Dynamic regulation of EZH2 from HPSc to hepatocyte-like cell fate

Currently, drug metabolization and toxicity studies rely on the use of primary human hepatocytes and hepatoma cell lines, which both have conceivable limitations. Human pluripotent stem cell (hPSC)—derived hepatocyte-like cells (HLCs) are an alternative and valuable source of hepatocytes that can overcome these limitations. EZH2 (enhancer of zeste homolog 2), a transcriptional repressor of the polycomb repressive complex 2 (PRC2), may play an important role in hepatocyte development, but its role during in vitro hPSC-HLC differentiation has not yet been assessed. We here demonstrate dynamic regulation of EZH2 during hepatic differentiation of hPSC. To enhance EZH2 expression, we inducibly overexpressed EZH2 between d0 and d8, demonstrating a significant improvement in definitive endoderm formation, and improved generation of HLCs. Despite induction of EZH2 overexpression until d8, EZH2 transcript and protein levels decreased from d4 onwards, which might be caused by expression of microRNAs predicted to inhibit EZH2 expression. In conclusion, our studies demonstrate that EZH2 plays a role in endoderm formation and hepatocyte differentiation, but its expression is tightly post-transcriptionally regulated during this process.


Introduction
Currently, primary human hepatocytes (PHHs) are the gold standard for in vitro drug toxicity and metabolization studies. Use of PHHs is however limited due to scarcity of donors, high inter-donor variability and rapid in vitro dedifferentiation [1]. Human pluripotent stem cells (hPSCs) have the capacity to differentiate into the three somatic germ layers and all cell types of the body, and are an alternative and renewable source of hepatocytes that could be used for drug toxicity and metabolization studies. hPSC-derived hepatocytes have many advantages over primary hepatocytes and hepatocellular carcinoma cell lines, as they could provide an unlimited supply of hepatocytes from a single donor, limiting inter-donor variability; as well as create cells from a diverse number of patients to study mechanisms underlying drug-induced liver injury Immunofluorescence and flow cytometry hESCS and/or differentiated cells were grown on glass slides and fixed with 4% paraformaldehyde (PFA), permeabilized with 0.2% Triton X-100 in PBS, blocked with 5% normal donkey serum (Jackson Laboratory), and stained overnight at 4˚C with OCT4 (0.4 μg/mL, Santa Cruz Sc-8628), TRA-1-60 (1 μg/ml, Millipore-Chemicon MAB4360), SOX17 (5 μg/mL, R&D AF1924), HNF4A (5 μg/mL, Abcam Ab41898), ALB (2.5 μg/mL, Dako A0001) and AAT (3.95 μg/mL, Dako A0012), or the relevant isotype controls in Dako diluent (Dako). Secondary antibodies were used at 1:500 dilution (species-specific AF555-conjugated immunoglobulin G, Alexa Fluor, Molecular Probes) and nuclei were visualized using Hoechst (Sigma-Aldrich). Signals were detected with an Axioimager.Z1 microscope and analyzed with the Axiovision software (Zeiss). The percentage of SOX17 and HNF4α positive cells was manually counted on five representative 10× images. For all pictures, the percentage of positive cells was contoured above the isotype level and three different differentiations were averaged. The histone modification H3K27me3 was stained and quantified as described [24]. For CXCR4/cKIT flow cytometry, cells were detached at day 4 with trypsin 0.05% and stained with 1μg/mL anti-CXCR4-PE and 2μg/ml anti-cKIT-APC antibody for 15 minutes at room temperature. Afterwards cells were washed and analyzed by flow cytometry analysis using a FACS-Canto (BD). For intracellular AAT flow cytometry staining, a single cell suspension was made by liberase treatment (Roche) followed by fixation with 4% PFA. Next, cells were permeabilized with 0.1% saponin and blocked with 10% goat serum (Dako). Afterwards cells were stained with 0.0625μg/ 200μL/10^6 cells anti-AAT antibody (Dako) or a rabbit IgG isotype control (BD Pharmingen) for 1h at RT. A secondary Alexa Fluor 647 antibody (1:1500) (Invitrogen) was used for 30 minutes at RT. Cells were analyzed by flow cytometry analysis using a FACS-Canto (BD) and analyzed with FACS Diva Software (BD Biosciences).

RNA extraction and quantitative reverse-transcription PCR (qRT-PCR)
For gene expression analysis, RNA was isolated from differentiated progeny cells by the GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich) following manufacturer's procedures. Genomic DNA was eliminated using the On-Column DNase I Digestion kit (Sigma-Aldrich). The Superscript III First-Strand synthesis system (Invitrogen) was used for subsequent cDNA synthesis. qPCR was performed with the Platinum SYBR green qPCR supermix-UDG kit (Invitrogen) in a ViiA 7 Real-Time PCR instrument (Thermo Fisher Scientific, Waltham, MA).
Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the housekeeping gene for normalization. Sequences of qRT-PCR primers are listed in Table 1. Relative expression to GAPDH was calculated as 2-ΔCt and relative gene expression as fold change was calculated as 2-ΔΔCt.

microRNA (miRNA) analysis
Total RNA samples, which include RNA from approximately 18 nucleotides (nt), were extracted and purified using miRNeasy Mini Kit (QIAGEN, ID: 217004) accordingly to the manufacturer's procedures.
Then, 1μg of RNA was polyadenylated and reverse transcribed (RT) using NCodeTM miRNA First-Strand cDNA Synthesis and qRT-PCR Kit (Invitrogen, MIRQ-10) with a universal primer. For the quantification miRNAs, real-time PCR for each microRNA assay was carried out in 10μl reaction mixture included 2 μl of diluted RT product, 2× SYBR green, and 0.5 μM forward primer and reverse primer. The reaction was incubated in ViiA 7 Real-Time PCR instrument (Thermo Fisher Scientific, Waltham, MA) in 384-well plates at 95˚C for 10 min, followed by 45 cycles of 95˚C for 15 sec and 60˚C for 30 sec. U6 was used as the housekeeping gene for normalization. Sequences of miRNA primers are listed in Table 2. Relative expression to U6 was calculated as 2-ΔCt and relative gene expression as fold change was calculated as 2-ΔΔCt.

Albumin ELISA
Enzyme-linked immunosorbent assay (ELISA) for ALBUMIN was performed according to the manufacturer's procedure (Bethyl, Montgomery, TX). Briefly, at day 16 of the hepatocyte differentiation protocol, supernatant was collected and incubated as described in [5].

Gene
Forward Reverse  [28]. Regular PCR was used to amplify specific fragments using 10-20 ng of converted DNA using forward 5'-TTGTATTT-TAGTTTGGATAATTAGAG-3'and reverse 5'-AAAAACCAAATTTAAACCAATTCAA-3' primers. The PCR products were cloned into a p-GEM easy vector (Promega) and 15-20 clones were sequenced. The methylation rate of the CpG pairs was quantified using QUMA software.

The expression of EZH2 decreases during hepatocyte differentiation
We first evaluated the transcript levels of the polycomb group (PcG) genes EZH2, EZH1, SUZ12, EED, RING1B and BMI1 in hPSCs, definitive endoderm (endo_d4), hepatic progenitors (HP_d8) and fetal hepatocytes (FH_d16) derived from hPSCs ( Fig 1A, CTCCCTGAAAGGTTCAGTAAA AAGAAGTCCTCCAAAGCG https://doi.org/10.1371/journal.pone.0186884.t001 Table 2. miRNAs expression primers.  Fig 1B). However, western blot analysis for EZH2 protein during hepatocyte differentiation demonstrated an increase in EZH2 between day 0 and day 4 of differentiation, and a decrease of EZH2 from day 8 to day 16 of differentiation ( Fig 1B). The low levels of EZH2 protein in undifferentiated hPSCs were not in accordance with the transcript levels detected by RT-qPCR ( Fig 1A, bottom part). Thus, during HLCs generation from PSCs, there appears to be a dynamic regulation of EZH2 protein.

Mature name Primer
To investigate whether epigenetic modifications modulate EZH2 gene expression during hPSC-HLCs differentiation, we studied the role of DNA methylation of the regulatory region upstream of the transcription start site (TSS) of EZH2, containing 33 CpG sites in a CpG island ( Fig 1C, upper part), in hPSCs, HP (day 8) and FH (day 16). Bisulfite sequencing demonstrated DNA hypomethylation of this region at all 3 steps of differentiation analyzed (Fig 1C, bottom  part). These results indicate that expression levels of EZH2 are not regulated by DNA methylation in the region of the TSS of EZH2, but suggest that EZH2 mRNA could be regulated by other epigenetic modifications, as the protein levels are most likely regulated by posttranslational modifications.

Definitive endoderm (DE) formation is improved by EZH2 overexpression
We recently described an efficient and very fast method to introduce doxycycline (doxy) inducible transgenes in hPSCs, by recombinase mediated cassette exchange (RMCE) in a flippase recognition target (FRT) flanked cassette in the adeno-associated virus integration site-1 (AAVS1) locus of hPSCs [26]. This allowed us to study the effect of inducible EZH2 overexpression on the differentiation potential of hPSCs towards HLCs. We inserted the amplified, human EZH2 transcript variant 1 (hEZH2 tv1, NM_004456) sequence into the "allin-one" inducible RMCE donor vector that contains the tetracycline response element (TRE) driving the expression of the transgene in reverse orientation to the CAGGS m2rtTA cassette (Fig 2A) and flanked by FRT sites. The cassette was introduced by flippase-mediated cassette exchange in the AAVS1 as described earlier [26]. The ihEZH2 inducible cell line (hPS-iEZH2) maintained the typical pluripotency characteristics of hPSCs as the expression of pluripotency genes OCT4 and TRA1-60 was comparable to wild-type hPSCs (Fig 2B).
To further investigate the regulatory role of EZH2 during hPSC differentiation into definitive endoderm (endo_d4) we induced the expression of EZH2 with doxy from day 0 to day 4 in hepatocytes differentiation protocol (Fig 1A, upper part). Both western blot ( Fig 2C) and RT-qPCR (Figs 3A, bottom part and 5A) analysis confirmed that addition of 5μg/ml doxy to the differentiation medium from day 0 to day 4 significantly induced the expression of EZH2 compared to the untreated control. In response to the higher levels of EZH2, a significantly more homogenous population of definitive endoderm cells was generated on day 4, demonstrated by flow cytometry analysis for CXCR4/cKIT double positive cells (75.2±0.91% in the presence of doxy and 53.9±3.29% in the absence of doxy, n = 3, p = 0.0172) (Fig 2D). This coincided with a significant increase in SOX17 positive cells (57.94±4.31% with doxy vs. 27.64 ±4.32% without doxy, n = 2, p = 0.00009), demonstrated by immunostaining ( Fig 2E) and also by RT-qPCR ( Fig 2F). In addition, transcripts for the definitive endoderm markers, FOXA2, EOMES, MIXL1, GOOSECOID and CXCR4, were up regulated in doxy-treated (Fig 2G) compared to untreated control cells. As expected, 8 days post differentiation (HP_d8), expression of the definitive endoderm markers FOXA2, EOMES, MIXL1, GOOSECOID and CXCR4 decreased, and no differences were seen between doxy-treated or untreated cells (Fig 2G). RT-qPCR also demonstrated that the key pluripotency markers OCT4 and NANOG decreased over-time (S2 Fig). In conclusion, EZH2 induction significantly improved definitive endoderm formation.

Regulation of EZH2 overexpression during hepatocyte differentiation
To address if induction of EZH2 also improved the generation of day 16 fetal hepatocytes from hPSCs, we differentiated hPS-iEZH2 with or without 5μg/ml doxycycline (doxy) from the start of differentiation until day 8 (Fig 3A, upper part). EZH2 was significantly induced in hPSCprogeny on day 4. Despite continued addition of doxy until day 8, EZH2 mRNA levels were almost zero beyond day 4, as analyzed by RT-qPCR with primers selective for the EZH2 variant 1 transgene transcript (Exogenous EZH2, Fig 3A, left) and primers that recognized both endogenous and transgenic EZH2 mRNA (Total EZH2, Fig 3A right). Expression of EZH1 mRNA did not change throughout differentiation and was not affected by doxy treatment (S2 Fig). Western blot confirmed that EZH2 protein levels were higher in doxy treated compared with control cells on day 4, but were no longer higher on day 8, and were not detectable on day 16 ( Fig 3B). Thus, despite addition of doxy between day 0 and day 8, EZH2 transcript and protein levels decreased progressively from day 4 onwards, suggesting that post-transcriptional regulation of EZH2 during differentiation might be responsible for EZH2 mRNA downregulation.
We also examined the effect of EZH2 overexpression on the differentiation of hPSCs towards fetal hepatocytes on day 16. Although EZH2 transcripts and protein were only enhanced during the initial 4 days of differentiation, we observed a significant increased expression of HNF3B and GSTp on day 8, and a significant increase in HNF1A, HNF3B, CEBPA, and CAR transcripts on day 16 (Fig 3C). We also found significantly increased transcript levels for ALB and AAT, two genes expressed in mature hepatocytes (Fig 3D). AFP, a typical hepatoblast/fetal hepatocyte gene, is continuously expressed as expected for the fetal nature of the hPSC-derived hepatocytes among samples (Fig 3D). We also monitored the expression of some additional fetal markers (CYP3A7, CYP1A2 and CKIT) and mature hepatocyte genes on day 8 and day 16 (CYP2A6, CYP2C9, CEBPB, GSTA1, APOA1, NTCP, SRBI, P300, MRP2, PROX1, PEPCK, PXR and HNF3G). Only the levels of CYP3A7, CYP2A6 and CYP2C9 were significantly influenced by overexpression of EZH2 between day 0 and day 4 (S3 Fig). On HP_d8, levels of HNF4A mRNA were not affected by addition of doxy (Fig 3E, left), but immunostaining demonstrated presence of significantly more HNF4A positive cells in the doxy treated cells (83.5±3.37% with doxy vs. 53.3±7.58% without doxy, n = 3, p = 0.0409) ( Fig  3E, right). On HP_d8 and FH_d16, immunostaining (Fig 3F and 3H) suggested that more doxy-treated progeny stained positive for ALBUMIN and AAT (n = 2). Consistently, ALBU-MIN secretion by doxy treated progeny was significantly increased on day 16 (Fig 3G), and flow cytometry analysis demonstrated that significantly more doxy treated cells stained positive for AAT (60.9 ±8.51% with doxy vs. 27.7 ±2.31 without doxy, n = 4, p = 0.015).
As EZH2 also plays a role in pancreas versus liver fate choice from endoderm [29,30], we also assessed the expression of PDX1, NGN3, NEUROD1, NKX6.1 and PAX6 during differentiation. We did not observe increased levels of these transcripts (S3 Fig) on day 8 and day 16 hPSC progeny in the presence of doxy compared with cultures without doxy, indicating that the improved hepatocyte specification as a result of EZH2 induction was specific.
Thus, enhanced expression of EZH2 between day 0 and day 4 significantly increased ALBUMIN secretion and significantly increased AAT expressing cells on d16 of differentiation, while not affecting pancreatic fate commitment. Interestingly, improved hepatic commitment on day 16 occurred despite the fact that EZH2 transcript and protein levels were only significantly elevated between day 0 and day 4.
Upon doxy-mediated induction of EZH2, H3K27me3 was significantly increased on day 4, day 8 and d16 hPSC progeny cells as was shown in Fig 4A (by immunofluorescence, left, and by densitometric analysis, right), while H3K27me3 significantly decreased on the endodermal markers (Fig 4B, left). The level of H3K27me3 on GAPDH, MYOD1 and HOXD11 remained unchanged between treated and doxy-treated cells on day 4 (Fig 4B, right). Moreover, a significant reduction of H3K27me3 on ALBUMIN (Fig 4C, left) and an increase on HOXD11 on day 16 (Fig 4C, right), a typical EZH2 target gene, suggested that EZH2 induction probably leads to the repression of not-hepatocyte genes while inducing endodermal and hepatocyte genes.  We have previously shown that inducible transgenes remained stably expressed throughout the differentiation when induction started from day 0 [26]. To further understand the kinetics of the decreased EZH2 expression from day 4 onwards despite constant administration of doxy until day 8 (Fig 5A, right), we monitored EZH2 transcript levels daily between day 4 and day 8. We demonstrated that EZH2 mRNA levels in doxy-treated cells gradually decreased between day 4 and day 8 (Fig 5A, left) while levels in untreated cells remained constant, suggesting some kind of post-transcriptional regulation mechanism that limited the total amount of EZH2. We therefore examined the levels of miRNAs known/predicted to have a link to EZH2 (see Fig 5C). miR-101 has been identified to directly target EZH2, acting as an EZH2 silencer involved in a negative feedback circuit with EZH2 [31,32]. In addition, miR-138 was recently reported to directly target EZH2 [33] and miR-214 has been shown to directly target EZH2 during myogenesis [34]. To identify additional candidate miRNAs that might target EZH2, we used data from published studies [35,36] and examined the online database miRDB (http://mirdb.org/miRDB/) for miRNA target prediction (S4 Fig). We monitored expression of 12 know/predicted miRNAs targeting EZH2. As shown in Fig 5B and S4B Fig, expression levels of all miRNAs analyzed were lower in day 4 hPSCs progeny (±doxy) compared with undifferentiated hPSCs. Transcript levels of miR-101, miR-138, miR-214 and miR-124 were significantly higher in day 16 progeny compared to day 8 progeny, which was more pronounced (albeit not significant) for cultures supplemented with doxy ( Fig 5B). The other candidate EZH2 binding miRNAs displayed a very similar pattern of expression: for example, the expression kinetics of miR-139 was similar to that of miR-101, and the expression pattern of miR-31 and miR-200b was comparable to that of miR-138 throughout differentiation (S4 Fig). miR-98, miR-125 and 181a were detectable only on day 16, and levels were similar between doxy treated and untreated cells. miR-217 was the only EZH2 binding miRNA that could not be detected at any point during differentiation (S4 Fig).
These studies demonstrated that miRNAs known/predicted to interact with EZH2 are highly up regulated during HLC differentiation and might explain a possible mechanisms for the loss of EZH2 mRNA and EZH2 protein expression from day 4 onwards, despite the transgenic overexpression of EZH2 until day 8 of differentiation.

Discussion
hPSCs can differentiate into all somatic cell types of the human body, including hepatocytes. Thus, hPSC-derived hepatocytes are an attractive alternative to PHHs to test the safety, efficacy, and metabolization of new chemical entities. To date, most hPSC differentiation protocols yield hepatocyte-like cells with phenotypic characteristics of fetal rather than mature hepatocytes [4,9,10,14,18,24,37,38].
It is well known that regulation of gene expression is modulated by epigenetic modification caused by DNA methylation, histone modifications and miRNAs. DNA and chromatin modulation in both non-coding and coding regions, tightly regulate gene expression in hPSCs and staining with DAPI (blue). On the right, more than 80% of HNF4A positive cells on HP_d8 EZH2 doxy induced cells were counted. Data as mean ± SEM of n ! 3 IEs. * p < 0.05. F. Immunofluorescence staining for ALBUMIN (red signal) at HP (day8) and FH (d16) in untreated (-) and EZH2 induced cells (+) (left part). Nuclei are staining with DAPI (blue). Data as representative images of n = 2 IES. G. ELISA for ALBUMIN secretion on FH_d16 of the hepatocyte differentiation protocol. EZH2 doxy induced cells (+) secrete significant amounts of albumin compared to untreated (-) cells. Data as mean ± SEM of n ! 3 IEs. ** p < 0.01. H. Immunofluorescence staining for AAT (red signal) at HP (day 8) and FH (day 16) showed abundant expression of the hepatocyte protein in doxy-treated cells compared to the untreated (left part). Nuclei are stained with DAPI (blue). Data as representative images of n = 2 IES. I. Intracellular flow cytometry analysis for AAT demonstrated that more then around 60% of EZH2 doxy induced cells (+) progeny were positive for AAT. Results represent the mean of three independent experiments ± SEM. ** p < 0.01. their progeny. Polycomb complexes in part modulate chromatin structure during lineage commitment [39,40]. EZH2, a core component of PRC2, represses the expression of genes by functioning as a methyltransferase for H3K27me3 [41][42][43]. We previously published that persistent H3K27me3 marking of regulatory gene regions of hepatocyte genes is detected during hepatocyte differentiation from hPSCs, and this despite increased gene transcription. H3K27me3 marking might therefore not have a determinant role in gene activity in later stages of hepatocyte differentiation [24]. It might be possible that hPSC display a distinct histone modification state on regulatory elements at initial stages, a sort of chromatin "pre-pattern" that may reflect  To unravel to role of EZH2 during hepatocyte commitment, we assessed EZH2 transcript and protein expression during hepatocyte differentiation from hPSCs. EZH2 protein levels increased significantly on day 4 and day 8, but decreased to nearly zero on day 16 of differentiation. To determine if differentiation would be enhanced if levels of EZH2 were further induced during the initial stages of differentiation, we inducibly overexpressed EZH2, by incorporating a doxy inducible EZH2 cassette by RMCE in the AAVS1 locus of hPSCs created in our lab [26]. When EZH2 expression was induced between day 0 and day 8 of the hepatocyte differentiation protocol, we demonstrated not only a more homogenous endoderm population on day 4, but also improved hepatoblast maturation on day 16. However, we demonstrated that doxy-mediated induction of EZH2 is associated with a significant increase in overall H3K27me3 of hPSC progeny on day 4, day 8 and day 16. In line with our previous study, high levels of H3K27me3 staining persisted until day 16, and this despite the further commitment of endoderm cells to hepatoblasts [24]. Interestingly, we demonstrated that ALBUMIN and AAT hepatocyte marker genes showed less H3K27me3 at the end of the differentiation. We hypothesize that EZH2 overexpression might induce not-hepatocyte specific genes to retain H3K27me3 in favor of the expression of hepatocyte specific genes expression.
The decrease in EZH2 expression from day 4 onwards despite continuous administration of doxy till day 8 was surprising. We previously described that inducible transgene expression from AAVS1 was inhibited during hepatic differentiation through an unknown mechanism compatible with TRE-silencing triggered by TRE inactivity. However, inducible expression was robust through the whole differentiation process when doxy was applied starting from day-2 or day 4 [26]. As expression of EZH2 was induced on day 0, the AAVS1-mediated inhibition is highly unlikely to explain the down-regulation of EZH2 from day 4 onwards. This loss, therefore suggested post-transcriptional modulation of EZH2 expression from day 4 onwards in the presence of doxycycline.
miRNAs are small non-coding RNAs that regulate the expression of more than 60% of protein coding genes in the human genome [44] by silencing target genes either via binding in a sequence specific manner messenger RNAs and cleaving them, or by inhibiting their translation [45]. miRNAs are known to regulate lineage-specific differentiation [46][47][48][49][50][51]. Expression of EZH2 can be modulated by direct binding of miRNAs, including miR-101, miR-138 and miR-214, to EZH2 [33,34,52]. In line with this, we observed that that these miRNAs as well as an additional 8 miRNAs predicted to bind EZH2, were significantly induced from day 4 onwards, which coincides with the progressive decrease of EZH2 mRNA and protein. Therefore, the loss of EZH2 from day 4 onwards despite transgenic overexpression until day 8, is likely due to miRNA mediated degradation, although further studies will be needed to demonstrate if miRNAs are the only responsible for this observation.

Conclusion
In conclusion, we demonstrate that overexpression of EZH2 early during the differentiation of hPSCs to hepatoblasts improved definitive endoderm formation and subsequent HLC generation. Surprisingly, despite doxy-mediated overexpression of EZH2 until day 8 of differentiation, transcript and protein levels of EHZ2 decreased precipitously from day 4 onwards. This was concomitant with increased levels of miRNAs known/predicted to inhibit EZH2 expression, which might be responsible for post-transcriptional regulation of EZH2 from day 4 onwards. In addition, despite the loss of EZH2 expression, overall H3K27me3 levels remained high until day 16 of differentiation. In conclusion, we demonstrate that EZH2, of which the expression is tightly post-transcriptionally regulated, has a role in endoderm formation and enhanced levels of the Polycomb gene leads to improved hepatocyte differentiation.