to form tubular structures according to the manufacturer’s protocol. Briefly, Matrigel solution (100 μL per one well of a 96-well plate) was added and allowed to solidify and polymerize at 37 °C and 5% CO₂ for 1 h. 10,000 cells per well were seeded on solidified Matrigel and incubated in 1:1 mix of HPC-CM and HUVECs culture medium in the presence or absence of 30nM pexidartinib for 12 hours. Quantitative analysis of the extent of tubulogenesis was performed using the Angiogenesis Analyzer for image J [27].

2.7. Enzyme-linked immunosorbent assay (ELISA)

To measure the concentrations of secreted paracrine factors in HPC-CM, we used the CSF1 ELISA Kit (EMCSF1, Thermo, MA, USA), the CSF2 ELISA Kit (BMS612, Invitrogen, MA, USA), the CCL2 ELISA Kit (BMS6005, Invitrogen, MA, USA), and the CXCL3 ELISA Kit (94566, Boster, CA, USA). The CSF1R ELISA Kit (ELH-GCSFR, RayBiotech, GA, USA) was used to measure the protein level of the CSF1 receptor (CSF1R) in HUVECs.

2.8. RNA isolation and quantitative reverse-transcription polymerase chain reaction analyses

Total RNA was extracted from HUVECs and HPCs using the ReliaPrep^™ RNA Tissue Miniprep System (Z6111, Promega, WI, USA). RNA was reverse transcribed using the High-Capacity RNA-to-cDNA^™ Kit (4387406, Applied Biosystems, CA, USA). Quantitative polymerase chain reaction (qPCR) was performed using the PowerUp^™ SYBR^™ Green Master Mix (A25741, Applied Biosystems, CA, USA) and the StepOnePlus^™ Real-Time PCR System (Applied Biosystems, CA, USA). Results were expressed as fold changes relative to TATA-Box binding protein (TBP). Primer sequences are shown in Supplementary Table S1.

2.9. Western blotting analyses

Protein extracts were prepared from cells using the RIPA buffer (786-489, G-Biosciences, MO, USA) supplemented with 1x protease inhibitor cocktail (78429, Thermo, MA, USA) and 1x phosphatase inhibitor cocktail (78420, Thermo, MA, USA). Lysates (20 μg of protein) were loaded on 4–20% gradient gels (3450033, Bio-Rad Laboratories, CA, USA) and transferred to polyvinylidene fluoride membranes (IB24001, Thermo, MA, USA) using the iBlot 2 Dry Blotting System (Thermo, MA, USA). The band intensity was quantified using Image J [26]. The following antibodies were used: Gapdh (ab9485, Abcam, Cambridge, UK), Hnf1α (ab272693, Abcam, Cambridge, UK), Hnf4α (PP-H1415-0C, R&D system, MN, USA), Prox1 (ab199359, Abcam, Cambridge, UK), Goat Anti-Rabbit IgG H&L (HRP) (ab97051, Abcam, Cambridge, UK), and Goat Anti-Mouse IgG H&L (HRP) (ab205719, Abcam, Cambridge, UK).

2.10. RNA sequencing (RNA-Seq)

HUVECs were treated with 1:1 mix of HPC-CM and HUVECs culture medium or 1:1 mix of vehicle serum-free medium and HUVECs culture medium for 48 hours (n = 4 per group), and total RNA was extracted using the ReliaPrep^™ RNA Tissue Miniprep System (Z6111, Promega, WI, USA) and constructed into libraries through the TruSeq RNA Library Prep Kit v2 (Ilumina, CA, USA) following manufacturer’s manuals. Sequencing data were generated via the Illumina Nextseq® 500 platform with 75-bp paired-end read length. Raw reads were processed using Kallisto, which uses pseudoalignment for accurate alignment (Hg38 with annotations provided by UCSC) and quantification with transcripts per million (TPM) as output. Differential analyses were constrained to reasonably expressed transcripts (TPM > 3 in 100% of samples in at least one condition), using a modified t-test with cutoffs of P < 0.05 and fold changes > 1.2 (N = 657 transcripts). Biological and pathway enrichments were identified using ToppGene.cchmc.org, which gathers ontological data from over 30 independent repositories [28].

2.11. Statistical analysis

To determine the significance of differences between groups, Student’s t-tests or one-way analysis of variance (ANOVA) followed by Duncan’s multiple range test were used, where applicable (SAS 9.4, CA, USA). A P-value of less than 0.05 was considered statistically significant.

3. RESULTS

3.1. HPCs enhance HUVECs viability in a paracrine manner

Based on our previous observation that HPCs expressing several paracrine factors accumulate in the peritumoral area and the expression levels of these factors by HPCs correlate with intratumoral cell proliferation and angiogenesis in the livers of patients with liver tumors [24], we hypothesized that HPCs secrete paracrine factors to communicate with HUVECs. To assess the paracrine effect on the cell viability of HUVECs, we used previously established murine HPCs [17] and prepared HPC-conditioned medium (HPC-CM) from undifferentiated HPCs. Equal numbers of HUVECs were seeded at 0 hr and incubated in HPC-CM or vehicle medium (serum-free medium used to prepare HPC-CM) for 24, 48, and 72 hours, respectively (Fig. 1A), and subjected to the MTT assay that is widely used to assess the number of viable cells based on mitochondrial activity [29]. HPC-CM treatment significantly promoted HUVECs viability compared to the vehicle serum-free medium-treated group at all time points. Furthermore, HPC-CM treatment maintained HUVECs viability for up to 48 hours, while the viability was reduced in the control group, indicating that HPC-CM protected HUVECs from serum deprivation-induced cell death.

Fig. 1.

HPCs enhance HUVECs viability and alter mRNA expression in a paracrine manner. A. MTT cell viability assay. Equal numbers of HUVECs were treated with HPC-CM or vehicle serum-free medium for 24, 48, and 72 hours, respectively. Data were expressed as mean ± standard error, n = 3 per group. Asterisks, statistically significant differences between time points (***, P < 0.001). Sharps, HPC-CM vs. serum-free medium at the same time point (###, P < 0.001). B. RNA-seq analysis using HUVECs treated with HPC-CM or vehicle serum-free medium for 48 hours was performed (n = 4 per group) and nine genes were selected based on statistically significant upregulation and their previously reported roles in liver diseases. C. Top upregulated biological processes.

We also performed RNA-seq analyses using HUVECs treated with HPC-CM or vehicle serum-free medium (n = 4 per group) to identify downstream molecular events. RNA-seq analyses identified 305 significantly upregulated genes and 352 downregulated genes in response to HPC-CM compared to the vehicle medium (Supplementary Table S2). Several upregulated genes were those reported to be associated with liver diseases including cirrhosis, alcoholic/non-alcoholic fatty liver disease, cell survival/growth, and HCC (Fig. 1B) [30–39]. These include fatty acid binding protein 4 (FABP4), lymphatic vessel endothelial hyaluronic acid receptor 1 (LYVE1), succinate receptor 1 (SUCNR1), insulin like growth factor binding protein 5 (IGFBP5), matrix Gla Protein (MGP), tumor necrosis factor ligand superfamily member 18 (TNFSF18), CD34 molecule (CD34), C-C motif chemokine ligand 2 (CCL2), and transforming growth factor beta 3 (TGFB3). Furthermore, top upregulated biological processes included vasculature development, cell migration, blood vessel morphogenesis, and angiogenesis (Fig. 1C). These data indicate that HPCs modulate viability and gene expression of HUVECs in a paracrine manner.

3.2. Differentiation medium and overexpression of Hnf4α induce differentiation of HPCs toward hepatocytes

To gain insight into the mechanism underlying the paracrine function of HPCs, we aimed to modulate the differentiation status of HPCs. First, we used a differentiation medium simulating liver maturation during development [17, 25]. This medium contains factors involved in liver development such as FGF4, HGF, and oncostatin M [40]. We previously demonstrated that treatment of HPCs with the differentiation medium is sufficient to induce glycogen deposition as well as global transcriptomic changes resembling those in primary hepatocytes, although the levels of some hepatocyte markers were induced but still remained low compared to primary hepatocytes [17]. Therefore, the present study used this method to assess whether incomplete differentiation is sufficient to affect the paracrine function of HPCs. Second, Hnf4α is a major regulator of hepatocyte differentiation and liver development [41], hence HPCs were transduced with lentiviral vectors to overexpress mouse Hnf4α and RFP (control), treated with puromycin to select transduced cells, and kept undifferentiated or treated with the differentiation medium. Therefore, we investigated and compared the following four groups throughout the study: Undifferentiated HPCs treated with the RFP-overexpressing vector (undifferentiated RFP-HPCs) or the Hnf4α-overexpressing vector (undifferentiated Hnf4α-HPCs), differentiated RFP-HPCs, and differentiated Hnf4α-HPCs. To confirm whether the mRNA levels of Hnf4a are upregulated upon treatment of HPCs with lentiviral vectors, we used qPCR (Fig. 2A). Hnf4a transcripts were detected in HPCs treated with the Hnf4α overexpression vector, both in undifferentiated and differentiated groups. Interestingly, while Hnf4a levels were not different between undifferentiated HPCs treated with RFP- and Hnf4α-overexpressing vectors, we detected a significant increase in the Hnf4a level upon treatment of Hnf4α-transduced HPCs with the differentiation medium.

Fig. 2.

Induction of hepatocyte marker expression in HPCs. A. qPCR analyses of the Hnf4a mRNA. HPCs transduced with RFP- or Hnf4α-expressing vectors were maintained in the HPCs culture medium (undifferentiated, black bars) or treated with the differentiation medium (differentiated, white bars). B. Western blotting analysis using an antibody that detects both P1 (53 kDa) and P2 (49~50 kDa) isoforms of Hnf4α. Glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used as the normalizing control. C. Quantification of western blot signals. D. qPCR analyses of hepatocyte markers. All data were expressed as mean ± standard error, n = 3 per group. Asterisks, undifferentiated vs. differentiated (**, P < 0.01; ***, P < 0.001). Sharps, undifferentiated RFP vs. undifferentiated Hnf4α (#, P < 0.05). Daggers, differentiated RFP vs. differentiated Hnf4α (†, P < 0.05; ††, P < 0.01).

The transcription of Hnf4a is regulated by two promoters, the proximal P1 promoter and the distal P2 promoter [42]. While P2-Hnf4α is more abundant than P1-Hnf4α in the mouse fetal liver, P1-Hnf4α becomes the predominant isoform after birth [43]. The lentiviral vector used to overexpress Hnf4a is designed based on the transcript NM_008261 which encodes the longest transcript variant leading to P1-Hnf4α expression. Our western blot data indicate that while transduction significantly induces the expression of P1-Hnf4α in both undifferentiated Hnf4α-HPCs and differentiated Hnf4α-HPCs compared to corresponding RFP controls, differentiated Hnf4α-HPCs showed the most dramatic induction as compared to other groups (Fig. 2B, C). Furthermore, a smaller-sized band was detected in all groups, suggesting that HPCs express the fetal isoform P2-Hnf4α, which levels were reduced in response to the differentiation medium (Fig. 2B, C). Of note, our qPCR primers detect both P1 and P2 transcripts, which explains a relatively moderate changes in the mRNA level as compared to the magnitude of P1-Hnf4α protein induction.

Next, to determine whether the differentiation medium and/or overexpression of Hnf4α efficiently differentiate HPCs toward hepatocytes, we measured the mRNA levels of hepatocyte makers. Alcohol dehydrogenase 1 (Adh1), albumin (Alb), arginase (Arg1), cytochrome P450, family 2, subfamily E, polypeptide 1 (Cyp2e1), fumarylacetoacetate hydrolase (Fah), fibrinogen alpha chain (Fga), 3-hydroxy-3-methylglutaryl-coenzyme A synthase 2 (Hmgcs2), tyrosine aminotransferase (Tat), and vitronectin (Vtn) were used as hepatocyte makers in this study [44–51]. The mRNA levels of Adh1, Alb, Arg1, Fah, Fga, Hmgcs2, Tat, and Vtn were significantly upregulated in differentiated RFP-HPCs as compared to undifferentiated RFP-HPCs (Fig. 2D). In differentiated Hnf4α-HPCs, the mRNA levels of Adh1, Alb, Cyp2e1, Fah, Fga, Hmgcs2, Tat, and Vtn were dramatically increased compared to undifferentiated Hnf4α-HPCs. Our data indicate that the differentiation medium alone is sufficient to induce Adh1, Arg1, Fah, Fga, and Hmgcs2 to the levels comparable to or higher than those of primary hepatocytes (Supplementary Table S3), and Hnf4α induces markers of hepatocytes in a culture medium-dependent manner. In addition, the mRNA levels of Adh1, Alb, Cyp2e1, Fah, Fga, Hmgcs2, and Vtn in differentiated Hnf4α-HPCs were higher than those of differentiated RFP-HPCs, suggesting that the combination of the differentiation medium and Hnf4α overexpression results in enhanced expression of some markers as compared to the differentiation medium alone. We also performed various assays to assess the induction of hepatocyte functions [52, 53]. Albumin production, intracellular glycogen contents, and LDL uptake were significantly induced in response to the differentiation medium and/or Hnf4α overexpression, although urea production remained unchanged (Fig. 3A–D). These results were in line with the qPCR data indicating that differentiated Hnf4α-HPCs express the highest levels of many hepatocyte markers compared to other groups, but the level of the urea cycle enzyme Arg1 was not induced in differentiated Hnf4α-HPCs compared to the RFP control (Fig. 2D), indicating that Hnf4α regulates some, but not all, aspects of liver functions. In summary, our data indicate that the differentiation medium and Hnf4α overexpression can be used to push the differentiation of HPCs toward hepatocytes.

Fig. 3.

Functional analyses of HPCs differentiated into hepatocytes. A. Concentrations of albumin secreted into the culture medium. HPCs transduced with RFP- or Hnf4α-expressing vectors were maintained in the HPCs culture medium (undifferentiated) or treated with the differentiation medium (differentiated). Data were expressed as means (SE), n = 3. Asterisks, undifferentiated vs. differentiated (**, P < 0.01; ***, P < 0.001). Sharps, vs. undifferentiated RFP (###, P < 0.001). Daggers, vs. differentiated RFP (†††, P < 0.001). B. Intracellular glycogen contents. C. Quantification of LDL-positive areas. D. Concentrations of urea secreted into the culture medium. E. qPCR analyses of liver-enriched transcription factors. F. Western blotting analyses of Hnf1α and Prox1. Gapdh was used as the normalizing control. G. Quantification of western blot signals. (B-E, G) All data were expressed as mean ± standard error, n = 3 per group. Asterisks, undifferentiated (black bars) vs. differentiated (white bars) (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Sharps, undifferentiated RFP vs. undifferentiated Hnf4α (#, P < 0.05; ##, P < 0.01; ###, P < 0.001). Daggers, differentiated RFP vs. differentiated Hnf4α (†, P < 0.05; †††, P < 0.001).

3.3. The expression level of liver-enriched transcription factors is upregulated in differentiated HPCs

To determine the molecular mechanisms by which the differentiation medium and overexpression of Hnf4α induce the markers of hepatocytes, we evaluated the mRNA levels of additional hepatic transcription factors. Presently, several liver-enriched transcription factors have been identified [54–56]. These factors regulate each other and form a transcriptional hierarchy for determination and maintenance of hepatic functions. Especially, Hnf4α is upstream of other factors and orchestrates liver development [57, 58].

Therefore, to assess the transcription factors downstream of Hnf4α, we determined the mRNA levels of liver-enriched transcription factors forkhead box A3 (Foxa3), hepatocyte nuclear factor 1 homeobox A (Hnf1a), and prospero homeobox 1 (Prox1) [59, 60]. Treatment of RFP- or Hnf4α-expressing HPCs with the differentiation medium increased the mRNA levels of Foxa3, Hnf1a, and Prox1 (Fig. 3E). Overexpression of Hnf4α in differentiated HPCs upregulated the mRNA levels of Foxa3, Hnf1a, and Prox1 as compared to RFP controls, in line with the expression pattern of the Hnf4a mRNA and P1 protein (Fig. 2A–C). Western blotting data indicated that the protein levels of Hnf1α and Prox1 are in good agreement with the qPCR data (Fig. 3F, G). These results indicate that while both the differentiation medium and Hnf4α overexpression increase the degree of hepatic maturation, Hnf4α combined with the differentiation medium is the most effective way to induce differentiation of HPCs and imply a potential involvement of Foxa3, Hnf1a, and Prox1 during this process.

3.4. Differentiation of HPCs toward hepatocytes reduces the paracrine enhancement of HUVECs viability

The next question we asked was whether hepatocytic differentiation of HPCs suppresses their paracrine function. To determine the effect of the differentiation medium and Hnf4α overexpression on HPC-CM-enhanced HUVECs viability, we prepared HPC-CM by treating the four groups of HPCs with the vehicle serum-free medium: undifferentiated RFP-HPCs, undifferentiated Hnf4α-HPCs, differentiated RFP-HPCs, and differentiated Hnf4α-HPCs. The MTT assay indicated that when RFP-expressing HPCs are treated with the differentiation medium prior to the preparation of the conditioned medium, this significantly lowered the ability of HPC-CM to promote HUVECs viability to the level comparable to the vehicle medium-treated group (Fig. 4A). Overexpression of Hnf4α in undifferentiated HPCs also lowered the ability of HPC-CM to enhance HUVECs viability compared to the undifferentiated RFP group. However, overexpression of Hnf4α in differentiation medium-treated HPCs did not further affect HUVECs viability compared to the differentiated RFP group, indicating that the differentiation medium maximally suppressed the paracrine function of HPCs. We also conducted the BrdU incorporation assay to measure the extent of cell proliferation. While we did not observe any effects upon overexpression of Hnf4α, the differentiation medium reduced proliferation of HUVECs in the RFP groups (Fig. 4B). In summary, our data indicate that incomplete differentiation of HPCs upon treatment with the differentiation medium is sufficient to inhibit the paracrine enhancement of HUVECs viability and proliferation.

Fig. 4.

Differentiation of HPCs toward hepatocytes suppresses the paracrine enhancement of HUVECs viability. A. MTT assay. HUVECs were treated for 48 hours with HPC-CM prepared from HPCs transduced with RFP- or Hnf4α-expressing vectors maintained in the HPCs culture medium (undifferentiated, black bars) or treated with the differentiation medium (differentiated, white bars). B. BrdU incorporation assay. All data were expressed as mean ± standard error, n = 3 per group. Asterisks, undifferentiated vs. differentiated (*, P < 0.05; ***, P < 0.001). Sharp, undifferentiated RFP vs. undifferentiated Hnf4α (#, P < 0.05). VM, vehicle serum-free medium used to prepare HPC-CM.

3.5. Differentiation of HPCs toward hepatocytes alters the mRNA levels of angiogenic paracrine factors

Having confirmed that the paracrine effect of HPC-CM is inhibited upon differentiation of HPCs, we aimed to investigate how the differentiation medium and overexpression of Hnf4α affect the mRNA levels of angiogenic paracrine factors in HPCs. Fourteen angiogenic factors selected based on their possible paracrine roles in animal models and association with liver disease, HCC, and cholangiocarcinoma have been investigated (Fig. 5) [61–71]: Angpt1, angiopoietin 2 (Angpt2), Csf1, colony stimulating factor 2 (Csf2), Ccl2, C-X-C motif chemokine ligand 3 (Cxcl3), C-X-C motif chemokine ligand 5 (Cxcl5), fibroblast growth factor 1 (Fgf1), fibroblast growth factor 18 (Fgf18), insulin-like growth factor (Igf), platelet-derived growth factor B (Pdgfb), Pdgfc, Vegf, and Vegfd. Treatment of RFP-expressing HPCs with the differentiation medium significantly downregulated the mRNA levels of Csf1, Csf2, Ccl2, Cxcl3, Cxcl5, Fgf1, and Fgf18. Overexpression of Hnf4α in addition to the differentiation medium further downregulated the levels of Csf1, Ccl2, and Cxcl5 compared to the differentiated medium alone. In summary, we found that several paracrine factors are downregulated upon induction of HPCs differentiation in line with the MTT data (Fig. 4A), implying that these paracrine factors mediate HPCs-enhanced HUVECs viability.

Fig. 5.

Alteration of the mRNA levels of angiogenic paracrine factors in response to induced differentiation of HPCs. HPCs transduced with RFP- or Hnf4α-expressing vectors were maintained in the HPCs culture medium (undifferentiated, black bars) or treated with the differentiation medium (differentiated, white bars), and subjected to qPCR analyses. Data were expressed as mean ± standard error, n = 4 per group. Asterisks, undifferentiated vs. differentiated (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Sharps, undifferentiated RFP vs. undifferentiated Hnf4α (#, P < 0.05). Daggers, differentiated RFP vs. differentiated Hnf4α (†, P < 0.05).

3.6. Secretion of paracrine factors by HPCs is inhibited upon differentiation

To demonstrate that HPCs secrete paracrine factors into HPC-CM, we selected CSF1, CSF2, CCL2, and CXCL3 for subsequent analyses, because their mRNA levels were dramatically downregulated upon differentiation (Fig. 5). Indeed, all four paracrine factors were detected in HPC-CM, and differentiation of HPCs significantly downregulated the protein levels of these factors secreted into HPC-CM (Fig. 6). While both CSF1 and CSF2 protein levels were low in HPC-CM collected from differentiated HPCs compared to undifferentiated HPCs in agreement with the MTT data (Fig. 4A), the absolute concentration of CSF2 was the lowest among selected paracrine factors. CCL2 was relatively abundant, but the differentiation status had little effect on the protein level. Therefore, to determine whether inhibition of a paracrine factor can block the effect of HPC-CM on HUVECs viability and mRNA expression, we selected CSF1 for subsequent investigation.

Fig. 6.

Protein levels of paracrine factors secreted by HPCs are altered in response to induced differentiation. Paracrine factor levels in HPC-CM were determined using ELISA. Data were expressed as mean ± standard error, n = 3 per group. Asterisks, undifferentiated (black bars) vs. differentiated (white bars) (*, P < 0.05; **, P < 0.01). Sharps, undifferentiated RFP vs. undifferentiated Hnf4α (#, P < 0.05). Daggers, differentiated RFP vs. differentiated Hnf4α (†, P < 0.05).

3.7. Inhibition of CSF1 signaling reduces the paracrine effect of HPCs on HUVECs

Recent studies reported that CSF1/CSF1 receptor (CSF1R) signaling is a key regulator of tumor-associated macrophage differentiation and survival [72, 73] and also promotes angiogenic functions of monocytes [74, 75]. Therefore, we used pexidartinib, an inhibitor of CSF1R [76], to determine whether CSF1 mediates the paracrine enhancement of HUVECs viability. Treatment of HUVECs with HPC-CM prepared from undifferentiated RFP-HPCs and undifferentiated Hnf4α-HPCs in the presence of 30 nM pexidartinib significantly reduced the cell viability compared to DMSO-treated control groups (Fig. 7A), while pexidartinib did not affect the protein level of CSF1R in HUVECs (Supplementary Fig. S1A). When HPC-CM collected from differentiated HPCs was used, pexidartinib had no further effect. This may be due to the fact that the differentiation medium already significantly reduced CSF1 levels (Fig. 6).

Fig. 7.

CSF1 is a mediator of HPCs-induced paracrine effects on HUVECs. A. MTT assay. HUVECs were treated for 48 hours with HPC-CM prepared from HPCs transduced with RFP- or Hnf4α-expressing vectors maintained in the HPCs culture medium (undifferentiated, black bars) or treated with the differentiation medium (differentiated, white bars). Asterisks, undifferentiated vs. differentiated (*, P < 0.05). Sharps, each black bar vs. undifferentiated RFP-DMSO (#, P < 0.05). Daggers, DMSO vs. Pexi (†, P < 0.05). DMSO, dimethyl sulfoxide. Pexi, 30 nM pexidartinib. B. qPCR analyses of HUVECs treated with HPC-CM for 48 hours. Asterisks, DMSO (black bars) vs. Pexi (white bars) (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Sharps, undifferentiated DMSO vs. differentated DMSO (##, P < 0.01; ###, P < 0.001). Daggers, undifferentiated Pexi vs. differentiated Pexi (†, P < 0.05; ††, P < 0.01; †††, P < 0.001). C. qPCR analyses of apoptosis markers using HUVECs treated with HPC-CM for 48 hours. Asterisks, DMSO (black bars) vs. Pexi (white bars) (*, P < 0.05; ***, P < 0.001). Sharps, undifferentiated DMSO vs. differentated DMSO (#, P < 0.05; ###, P < 0.001). Dagger, undifferentiated Pexi vs. differentiated Pexi (†, P < 0.05). D. Representative images of TUNEL-positive HUVECs treated with vehicle medium (VM, left panel) and HPC-CM prepared from the undifferentiated DMSO group (right panel). E. Quantification of the percentage of TUNEL-positive HUVECs. Asterisks, DMSO vs. Pexi (**, P < 0.01). Sharps, vs. VM (#, P < 0.05). Daggers, undifferentiated DMSO vs. differentiated DMSO (††, P < 0.01). All data were expressed as mean ± standard error, n = 3 per group.

We also investigated whether pexidartinib abolishes HUVECs gene expression induced by HPC-CM in Fig. 1B. Consistent with our qPCR and ELISA data showing that treatment of HPCs with the differentiation medium inhibits expression of several paracrine factors (Fig. 5 and 6), the mRNA levels of LYVE1, SUCNR1, IGFBP5, and MGP were low in HUVECs treated HPC-CM from differentiated HPCs compared to undifferentiated HPCs within the DMSO groups (Fig. 7B). In addition, the mRNA levels of FABP4, LYVE1, SUCNR1, IGFBP5, MGP, and CCL2 were downregulated in response to pexidartinib treatment within the undifferentiated groups. To investigate the effect of CSF1R inhibition on the levels of cell proliferation and apoptosis markers, we also determined the mRNA levels of proliferating cell nuclear antigen (PCNA), BCL2 antagonist/killer 1 (BAK1), and BCL2 associated X (BAX). As shown in Fig. 7C, the levels of apoptosis markers were high in HUVECs treated with HPC-CM from differentiated HPCs compared to undifferentiated HPCs within the DMSO groups, and inhibition of CSF1R also promoted mRNA expression of apoptosis markers. As the MTT assay measures the number of viable cells based on the total mitochondrial activity and does not distinguish proliferation, cell death, and metabolic activity [29, 77], we conducted the TUNEL assay to measure the extent of apoptosis. Our data indicated a dramatic reduction of apoptosis in HUVECs treated with HPC-CM prepared from undifferentiated HPCs compared to vehicle medium-treated HUVECs (Fig. 7D, E). Inhibition of CSF1R significantly increased the percentage of TUNEL-positive cells and the effect was comparable to differentiation of HPCs, in line with the MTT data in Fig. 7A.

We then performed the tubulogenesis assay to examine the paracrine effect of HPCs on HUVECs’ ability to form tubular structures. We quantified the extent of tubulogenesis using the Angiogenesis Analyzer, which models the tubular structure at multiple levels and captures series of vectorial elements (Fig. 8A) [27]. Our results indicated that HPC-CM harvested from undifferentiated HPCs induced tubulogenesis capabilities of HUVECs as compared to the vehicle medium (Fig. 8B). We observed a significant increase of multiple parameters in response to HPC-CM as compared to the vehicle medium in terms of the number of nodes, master junctions, and meshes, as well as the number and length of master segments, and these effects were dramatically suppressed by pexidartinib treatment in the undifferentiated groups. Differentiation of HPCs also suppressed the ability of HPC-CM to increase the number of junctions, meshes, and master segments. In summary, these results suggest that CSF1 partly mediates pleiotropic and paracrine effects of HPCs on HUVECs, and imply a potential role of HPCs-secreted CSF1 in the progression of liver diseases.

Fig. 8.

HPCs-induced tubulogenesis of HUVECs is dependent on CSF1. A. A representative image of HUVECs induced for tubulogenesis in the presence of HPC-CM prepared from undifferentiated HPCs and elements measured using the Angiogenesis Analyzer are shown. B. Quantification of the angiogenic parameters of HUVECs induced for tubulogenesis in the presence of HPC-CM prepared from undifferentiated HPCs and differentiated HPCs. Data were expressed as mean ± standard error, n = 3 per group. Asterisks, DMSO (black bars) vs. Pexi (white bars) (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Sharps, vs. vehicle medium (VM) (#, P < 0.05; ##, P < 0.01; ###, P < 0.001). Daggers, undifferentiated DMSO vs. differentiated DMSO (†, P < 0.05; ††, P < 0.01). DMSO, dimethyl sulfoxide. Pexi, 30 nM pexidartinib.

4. DISCUSSION

This study revealed several novel findings that; (a) HPCs promote endothelial cell viability and proliferation and reduce apoptosis in a paracrine manner; (b) HPCs-secreted factors stimulate HUVECs to express genes involved in several aspects of endothelial biology; (c) the CSF1/CSF1R signaling axis partly mediates the paracrine effects of HPCs on endothelial cells, and; (d) imperfect differentiation of HPCs toward hepatocytes induced by the differentiation medium is sufficient to inhibit the paracrine function of HPCs. Of note, it has been observed that stem/progenitor cells in other tissues including bone marrow and skeletal muscle secrete paracrine factors [78], and monocyte chemotactic protein secretion by pre-adipocytes is reduced upon differentiation of pre-adipocytes into adipocytes [79] implying that the inverse relationship between differentiation and paracrine factor secretion is a common feature of stem/progenitor cells.

To dissect the relationship between hepatocytic differentiation of HPCs and their paracrine communication with endothelial cells, two protocols for the induction of HPCs differentiation were used; the differentiation medium and overexpression of Hnf4α. Our data suggest that the combination of the differentiation medium and Hnf4α overexpression is the most effective way to induce hepatocyte markers in this study. Interestingly, treatment of HPCs with the differentiation medium alone was sufficient to abolish the paracrine effect on HUVECs viability, although overexpression of Hnf4α also altered the mRNA levels of some paracrine factors. Due to the slow proliferation rate of HUVECs [80], the MTT assay may not have fully reflected the alteration in the mRNA levels. Another possible explanation is that the conversion of HPCs into mature hepatocytes is a multiple-step process, and the early-stage commitment of HPCs toward the hepatocytic fate is sufficient to inhibit the paracrine function. This is in accordance with our data that although induced differentiation significantly inhibited the paracrine function of HPCs and upregulated expression of Adh1, Arg1, Fah, Fga, and Hmgcs2 to the levels comparable to primary hepatocytes, the mRNA levels of other hepatocyte markers and albumin secretion were still lower than those of primary hepatocytes. Further research is needed to determine which stages of HPCs differentiation and functions are regulated by the individual component of the medium and Hnf4α. Indeed, the expression of VEGF was elevated in response to the differentiation medium, which contains HGF previously reported to induce VEGF expression [81]. Additional subjects of future studies would include the mechanism underlying the regulation of P1- and P2-Hnf4α expression in response to culture media and differential roles of these isoforms in differentiation and functions of HPCs for the following reasons: (a) while we did not detect changes in Hnf4a mRNA levels in response to the Hnf4α-expressing vector in undifferentiated HPCs, there was a dramatic induction of Hnf4a mRNA and P1 protein isoform in the presence of the differentiation medium compared to the RFP control; (b) the differentiation medium induced the protein level of P1 isoform and reduced the P2 protein level.

CSF1 not only regulates macrophage differentiation and recruitment, but also educates monocytes/macrophages toward angiogenic phenotypes [74, 75]. Furthermore, the CSF1/CSF1R axis plays an important role in tumor-promoting macrophages as well as liver cancer growth [61, 82, 83]. However, whether HPCs can be a source of CSF1 and the events downstream of HPCs-produced CSF1 were unknown. Here, our qPCR and ELISA data show that CSF1 in HPC-CM decreases expression of apoptosis markers in HUVECs. Inhibition of CSF1R significantly suppressed HPC-CM-enhanced viability and tubulogenesis of HUVECs as well as expression of genes involved in liver disease and endothelial cell biology. This indicates the involvement of CSF1 in the paracrine action of HPCs on HUVECs and suggests that CSF1 is a potential therapeutic target. Data from the Human Protein Atlas indicate that the CSF1R transcript is detected in not only Kupffer cells but also sinusoid endothelial cells and vascular endothelial cells in human liver tissues (Supplementary Fig. S1B) [84]. Our ELISA data confirmed expression of the CSF1R protein in HUVECs (Supplementary Fig. S1A), in line with the observation made by Ao and colleagues indicating that HUVECs express this gene [61]. In addition, the CSF1R protein was detected in the capillary endothelial cells of the mouse central nervous system and in the mouse brain capillary endothelial cell line MBEC4 [85], and the existence of CSF1R lineage endothelial cells has been reported in mouse glioma [86], implying a potential role of CSF1R signaling in the endothelial compartment of various tissues.

Although this study mainly focused on the role of HPCs-secreted factors in communication with endothelial cells, a future area of investigation would be to track in vivo roles of HPCs in several pathogenic mechanisms, since the paracrine factors expressed by HPCs are also involved in fibrosis, inflammation, and tumorigenesis in addition to angiogenesis. Our study also indicates that differentiation of HPCs not only upregulates hepatocyte markers but also downregulates CSF1 and other angiogenic factors. These results imply that in addition to direct inhibition of HPCs-secreted factors, induction of HPCs differentiation can be a valid strategy for enhancement of liver gene expression as well as inhibition of HPCs-enhanced pathogenic progression.

5. CONCLUSION

This study demonstrates that HPCs communicate with endothelial cells in a paracrine manner and reveals the CSF1/CSF1R axis as a potential therapeutic target. Our results also provide fundamental insight into progenitor cell biology, i.e. an inverse relationship between differentiation and paracrine function.

Highlights.

• Hepatic progenitor cells (HPCs) enhance endothelial viability in a paracrine manner

• Paracrine factors secreted by HPCs induce endothelial cell gene expression

• Differentiation of HPCs toward hepatocytes inhibits their paracrine function

• The CSF1/CSF1R axis mediates the paracrine effects of HPCs on endothelial cells