Silencing of hepatic fate-conversion factors induce tumorigenesis in reprogrammed hepatic progenitor-like cells

Mouse embryonic fibroblasts (MEF) were prepared from 13.5-day post-coitum embryos. MEF were grown in DMEMc (Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10 % fetal bovine serum (FBS) and 2 mM Glutamax). In the reprogramming experiments two different media were used: hepatocyte conditioned medium (HCM) I and HCM II. HCM I is composed of IMDM:F12 (1:3), supplemented with 10 % FBS, 2 mM l-glutamine, penicillin/streptomycin, 10 ng/ml epidermal growth factor (EGF), 100 ng/ml fibroblast growth factor (FGF)2, 50 ng/ml vascular endothelial growth factor (VEGF), and 100 ng/ml transforming growth factor (TGF)β. HCM II is composed of IMDM:F12 (1:3), supplemented with 10 % FBS, 2 mM l-glutamine, penicillin/streptomycin, 10 ng/ml hepatocyte growth factor (HGF), and 10 ng/ml Oncostatin M. All media was purchased from Invitrogen (www.thermofisher.com↗). Growth factors were purchased from R&D Systems (www.rndsystems.com↗). iHepL cells exhibited enhanced attachment to the culture dishes and needed trypsinization for 30 min at 37 °C for passaging.

All cells were maintained at 37 °C with 5 % CO₂ and were regularly examined with an Olympus CKX41 microscope. Images were taken on an Olympus FV1000 confocal mounted on an IX81 inverted microscope.

The retroviral constructs pMIGR1-Hhex, pMIGR1-Hnf1a, and pMIGR1-Hnf6a were generated by polymerase chain reaction (PCR) amplification of the cDNAs (see Additional file 1: Table S1 for oligo sequence) followed by subcloning into the XhoI-EcoRI restriction sites of pMIGR1 [16]. All constructs were verified by sequencing. pBabe-Foxa2, pBabe-Hnf4a, and pBabe-Gata4 are derivatives of the pBabe-puro retroviral vector [17] donated by Dr. Ken Zaret (University of Pennsylvania, Philadelphia, PA, USA). The plasmids encoding the reprogramming factors pMXs-Oct4, pMXs-Sox2, pMXs-Klf4, and pMXs-cMyc were purchased from Addgene (Cambridge, MA, USA; www.addgene.com↗) [18]. A summary of the retroviral plasmids is shown in Additional file 1 (Table S2). Ecotropic retroviruses were generated in 293 T cells as described elsewhere [19]. MEF were infected with equal volumes of each retrovirus.

Mice hepatocytes were isolated using a two-step perfusion technique as previsouly described [20]. Briefly, the liver was pre-perfused through the portal vein with calcium-free buffer (118 mM NaCl, 4.7 mM KCl, 1.2 mM H₂KPO₄, 1.2 mM Mg₂SO₄, 25 mM HNaCO₃, 10 mM glucose, 0.5 mM EGTA, pH 7.4) and then perfused with the same buffer containing 2.5 mM CaCl₂ and 125 U/ml collagenase IV. Once the enzymatic digestion was completed, the liver was transferred to a petri dish and the cells were gently dispersed with a blunt tool. Cells were collected by low-speed centrifugation. Viability of isolated hepatocytes was around 90 % as determined by Trypan blue.

Approximately 10⁶ early passage (passage 2 or 3) MEF were seeded on a 10-cm dish containing DMEMc. One day later, cells were infected with indicated viruses supplemented with 4 μg/ml polybrene for 24 h. Seventy two hours later, media were changed to HCM I and the culture continued for an additional 18 days. Culture medium was refreshed every day. Cells were frozen at this stage at 1.5 × 10⁶ cells per vial to generate cell stocks for future use. The final differentiation step was achieved by thawing one vial into a 10-cm dish containing HCM II media with further culturing for 6 days. Cells should be split if confluent by using TrypLE™ Select. If needed, cells can be maintained for longer periods by regular passaging. Phenotypic characterization of iHepL cells was performed at confluency. All the data shown in the paper come from the average of two biological replicates from each of the three infections.

To obtain isogenic cell clones, cells were seeded at low density and colonies were transferred to 96-well plates and expanded. Data depicted in the figures in Additional filewere obtained from two biological replicates from clone 2 (isolated from infection #1), clone 6 (isolated from infection #2), and clone 15 (isolated from infection #3). 1

Total RNA was extracted using the RNeasy mini kit (www.qiagen.com↗) and reverse-transcribed using Moloney Murine Leukemia Virus reverse transcriptase (www.thermofisher.com↗) according to the manufacturer’s protocol. Genomic DNA was extracted from paraffin-embedded tumors and tissues using the Qiamp DNA FFPE Tissue Kit (www.qiagen.com↗). PCR amplification was performed using the expand high fidelity PCR system (Roche, Basel, Switzerland; www.lifescience.roche.com↗) following the manufacturer’s instructions on a Light Cycler 480 II Real-Time PCR System using the Light Cycler 480 SYBR Green I Master. The specificity of the amplified PCR products was confirmed by analysis of the melting curve and agarose gel electrophoresis. Primers used for the quantitative PCR (qPCR) are shown in Additional file 1: Tables S3 and S4. Primers designed to measure endogenously expressed transcription factors were validated by running a PCR assay using the corresponding retroviral shuttle plasmids. The relative expression of each mRNA was normalized against mouse beta Actin (Actb). The relative gene copy number in genomic DNA was estimated as described elsewhere [21] using a fragment amplified from intron 1 of Ccnd1for normalization. Primers used to amplify exogenous genes were validated using the corresponding retroviral plasmids.

MicroRNA was extracted with an adapted protocol from Qiagen using the miRNeasy kit which consisted of precipitating with 1.5 volumes of 100 % ethanol instead of 1 volume of 70 % ethanol. miR122-specific reverse transcription stem loop was performed using the primer 5′-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACCAAACA-3′ as described in [22]. qPCR was run using miR-122 specific primers 5′-TTGGAGCTCCCTTTTTGCTA-3′ and 5′-CACCATGCCTGGCTAATTTT-3′.

Cells were fixed with 4 % paraformaldehyde at room temperature, and then washed three times with phosphate-buffered saline (PBS). Cells were blocked with blocking buffer (5 % normal donkey serum, 0.3 % Triton X-100 in PBS) for 60 min at room temperature and then incubated with primary antibodies at 4 °C overnight in 0.1× blocking buffer. The next day cells were washed three times with PBS, and then incubated with Alexa Fluor conjugated secondary antibodies at 1/500 dilution (Life Technologies) for 60 min at room temperature in the dark. Nuclei were stained with DAPI (200 ng/ml for 15 min). The high content screening imaging station Scan® from Olympus was used to analyze the number and intensity of cell staining. Total cell number was calculated from the number of DAPI-positive nuclei (software used).

Sections of the paraffin-embedded tissues were dewaxed and rehydrated through ethanol:water series and unmasked by boiling in sodium citrate buffer (10 mM, pH 6.0, for 10 min). Immunostaining was performed as described above, but the blocking buffer contained 10 % normal donkey serum in PBT (0.1 % Tween 20 in PBS). Slides were mounted with fluorescence mounting media (Dako). Antibodies used for immunofluorescence staining are described in Additional file(Table S5). 1

Adherent cells were washed with PBS and detached by treatment with trypsin (Life Technologies). Cells fixed with 4 % paraformaldehyde for 10 min at 4 °C were permeabilized with 90 % methanol for 30 min at 4 °C. After blocking with 0.5 % ovalbumin, cells were incubated with the primary antibody or normal serum (control) for 60 min at room temperature. Cells were then incubated with secondary antibody for 30 min in the dark. Cells were analyzed by FACSCanto Flow Cytometer (Becton Dickinson). Nuclei were stained with DAPI (200 ng/ml in PBS) for 15 min in the dark. Primary antibodies used for flow cytometry are described in Additional file(Table S5). To determine the percentage of live green fluorescent protein (GFP)-positive cells, trypsinized cells were resuspended in complete DMEM containing 50 μg/ml 7-amino actinomycin D (7-AAD) and incubated for 30 min in the dark. Data were analyzed using FACSDiva software (Becton Dickinson). Doublets were eliminated using a pulse geometry gate (FSC-H x FSC-A). 1

Confluent cells were incubated with fresh HCM II medium supplemented with 1 mg/ml indocyanine green at 37 °C for 30 min. Once pictures were taken, cells were washed with PBS and cultured in media without indocyanine green. Cells were stained by periodic acid-Schiff (PAS; Sigma) following the manufacturer’s instructions. To induce glycogen depletion by starvation, iHepL cells cultured in six-well plates were incubated with Krebs-Henseleit solution without glucose. Cells were fixed at 0, 1.5, 3, and 6 hours and stained with PAS to estimate intracellular glycogen content. To induce glycogen depletion with glucagon, cells were incubated in HCM II media containing 300 ng/ml glucagon.

Cyp450 activities were measured essentially as described previously [23]. Briefly, confluent iHepL cells cultured in six-well plates were incubated in media containing eight substrates for 4 h at 37 °C. Metabolites were measured in the supernatant by liquid chromatography mass spectrometry (LC-MS/MS). Total cell protein was used to normalize the data. Isolated primary hepatocytes and wild-type MEF were used as a positive and negative control, respectively. Formation of estradiol-3-glucuronide and estradiol-17-glucuronide was measured in cells incubated with β-estradiol 100 μM for 2 h as described previously [24].

The Animal Care and Ethics Committee from CIEMAT specifically approved this study (approval number HEM4-12). The mouse strain used was NSG (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ). Anesthesia was induced by intraperitoneal injection of a mixture of ketamine (125 mg/kg) and medetomidine (10 mg/kg). To provide a proliferative stimulus to hepatic cells, a 45 % partial hepatectomy (PH) was conducted, removing the right and left lateral lobules through a transversal midline incision in the abdomen of NSG mice. Immediately after PH, the animals were placed in a supine position inside a lead box where a hole was made to allow the localized irradiation of the median hepatic lobule. Two 3 × 5-cm lead shields, each 2 mm thick, were fitted under the liver to protect the stomach and intestines. The liver was irradiated using Philips MG324 X-ray equipment (Philips, Hamburg, Germany) to deliver a dose of 41.05 Gy, 260 kV, 12.3 mA, at a dose rate of 4.105 Gy/min. The abdomen was then closed in two layers. Anesthesia was reverted by intraperitoneal injection of atipamezol (50 mg/kg). Cell transplantation was performed 4 days after PH and hepatic irradiation. Mice were anesthetized and the spleen was exposed by a left flank incision. Homeostasis was achieved by ligation of the splenic tip using a 2-0 silk suture. To validate our model, 10⁶ hepatocytes freshly isolated by collagenase perfusion of the liver from ubiquitously expressing mRFP mice were injected (data not shown). In subsequent experiments, 10⁶ iHepL cells were injected. Three weeks after transplantation, mice were sacrificed and pieces of the liver were fixed in formalin and histologically evaluated after paraffin-embedding.

Male, 6- to 8-week-old NSG (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ) were kept at least 1 week before experimental manipulation. Cultured iHepL cells (1.5 × 10⁶/injection) in 150 μl DPBS were implanted subcutaneously in the right flank region of mice (n = 5). Tumor progression was followed for up to 6 weeks. Upon sacrifice, primary tumors were removed, formalin-fixed, and histologically evaluated.

All in vitro experiments were conducted a minimum of three times. Data are expressed as the mean ± standard deviation. Statistical significance was calculated by unpaired Student’s t test.

Considering that reprogramming is a stochastic process, we reasoned that initial cell reprogramming by expression of OSK/M could be diverged towards the hepatic cell fate by co-expressing key determinants of hepatic cell fate [25]. MEF were infected with retroviral vectors expressing Gata4, Foxa2, Hnf4a, Hhex, Hnf1a, and Hnf6a, with or without OSK/M, and cultured with DMEMc for 2 days. Then, cells were split into two gelatine-coated 10-cm plates; one plate was maintained in DMEMc and the other in HCM I containing essential growth factors for maintenance of hepatic lineage. Leukemia inhibitory factor (LIF) was not included in the reprogramming media to hamper commitment to the pluripotent fate.

When looking at the reprogramming plates at day 7, we identify groups of epithelial cells with flattened, enlarged, and polygonal shapes between fibroblasts (Fig. 1a and Additional file 1: Figure S1A). The morphology of iHepL cells resembled that of hepatic endoderm, not mature polygonal hepatocytes. At later stages, cells with fibroblastic morphology died out while the cultures became homogeneously formed by epithelial cells. We extracted RNA at day 14 to check if any plate had initiated the hepatic differentiation program by measuring the expression of early hepatic markers Alb, Foxa1, Foxa2_endo, Tat, Slc10a1 (Ntcp), and Cyp7a1 (Fig. 1b). The combination of hepatic factors and OSKM supplemented with the incubation of the transduced cells in HCM I showed the best induced hepatic-gene expression (Additional file 1: Figure S1B). Concomitantly, we did not detect expression of Nanog and Lin28, confirming that infected cells did not commit to the pluripotent fate (Fig. 1c). It is noteworthy that retroviral combinations that did not include OSK/M became quiescent and did not allow us to obtain frozen cellular stocks. Thus, the presence of reprogramming factors in the cocktail not only seems to improve the expression of early hepatic markers, but also provides the proliferative stimuli necessary for both reprogramming and cell growth [26]. We selected the retroviral cocktail containing OSKM with Gata4, Foxa2, Hnf4a, Hhex, Hnf1a, and Hnf6a for future improvement of our differentiation protocol.

Having successfully initiated the earliest hepatocyte reprogramming of MEF, we added an additional step in order to improve our differentiation protocol (Fig. 1d). Briefly, infected MEF were cultured in DMEMc for 3 days and then cells were switched to HCM I and maintained for 18 days. At this stage, cells could be maintained in culture at low density for several passages in HCM I media without any further differentiation (expansion phase), frozen in liquid nitrogen, or continue to the final differentiation process. The efficiency of conversion to early iHepL was approximately 0.06 %. To induce terminal differentiation, cells (frozen or in culture) were cultured at high confluence in HMC II media for 6 additional days. Early iHepL cells (after 21 days) were highly proliferative in culture (Fig. 1e), probably due to the expression of Myc (Fig. 1f).

iHepL cells displayed a typical epithelial morphology, showing a significant increase in the epithelial markers Cdh1 and Ocln (Fig. 2a and d) and downregulation of the mesenchymal markers Snail1, Zeb1, and Thy1(Fig. 2b).

Our protocol is not favorable for the establishment or maintenance of pluripotency since we do not include LIF in our reprogramming media. However, to exclude the possibility that we carried over cells from colonies in an intermediate reprogramming state, we measured the expression levels of critical genes in the pluripotent transcriptional network. We could not detect expression of Nanog, Lin28, endogenous Oct4, or endogenous Sox2 by qRT-PCR (Fig. 2c and Additional file 1: Figure S2A). Moreover, exogenous Oct4 and Sox2 could not be detected. Absence of Nanog and Ssea1 were confirmed by immunocytochemistry and flow cytometry (Fig. 2d and Additional file 1: Figure S2B).

Next, we examined endogenous hepatic transcription network activation in iHepL cells. The qRT-PCR results showed that the endogenous expression of Foxa1, Foxa2, Foxa3, Cebpa, and Hhex were activated (Fig. 2e). Hnf4a, a core transcription factor involved in the hepatic cross-regulatory network and hepatocyte maturation [27], was endogenously expressed at high levels. iHepL cells significantly expressed genes such as Afp, Slc4a2, MaoB, Fmo1, or Nat2. In contrast, the expression of genes characteristic of mature hepatocytes was not homogeneous. While Cyp7a1, Cyp2c39, or Ugt1a1 were expressed at significant levels compared to primary cultured mouse hepatocytes, other Cyp450 isozymes, Tat, Slc10a1 (Ntcp), Aat1, or Otc were absent or hardly detectable. Hepatic-specific microRNA miR122, highly and specifically expressed during embryonic development of the liver [28], was expressed at hepatocyte-specific levels in iHepL cells.

Immunofluorescence analysis revealed high expression of multiple transcription factors (Hnf4, Hnf6, Hnf1, and Sox9) as well as hepatic enzymes such as Glucokinase (Gck), Glycogen synthase 2 (Gys2), and Albumin (Fig. 3a). The percentage of iHepL cells expressing those proteins as well as Ugt1a1 was between 60 and 90 %, showing a striking homogeneity of the cells (Fig. 3b). In fact, when individual iHepL cells were isolated and expanded, they displayed similar expression patterns (Additional file 1: Figure S3A). Interestingly, an average 99.4 ± 0.6 % of live iHepL cells were GFP-positive (Additional file 1: Figure S3B).

Having successfully proven the expression of the hepatic transcriptional program, we explored the ability of iHepL cells to perform specific hepatocyte tasks. Indocyanine green uptake, a clinically used test for basic liver function [29], was performed by most of the iHepL cells (Fig. 4a). PAS staining of the adherent cultures revealed that most of the cells stored glycogen and depleted it in response to glucagon (Fig. 4a and b). Moreover, when iHepL cells grown in complete media were changed to media without glucose, glycogen stores were also depleted (Fig. 4c). However, glycogen depletion was not followed by a significant secretion of glucose into the media, a role that depends on the activity of glucose-6-phosphatase. Similarly, iHepL cells could not generate urea when cells were incubated with ammonium chloride as a nitrogen source, an activity that relies on an active urea cycle (data not shown).

Hepatic Cyp450 are major enzymes accounting for drug detoxification expressed at high levels only in fully mature postnatal hepatocytes. iHepL cells expressed some phase I and phase II drug metabolic genes (Cyp2c39, Ugt1a1, Fmo1, and Nat2) at the mRNA level (Fig. 2a). To evaluate the acquisition of drug-metabolizing capabilities, we incubated iHepL cells with a cocktail of Cyp450 substrates [23] and monitored the formation of the metabolites by ultra-performance liquid chromatography-tandem mass spectrometry. iHepL cells were able to metabolize phenacetin, bufuralol, diclofenac, midazolam, mephenytoin, and bupropion, but the rate was between 30- and 500-fold lower than 24-h primary cultured mouse hepatocytes (Additional file 1: Figure S4). Remarkably, UDP-glucuronosyl transferase activity in iHepL cells (measured by glucuronidation of β-estradiol) was in the range of primary cultured hepatocytes.

The absence of mature hepatic functions prompted the investigation of whether iHepL cells were, in fact, hepatic progenitors or immature hepatocytes. Hepatic progenitors are bipotential, i.e., they can differentiate towards hepatocytes or cholangiocytes. Previous studies have shown that hepatic progenitors isolated from adult liver (Lgr5-positive [30], Foxl1-positive [15] and M + 133 + 26– cells [31]) or obtained by fibroblast reprogramming (iHepSC [9]) are enriched in duct-specific markers. Analysis by qRT-PCR confirmed that our iHepL cells express high mRNA levels of multiple duct cell markers (Fig. 4d). Moreover, iHepL cells also expressed Cd24a and Foxl1, specific lineage markers of bipotential hepatic progenitors in the adult liver (Fig. 4e). Lgr5 expression was not homogeneous, being only present in specific cells, even in cells expanded from a single cell-colony (Fig. 4f and Additional file 1: Figure S5). We concluded that iHepL cells do not functionally resemble mature hepatocytes, but bipotential hepatic progenitor cells.

In order to test the bipotentiality of iHepL cells in vivo, we performed intrasplenic transplantation into five mice subjected to liver irradiation and partial hepatectomy. Liver irradiation limits the proliferation of resident hepatocytes while hindering the immune response of the acceptor mice [32]. Partial hepatectomy provides the needed mitotic inducer in the donor hepatocytes. In order to track iHepL cells in vivo, we took advantage of maintained GFP expression from the bicistronic pMIGR1 retroviral vector (GFP-positive cells; Fig. 3a and Additional file 1: Figure S3).

All five mice transplanted with iHepL survived and show different degrees of colonization, estimated to be between 5 % and 25 % based on GFP-positivity in serial sections (Additional file 1: Figure S6). Of note, our damage model provides no stimulus for expansion of the cells after engraftment, limiting the degree of colonization. Fluorescence above the general background was detected in big clusters, small clusters, and sometimes 1-3 isolated iHepL cells infiltrating into the surrounding mouse liver parenchyma (Fig. 5a). Contribution of iHepL cells to the bile duct network was limited and, commonly, big clusters of GFP-positive cells surrounding GFP-negative cells forming multiple duct structures could be observed (Fig. 5a). Immunostaining with Hnf4a, Hnf1a, Albumin, Haptoglobin, and Glucokinase antibodies confirmed the differentiation of iHepL cells toward the hepatocyte lineage (Fig. 5b). GFP-positive iHepL-derived cells fully integrated into the E-cadherin network of the liver parenchyma. Binucleation, a hallmark of hepatocyte maturation, was also commonly detected in GFP-positive cells. IHepL-derived cholangiocytes were identified by immunostaining with Sox9 and Ck17-19 antibodies (Fig. 5b). Approximately 2 % of iHepL-derived cholangiocytes were counted. Thus, iHepL cells display characteristics of bipotential hepatic cells and are able to differentiate into hepatocytes or cholangiocytes in vivo. Most of the iHepL cells in recipient mouse livers had stopped proliferation, similar to hepatocytes in wild-type mice [33], as assessed by the expression of the proliferation marker Ki67 and phospho-Histone H3 (Additional file 1: Figure S6 and data not shown).

In hematoxylin-eosin stained sections, no major abnormalities were observed in most of the mice, except for a dysplastic cell mass found in one of the livers. Nuclei within the cell mass were very heterogeneous in size and shape, and cells infiltrated to the liver parenchyma (eosinophilic, orange staining), suggesting a cell malignancy. The presence of dysplasia in bile cholangioles suggest that the origin of the tumor might be the bile duct tree (Fig. 6a). It is noteworthy that the cells within the tumor were GFP-negative (Fig. 6b). Tumor cells were highly proliferative based on phospho-Histone H3 (pH3) immunostaining; however, GFP-positive cells present in the surrounding tissue were not proliferating. Tumor cells were also negative for E-cadherin. In fact, an E-cadherin staining pattern was only maintained in GFP-positive cells found at the interface between normal and tumorigenic tissue. In conclusion, the absence of mature ectodermal, mesodermal, and endodermal structures, the lack of encapsulation, and the high proliferation rate of the cells exclude that tumor cells were a teratoma.

The absence of GFP in the tumor cells did not clarify the cellular origin of the tumor. In order to investigate whether the tumor originated from our iHepL cells, we ran conventional tumorigenic analysis by subcutaneous injection in four mice. After 45 days, all the animals developed a solid dysplastic cell aggregate mainly composed of highly proliferative undifferentiated cells (Fig. 6c). Again, cells were GFP-negative, recapitulating the features of the cell mass found in one of the livers. We could not detect any evidence of mature ectodermal, mesodermal, or endodermal structures (most frequently skin, hair follicles, sebaceous and sweat glands, cartilage, adipose tissue, and glia) characteristic of teratomas typically originated from embryonic stem cell implanted in nude mice. To get a deeper insight into the tumor cell origin, we performed a comparison of transgene copy number in genomic DNA isolated from tumors and iHepL (Fig. 6d). Fifty percent or more of the tumors contain the same copy number for GFP, Hnf6a_exo, Hhex_exo, Oct4_exo, Sox2_exo, and Klf4_exo than iHepL cells (Additional file 1: Table S6). Moreover, iHepL and tumor cells do not contain the transgenes for Oct4 and Sox2 in their genomic DNA, confirming the lack of exogenous expression. These analyses exclude the possibility that partially pluripotent iPSC are the origin of tumors and a possible transition through a transient pluripotent state before committing to the hepatic lineage.

FS was the recipient of a pre-doctoral fellowship from the Ministerio de Ciencia e Innovacion. This work was supported by the Ministerio de Ciencia e Innovacion (grant number SAF2011-29718 and SAF2014-51991) to RB and Fondo de Investigaciones Sanitarias, Instituto de Salud Carlos III (grant number RETICS-RD12/0019/0023) and Fundación Eugenio Rodríguez Pascual to JCS.

FS designed and performed the experiments, and drafted and revised the manuscript. MG-B designed and performed the engraftment and xenograft experiments, and drafted and revised the manuscript. MB carried out the experiments and data analysis and revised the manuscript. JT, JVC, and JCS provided technical support and revised and edited the manuscript. RB conceived of and designed the experiments and data analysis, and revised and edited the manuscript. All authors read and approved the final manuscript.

The authors declare that they have no competing interests.